Optimizing fNIRS Spatial Resolution: A Comprehensive Guide to Advanced Optode Placement Strategies

Abstract

This article provides a systematic framework for optimizing functional near-infrared spectroscopy (fNIRS) optode placement to enhance spatial resolution and data quality. Targeting researchers and drug development professionals, we explore foundational principles of light transport and neurovascular coupling, detail methodological advances from toolbox-guided to MRI-informed personalized montages, and address key troubleshooting challenges in signal quality and reproducibility. The content synthesizes current evidence on validation protocols and comparative performance of different placement strategies, offering practical guidance for improving the reliability and precision of fNIRS measurements in both research and clinical applications.

Understanding the Core Principles of fNIRS Light Transport and Spatial Specificity

Troubleshooting Guides

Guide 1: Poor Spatial Specificity and Inaccurate Brain Region Targeting

Problem: Your fNIRS signals lack anatomical precision, making it unclear which brain regions are actually being measured. This is often caused by using standard cap placements that don't account for individual anatomical differences [1] [2].

• Check: Verify optode positioning relative to individual scalp and cranial landmarks rather than relying solely on standard EEG positions [2].
• Solution: Implement subject-specific optode placement using neuroimaging guidance. Probabilistic approaches using anatomical MRI with fMRI maps from independent datasets can significantly improve targeting without requiring individual fMRI scans [2].
• Advanced Solution: For critical applications, use individual fMRI data to guide optode placement targeting functionally defined regions rather than just anatomically defined areas [2].

Guide 2: Low Signal-to-Noise Ratio and Poor Signal Quality

Problem: Signals are contaminated by physiological noise or motion artifacts, particularly problematic for real-time applications like brain-computer interfaces [1].

• Check: Ensure proper optode-scalp coupling and monitor for motion artifacts during data collection [1].
• Solution: Incorporate short-separation channels (typically 8mm) to regress out systemic physiological noise from superficial tissues [3].
• Advanced Solution: Implement high-density arrays with overlapping, multidistance channels to improve depth sensitivity and signal specificity [3].

Guide 3: Inconsistent Results Across Repeated Measurements

Problem: Findings cannot be reliably reproduced across multiple sessions with the same subject, limiting research validity [4].

• Check: Document and minimize cap placement shifts between sessions using digitized optode positions [4].
• Solution: Use source localization techniques with anatomically specific head models rather than relying solely on channel-level data [4].
• Advanced Solution: Focus on oxygenated hemoglobin (HbO) measurements, which demonstrate higher reproducibility across sessions compared to deoxygenated hemoglobin (HbR) [4].

Frequently Asked Questions (FAQs)

Q1: What is the fundamental advantage of high-density fNIRS arrays over traditional sparse arrays?

High-density (HD) arrays with overlapping, multidistance channels provide superior spatial resolution, better depth sensitivity, and improved localization accuracy compared to traditional sparse arrays with standard 30mm spacing [3]. HD arrays particularly excel at detecting and localizing brain activity during lower cognitive load tasks where sparse arrays may fail, and they demonstrate significantly better inter-subject consistency in localization [3].

Q2: How much can individual anatomical variability affect fNIRS signal quality?

Individual anatomical differences significantly impact fNIRS signal quality and sensitivity. Studies show that approaches incorporating individual anatomical information (probabilistic methods with MRI data) consistently outperform literature-based placement approaches that assume standardized anatomy [2]. Vascular structures, which are highly variable between individuals, particularly influence light sensitivity profiles and signal quality [2].

Q3: What practical methods can improve optode placement without requiring expensive MRI scans for every subject?

The probabilistic approach (PROB) provides an excellent balance between practicality and performance. This method uses individual anatomical data (which could be obtained from atlases) combined with probabilistic fMRI activation maps from independent datasets [2]. Research demonstrates this approach performs nearly as well as methods requiring individual fMRI data while being more practical and cost-effective [2].

Q4: How does optode placement reproducibility affect measurement consistency across sessions?

Even minor shifts in optode positioning (≥1cm) between sessions significantly reduce spatial overlap and measurement consistency [4]. Increased cap placement shifts correlate strongly with decreased reproducibility, highlighting the critical importance of consistent, precise optode placement across repeated measurements [4].

Table 1: Comparison of fNIRS Array Configurations and Performance Characteristics

Array Type	Spatial Resolution	Depth Sensitivity	Localization Accuracy	Setup Complexity	Optimal Use Cases
Sparse Arrays (30mm spacing)	Limited [3]	Poor without short-separation channels [3]	Low; may average signals from multiple regions [3]	Low; faster setup [3]	Detecting presence of activation during high cognitive load tasks [3]
High-Density Arrays (Multidistance, overlapping)	High [3]	Excellent with proper channel combinations [3]	Superior; can differentiate nearby regions [3]	High; longer setup time [3]	Precise localization, especially for lower cognitive load tasks [3]
Short-Separation Enhanced	Moderate	Improved with superficial signal regression [3]	Moderate	Moderate	Applications requiring noise reduction without full HD complexity [3]

Table 2: Performance Comparison of Optode Placement Guidance Approaches

Guidance Approach	Anatomical Specificity	Functional Precision	Practical Implementation	Relative Performance
Literature-Based (LIT)	Low	Low	High (minimal requirements)	Baseline/reference [2]
Probabilistic (PROB)	High (using anatomical data)	Moderate (group fMRI maps)	Moderate	Significantly outperforms LIT [2]
Individual fMRI (iFMRI)	High	High	Low (requires individual fMRI)	Similar to PROB and fVASC [2]
Full Vascular (fVASC)	Highest (includes vasculature)	High	Lowest (multiple scans needed)	Similar to PROB and iFMRI [2]

Experimental Protocols

Protocol 1: Subject-Specific Optode Placement Using Probabilistic Approach

This methodology optimizes optode placement using anatomical data and probabilistic functional maps, balancing performance and practicality [2].

Materials Required:

• Individual anatomical MRI data or appropriate head atlas
• Probabilistic fMRI activation maps from independent datasets
• fNIRS optode configuration software (e.g., AtlasViewer, fOLD toolbox)
• Neuronavigation system (for precise placement)

Procedure:

• Data Acquisition: Obtain individual anatomical MRI scans or select appropriate head atlas based on participant demographics.
• ROI Definition: Identify target regions using probabilistic fMRI activation maps from independent studies performing similar tasks.
• Sensitivity Modeling: Conduct photon migration simulations using Monte Carlo methods to model light sensitivity profiles [5].
• Layout Optimization: Use optimization algorithms to determine optode positions that maximize sensitivity to target ROIs while maintaining practical constraints (typical source-detector distance: 25-40mm) [2].
• Placement Verification: Use neuronavigation to ensure accurate translation of optimized positions to physical optode placement on the scalp.

Protocol 2: Performance Comparison Between Sparse and High-Density Arrays

This protocol directly evaluates the benefits of HD arrays for specific research applications [3].

Materials Required:

• fNIRS system capable of both sparse and high-density configurations
• Custom cap design allowing both array types
• Task paradigm with varying cognitive loads (e.g., Stroop task)
• Analysis pipeline for both channel-space and image-space reconstruction

Procedure:

• Probe Design: Create matched field-of-view sparse and HD arrays. Sparse arrays should follow standard 30mm grid patterns, while HD arrays should implement overlapping, multidistance channels [3].
• Experimental Design: Implement tasks with varying cognitive demands (e.g., congruent vs. incongruent Stroop conditions) [3].
• Data Collection: Collect data from the same participants using both array types, counterbalancing order across participants.
• Signal Processing: Apply identical preprocessing including short-separation regression, filtering, and motion artifact correction [3].
• Image Reconstruction: Perform image reconstruction for HD-DOT data to compare localization accuracy [3].
• Statistical Comparison: Quantitatively compare activation strength, localization precision, and inter-subject consistency between array types.

Signaling Pathways and Workflows

Optode Placement Optimization Workflow

fNIRS Probe Design Decision Framework

Research Reagent Solutions

Table 3: Essential Tools and Software for Optimal Optode Placement

Tool Category	Specific Examples	Primary Function	Implementation Considerations
Placement Optimization Software	fOLD Toolbox [5], AtlasViewer [6]	Automates optode position decision based on ROI sensitivity profiles	fOLD uses photon transport simulations on head atlases; AtlasViewer enables probe design visualization [5] [6]
Neuronavigation Systems	Commercial neuronavigation platforms	Precisely translates virtual optode positions to physical scalp locations	Critical for implementing subject-specific layouts; requires training for proper operation [2]
Head Modeling Resources	Colin27 Atlas [5], SPM12 Tissue Probability Maps [5]	Provides anatomical templates for photon migration simulations	SPM12 template based on 549 subjects offers population-representative modeling [5]
Monte Carlo Simulation Tools	Monte Carlo eXtreme (MCX) [5]	Models photon transport through head tissues for sensitivity profiles	Computationally intensive; benefits from GPU acceleration [5]
Multimodal Integration Tools	Custom co-localization designs [6]	Enables combined fNIRS-EEG measurements with minimal compromise	Allows electrodes and optodes to share positions; reduces coverage tradeoffs [6]

## FAQs on Fundamental Principles

How does light transport in tissue affect fNIRS measurement sensitivity? When near-infrared light travels from a source to a detector on the scalp, it is scattered and absorbed by different tissue layers. The resulting measurement sensitivity at a specific brain location depends on the photon fluence from the source and the detector, which forms a spatial sensitivity profile [5]. Deeper brain structures typically show lower sensitivity as fewer photons reach and return from these depths.

What is the relationship between source-detector distance and penetration depth? Increasing the source-detector distance generally increases penetration depth, but with a trade-off. Longer distances (typically 30-45 mm) allow light to sample deeper cortical gray matter, but the signal strength diminishes significantly. Shorter distances (e.g., 8-15 mm) are predominantly sensitive to systemic physiological noise in the scalp and skull, which is why they are used as reference channels to clean data from long-distance channels [3] [7].

Why is a multi-distance probe configuration beneficial? A multi-distance configuration uses channels of varying lengths (e.g., 28.2 mm, 40 mm, and 44.7 mm) simultaneously [7]. This design improves spatial resolution in both depth and lateral dimensions. Shorter channels characterize superficial signals, while longer channels probe cerebral tissue. Combining them allows for better separation of brain activity from extracerebral contamination and provides a more accurate tomographic image of cortical activation [7].

## Troubleshooting Guides

Problem: Poor Signal-to-Noise Ratio (SNR)

A weak optical signal can make it difficult to distinguish brain activity from noise.

• Potential Solution: Verify Optode Contact and Distance

  • Action: Ensure all optodes have firm, consistent contact with the scalp. Verify that source-detector distances are within the optimal range (typically 25-45 mm for adult cortical measurements) [7]. Distances that are too short will not probe the brain, while those that are too long will result in a very weak signal.
  • Check: Use a photodetector to measure the intensity of light received at each detector position during setup. Excessively low values indicate poor contact or excessive distance.

• Potential Solution: Implement Short-Separation Regression

  • Action: Incorporate short-separation channels (e.g., 8-15 mm) into your probe layout. Use the signal from these channels as a regressor in your analysis to remove systemic physiological noise (e.g., from scalp blood flow) from the standard long-distance channels [3].
  • Check: The correlation between the short-separation channel and the long-channel signals should be significantly reduced after regression, indicating successful removal of superficial noise.

Problem: Inaccurate or Inconsistent Spatial Localization

Measurements do not align with the expected brain region, or results vary greatly between subjects with the same cap placement.

• Potential Solution: Use Anatomical Guidance for Probe Placement

  • Action: Move away from manual cap placement. Use a transcranial brain atlas and a navigation system (e.g., using a portable digitizer) to position optodes precisely on each individual subject's scalp based on their anatomical landmarks [8]. Software tools like AtlasViewer and the fOLD toolbox can convert desired brain regions-of-interest into optimal optode positions [5] [9].
  • Check: After data collection, project the measured channel locations onto an individual or standard brain atlas (e.g., Colin27) to confirm the brain regions each channel is actually sampling [9].

• Potential Solution: Adopt a High-Density (HD) Array

  • Action: If your research question requires precise localization, consider using a high-density array with overlapping, multi-distance channels. While it requires more optodes and complex setup, HD-DOT provides superior spatial resolution and localization accuracy compared to traditional sparse arrays [3].
  • Check: Compare the reconstructed activation images from a sparse array and an HD array for the same task. The HD array should show more focal and anatomically plausible activation patterns, particularly for tasks with lower cognitive load [3].

Problem: Data Cannot Be Compared Across Subjects or with fMRI

fNIRS results are difficult to interpret in a standard brain space or relate to the broader neuroimaging literature.

• Potential Solution: Register Data to a Standard Brain Space
  • Action: Transform your fNIRS data into a standard coordinate system like Montreal Neurological Institute (MNI) space. This involves co-registering the measured optode positions with a head model (either from individual MRI or a standard atlas like Colin27) [10] [9]. Tools like AtlasViewer automate this process.
  • Check: Ensure that the fNIRS activation maps you generate are overlaid on a standard brain template, allowing for direct comparison with fMRI or PET findings [9].

## Quantitative Data for Probe Design

The tables below summarize key quantitative relationships to guide your experimental design.

Table 1: Impact of Source-Detector Distance on fNIRS Measurements

Source-Detector Distance	Primary Sensitivity	Key Advantages	Key Limitations
Short (e.g., 10-15 mm)	Scalp and skull layers [7]	Essential for superficial noise regression [3]	Insensitive to brain activity
Standard (e.g., 30 mm)	Superficial cortex (gray matter)	Good balance of signal strength and brain sensitivity	Limited depth resolution; poor differentiation of adjacent regions [3]
Long (e.g., 40-45 mm)	Deeper cortical layers	Increased sensitivity to a larger brain volume	Very weak signal strength; lower SNR [7]

Table 2: Comparison of Sparse vs. High-Density (HD) fNIRS Arrays

Characteristic	Sparse Array (30 mm grid)	High-Density (HD) Array
Typical Layout	Non-overlapping, grid pattern	Overlapping, multi-distance, often hexagonal pattern [3]
Spatial Resolution	Limited, coarse [3]	Improved, can localize within a cortical gyrus [9]
Spatial Localization	Poor, prone to averaging signals from multiple regions [3]	Superior, accurately localizes functional activity [3]
Depth Sensitivity	Limited without short separations	Improved with multiple distance channels [3]
Setup Complexity & Cost	Lower	Higher (more optodes, longer setup) [3]
Best Suited For	Detecting presence/absence of activation in a broad area [3]	Studies requiring precise localization and tomographic imaging [3]

## Experimental Protocols

Protocol 1: Validating Probe Sensitivity Using Photon Migration Simulation

This protocol uses the fNIRS Optodes' Location Decider (fOLD) toolbox to determine the optimal probe arrangement for targeting specific brain regions before an experiment [5].

• Define Regions-of-Interest (ROIs): Identify the brain areas you wish to target based on your experimental hypothesis (e.g., dorsolateral prefrontal cortex for a working memory task).
• Select a Head Atlas: Choose a digital head model (e.g., the Colin27 atlas or the SPM12 atlas based on 549 subjects) [5].
• Run Monte Carlo Simulations: Use software like Monte Carlo eXtreme (MCX) to simulate photon transport from potential source and detector positions on the scalp. This calculates the normalized sensitivity profile for each possible channel, showing which brain voxels it samples [5].
• Calculate Brain Sensitivity: For each channel, sum the normalized sensitivity across all voxels classified as gray and white matter. This gives a brain sensitivity metric [5].
• Optimize Probe Design: The fOLD toolbox automatically selects the optode locations from a set of predefined positions (e.g., based on the 10-10 system) that maximize the anatomical specificity to your predefined ROIs [5].

Protocol 2: Group-Level Analysis in Sensor Space

This protocol ensures that data from multiple subjects, with inherent variability in head shape and optode placement, can be accurately combined for a group-level inference [10].

• First-Level (Within-Subject) Analysis: For each subject, use a General Linear Model (GLM) to fit the fNIRS data (Δ[HbO] and Δ[HbR]) and generate a statistical parametric map (e.g., t-values) for each channel.
• Spatial Interpolation to a Canonical Scalp Surface: To address misalignment of channel locations between subjects, interpolate the channel-specific contrast values onto a standardized 2D or 3D representation of the scalp surface [10].
• Second-Level (Between-Subject) Analysis: Apply a random-effects analysis (e.g., a one-sample t-test) to the interpolated contrast images from all subjects. This allows you to make inferences about the population from which your subjects were drawn [10].
• Statistical Inference: Assess the significance of regional effects using Random Field Theory (RFT) to control for multiple comparisons across the sensor space [10].

## Essential Visualizations

Diagram 1: Photon migration and sensitivity.

Diagram 2: Probe optimization workflow.

## The Scientist's Toolkit

Table 3: Essential Research Reagents & Computational Tools

Item / Solution	Function / Purpose
Monte Carlo eXtreme (MCX)	Software for simulating photon transport in tissue. It models the scattering and absorption of light to generate sensitivity profiles for optode pairs [5].
fOLD Toolbox	A toolbox that automatically decides optode locations to maximize anatomical specificity to pre-defined brain regions-of-interest, based on MCX simulations [5].
AtlasViewer	Open-source software for designing fNIRS probes, co-registering them with head atlases (e.g., Colin27), and visualizing sensitivity profiles and reconstructed images [9].
Colin27 & SPM12 Head Atlases	Digital, segmented models of the human head (scalp, skull, CSF, gray/white matter). Used as anatomical priors for photon migration simulations when individual MRI is unavailable [5].
Short-Separation Channels	Reference channels with a small source-detector distance (e.g., 8 mm). Their signal is used to regress out the confounding hemodynamics from superficial tissues, improving the quality of brain signals [3].

Frequently Asked Questions (FAQs)

FAQ 1: Why is accurate tissue segmentation critical for fNIRS optode placement? Accurate tissue segmentation is fundamental because the different tissues of the human head (scalp, skull, cerebrospinal fluid - CSF, gray matter, and white matter) possess distinct optical properties, primarily their absorption and scattering coefficients [5]. When designing an optode layout, researchers use computational models of photon migration to predict the sensitivity profile of each source-detector channel to the underlying brain cortex. These simulations rely on an anatomically accurate head model. Incorrect tissue segmentation can lead to flawed sensitivity profiles, resulting in optode placements that do not optimally target the intended cortical regions-of-interest, thereby compromising the spatial resolution and anatomical specificity of the fNIRS measurements [5] [2] [11].

FAQ 2: What are the typical optical properties used for the five key tissues in simulations? The following table summarizes standard optical properties (absorption coefficient μa, scattering coefficient μs, and anisotropy g) for the five tissues at a common near-infrared wavelength, as used in Monte Carlo simulations for photon transport [5].

Table 1: Typical Optical Properties of Head Tissues for fNIRS Simulations

Tissue Type	Absorption Coefficient μa (mm⁻¹)	Scattering Coefficient μs (mm⁻¹)	Anisotropy (g)
Scalp	0.018	7.8	0.89
Skull	0.016	9.4	0.89
CSF	0.004	0.3	0.89
Gray Matter	0.036	9.2	0.89
White Matter	0.014	1.5	0.89

FAQ 3: How does the CSF layer impact fNIRS signal quality? The CSF layer presents a unique challenge due to its low scattering property (μs = 0.3 mm⁻¹) [5]. Because it scatters light less than surrounding tissues, CSF can act as a "light guide," potentially channeling photons away from the underlying cortical gray matter. This effect can create a confounding signal and reduce the sensitivity of fNIRS measurements to the targeted brain activation, leading to an underestimation of the true hemodynamic response [5]. Accurate segmentation and modeling of the CSF layer are therefore essential for correcting this effect and improving the quantitative accuracy of fNIRS.

FAQ 4: What is the practical impact of using a population head atlas versus subject-specific anatomy? Using a population head atlas (e.g., the SPM12 atlas based on 549 subjects) is a robust and common approach when subject-specific structural MRI is unavailable [5]. However, research shows that using subject-specific anatomical data for segmentation and probe placement optimization can lead to measurable improvements. One study found that approaches using individual anatomical data (probabilistic, individual fMRI, or vascular approaches) outperformed a standard literature-based approach in terms of fNIRS signal quality and sensitivity to brain activation [2]. Subject-specific models account for individual variations in head anatomy, cortical folding, and tissue thickness, which can significantly influence photon path and measurement sensitivity [2] [11].

Troubleshooting Guides

Issue 1: Poor Spatial Specificity and Localization Accuracy

Problem: The fNIRS measurements do not reliably localize activity to the intended brain region, or the detected signals are blurred and lack specificity.

Potential Causes and Solutions:

• Cause 1: Inaccurate head model segmentation.
  • Solution: Validate your segmentation pipeline. Use established software like SPM12 for tissue segmentation, which generates probability maps for each tissue [5]. Ensure proper post-processing, such as smoothing and hole-filling, to create a continuous segmentation volume. When possible, move from a generic atlas to a subject-specific anatomical model for improved accuracy [2] [11].
• Cause 2: Suboptimal optode layout.
  • Solution: Utilize computational tools for optode layout optimization. Toolboxes like the fNIRS Optodes' Location Decider (fOLD) can automatically determine optode positions from a set of predefined locations (e.g., the 10-10 system) to maximize sensitivity to your target regions-of-interest [5]. For higher specificity, consider designing high-density (HD) arrays with overlapping, multi-distance channels, which have been shown to provide superior localization compared to traditional sparse arrays [3].
• Cause 3: Thick CSF layer confounding the signal.
  • Solution: Since the low-scattering CSF layer can guide light away from the cortex, ensure it is correctly included in your photon transport model. Using subject-specific MRI data allows for the most accurate modeling of this layer's thickness and impact [5].

Issue 2: Low Signal-to-Noise Ratio (SNR) in Measurements

Problem: The measured fNIRS signals are weak and dominated by noise, making it difficult to detect task-related hemodynamic changes.

Potential Causes and Solutions:

• Cause 1: Inadequate optode-scalp coupling.
  • Solution: Ensure a stable and high-quality optical contact. For prolonged acquisitions, consider using a clinical adhesive like collodion to glue optodes directly to the scalp. This method, common in clinical EEG, can maintain excellent signal quality for many hours, even with hair present [11].
• Cause 2: Systemic physiological noise from superficial tissues.
  • Solution: Incorporate short-separation channels into your optode layout. These channels (typically with a source-detector distance of 8-15 mm) are primarily sensitive to the scalp and skull. Their signals can be used as regressors to remove systemic physiological noise (e.g., from blood pressure changes) from the standard long-separation channels that measure brain activity [12] [3].
• Cause 3: Sub-optimal source-detector distance.
  • Solution: The source-detector distance is a critical parameter. Distances that are too short will not penetrate sufficiently to the cortex, while distances that are too long will result in a very weak detected signal. Adhere to the typical range of 25-40 mm to ensure a reasonable SNR while maintaining cortical sensitivity [2] [11].

Experimental Protocols & Workflows

Protocol 1: Standardized Tissue Segmentation from T1-Weighted MRI

This protocol outlines the steps for segmenting head tissues from a T1-MRI volume, as derived from established methodologies [5].

Objective: To create a segmented head model comprising five tissues (scalp, skull, CSF, gray matter, white matter) for use in photon transport simulations.

Materials:

• T1-weighted anatomical MRI volume of the subject.
• Segmentation software (e.g., SPM12).

Procedure:

• Preprocessing: Load the T1 image into your segmentation software.
• Segmentation: Run the segmentation algorithm (e.g., in SPM12 with default parameters). This will generate probability maps for each of the five tissues. Each voxel in these maps contains the probability (0-1) of belonging to a specific tissue.
• Hard Segmentation: Create a single, definitive segmentation volume by assigning each voxel to the tissue type for which it has the highest probability, provided that probability exceeds a threshold (e.g., >0.2). This helps manage boundary voxels.
  • Assignment values: Scalp (1), Skull (2), CSF (3), Gray Matter (4), White Matter (5). Voxels not meeting the threshold are assigned to air (0).
• Post-processing: Smooth the resulting segmented image (e.g., with a 2mm FWHM Gaussian kernel) to eliminate any single-voxel "holes" within tissues that may have been created during the thresholding process.
• Verification: Visually inspect the segmented tissues against the original MRI to ensure anatomical plausibility.

Protocol 2: Photon Transport Simulation for Sensitivity Profile Calculation

Objective: To compute the sensitivity profile (or "banana-shaped" photon path) for a given source-detector pair on the segmented head model.

Materials:

• Segmented head model (from Protocol 1).
• Photon transport simulation software (e.g., Monte Carlo eXtreme - MCX).
• Optical properties for each tissue (see Table 1).

Procedure:

• Setup: Convert the segmented head model into a simulation-compatible format (e.g., a *.bin file for MCX).
• Configure Simulation: Create an input file specifying:
  • The position of the source optode in voxel coordinates.
  • The initial photon direction (typically towards the center of the head).
  • The number of photons to simulate (e.g., 10⁸ for robust statistics).
  • The optical properties (μa, μs, g) for each tissue label.
  • Simulation volume dimensions and voxel size.
• Run Simulation: Execute the Monte Carlo simulation for the source optode.
• Adjoint Simulation: Run a second simulation, using the detector optode position as the source (the "adjoint" field).
• Calculate Sensitivity: The sensitivity of the channel (source-detector pair) is calculated as the voxel-wise product of the normalized photon fluence from the source simulation and the adjoint simulation [5]. This resulting sensitivity map indicates the probability that a photon traveled through each voxel in the volume.

Workflow Visualization

Diagram 1: fNIRS Optode Optimization Workflow

The Scientist's Toolkit

Table 2: Essential Research Reagents and Solutions for fNIRS Tissue Segmentation Studies

Item Name	Function/Brief Explanation	Example Use Case
SPM12 Software	A statistical software package used for the segmentation of T1-MRI images into probabilistic tissue maps (scalp, skull, CSF, gray/white matter).	Automated tissue classification to create a head model for photon migration simulations [5].
Colin27 & SPM12 Head Atlases	Standardized, high-resolution digital head models. Colin27 is based on 27 scans of one individual; the SPM12 atlas is based on 549 subjects. Used when subject-specific MRI is unavailable [5].	Provides a generic but robust anatomical model for simulating optode sensitivity and designing initial probe layouts [5].
Monte Carlo eXtreme (MCX)	A GPU-accelerated software for simulating photon transport in turbid media (like biological tissues). Crucial for modeling light propagation in complex, layered head models [5].	Calculating the sensitivity profile and photon fluence for each source and detector optode placed on the scalp [5].
fOLD Toolbox	The fNIRS Optodes' Location Decider is a toolbox that uses pre-computed sensitivity profiles to automatically decide optode positions that maximize sensitivity to user-defined brain regions-of-interest [5].	Translating a target brain region (e.g., dorsolateral prefrontal cortex) into a practical and optimal optode layout on a measurement cap [5].
Collodion Adhesive	A water-resistant, quick-drying adhesive used to firmly attach fNIRS optodes to the scalp.	Ensuring stable optical coupling for prolonged fNIRS recordings, which is critical for high signal quality and data reliability [11].
3D Neuronavigation System	A device that tracks the 3D position of instruments in real-time relative to the subject's anatomy, often co-registered with their MRI.	Guiding the precise placement of optodes on the scalp according to the planned, optimized coordinates from software like fOLD [11].

Frequently Asked Questions (FAQs)

Q1: What is a normalized sensitivity profile in fNIRS, and why is it critical for my research? A normalized sensitivity profile, often expressed as a percentage, quantifies the relative contribution of a specific voxel (a 3D pixel in the brain) to the total fNIRS signal measured by a source-detector channel. It is calculated by normalizing the sensitivity of all voxels so that their sum equals one (or 100%) [5]. This normalization is crucial because it allows for a standardized comparison of sensitivity between different channels, subjects, or experimental setups. It directly informs how well your optode placement targets a specific region of interest (ROI), which is fundamental for achieving high spatial resolution and accurate interpretation of hemodynamic activity [2] [11].

Q2: I have my Monte Carlo simulation results. What are the concrete mathematical steps to calculate normalized sensitivity? The procedure involves a few key steps, moving from the raw output of the Monte Carlo simulation to the final normalized map [5]:

• Obtain the Fluence Fields: Run separate Monte Carlo simulations for each source and detector optode position. The output for each is a 3D volume called the photon fluence (Φ), which represents the light energy distribution.
• Calculate Channel Sensitivity: For a given source-detector pair (a channel), the sensitivity profile is computed as the voxel-wise product of the source's fluence field (Φsource) and the detector's adjoint field (Φdetector). The adjoint field is effectively the fluence field for the detector acting as a source.
• Normalize the Sensitivity: The final step is to normalize this sensitivity volume. This is done by dividing the sensitivity value at each voxel by the sum of the sensitivity values across all voxels in the volume. This yields the normalized sensitivity, where the sum of all voxels is 1.

The mathematical expression is [5]: normSens_i = (Φ_source_i × Φ_detector_i) / Σ(Φ_source × Φ_detector) Where i represents a specific voxel and the denominator is the sum over all voxels.

Q3: My normalized sensitivity values seem extremely low. Is this expected? Yes, this is a common and expected observation. Since the normalization is performed over the entire head volume—including the scalp, skull, and cerebrospinal fluid (CSF), which light passes through to reach the cortex—the sensitivity values assigned to any small region, including the cortical grey matter, will be very small [13]. One study reported that the first quartile of channels (those with the shortest source-detector distances) accounted for only about 0.391% of the total normalized sensitivity profile, highlighting how the signal is distributed across a vast volume [13]. The critical factor is the relative sensitivity within your target ROI compared to surrounding areas.

Q4: How does anatomical variability between subjects impact my normalized sensitivity profiles? Anatomical variability has a profound impact [13]. Differences in the thickness of the scalp, skull, and CSF, as well as the unique folding patterns of gyri and sulci, significantly alter the path of light and thus the sensitivity profile. Research has shown high dispersion of sensitivity profiles among subjects when using subject-specific anatomy (SSA) compared to results from using a standard atlas-based anatomy (ABA) [13]. This means that using a single, generic head model (like Colin27) for probe design may lead to suboptimal and highly variable sensitivity across a study cohort. For research demanding high spatial specificity, using individual anatomical MRI data is recommended [2] [13].

Troubleshooting Common Issues

Problem: Inconsistent or Non-Reproducible Sensitivity Profiles Across Simulation Runs

Potential Cause	Solution
Insufficient Photons	A low number of photons in the Monte Carlo simulation leads to a poor signal-to-noise ratio.	Increase the number of photons simulated. Studies often use 10^8 photons per simulation to ensure stable results [5] [14].
Incorrect Co-registration	The optode positions are not accurately mapped to the anatomical model's scalp surface.	Use a neuronavigation system or validated co-registration algorithms to ensure the virtual optode placement matches the intended physical (or 10-5/10-10 system) locations on the specific head model [15] [11].
Inaccurate Tissue Segmentation	Errors in segmenting the MRI into different tissue types (scalp, skull, CSF, grey/white matter) propagate into flawed optical property assignments.	Use well-validated segmentation software (e.g., Freesurfer, SPM) and manually check the results, especially at tissue boundaries. Ensure the segmentation probability threshold is appropriate (e.g., >0.2) [5].

Problem: Poor Sensitivity to the Targeted Cortical Region

Potential Cause	Solution
Suboptimal Optode Placement	The chosen source-detector pairs are not sufficiently sensitive to the underlying region of interest (ROI).	Use an optimal montage tool (e.g., within the NIRSTORM toolbox) that automatically calculates optode positions to maximize sensitivity to a user-defined ROI while respecting constraints like source-detector distance [15] [11].
Excessive Source-Detector Distance	While a larger distance increases penetration depth, it also exponentially decreases signal strength.	Constrain the source-detector distance to a range that provides a good trade-off, typically between 25 mm and 40 mm, to ensure a reasonable signal-to-noise ratio [2].
Ignoring Subject Anatomy	Using a generic atlas head model for a population with high anatomical variability.	Incorporate subject-specific anatomical MRI data to compute the sensitivity profiles. If this is not feasible, using a probabilistic approach that combines individual anatomy with functional maps from an independent dataset can be a good compromise [2] [13].

Experimental Protocols & Methodologies

Protocol 1: Core Workflow for Generating a Normalized Sensitivity Map

The following diagram illustrates the end-to-end pipeline for calculating normalized sensitivity profiles, integrating common tools and steps from the literature.

Diagram 1: Workflow for generating a normalized sensitivity map.

Detailed Steps:

• Acquire and Segment Anatomical Data: Begin with a high-resolution T1-weighted MRI. Process this volume using segmentation software (e.g., Freesurfer, SPM12) to generate distinct 3D maps for each tissue type: scalp, skull, cerebrospinal fluid (CSF), grey matter (GM), and white matter (WM) [13] [5]. The accuracy here is paramount.
• Define Optode Positions: Determine the coordinates for your sources and detectors. This can be done using standard systems like the 10-5 international system [13] [5] or by defining custom positions. These coordinates must be co-registered to the subject's native MRI space or a standard template space (e.g., MNI).
• Set Up Monte Carlo Simulation: Configure the simulation using a tool like MCXlab [15] [14]. Key parameters include:
  • Photon Count: Set to a high number, typically 10^8, to ensure results are robust and not noisy [5] [14].
  • Optical Properties: Define the absorption coefficient (μa), scattering coefficient (μs), anisotropy factor (g), and refractive index (n) for each tissue type at your desired wavelength (e.g., 800 nm) [5] [14].
  • Volume: Input the segmented tissue map, where each voxel is labeled according to its tissue type.
• Run Simulations and Calculate Sensitivity: Execute a simulation for each source and detector optode. The sensitivity for a specific channel is computed by taking the voxel-wise product of the fluence field from the source and the fluence field from the detector (the adjoint field) [5].
• Perform Normalization: Finally, normalize the resulting 3D sensitivity map. Sum the sensitivity values across every voxel in the entire volume, then divide the value at each individual voxel by this total sum. This yields the normalized sensitivity map where the sum of all voxels is 1 [5].

Protocol 2: Quantitative Analysis of Sensitivity Profiles

Once normalized sensitivity maps are generated, their properties can be quantified to evaluate an optode layout. The following table summarizes key metrics and findings from relevant studies.

Table 1: Quantitative Metrics from fNIRS Sensitivity Profile Analyses

Metric	Description	Exemplary Finding / Value
Sensitivity by Depth	The proportion of the total normalized signal originating from different depth ranges (quartiles) from the scalp.	~70% of the normalized signal originated from the first two depth quartiles (gyri), with the first quartile (depth < ~11.8 mm) contributing 0.391% and the second (depth < ~13.6 mm) contributing 0.292% of the total profile [13].
Spatial Spread (FWHM)	The full width at half maximum of the sensitivity profile, measuring its spatial spread on the cortex.	In the source-detector direction, the spatial spread was broad (20.95 mm FWHM), while it was steeper in the transversal direction (6.08 mm FWHM) for the first depth quartile [13].
Number of Photons	The number of photon packets launched per simulation to achieve stable results.	Standard practice is to use 100 million (10^8) photons per optode simulation [5] [14].

The Scientist's Toolkit: Essential Research Reagents & Solutions

Table 2: Essential Tools and Software for fNIRS Sensitivity Modeling

Tool / Solution	Function	Key Features & Notes
MCX / MCXlab	A Monte Carlo simulation platform for photon transport in 3D media, accelerated by Graphics Processing Units (GPUs) [5] [14].	Considered a gold standard for accurate simulations. Function: Dramatically reduces computation time for simulating millions of photons.
NIRSTORM	An extension for the Brainstorm software dedicated to fNIRS analysis and, crucially, optimal montage design [15].	Function: Provides a user-friendly interface to compute optimal optode placements that maximize sensitivity to a target ROI. It can leverage pre-computed fluence fields.
fOLD Toolbox	The fNIRS Optodes' Location Decider provides a first-order approach to guide optode placement based on brain regions-of-interest using pre-computed sensitivity profiles on head atlases [5].	Function: Offers a simpler, atlas-based solution for initial experimental design when subject-specific MRI is unavailable.
COLIN27 / ICBM152 Atlases	Standardized, high-resolution head models derived from MRI averages of one and 152 subjects, respectively [16] [5].	Function: Serve as a realistic generic head model for simulations when individual anatomy is not available. Note: Results will not account for inter-subject anatomical variability [13].
IBM ILOG CPLEX	Optimization software used by tools like NIRSTORM to solve the mixed linear integer programming problem of finding the best optode positions under constraints [15].	Function: Powers the computational optimization behind personalized montage design.

Key Concepts and Common Questions (FAQs)

What is neurovascular coupling and why is it fundamental to fNIRS measurements? Neurovascular coupling (NVC) is the physiological mechanism that links local neural activity to subsequent changes in cerebral blood flow and blood oxygenation. When a brain region becomes active, it triggers a process called functional hyperemia: cerebral blood vessels dilate, leading to an oversupply of oxygenated blood to the activated area. Consequently, this increases oxygenated hemoglobin (HbO) and decreases deoxygenated hemoglobin (HbR) in the local vasculature. fNIRS measures these hemodynamic changes by detecting near-infrared light absorption, providing a proxy for underlying neural activity [17].

What are the most common sources of error in fNIRS signals? fNIRS signals are susceptible to several sources of contamination, which can be categorized as follows [1]:

• Physiological Noise: This includes artifacts from the cardiac pulse (~1 Hz), respiration (~0.3 Hz), and blood pressure waves (Mayer waves, ~0.1 Hz). These systemic signals originate from both cerebral and extracerebral tissues.
• Motion Artifacts: Sudden movements of the optodes relative to the scalp can cause significant signal disruptions. While fNIRS is more tolerant to motion than fMRI, artifacts remain a major challenge.
• Poor Optode Coupling: Inadequate contact between the optodes and the scalp, often due to hair, is a frequent cause of low signal-to-noise ratio.
• Systemic Confounds: Changes in systemic physiology (e.g., blood pressure, heart rate, CO₂ levels) that are unrelated to the neural task can profoundly influence the fNIRS signal.

How can I improve the spatial specificity of my fNIRS setup? Achieving good spatial specificity requires careful attention to optode placement and design [1] [2] [11]:

• Personalized Montages: Use individual anatomical (MRI) data to design optode layouts that maximize sensitivity to your specific Region of Interest (ROI). This is superior to using standard cap positions based on the 10-20 system alone.
• Source-Detector Distance: Maintain a distance of 25-40 mm between sources and detectors. This range is a trade-off: it ensures sufficient light penetration to the cortex while maintaining a reasonable signal-to-noise ratio.
• Algorithmic Optimization: Employ computational methods that use light sensitivity profiles (often calculated via Monte Carlo simulations) to determine the optode positions that provide the best sensitivity to the target cortical areas.

Troubleshooting Guide: Common fNIRS Issues and Solutions

Table 1: Troubleshooting Common fNIRS Experimental Issues

Problem Category	Specific Symptom	Potential Cause	Recommended Solution
Signal Quality	No signal or very low-intensity signal on all/many channels.	Poor optode-scalp coupling; broken or unplugged fiber optics; incorrect source-detector distance.	Check optical coupling and reposition optodes; ensure all cables are securely connected; verify source-detector distance is within 25-40 mm [1] [2].
Signal Quality	High-frequency noise overwhelming the signal.	Physiological noise (heartbeat, respiration); electronic interference.	Apply band-pass filtering in real-time or offline (e.g., 0.01 - 0.5 Hz to remove cardiac and respiratory signals); use shorter ground leads and separate power cables from data cables [1].
Signal Quality	Slow, drifts or large, sporadic spikes in the signal.	Participant movement (Motion Artifacts); changes in systemic physiology.	Use motion correction algorithms (e.g., PCA, wavelet-based methods); employ a water-resistant adhesive (e.g., collodion) to secure optodes for prolonged recordings [1] [11].
Spatial Specificity	Inability to reliably detect activation in the target brain region.	Generic, non-optimized optode layout; inaccurate placement on the scalp.	Implement a personalized montage design using subject-specific MRI data and probabilistic maps of brain activation [2] [11].
Spatial Specificity	Low reproducibility of results across multiple sessions.	Inconsistent optode placement between sessions.	Use a 3D neuronavigation system to guide precise and repeatable optode placement for each session [11].
Experimental Design	Weak or absent hemodynamic response to a stimulus.	Stimulus properties (intensity, duration) are insufficient to evoke a robust hemodynamic response.	Optimize stimulus parameters based on prior literature. For auditory stimuli, higher intensities (e.g., 70-90 dB SPL) are more effective at modulating the hemodynamic response [17].

Featured Experimental Protocol: Investigating Intensity-Dependent Auditory Activation

This protocol details a combined EEG-fNIRS experiment to study neurovascular coupling, based on a study that explored how sound intensity modulates cortical activation [17].

Objective: To analyze the topographical effect of auditory stimulus intensity on cortical activation and explore neurovascular coupling between auditory-evoked potentials (AEPs) and fNIRS hemodynamic signals.

Participants:

• 40 healthy volunteers.
• Demographic: 13 males, 27 females; mean age = 22.27 ± 3.96 years [17].

Equipment and Reagents:

• fNIRS System: A continuous-wave fNIRS device with sources and detectors integrated into a single cap.
• EEG System: Electroencephalography system synchronized with the fNIRS device.
• Stimulus Presentation: Headphones for auditory stimulus delivery.
• Adhesive: Collodion or a similar clinical adhesive for securing optodes for prolonged, stable recordings [11].

Procedure:

• Optode Placement: Place fNIRS optodes and EEG electrodes on the scalp according to the experimental montage, targeting the auditory cortex (e.g., superior temporal gyrus) and prefrontal regions (e.g., superior and inferior frontal gyri). Using collodion to fix optodes is recommended for optimal signal quality [11].
• Stimulus Design:
  • Stimuli: Complex tones across seven frequencies (range: 400–2750 Hz).
  • Intensities: Three intensity levels: 50-dB, 70-dB, and 90-dB SPL.
  • Paradigm: Present tones in blocks of five for each intensity level, with intensities randomized across the experiment.
• Data Acquisition: Simultaneously record EEG and fNIRS data while participants listen to the auditory stimuli. The total recording session typically lasts 1-2 hours.
• Data Analysis:
  • EEG Processing: Extract AEP components (N1 and P2 amplitudes) from the EEG data.
  • fNIRS Processing: Convert raw light intensity changes into concentration changes of HbO and HbR.
  • Statistical Analysis: Use PERMANOVA to assess the effect of intensity on hemodynamic activity. Perform Spearman correlations on the residuals of AEPs and fNIRS responses to isolate stimulus-specific neurovascular coupling [17].

Research Reagent Solutions

Table 2: Essential Materials for fNIRS Experiments

Item	Function in the Experiment	Technical Specifications / Examples
fNIRS System	Measures hemodynamic changes by emitting near-infrared light and detecting its absorption after passing through brain tissue.	Continuous-wave systems are common; wavelengths typically 650-1000 nm [17].
EEG System	Provides direct measurement of neural electrical activity with high temporal resolution, complementing fNIRS.	Used to record Auditory Evoked Potentials (AEPs) like N1 and P2 components [17].
Neuronavigation System	Guides precise and reproducible placement of fNIRS optodes on the scalp based on individual anatomical MRI data.	Critical for implementing personalized optode montages [11].
Clinical Adhesive (Collodion)	Secures optodes to the scalp with a water-resistant bond, ensuring stable optical coupling and reducing motion artifacts for long durations.	Enables recordings of several hours with excellent signal quality, even with challenging hair types [11].
Anatomical & Functional MRI Data	Provides subject-specific information for designing optimal optode layouts and accurately localizing the measured brain activity.	Used to compute light sensitivity profiles and define target Regions of Interest (ROIs) [2] [11].

Table 3: Correlations Between AEP Components and fNIRS Hemodynamic Responses [17]

AEP Component	Hemoglobin Species	Correlated Brain Regions (Change in Activity)
N1 Amplitude	HbO	↑ in Superior Temporal Gyrus (STG) & Superior Frontal Gyrus (SFG); ↓ in Inferior Frontal Gyrus (IFG)
N1 Amplitude	HbR	↑ near Supramarginal Gyrus (SMG)
P2 Amplitude	HbO	↑ in Superior Frontal Gyrus (SFG) & Inferior Frontal Gyrus (IFG)
P2 Amplitude	HbR	↑ in Supramarginal Gyrus (SMG), Angular Gyrus (AnG), SFG, and IFG

Diagram 1: Neurovascular Coupling Mechanism

Diagram 2: Optode Placement Optimization Workflow

Frequently Asked Questions (FAQs)

Q1: What are the fundamental spatial resolution and depth penetration limits of fNIRS technology?

fNIRS is fundamentally limited to measuring brain activity in the superficial cortex, typically at depths of 1.5 to 2 centimeters from the scalp [18]. Its spatial resolution is generally in the range of 1 to 3 centimeters, which is sufficient for localizing activity to broad cortical areas but insufficient for probing deeper subcortical structures like the hippocampus or amygdala [19] [12]. This creates a "superficial cortex bias," meaning fNIRS data inherently reflects activity in the brain's outer layers.

Q2: How does optode placement and array density affect the spatial resolution and quality of my data?

Optode placement and array density are critical factors. Inconsistent optode placement across sessions significantly reduces the reproducibility and spatial overlap of the measured signals [4]. Furthermore, traditional sparse arrays (e.g., with 30 mm channel spacing) have limited spatial resolution and sensitivity compared to modern high-density (HD) arrays. HD arrays with overlapping, multi-distance channels provide superior localization, sensitivity, and inter-subject consistency [3].

Q3: Why is it challenging to reliably target the same brain region across multiple experimental sessions?

Achieving consistent spatial targeting is difficult due to several factors:

• Variability in Cap Placement: Even slight shifts in the fNIRS cap position on the scalp between sessions change the underlying cortical regions being measured [12].
• Limited Anatomical Information: Standard fNIRS setups lack real-time anatomical information, making it hard to verify which brain region is being targeted [20].
• Individual Anatomical Differences: Head shape and brain anatomy vary significantly between individuals, so the same scalp coordinates can correspond to different cortical areas in different people [20].

Q4: What is the functional consequence of fNIRS's superficial cortex bias for cognitive neuroscience research?

The primary consequence is that fNIRS cannot directly investigate the functions of deep brain structures. Research questions and experimental designs must be framed around hypotheses concerning the cerebral cortex. This limitation necessitates using complementary techniques like fMRI or careful task design to infer the role of cortical-subcortical networks [19].

Troubleshooting Guides

Guide 1: Improving Spatial Specificity and Targeting

Problem: Inability to precisely and consistently target a specific Region of Interest (ROI), especially across multiple sessions or subjects.

Solution: Implement probabilistic spatial registration techniques to coregister fNIRS data with standard brain anatomy.

• Step 1: Digitize Optode Positions. Use a 3D digitizer to record the precise locations of your optodes on the participant's scalp relative to standard landmarks (e.g., nasion, inion, preauricular points) [20].
• Step 2: Coregister with Anatomical Data. Use software to map the digitized scalp positions onto a subject-specific MRI or a standard atlas brain (e.g., MNI space). This step converts "scalp channel" data into an anatomically defined "cortical location" [20].
• Step 3: Validate ROI Targeting. Perform a sensitivity analysis to confirm that your fNIRS setup adequately probes your intended region, adjusting optode placement if necessary [21].

The following workflow outlines the coregistration process for improving spatial specificity.

Guide 2: Mitigating Signal Contamination from Superficial Tissues

Problem: The fNIRS signal is contaminated by systemic physiological noise (e.g., from scalp blood flow), which can be confounded with the cerebral hemodynamic response.

Solution: Integrate short-separation channels into your optode array and use them as regressors to remove superficial contamination.

• Step 1: Hardware Setup. Incorporate short-separation detectors (typically 8 mm or less from the source) into your probe layout. These channels are predominantly sensitive to systemic activity in the scalp and skull [3].
• Step 2: Data Processing. During preprocessing, use the signal from the short-separation channels as a nuisance regressor in a general linear model (GLM) or adaptive filter.
• Step 3: Signal Extraction. Subtract the superficial component estimated from the short-separation signal from the standard long-separation channel signals. This yields a cleaner signal that is more specific to cerebral brain activity [3].

Guide 3: Selecting an Optimal fNIRS Array for Your Spatial Resolution Needs

Problem: Choosing between a sparse and a high-density (HD) fNIRS array for a specific research application.

Solution: Base your decision on the specific goals of your study, weighing the trade-offs between spatial resolution, setup time, and cost. The following table summarizes the key differences.

• Sparse Arrays (e.g., 30 mm grid): Best suited for studies aiming to detect the presence or absence of activation in a broad cortical area with minimal setup complexity [3].
• High-Density (HD) Arrays: Essential for studies requiring precise localization of brain activity, differentiating between adjacent functional areas, or examining functional connectivity within a region [3].

Table 1: Quantitative Comparison of Sparse vs. High-Density (HD) fNIRS Arrays

Feature	Sparse Array	High-Density (HD) Array
Typical Channel Spacing	~30 mm	Multiple, overlapping distances (e.g., 10-45 mm) [3]
Spatial Resolution	Lower (1-3 cm) [19]	Higher (sub-centimeter potential with image reconstruction) [3]
Depth Sensitivity	Limited, improved with short-separation channels	Superior due to multi-distance measurements [3]
Localization Accuracy	Poor; difficult to differentiate adjacent regions [3]	Excellent; enables precise mapping of activated regions [3]
Key Advantage	Faster setup, lower computational demand [3]	Improved sensitivity, specificity, and inter-subject consistency [3]
Best For	Detecting task-evoked activity in a broad field-of-view [3]	Studies requiring precise localization, especially for low cognitive load tasks [3]

The Scientist's Toolkit: Essential Reagents & Materials

Table 2: Key Materials and Solutions for fNIRS Spatial Resolution Research

Item	Function/Explanation
3D Digitizer	A magnetic or optical device to record the precise 3D locations of fNIRS optodes on the scalp, which is the first critical step for anatomical coregistration [20].
Individual T1-Weighted MRI	Provides subject-specific anatomical information. It is the gold standard for coregistering fNIRS channel locations to the underlying cortical anatomy for individual analysis [20].
Probabilistic Atlas/MRI Database	Enables "MRI-free" spatial registration by mapping fNIRS data to a standard brain (e.g., MNI space) using probabilistic information from a database of many MRIs. Crucial for studies without access to MRI [20].
High-Density (HD) fNIRS Probe	A custom or commercial probe layout with many overlapping source-detector pairs at multiple distances. This hardware is foundational for improving spatial resolution and depth localization via Diffuse Optical Tomography (DOT) [3].
Short-Separation Channels	Optode pairs placed at a minimal distance (e.g., 8 mm) to selectively measure hemodynamic signals from superficial tissues (scalp, skull). Their signal is used as a regressor to remove non-cerebral physiological noise from standard channels [3].
Coregistration Software	Software packages (e.g., NIRS Brain AnalyzIR, AtlasViewer, MNE-NIRS) that implement algorithms for mapping scalp-based fNIRS data onto cortical surfaces, a mandatory step for accurate spatial interpretation [20].

Advanced Methodologies for Precision Optode Arrangement and Probe Design

Troubleshooting Guides & FAQs

Frequently Asked Questions

What is the fOLD toolbox and what problem does it solve? The fNIRS Optodes' Location Decider (fOLD) is a toolbox designed to help researchers determine optimal optode placements on the scalp to maximize sensitivity to specific, pre-defined brain Regions of Interest (ROIs) [22]. It addresses the key challenge in fNIRS research that the scalp-channel correspondence and sensitivity profiles change with age, ensuring that the measured signals originate from the intended cortical areas [22].

My research involves infants. Can I use the standard fOLD toolbox? For developmental research involving infants or children, you should use the devfOLD toolbox, an extension of the original fOLD [22]. The devfOLD provides age-specific channel-to-ROI specificity estimates computed using realistic head models from infant, child, and adult age groups, as channel sensitivity profiles and scalp-to-cortex mapping differ significantly across ages [22].

How does fOLD's approach differ from simple spatial projection? fOLD uses photon transport simulations, which model how near-infrared light propagates through the heterogeneous tissues of the head, to quantify a channel's sensitivity to an ROI [22]. This is more accurate than simple spatial projection (which defines a channel as a point equidistant from the source and detector) because it accounts for light-tissue interaction and provides a more realistic sensitivity profile [22].

Why is my channel-to-ROI specificity value low even when my channel is directly over the ROI? Low specificity can result from a suboptimal source-detector separation distance [22]. Specificity is defined as a channel's sensitivity to a target ROI relative to its sensitivity to the whole brain. Even with good placement, an inappropriate separation distance can lead to a shallow or overly broad sensitivity profile, reducing the proportion of sensitivity focused on your ROI.

How can I improve the reproducibility of my fNIRS measurements across multiple sessions? Reproducibility is highly dependent on consistent optode placement [4]. Increased shifts in optode position between sessions correlate with reduced spatial overlap of the measured brain activity [4]. Using digitized optode positions for each session to improve source localization can significantly enhance reliability [4].

Common Error Messages and Solutions

Error / Issue	Possible Cause	Solution
Low specificity for all suggested channels.	The target Region of Interest (ROI) may be too small or located in a deep cortical fold.	Consider increasing the source-detector separation distance within a safe limit (e.g., up to 35-40 mm for adults) to achieve a deeper penetration depth [22].
Inconsistent channel-to-ROI mapping in developmental groups.	Using an adult head model for infant or child studies.	Switch to the devfOLD toolbox and select the appropriate age-specific head model for your population [22].
Poor reproducibility of HbO/HbR signals across sessions.	Inconsistent optode placement or lack of anatomical registration.	Use digitization to record optode positions for each session and employ source localization with an anatomically specific model during analysis [4].
Signals are contaminated by systemic physiological noise.	Insufficient signal processing for extracerebral artifacts.	Apply real-time preprocessing techniques, such as short-channel regression or PCA-based filtering, to remove systemic noise [1].

Experimental Protocols & Methodologies

Protocol 1: Establishing Channel-to-ROI Specificity with fOLD

This methodology outlines how the fOLD toolbox computes its core metrics, which are essential for planning an experiment [22].

• Head Model Selection: The process begins with selecting an appropriate head model. The standard fOLD uses an adult model (Colin27), while devfOLD offers age-specific models for infants and children [22].
• Photon Transport Simulation: For a given source-detector pair, a Monte Carlo simulation models the propagation of near-infrared photons through the multi-layered head tissues (scalp, skull, CSF, cortex). This calculates the spatial sensitivity profile, often expressed as a "banana-shaped" region [22].
• Specificity Calculation: The toolbox calculates the channel-to-ROI specificity. This is a normalized metric, typically defined as the sensitivity of the channel to the user-specified ROI relative to its total sensitivity to the entire brain [22].
• Output and Configuration: Researchers receive a list of potential channels (source-detector pairs) ranked by their specificity for the target ROI. This allows for the selection of an optode configuration that maximizes the signal from the brain area of interest.

Protocol 2: Validating fNIRS Reproducibility for Multi-Session Studies

This protocol is based on a study that quantified the within-subject reproducibility of fNIRS signals over ten sessions [4].

• Participant & Setup: Participants complete multiple testing sessions on separate days. A high-density fNIRS cap (e.g., 102 channels) covering the entire head is used.
• Task Design: Participants perform blocked or event-related tasks known to activate specific regions, such as a finger-tapping motor task or a visual stimulation task.
• Optode Digitization: In each session, the 3D positions of the optodes on the participant's scalp are digitized using a stylus and position tracker.
• Data Acquisition: fNIRS data (changes in HbO and HbR) are collected throughout the task performance.
• Data Analysis:
  • Channel-Level Analysis: Activation maps are created for each session. Reproducibility is quantified as the percentage of sessions in which a channel shows significant task-related activity.
  • Source-Level Analysis: Using the digitized optode positions and the participant's MRI (or a default head model), the scalp-measured signals are projected onto the cortical surface to create source-reconstructed activation maps. The overlap of these source maps across sessions is then calculated.
• Outcome: The study found that HbO signals were significantly more reproducible than HbR. Furthermore, increased shifts in optode placement between sessions reduced spatial overlap, highlighting the need for consistent placement and digitization [4].

Workflow Visualization

fOLD Implementation Workflow

fOLD Implementation Workflow

Signal Quality Assurance Pathway

Signal Quality Assurance Pathway

The Scientist's Toolkit: Essential Research Reagents & Materials

Key Software Tools for fNIRS Optode Placement and Analysis

Tool Name	Primary Function	Relevance to fOLD & Spatial Specificity
fOLD / devfOLD	Provides pre-calculated channel-to-ROI specificity to guide optode configuration for standard 10-10/10-5 positions without requiring individual MRIs [22].	Core toolbox for this article; enables informed design of channel arrangement for a target ROI.
AtlasViewer	Allows for visualization and manual optimization of custom optode layouts on a head model, with photon migration simulation to evaluate sensitivity profiles [22].	Complements fOLD by allowing iterative, manual placement and 3D visualization of sensitivity "bananas".
Array Designer	An optimization algorithm that automatically determines source-detector arrangements to maximize sensitivity and coverage for a user-specified ROI [22].	An alternative to fOLD for designing custom holder configurations beyond standard 10-10 positions.
PHOEBE	Graphical software that measures and displays, in real-time, the optical coupling between fNIRS optodes and the scalp during headgear placement [23].	Ensures good scalp coupling in practice, which is critical for achieving the signal quality assumed by fOLD.
Homer3 / Homer2	A comprehensive, widely-used software suite for fNIRS data preprocessing and analysis, including functions for signal filtering, GLM analysis, and 3D visualization [23].	Used for downstream processing of data collected based on an fOLD-informed montage.
NIRSite	An MNI-based montage creator from NIRx that allows for manual and imported coordinate registration and digitization in 2D and 3D [23].	Helps translate the fOLD-recommended channel locations to a practical montage file for acquisition systems.

Quantitative fNIRS Specificity and Reproducibility Data

Metric	Value / Finding	Experimental Context & Implication
HbO vs. HbR Reproducibility	HbO significantly more reproducible than HbR (F(1, 66) = 5.03, p < 0.05) [4].	Based on multi-session visual/motor tasks; suggests prioritizing HbO for longitudinal studies.
Impact of Optode Shift	Increased shifts in optode position correlate with reduced spatial overlap across sessions [4].	Underscores the critical need for consistent cap placement or digitization for reliable results.
Source Localization Benefit	Source localization (using digitized positions) improves reliability to capture brain activity vs. channel-level analysis [4].	Recommends investing time in digitizing optodes and using anatomical models for analysis.
Example Specificity Channels	Channels at F7, F8, F5, F6, FC5, FC6 showed consistent sensitivity to Inferior Frontal Gyrus across infant, child, and adult groups [22].	Demonstrates that some scalp locations provide stable ROI mapping across ages, while others do not.

This guide provides essential troubleshooting information for researchers using the International 10-10 and 10-5 systems to optimize optode placement for functional Near-Infrared Spectroscopy (fNIRS) experiments.

Troubleshooting FAQs: Spatial Specificity and Reproducibility

1. How can I improve the consistency of optode placement across multiple experimental sessions?

• Challenge: Variations in cap placement and difficulty in relocating exact scalp positions can reduce the spatial overlap of measured signals across sessions [1] [12].
• Solutions:
  • Use Detailed System Rules: Adhere to a strictly defined and unambiguous version of the 10-10 or 10-5 system. Precise rules yield precise and reproducible landmark positions on the scalp [24] [25].
  • 3D Digitization: Use a digitization pen to record the 3D coordinates of your optodes relative to cranial landmarks (nasion, inion, pre-auricular points) in each session. This allows you to quantify and account for placement shifts [4].
  • Leverage Probabilistic Registration: Use available tools that provide Montreal Neurological Institute (MNI) standard brain coordinates for 10-10 and 10-5 positions. This facilitates the co-registration of your fNIRS data with a standard brain atlas, improving inter-subject and cross-session consistency [24] [5].

2. What is the effective spatial resolution of the 10-10 and 10-5 systems? The systems provide a theoretical framework for many positions, but their effective resolution depends on avoiding overlap between adjacent measurement points.

• 10-10 System: Positions can be well-separated on the scalp without overlapping, making them highly effective for standard high-density setups [24].
• 10-5 System: Offers over 300 potential scalp locations. However, a study evaluating 329 positions found that about 241 could be set effectively without overlapping with a neighbor, defining the practical upper limit for multi-subject studies [24] [25].

3. How does optode density impact signal quality and localization? The choice between a sparse array (e.g., based on 10-10) and a high-density (HD), overlapping array (e.g., based on 10-5) involves a key trade-off.

• Sparse Arrays (e.g., 30mm spacing): Are simpler and faster to set up but have limited spatial resolution and sensitivity. They may miss active brain regions or average signals from multiple areas [3].
• High-Density Arrays: Use multiple, closely-spaced source-detector distances, including short-separation channels. HD arrays provide superior sensitivity, better depth resolution, and significantly improved localization of brain activity, though they require more resources and complex data processing [3].

The table below summarizes a quantitative comparison from a 2025 study:

Feature	Sparse Array	High-Density (HD) Array
Spatial Localization	Limited and less specific [3]	Superior and more precise [3]
Sensitivity to Brain Activity	Lower; may only detect high-cognitive-load tasks [3]	Higher; can detect activity even during lower-load tasks [3]
Inter-subject Consistency	Lower reproducibility due to non-uniform sensitivity [3]	Improved consistency and signal reproducibility [3]
Setup Time & Complexity	Lower	Higher (more optodes, more complex caps) [3]
Data Processing	Simpler	More complex (requires image reconstruction) [3]

4. Which hemodynamic signal is more reproducible across sessions? Evidence suggests that changes in oxygenated hemoglobin (Δ[HbO]) are significantly more reproducible over multiple sessions than changes in deoxygenated hemoglobin (Δ[HbR]) [4]. Focusing your analysis on HbO may lead to more reliable results in longitudinal studies.

Essential Experimental Protocols

Protocol 1: Validating and Digitizing Optode Placement

Purpose: To ensure accurate and reproducible optode positioning session-to-session. Materials: fNIRS cap (10-10 or 10-5 layout), 3D digitizer, measurement tape. Methodology:

• Landmark Identification: Precisely locate and mark the nasion, inion, and left/right pre-auricular points on the participant's scalp [24] [26].
• Cap Placement: Secure the fNIRS cap on the participant's head, aligning its reference points with the anatomical marks.
• Optode Digitization: Using the 3D digitizer, record the spatial coordinates (x, y, z) of every source and detector optode.
• Verification: Use software to project the digitized positions onto a template MRI (e.g., in MNI space) to verify they are over the intended brain regions [4] [5].

Protocol 2: Assessing Systemic Physiological Noise

Purpose: To identify and separate non-neural physiological signals (e.g., heart rate, blood pressure) from the task-evoked hemodynamic response. Materials: fNIRS system, short-separation channels, physiological monitors (e.g., pulse oximeter, blood pressure cuff, respiratory belt). Methodology:

• Data Acquisition: Simultaneously record fNIRS signals (including short-separation channels), heart rate, blood pressure, and respiration during both rest and task conditions.
• Signal Processing: Apply a band-pass filter (e.g., 0.5 - 2.0 Hz) to the fNIRS data to isolate cardiac and respiratory frequencies [27].
• Noise Regression: Use the short-separation channels and/or the directly recorded physiological signals as regressors in a General Linear Model (GLM) to remove these noise components from the long-channel fNIRS data [27].

Research Reagent Solutions

The table below lists key tools and their functions for conducting high-quality fNIRS research.

Tool / Material	Function in Research
10-10 / 10-5 System Cap	Provides a standardized grid for positioning optodes on the scalp, ensuring reproducibility and allowing comparison with published literature [24] [26].
3D Digitizer	Records the precise 3D location of optodes co-registered with anatomical landmarks, enabling accurate mapping of fNIRS channels to brain anatomy [4].
fOLD Toolbox	A software toolbox that uses photon migration simulation on head atlases to recommend optode placements that maximize sensitivity to specific, pre-defined brain Regions of Interest (ROIs) [5].
Short-Separation Channels	Source-detector pairs placed 8-15 mm apart. They are primarily sensitive to systemic physiological noise in the scalp and are used as regressors to clean these artifacts from standard channels [3].
Montreal Neurological Institute (MNI) Template	A standardized coordinate system and brain atlas. Co-registering fNIRS data to MNI space allows for group-level analysis and comparison with other neuroimaging studies (e.g., fMRI) [24] [5].

Workflow Visualization

The following diagram illustrates a recommended workflow for planning and executing an fNIRS experiment with high spatial specificity.

fNIRS Experimental Workflow for Spatial Specificity

The diagram below outlines the process for using the fNIRS Optodes' Location Decider (fOLD) toolbox to inform probe design.

fOLD Toolbox Logic for Probe Design

Core Concepts and Key Advantages

FAQ: What is MRI-informed optode placement and what are its primary benefits?

MRI-informed optode placement is a advanced procedure for designing personalized functional near-infrared spectroscopy (fNIRS) montages that optimize sensitivity to specific, individual brain regions of interest (ROIs). This method uses subject-specific anatomical (and sometimes functional) MRI data to computationally determine the best positions on the scalp for light sources and detectors [11] [2]. The core benefit is enhanced spatial specificity, ensuring that the fNIRS setup is precisely targeted to the cortical areas most relevant to your investigation, which is crucial for both clinical applications and basic research [12] [2].

FAQ: How does this personalized approach improve upon standard cap-based methods?

Standardized caps place optodes at predetermined locations based on average head anatomy. In contrast, MRI-informed placement accounts for individual variability in brain structure, skull thickness, and the exact folding of the cortical surface [11]. This leads to two major improvements:

• Increased Signal Quality: By optimizing for a target region, the method maximizes the sensitivity of the measurements to brain activity in that area, improving the signal-to-noise ratio (SNR) [2].
• Reduced Optode Count: Personalized montages can achieve accurate local reconstructions with fewer optodes than ultra-high-density grids, reducing setup time and improving subject comfort for prolonged investigations [11].

Technical Specifications and Performance Data

The performance of different levels of MRI-informed montages has been quantitatively evaluated. The table below summarizes a comparison from a study that tested approaches with incrementally more individual data [2].

Table 1: Comparison of MRI-Informed Montage Approaches

Approach Name	Data Utilized	Key Advantage	Reported Outcome
Literature-Based (LIT)	Published coordinates and templates; no individual MRI.	Low resource requirement; no need for MRI scan.	Served as a baseline; was outperformed by MRI-informed approaches [2].
Probabilistic (PROB)	Individual anatomical MRI + probabilistic fMRI maps from an independent group.	Balances personalization with resource constraints; no need for subject-specific functional scans.	Similar signal quality and sensitivity to more informed approaches (iFMRI, fVASC) [2].
Individual fMRI (iFMRI)	Individual anatomical and task-based functional MRI (fMRI).	Targets the functionally active region for a specific task in the individual.	High signal quality and sensitivity; comparable to PROB and fVASC approaches [2].
Vascular (fVASC)	Individual anatomical, functional, and vascular MRI.	Accounts for the influence of individual vascular structures on light propagation.	High performance, but similar overall outcomes to PROB and iFMRI [2].

Table 2: Key Constraints for Computational Montage Optimization

Optimization Parameter	Typical Constraint	Functional Rationale
Source-Detector Distance	25-40 mm [11] [2]	Balances sufficient light penetration depth with an acceptable signal-to-noise ratio.
Number of Sources/Detectors	User-defined (e.g., 3 sources, 7 detectors) [15]	Limits the montage size for practicality and comfort while ensuring adequate coverage.
Adjacency Constraint	Minimum number of channels per source (e.g., 7) [15]	Ensures spatial overlap of measurements, which is critical for later 3D reconstruction of the hemodynamic activity [11].

Experimental Protocols and Workflows

FAQ: What is the typical end-to-end workflow for implementing a personalized, MRI-informed fNIRS montage?

The following diagram illustrates the multi-stage workflow for creating and using a personalized fNIRS montage.

Workflow for Personalized fNIRS Montage

Troubleshooting Guide: My optimized montage does not cover the target region correctly. What should I check?

• Problem: Inaccurate Head Model.
  • Solution: Verify the quality of the MRI segmentation. Ensure that the scalp, skull, CSF, and gray/white matter surfaces are accurately defined, as errors here propagate to flawed sensitivity profiles [11] [15].
• Problem: Poor ROI Definition.
  • Solution: Double-check the definition of your target region. The ROI should be a set of vertices on the cortical surface, which can be defined manually, using an anatomical atlas (e.g., Destrieux, Desikan-Killiany), or based on a functional localizer from an individual fMRI scan [2] [15].
• Problem: Overly Restrictive Search Space.
  • Solution: In the optimization software, ensure the "search space" for possible optode positions on the scalp is sufficiently large and appropriately centered over the projection of your cortical ROI. A common practice is to use all scalp vertices within a 4 cm radius of the ROI [15].

The Scientist's Toolkit

Table 3: Essential Research Reagents and Solutions

Item / Software	Critical Function	Implementation Notes
High-Res T1-Weighted MRI	Provides individual anatomical data for head model creation.	Essential for all personalized approaches. Protocol should ensure clear contrast between tissue types [11] [2].
fNIRS Analysis Suite with Montage Optimization	Computes light propagation and solves the optimization problem.	Software like the NIRSTORM plugin for Brainstorm is specifically designed for this task and integrates with the IBM ILOG CPLEX solver [15].
3D Neuronavigation System	Precisely guides optode placement on the scalp according to the computed optimal positions.	Crucial for translating the virtual montage from software to the subject's head with high accuracy [11] [2].
Clinical Adhesive (Collodion)	Secures optodes for prolonged recordings.	Enables maintenance of excellent optical coupling for several hours, which is vital for clinical or long-duration studies [11].
Task-Based fMRI Protocol	(For iFMRI approach) Identifies the exact functional ROI for a given task in the individual.	The functional localizer task (e.g., finger tapping, mental calculation) must be carefully designed to robustly activate the target network [2].

Advanced Optimization and Analysis

FAQ: What happens after I collect data with an optimized montage? How is the data analyzed?

The final step in the procedure is the local reconstruction of hemodynamic activity along the cortical surface using inverse modeling [11]. Unlike the simplified modified Beer-Lambert law, this approach uses the sensitivity profiles from your personalized head model to tomographically reconstruct changes in hemoglobin concentration (Δ[HbO] and Δ[HbR]) directly on the cortex. This corrects for partial volume effects and improves quantitative accuracy, providing a more precise map of brain activity [11].

The logical relationship between the optimization constraints and the final analytical outcome is shown below.

From Montage Design to Analysis Outcome

Frequently Asked Questions (FAQs)

What is probabilistic placement for fNIRS and why is it important? Probabilistic placement is a method for determining where to place fNIRS optodes on the scalp. It uses pre-existing fMRI activation maps from many individuals to identify locations with the highest probability of measuring activity from a specific brain network or region of interest (ROI). This approach is crucial because it addresses the challenge of inter-subject variability; brain regions are not in the exact same spot in every person. By using probabilistic maps, you can target optode placements to locations that most consistently correspond to your ROI across a population, thereby improving the anatomical specificity and reliability of your fNIRS measurements [1] [28] [29].

My research focuses on the language network. Are there specific atlases available? Yes. You can use the Language Atlas (LanA), a probabilistic functional atlas created from fMRI data of over 800 individuals performing a language localizer task. This atlas allows you to estimate, for any location in a standard brain space, the probability that it falls within the language network. It is available for both volume-based (MNI) and surface-based (FSaverage) brain templates, facilitating its use with common neuroimaging software and data repositories [29].

I have access to an HD-fNIRS system. How does probabilistic placement benefit me? While High-Density (HD) fNIRS arrays inherently offer better spatial resolution and localization than traditional sparse arrays, probabilistic placement enhances their effectiveness. The probabilistic maps guide the arrangement of your dense optode grid to ensure it optimally covers the cortical network you intend to study. Research shows that HD arrays with overlapping, multi-distance channels provide superior sensitivity and localization, particularly for detecting activity during tasks with lower cognitive load. Using a probabilistically-informed HD design maximizes your chances of accurately capturing and localizing brain activity [3].

What are the main sources of variability when trying to reproduce fNIRS results? A major international initiative (the fNIRS Reproducibility Study Hub) found that agreement on results is highest when hypotheses are strongly supported by existing literature. The primary sources of analytical variability include:

• How poor-quality data are handled and which data is included.
• The methods used to model hemodynamic responses.
• The specific choices made during statistical analysis. Teams with greater fNIRS experience and higher analysis confidence showed better agreement. This underscores the importance of clear methodological and reporting standards [30].

Troubleshooting Common Experimental Issues

Problem: Inconsistent fNIRS Signals Across Multiple Sessions

Symptoms: You are running a longitudinal study, but the fNIRS signals from the same putative brain region are not consistent across different sessions.

Possible Causes and Solutions:

• Cause: Inconsistent optode placement. Even small shifts in cap placement between sessions can significantly reduce the spatial overlap of the measured brain areas.
  • Solution: Use a probabilistic atlas to define target locations based on the 10-5 system for higher precision. Employ digitization tools to record the exact optode positions in each session and verify placement consistency [1] [4].
• Cause: Lower reproducibility of Deoxygenated Hemoglobin (HbR).
  • Solution: Focus your analysis on Oxygenated Hemoglobin (HbO) signals, as studies have shown that task-related changes in HbO are significantly more reproducible over multiple sessions than changes in HbR [4].
• Cause: Inadequate anatomical specificity in the measurement.
  • Solution: Move your analysis from channel-space to source-space. Using individual head models and source localization techniques can improve the reliability of capturing brain activity by providing anatomically specific information [4].

Problem: Weak or Unclear Activation in fNIRS Data

Symptoms: Your fNIRS data shows a weak, noisy, or statistically non-significant hemodynamic response to the task, even though the task is known to elicit a robust brain response.

Possible Causes and Solutions:

• Cause: The fNIRS channels are not optimally positioned over the functional region of interest.
  • Solution: Implement a probabilistic placement approach for your next study. Use a toolbox like fOLD (fNIRS Optodes' Location Decider) to determine optode locations that maximize sensitivity to your target ROIs, based on photon migration simulations [5].
• Cause: Using a low-density (sparse) array for a complex or subtle cognitive task.
  • Solution: If possible, upgrade to an HD-fNIRS array. Quantitative comparisons show that HD arrays outperform sparse arrays in detecting and localizing brain activity, especially for tasks with lower cognitive load. If you must use a sparse array, be aware that it may only be suitable for detecting strong, unambiguous activation [3].
• Cause: The target brain region exhibits high inter-subject variability.
  • Solution: Consult probabilistic maps to identify "core" regions of a network that have high inter-subject consensus. Focusing your analysis on these high-probability areas reduces noise introduced by averaging across individuals with different functional topographies [28].

Table 1: Comparison of fNIRS Array Types

Array Type	Typical Channel Spacing	Key Advantages	Key Limitations	Best Use Cases
Sparse (Low-Density)	~30 mm, non-overlapping	Wider field-of-view, faster setup, lower cost [3]	Limited spatial resolution and sensitivity; poor localization; may miss subtle activity [3]	Detecting presence/absence of strong activation in broad areas [3]
High-Density (HD-DOT)	Multiple, overlapping distances (e.g., 10-45 mm)	Superior spatial resolution, sensitivity, and localization; better inter-subject consistency [3]	Higher cost; longer setup time; more complex data processing [3]	Studies requiring precise localization, especially for subtle cognitive tasks [3]

Table 2: Features of Publicly Available Probabilistic Resources

Resource Name	Description	Based On	Key Application
Language Atlas (LanA) [29]	Probabilistic atlas for the language network	Precision fMRI data from >800 individuals	Interpreting group activations or lesions; selecting units for analysis in language studies [29]
Probabilistic Functional Maps [28]	Maps for 14 functional networks (e.g., Frontoparietal, Default Mode)	Highly sampled fMRI data from multiple datasets (HCP, MSC, etc.)	Seeding group analyses; focusing on high-consensus network regions [28]
fOLD Toolbox [5]	fNIRS Optodes' Location Decider	Photon transport simulations on head atlases	Guiding optode placement during fNIRS experimental design to maximize anatomical specificity [5]

Detailed Experimental Protocols

Protocol: Utilizing the fOLD Toolbox for Probe Design

The fNIRS Optodes' Location Decider (fOLD) is a toolbox that helps decide optode positions based on a set of brain regions-of-interest (ROIs) [5].

• Input ROIs: Define your target brain regions in a standard space (e.g., MNI coordinates).
• Simulate Sensitivity: fOLD uses Monte Carlo simulations of photon transport on segmented head atlases (like Colin27) to calculate the sensitivity profile of potential fNIRS channels [5].
• Calculate Specificity: For each possible channel (source-detector pair), the toolbox calculates its specificity to your predefined ROIs.
• Decide Optode Layout: The algorithm automatically selects the set of optode positions from predefined 10-5 or 10-10 system locations that maximizes the anatomical specificity to your ROIs [5].
• Output: The final output is a recommended optode arrangement tailored to your experiment's hypotheses.

Protocol: Creating a Probabilistic Map from fMRI Data

This protocol describes the general workflow for generating a probabilistic functional map, as implemented in software like BrainVoyager and used in foundational studies [28] [31].

• Data Acquisition: Collect a large amount of high-quality fMRI data per subject (e.g., >20 minutes of resting-state or multiple task runs) to ensure reliable individual-level network estimates [28].
• Individual Network Identification: For each subject, identify functional networks using a template-matching procedure with high-quality group-average network templates or functional localizer tasks [28] [29].
• Spatial Normalization: Normalize each individual's brain and their functional map to a common template space (e.g., MNI, FSaverage).
• Overlay and Count: For each voxel or vertex in the common space, count the number of subjects for whom that location is significantly active or assigned to the network of interest.
• Calculate Probability: Express the count as a percentage of the total number of subjects, creating a map where every brain location has a value from 0-100% representing the group consensus [31].

Probabilistic Map Creation Workflow

Table 3: Key Resources for Probabilistic fNIRS Studies

Resource / Solution	Function / Description	Example in Research
Probabilistic Brain Atlases [28] [29]	Quantitative maps of inter-subject consensus for functional networks.	LanA atlas for language; 14-network maps for cognitive systems [28] [29].
fOLD Toolbox [5]	Computes optimal fNIRS optode positions to maximize sensitivity to target ROIs.	Informs probe design before data collection using photon migration models [5].
High-Density fNIRS Arrays [3]	Optode layouts with overlapping, multi-distance channels for improved resolution.	HD-DOT systems for superior localization of prefrontal cortex activity during Stroop tasks [3].
Digitization Systems	Records the 3D coordinates of optodes on a subject's head.	Used to verify consistent cap placement across sessions and for accurate source localization [4].
Cortex-Based Alignment Tools	Advanced normalization technique that aligns brains based on cortical folding patterns.	Improves the accuracy of group-level analysis and probabilistic map creation in surface-based space [31].

Technical Troubleshooting Guide

Frequently Asked Questions (FAQs)

Q1: The optimization algorithm fails to converge on a feasible optode layout. What are the primary causes? Infeasible solutions commonly result from overly restrictive constraints. Key factors to investigate include:

• Excessive Minimum Adjacency Number: Over-constraining the minimum number of channels per source can make a solution impossible. Relax this requirement and verify the algorithm's parameters [32].
• Conflicting Hardware and Coverage Demands: Requesting high sensitivity to a large or deep brain region with a very limited number of optodes is often physically infeasible. Reduce the size of the target region or increase the number of available optodes [33] [32].
• Incompatible Positioning Constraints: Constraints that reserve specific positions for EEG electrodes might conflict with the need to create valid source-detector pairs. Review the reserved positions to ensure sufficient adjacent locations remain for fNIRS optodes [33] [32].

Q2: Our experimentally measured fNIRS signal quality is poor despite using an optimized montage. Why? An optimized layout ensures sensitivity to the target brain region but does not immunize the signal against other sources of contamination.

• Insufficient Real-Time Preprocessing: fNIRS signals are susceptible to cerebral and extracerebral systemic noise and motion artifacts. Implement robust real-time preprocessing (e.g., motion artifact correction, filtering) to ensure the system operates on brain activity and not noise [1] [12].
• Poor Optode-Scalp Contact: Physical issues like hair blocking the light path or poor contact can severely degrade signal quality. This is a common challenge that optimization software cannot overcome [18].
• Subject-Specific Anatomical Variations: The optimization might be based on a standard head atlas, but individual factors like skull thickness, CSF volume, and hair pigmentation can negatively affect signal quality. Using subject-specific anatomical data from MRI can mitigate this [34] [2].

Q3: How can we ensure consistent targeting of the same brain region across multiple sessions? Reproducibility is a recognized challenge in fNIRS.

• Use Neuronavigation: Employ neuronavigation systems to coregister optode positions with individual MRI data for highly repeatable placements across sessions [2].
• Digitized Optode Positioning: Digitizing the actual optode positions after each setup allows researchers to quantify and account for placement shifts during data analysis, improving the accuracy of source localization [4].
• Standardized Caps with Dense Arrays: Using caps with high-density holder arrangements (e.g., 10/05 system) provides a standardized framework. Even with slight shifts, a dense array ensures some channels will maintain sensitivity to the target region [33] [4].

Q4: Which fNIRS chromophore (HbO or HbR) should be prioritized for neurofeedback? While both can be used, the choice involves a trade-off.

• Higher Signal-to-Noise of HbO: Oxygenated hemoglobin (Δ[HbO]) typically exhibits a larger amplitude change during brain activation and is often more reproducible across sessions, making it a common choice [34] [4].
• Greater Specificity of HbR: Deoxygenated hemoglobin (Δ[HbR]) may provide a more direct correlate of the fMRI BOLD signal and can offer improved spatial specificity [35].
• Combined Use: For increased robustness and validity, some approaches combine both Δ[HbO] and Δ[HbR] for neurofeedback [34].

Quantitative Data and Specifications

Table 1: Key Performance Metrics in fNIRS Optode Positioning

Parameter	Typical Value/Range	Implication for Optimization	Source
Spatial Resolution	~1 cm³	Limits the fineness of discriminable activation features; MLIP must work within this physical constraint.	[34]
Penetration Depth	1.5 - 2 cm	Optimization is only relevant for targeting superficial cortical layers.	[34] [18]
Optimal Source-Detector Distance	25 - 40 mm	A key constraint in MLIP to ensure a reasonable trade-off between penetration depth and signal-to-noise ratio.	[2] [33]
Temporal Resolution	~10 Hz (typically)	Much higher than fMRI; allows for rapid feedback in BCI/neurofeedback applications.	[1] [12]
Sensitivity Profile Depth	Approx. half the SD distance	Informs the forward model for light propagation used in the MLIP optimization.	[33]

Table 2: Comparison of Optode Layout Design Approaches

Approach	Key Inputs	Advantages	Limitations / Challenges
Literature-Based (LIT)	Prior published studies and standard EEG 10-20/10-05 systems.	Simple, fast, no need for additional neuroimaging data.	Suboptimal for individuals due to anatomical/functional variability; can lead to inconsistent results [2].
Probabilistic (PROB)	Individual anatomical MRI + probabilistic fMRI maps from an independent group.	Accounts for individual anatomy without requiring subject-specific fMRI; outperforms LIT approach [2].	Functional maps are not subject-specific, which may reduce targeting precision [2].
Individual fMRI (iFMRI)	Individual anatomical and functional MRI data.	Accounts for individual functional localization; high spatial specificity [2].	Requires costly and time-consuming fMRI data acquisition for each subject [2].
MLIP-based Optimal Montage	Target ROI, number of optodes, adjacency constraints, individual or atlas-based head model.	Formally optimizes sensitivity; can incorporate EEG positioning constraints; enables high-density designs [33] [32].	Computationally intensive; requires expertise in optimization; solution feasibility depends on constraint tuning [33] [32].

Experimental Protocols & Methodologies

Protocol 1: Formulating and Solving the MLIP for Optode Montage

This protocol details the core computational method for determining the optimal optode positions [33] [32].

1. Problem Definition and Inputs:

• Objective: Select a set of positions for sources (( S )) and detectors (( D )) from a finite set of possible holder positions (( P )) on an EEG/fNIRS cap.
• Goal: Maximize the total sensitivity of all created channels to a pre-defined Target Region of Interest (ROI).
• Key Inputs:
  • A target cortical ROI, often defined from an anatomical or functional MRI.
  • A sensitivity profile for every potential channel (every possible source-detector pair in ( P )). This is computed beforehand using Monte Carlo simulations on a head model (atlas-based or individual) to model light propagation [5].
  • The number of available sources and detectors.
  • Constraints (e.g., minimum source-detector distance, minimum number of channels per source "adjacency number").

2. Mathematical Formulation (Mixed Linear Integer Programming): The problem is formalized as follows [33]:

• Variables: Binary decision variables ( si, dj \in {0,1} ) indicating whether a source (or detector) is placed at position ( i ) (or ( j )).
• Objective Function: Maximize ( \sum{i \in P} \sum{j \in P} sens{ij} \cdot c{ij} ), where ( sens{ij} ) is the precomputed sensitivity of the channel between positions ( i ) and ( j ) to the target ROI, and ( c{ij} ) is a binary variable indicating whether a channel is formed between ( i ) and ( j ).
• Constraints:
  • Number of Optodes: ( \sum si = Ns ) and ( \sum dj = Nd ).
  • Channel Formation: A channel ( c{ij} ) can only be 1 if both a source is placed at ( i ) AND a detector is placed at ( j ). This is a linear constraint: ( c{ij} \leq si ) and ( c{ij} \leq dj ).
  • Distance: ( c{ij} = 0 ) if the distance between ( i ) and ( j ) is outside the acceptable range (e.g., 25-40 mm).
  • Adjacency: To encourage dense sampling, a constraint can enforce that each selected source must form at least ( K ) channels: ( \sum{j} c{ij} \geq K \cdot s_i ) for all ( i ).

3. Solution and Output:

• The MLIP problem is solved using optimization solvers (e.g., CPLEX).
• The output is the optimal set of positions for sources and detectors that maximizes sensitivity to the ROI under the specified constraints.

Protocol 2: Experimental Validation of Optimized Montages

This protocol validates the performance of an MLIP-designed montage against other approaches [2].

1. Participant and Data Acquisition:

• Recruit participants for multiple sessions involving MRI, neuronavigation, and fNIRS.
• f/MRI Session: Acquire individual anatomical, functional (fMRI during relevant tasks), and vascular MRI data.
• Neuronavigation Session: Coregister the participant's head with their MRI to guide precise optode placement.

2. Montage Design and Comparison:

• Design several optode layouts for the same participant using different approaches:
  • LIT: Based on literature review.
  • PROB: Using individual anatomy and probabilistic fMRI maps.
  • iFMRI: Using individual anatomy and fMRI.
  • MLIP (fVASC): Using the MLIP method with individual anatomical, functional, and vascular data.
• Keep the number of channels constant (e.g., two channels sharing a common source) for a fair comparison.

3. fNIRS Data Collection and Analysis:

• In the fNIRS session, participants perform mental-imagery tasks (e.g., mental calculation, motor imagery).
• Acquire fNIRS data for each of the designed layouts.
• Primary Outcome Measures:
  • Signal-to-Noise Ratio (SNR): Quantifies the quality of the raw fNIRS signal.
  • Sensitivity to Task: Statistical power of the detected hemodynamic response (e.g., amplitude of Δ[HbO] during task vs. rest).

4. Validation Conclusion:

• Studies show that individualized approaches (PROB, iFMRI, MLIP) consistently outperform the literature-based (LIT) approach in both SNR and sensitivity [2].
• The MLIP method provides a flexible framework to incorporate various constraints and achieve performance on par with other individualized methods.

The Scientist's Toolkit

Table 3: Essential Research Reagents and Software Solutions

Tool / Reagent	Type	Primary Function	Relevance to MLIP Optimization
fOLD Toolbox	Software Toolbox	Automatically decides optode locations from a set of predefined positions (10-10/10-05 system) to maximize sensitivity to a brain ROI [5].	Provides a first-order, atlas-based approach for defining target sensitivities, which can serve as input for more advanced MLIP optimization.
NIRSTORM	Software Plugin (for Brainstorm)	Provides a comprehensive fNIRS analysis environment, including an "optimal montage" module that uses MLIP to design subject-specific optode layouts [32].	A key implementation that integrates the MLIP methodology directly into a user-friendly, multimodal neuroimaging software platform.
Monte Carlo eXtreme (MCX)	Software Simulator	Simulates photon migration in complex biological tissues using GPU acceleration [5].	Critical for generating the sensitivity profiles for every potential channel, which form the core input data for the MLIP objective function.
CPLEX Optimizer	Software Library	A high-performance solver for mathematical optimization problems, including Mixed Integer Programming (MIP) [32].	The computational engine used to solve the numerically complex MLIP problem to find the global optimum for the optode layout.
Subject-Specific MRI Data	Data	Provides individual anatomical information (tissue segmentation, cortical structure).	Greatly improves the accuracy of the light model (forward problem) used in Monte Carlo simulations, leading to a more robust and personalized MLIP optimization [2].
Digitized Optode Positions	Data/Technique	Records the 3D coordinates of optodes after placement on the scalp.	Crucial for validating the accuracy of the implemented montage and for improving the precision of source localization during data analysis [4].

Frequently Asked Questions (FAQs)

Q1: What is photon transport modeling in fNIRS and why is it important?

Photon transport modeling refers to the simulation of photon movement through biological tissue. In functional near-infrared spectroscopy (fNIRS), it models how photons propagate from source optodes through various head tissues to detector optodes. This modeling generates sensitivity profiles (Jacobian matrices) that are essential for creating 3D voxel-based image reconstruction of brain activity. The accuracy of these models directly impacts the quality of brain activity localization and quantification, making them fundamental for optimizing fNIRS spatial resolution [36].

Q2: What are the main computational methods used for photon transport simulation?

The primary method for photon transport simulation is Monte Carlo modeling, which tracks the statistical path of numerous photon packets as they travel through tissue. These simulations account for scattering events, absorption, and tissue heterogeneity. Monte Carlo methods can be accelerated using graphics processing units (GPUs) for faster computation. The technique requires input parameters including tissue scattering and absorption properties, optode placement coordinates, and head anatomy [36].

Q3: Why is my probe misaligned or floating when I try to run sensitivity simulations?

This common issue typically occurs when the probe coordinates don't properly align with the scalp surface in your simulation environment. As reported in user forums, warnings like "Probe might be misaligned with head or too far from surface to project correctly" indicate coordinate system mismatches. Solutions include:

• Verifying landmark consistency (NZ, LPA, RPA, CZ, IZ) between MRI data and probe placement files
• Checking for left-right flipping issues in coordinate systems
• Ensuring proper scalp surface extraction from MRI data
• Using neuronavigation techniques for real-world validation [37]

Q4: How does anatomical modeling accuracy affect fNIRS sensitivity profiles?

Anatomical modeling accuracy significantly impacts sensitivity profile calculations. Studies demonstrate noticeable discrepancies in brain partial pathlengths when using approximated anatomies:

• Voxel-based brain segmentation: Shows -1.5% to 23% variation compared to accurate mesh models
• Layered-slab brain model: Shows 36% to 166% variation compared to accurate mesh models

These errors arise from inadequate representation of complex cortical surfaces and tissue boundaries, particularly problematic in areas with low-scattering tissues like cerebrospinal fluid (CSF) [38].

Q5: What are the advantages of mesh-based over voxel-based head models?

Mesh-based head models (tetrahedral or triangular) provide superior boundary accuracy and computational efficiency compared to voxel-based models:

Table: Model Comparison for fNIRS Photon Transport Simulation

Feature	Voxel-Based Models	Mesh-Based Models
Boundary representation	Terraced/staircased	Smooth and curved
Memory efficiency	Lower (uniform grid)	Higher (adaptive elements)
Accuracy near tissue boundaries	Limited	Superior
Handling complex anatomy	Limited resolution	Excellent flexibility
Computational demand for same accuracy	Higher	Lower
Representation of CSF layers	Poor	Accurate

Mesh models particularly excel at modeling photon reflection/transmission characteristics near tissue-air boundaries and low-scattering regions [38].

Troubleshooting Guides

Common Simulation Errors and Solutions

Table: Monte Carlo Simulation Troubleshooting

Error Symptom	Potential Causes	Solutions
Poor photon convergence	Insufficient photon packets	Increase packet count (typically 1-10 million)
	Inaccurate tissue optical properties	Verify absorption/scattering coefficients
Abnormal sensitivity patterns	Misaligned probe coordinates	Use neuronavigation validation [11]
	Incorrect head model registration	Check landmark consistency (LPA, RPA, NZ)
Excessive computation time	Inefficient algorithm	Utilize GPU-accelerated Monte Carlo
Overly dense mesh	Implement mesh simplification where possible
Inaccurate depth sensitivity	Improper source-detector distance	Maintain 25-40mm range for optimal SNR [2]
	Simplified anatomy	Use subject-specific MRI models [38]

Optimization Strategies for Personalized Montages

Problem: Inefficient optode placement targeting specific brain regions.

Solution: Implement mathematical optimization for personalized montages:

• Define Target Regions: Identify specific cortical volumes of interest (VOIs) based on your research question [11].

• Generate Sensitivity Profiles: Use Monte Carlo simulations with accurate head models to compute light sensitivity profiles [36].

• Formulate Optimization Problem: Apply mixed linear integer programming with functional constraints to determine optode positions that maximize sensitivity to target VOIs [11].

• Experimental Validation: Verify placement accuracy using 3D neuronavigation devices and validate with motor tasks like finger opposition [11].

This approach provides spatial resolution approaching ultra-high-density montages while significantly reducing optode count, enabling more practical clinical applications [11].

Experimental Protocols

Protocol 1: Mesh-Based Head Model Generation

Purpose: Create high-quality tetrahedral mesh models from MRI data for accurate photon transport simulation [38].

Materials and Software:

• T1-weighted MRI data
• Brain segmentation software (FreeSurfer, SPM, FSL)
• Brain2Mesh or Iso2Mesh toolbox (open source)
• Computational resources (recommended: 8GB+ RAM)

Procedure:

• Segment Anatomical Data: Process MRI volumes to identify tissue types (scalp, skull, CSF, gray matter, white matter).
• Generate Surface Meshes: Convert segmented volumes to multi-layered surfaces.
• Create Tetrahedral Mesh: Use surface-based meshing pipeline to produce volumetric mesh.
• Assign Optical Properties: Apply wavelength-specific absorption and scattering coefficients to each tissue type.
• Quality Control: Verify mesh quality and boundary smoothness.

Typical Processing Time: Several minutes to hours depending on data complexity [38].

Protocol 2: Subject-Specific Optode Layout Optimization

Purpose: Design personalized fNIRS optode layouts using different levels of individual MRI information [2].

Experimental Design: Compare four approaches with incremental MRI information:

• Literature-Based (LIT): Use literature review to guide layout design (no individual MRI data).
• Probabilistic (PROB): Employ individual anatomical data with probabilistic fMRI maps from independent datasets.
• Individual fMRI (iFMRI): Use individual anatomical and fMRI activation data.
• Vascular (fVASC): Incorporate individual anatomical, functional, and vascular information.

Implementation:

• Data Acquisition: Collect anatomical, functional, and vascular MRI data.
• Monte Carlo Simulation: Compute light sensitivity profiles using appropriate head model.
• Layout Optimization: Algorithmically determine optode positions constrained by:
  • Inter-optode distance: 25-40mm range
  • Channel count: Minimum two channels sharing a common source
  • Target region sensitivity maximization
• Validation: Compare signal quality and brain activation sensitivity across approaches.

Key Finding: PROB, iFMRI, and fVASC approaches outperform LIT, with similar performance among the three informed approaches [2].

Research Reagent Solutions

Table: Essential Tools for fNIRS Photon Transport Modeling

Tool/Resource	Type	Function	Availability
Brain2Mesh	Software toolbox	Generates high-quality brain mesh models from MRI	Open source [38]
Iso2Mesh	Software toolbox	General-purpose mesh generation for biomedical data	Open source [38]
AtlasViewer	Software package	Visualization and probe placement validation	Open source [37]
Monte Carlo eXtreme (MCX)	Simulation software	GPU-accelerated Monte Carlo photon transport	Open source [38]
Collodion adhesive	Material	Secures optodes for prolonged measurements with good optical contact	Commercial [11]
3D Neuronavigation device	Hardware	Guides precise optode placement according to simulation	Commercial [11]

Workflow Diagrams

Photon Transport Modeling Workflow

Computational Model Classification

Frequently Asked Questions (FAQs)

FAQ 1: What is the fundamental principle behind source-detector separation distance? The separation distance between a light source and a detector on the scalp determines the depth and volume of tissue sampled. Near-infrared light follows a banana-shaped path between the source and detector [39]. Shorter distances (e.g., 8 mm) primarily sample extracerebral layers like the scalp, while longer distances (typically 30 mm in adults) allow the light to penetrate the superficial cortex (1-2 cm depth) [40] [1]. The spatial resolution is on the order of 2.5–3 cm [40].

FAQ 2: How does increasing the number of measurement channels improve my data? A higher channel count, achieved through high-density arrays with multiple overlapping source-detector pairs, improves spatial sampling. This allows for better localization of brain activity and enables advanced analysis like image reconstruction, which can map activation patterns to specific anatomical structures [9]. High-density measurements with minimum separations of 13 mm or less can localize functional activation to within the width of a cortical gyrus [9].

FAQ 3: Why is my signal weak even with an appropriate source-detector distance? Weak signals can stem from several issues:

• Poor Optode-Scalp Coupling: Inadequate contact, often due to hair underneath the optode, drastically reduces signal quality.
• Insufficient Light Intensity: The source may not be emitting enough power to penetrate the tissue and return a measurable signal to the detector.
• Optode Displacement: Even small shifts in optode position between sessions can significantly reduce signal reproducibility and strength [4].

FAQ 4: What are the most effective methods for regressing physiological noise? Short-channel regression (SCR) is a highly effective method. It uses dedicated short-separation channels (∼8 mm) to directly measure systemic physiological noise from the scalp. This signal is then used as a regressor to remove similar noise components from the long-separation channels that contain both brain and noise signals [41]. This technique improves statistical robustness, even in cognitive tasks with minimal motion [41].

Troubleshooting Guides

Problem 1: Poor Spatial Specificity and Inability to Localize Activation

Symptoms	Possible Causes	Solutions
Blurred or diffuse activation maps.	Low channel density; limited head coverage.	Implement high-density probe designs with overlapping measurements [9].
Activation appears in anatomically implausible locations.	Lack of anatomical guidance for optode placement; inaccurate registration.	Use digitization tools to record 3D optode positions and coregister them with a digital brain atlas (e.g., using AtlasViewer software) [9].
Inconsistent localization of the same region across sessions.	Variations in cap placement on the head between sessions [4].	Use individual head molds or anatomical landmarks for highly reproducible placement.

Step-by-Step Protocol: Anatomical Registration with AtlasViewer

• Digitize Optode Positions: After placing the fNIRS cap on the participant, use a digitization pen to record the 3D locations of all sources, detectors, and key anatomical reference points (e.g., nasion, inion, pre-auricular points).
• Load Data into AtlasViewer: Import the digitized positions into the open-source AtlasViewer GUI (part of the HOMER2 software package) [9].
• Coregister with Atlas: The software will align the digitized probe layout with a canonical head atlas (e.g., Colin27).
• Validate Sensitivity Profile: Examine the calculated sensitivity profile to ensure your channels are sensitive to your targeted region of interest (ROI), such as the dorsolateral prefrontal cortex (DLPFC).

Problem 2: Low Signal-to-Noise Ratio (SNR) Contaminated by Physiological Artifacts

Symptoms	Possible Causes	Solutions
Signal contains clear oscillations at heart rate (~1 Hz) or respiration rate (~0.3 Hz).	Physiological noise from cardiac pulsation, respiration, and blood pressure waves [1].	Apply real-time or offline band-pass filtering (e.g., 0.01 - 0.2 Hz) to isolate the hemodynamic response.
Large, slow drifts in the baseline signal.	Systemic physiological changes (e.g., Mayer waves).	Use short-channel regression (SCR) to remove systemic artifacts [41].
Sharp, high-amplitude spikes in the data.	Motion artifacts from sudden head movements.	Employ motion correction algorithms (e.g., wavelet-based, correlation-based signal improvement).

Step-by-Step Protocol: Implementing Short-Channel Regression

• Probe Design: Integrate short-separation channels (SSCs) into your probe layout. These are typically source-detector pairs with a distance of 8 mm and are interspersed among the long channels [41].
• Data Acquisition: Record data from both long channels (e.g., 30 mm) and SSCs simultaneously.
• Signal Processing: For each long channel, identify the nearest SSC.
• Regression: Use a general linear model (GLM) where the signal from the SSC is used as a nuisance regressor to remove the shared physiological noise from the long-channel signal. This isolates the cerebral component of the hemodynamic response.

Experimental Protocols for Validation

Protocol: Validating Configuration Sensitivity to Cognitive Load

This protocol uses the well-established N-Back task to test if your fNIRS configuration can detect graded changes in brain activity.

• Objective: To verify that the fNIRS setup can detect a linear increase in prefrontal cortex activation with increasing working memory load [40] [41].
• Task Design:
  • Use the N-Back task with at least three load conditions (e.g., 0-Back, 1-Back, 2-Back).
  • In the 0-Back condition, participants identify a pre-specified target (e.g., the letter "X").
  • In the 1-Back condition, participants identify if the current stimulus matches the one immediately prior.
  • In the 2-Back condition, participants identify if the current stimulus matches the one two steps back.
  • Each block should consist of 10-15 stimuli, with 3-4 targets per block. Each stimulus is displayed for 2 seconds followed by a short inter-stimulus interval [41].
• fNIRS Configuration:
  • Target Region: Prefrontal cortex, specifically the dorsolateral prefrontal cortex (DLPFC).
  • Probe Layout: Ensure multiple long-separation channels (∼30 mm) cover the DLPFC, interspersed with short-separation channels (∼8 mm) for SCR.
• Expected Outcome: A statistically significant linear increase in oxygenated hemoglobin (Δ[HbO]) and a decrease in deoxygenated hemoglobin (Δ[HbR]) in the DLPFC as the N-Back level increases. SCR should enhance this statistical effect [41].

Protocol: Assessing Test-Retest Reproducibility

This protocol is critical for longitudinal studies or clinical applications where measurements are taken over multiple sessions.

• Objective: To determine the consistency of fNIRS measurements across multiple sessions with the same participant [4].
• Task Design:
  • Use a simple, robust functional task like a finger-tapping Motor task or a visual stimulus.
  • Perform the experiment on the same participant across multiple days (e.g., 10 separate sessions) [4].
• Key Measures:
  • Spatial Overlap: Quantify the overlap of significant activation maps across sessions.
  • Reproducibility Percentage: Calculate the percentage of channels or vertices that show significant task-related activity across all sessions [4].
• Critical Factor: Minimize shifts in optode placement. Increased shifts correlate with reduced spatial reproducibility [4].
• Analysis Recommendation: For the highest accuracy, use source localization/image reconstruction techniques instead of traditional channel-space analysis, as they are more robust to variability in head size and optode placement [4] [42].

The Scientist's Toolkit: Essential Research Reagents & Materials

Item	Function & Application
Short-Separation Channels (SSCs)	Optode pairs placed 8-15 mm apart to directly sample and regress out systemic physiological noise from scalp layers, significantly improving signal validity [41].
3D Digitizer	A magnetic or optical system to record the precise 3D locations of fNIRS optodes and anatomical landmarks on the head. Essential for accurate anatomical coregistration [9].
Digital Brain Atlas (e.g., Colin27)	A canonical model of the head and brain used to estimate the sensitivity profile of a given probe layout and to reconstruct functional images when subject-specific MRI is unavailable [9].
AtlasViewer / HOMER2 Software	An open-source analysis package for designing fNIRS probes, coregistering them with anatomy, and visualizing brain sensitivity and reconstructed images [9].
Individualized Head Molds	3D-printed or custom-fitted molds that hold the fNIRS probe in a fixed, subject-specific position. Crucial for maximizing reproducibility across longitudinal sessions [4] [43].

Quick-Reference Data Tables

Table 1: Standard fNIRS Configuration Parameters

Parameter	Typical Value (Adults)	Functional Role
Source-Detector Distance (Long)	25 - 35 mm	Determines sampling depth, allowing measurement of cortical hemodynamics [40] [39].
Source-Detector Distance (Short)	8 - 15 mm	Samples extracerebral layers (scalp) for noise measurement and regression [41].
Temporal Resolution	~10 Hz (up to 100 Hz)	Allows separation of cardiac (~1 Hz) and respiratory (~0.3 Hz) signals from the hemodynamic response [1].
Penetration Depth	1 - 2 cm	Limits measurement to the superficial cerebral cortex [40] [1].

Table 2: Impact of Common fNIRS Artifacts and Mitigation Strategies

Artifact Type	Source / Cause	Recommended Mitigation Strategy
Motion Artifact	Sudden head movement, causing optode displacement.	Wavelet-based filtering; spline interpolation; SCR can also help [1] [41].
Cardiac Pulsation	Arterial pulse in scalp and brain.	Band-pass filtering; use of SCR with SSCs [1] [41].
Respiratory Oscillation	Systemic blood pressure changes from breathing.	Band-pass filtering; use of SCR with SSCs [1] [41].
Scalp Hemodynamics	Systemic blood flow changes in skin and skull.	Short-Channel Regression (SCR) is the most direct and effective method [41].

Addressing Practical Challenges and Implementing Optimization Strategies

Frequently Asked Questions (FAQs)

FAQ 1: Why is fNIRS signal quality and reproducibility a concern, and what are the key factors affecting it? fNIRS signal quality and reproducibility are critical because they determine the reliability of the brain activity measurements. Several factors can affect them. Key findings indicate that oxyhemoglobin (HbO) is significantly more reproducible over multiple sessions than deoxyhemoglobin (HbR) [4]. Furthermore, the reproducibility of results can vary significantly with data quality, the choice of analysis pipeline, and the level of researcher experience [30]. Another major factor is optode placement: increased shifts in optode position between sessions correlate with less spatial overlap in the measured signals [4]. Using source localization techniques can improve the reliability of captured brain activity [4].

FAQ 2: What are motion artifacts and how do they affect my fNIRS signal? Motion artifacts (MAs) are unwanted changes in the fNIRS signal caused by participant movement rather than neural activity. They significantly deteriorate the measurement and reduce the signal-to-noise ratio (SNR) [44]. These artifacts occur primarily due to two reasons: optode movement relative to the skin, which changes the light path, and redistribution of blood in tissue from physical movements [45]. They manifest in the signal as:

• Spikes: Sudden, brief signal changes.
• Baseline shifts: Sustained changes in signal level.
• Oscillatory artefacts: Periodic fluctuations from repetitive movements like breathing [45].

FAQ 3: Which cortical regions are most susceptible to motion artifacts? The susceptibility to motion artifacts is not uniform across the head. Evidence suggests that the occipital and pre-occipital regions are particularly susceptible to MAs following upward or downward head movements. In contrast, the temporal regions are most affected by lateral movements, such as bending the head left or right [46]. This highlights the importance of a secure cap fit in these areas.

FAQ 4: How can I improve the spatial specificity of my fNIRS measurements? Improving spatial specificity involves precise targeting of cortical regions. Using 3D-printed optode holders ensures precise array geometry relative to brain anatomy, thereby improving signal quality [47]. For experimental design, the fNIRS Optodes' Location Decider (fOLD) toolbox can be used to automatically decide optode locations that maximize the anatomical specificity to pre-defined brain regions-of-interest, based on photon transport simulations [5]. Additionally, employing source localization techniques, especially with digitized optode positions, enhances the anatomical specificity and reliability of the captured brain activity [4] [1].

Troubleshooting Guides

This section provides structured guides to diagnose and resolve common signal quality issues.

Poor Optical Coupling

• Problem: Weak signal strength, poor Scalp Coupling Index (SCI), or inconsistent measurements across channels.
• Symptoms: Low-intensity signals, high-frequency noise dominating the raw signal, or channels frequently being flagged as "bad".
• Solutions:

Solution	Description	Key Benefit
Secure Optode Placement	Select a well-fitting headcap. Part hair between the optode and scalp using a tool. Ensure firm scalp contact [45].	Maximizes light transmission and reduces motion-induced signal dropout.
Use 3D-Printed Holders	Implement custom 3D-printed optode holders to maintain precise and consistent array geometry [47].	Improves inter-session reproducibility and overall signal quality.
Apply Spring-Loaded Holders	Utilize spring-loaded optode holders that automatically adjust pressure to maintain optimal contact with the scalp, especially useful with varying hair types [45].	Maintains consistent coupling under minor movement.
Monitor Signal Quality	Employ metrics like the Scalp Coupling Index (SCI) and Coefficient of Variation (CV) to quantitatively assess and verify signal quality during setup [47] [48].	Allows for proactive correction of poorly coupled optodes before data collection.

Motion Artifact Mitigation

A combined approach of prevention, detection, and correction is most effective.

• Problem: Signal corruption due to participant head, body, or facial movements.
• Symptoms: Sudden, large-amplitude spikes; sustained shifts in the signal baseline; or high-frequency oscillatory patterns not linked to the task [45].
• Solutions:

Category	Method	Principle	Best For
Prevention	Task Design	Minimize non-essential movement in the experimental protocol [45].	All studies, especially clinical.
Prevention	Participant Instruction	Give clear guidance to minimize unnecessary movements like frowning or talking [45].	All studies.
Prevention	Secure Cap & Optode Fit	As outlined in the previous table, a secure fit is the first line of defense [45].	All studies.
Hardware-Based Correction	Accelerometer	Use an auxiliary accelerometer to record head motion. This signal is then used in methods like Adaptive Filtering or Active Noise Cancelation (ANC) to remove motion-related noise [44].	Real-time applications where motion is predictable.
Algorithmic Correction	Moving Average / Spline Interpolation	Identifies motion artifacts based on moving standard deviation and corrects the signal using spline interpolation [44] [45].	Offline analysis of data with distinct motion spikes.
Algorithmic Correction	Wavelet-Based Filtering	Uses wavelet transformation to decompose the signal and isolate and remove components corresponding to motion artifacts [45].	Offline analysis, effective for various artifact types.
Algorithmic Correction	PCA/ICA	Methods like Principal Component Analysis (PCA) or Independent Component Analysis (ICA) separate the fNIRS signal into statistical components, allowing for the removal of those correlated with motion [45].	Data with strong periodic artifacts or multiple noise sources.

The following diagram illustrates the decision workflow for addressing motion artifacts, integrating the methods from the table above.

Experimental Protocols for Enhanced Signal Quality

Standardized fNIRS Preprocessing Workflow

A robust and standardized preprocessing pipeline is fundamental for enhancing signal quality and ensuring reproducible results. The following protocol, adapted from the MNE-Python tutorial, provides a step-by-step guide [48].

Detailed Methodology:

• Raw Intensity to Optical Density: Convert the raw light intensity measurements to optical density. This step calculates the attenuation of light, which is more stable for subsequent processing [48].
• Quality Assessment with Scalp Coupling Index (SCI): Quantify the coupling between each optode and the scalp. The SCI identifies the presence of the cardiac signal in the data. Channels with an SCI less than 0.5 are typically marked as bad and excluded from further analysis, as this indicates poor optode-scalp contact [48].
• Conversion to Hemoglobin Concentration: Apply the Modified Beer-Lambert Law (using a pathlength factor, e.g., 0.1) to convert optical density data into relative changes in oxygenated (HbO) and deoxygenated (HbR) hemoglobin concentration [48].
• Filtering: Apply a zero-phase band-pass filter (e.g., 0.05 Hz to 0.7 Hz) to the hemoglobin signals. This removes high-frequency noise (like the heart rate around 1 Hz) and very low-frequency drifts, isolating the hemodynamic response [48].
• Epoching and Artifact Rejection: Extract epochs of data time-locked to experimental events. During this step, implement automated rejection criteria to discard epochs with excessive artifacts. A common criterion is to reject epochs where the HbO concentration change exceeds 80e-6 (or another justified threshold) [48].

Protocol for Characterizing Motion Artifacts

To objectively characterize and validate motion artifact correction algorithms, a protocol using computer vision can be employed [46].

• Experimental Setup: Participants perform controlled head movements along three rotational axes (vertical, frontal, sagittal) at varying speeds (fast, slow) and types (half, full, repeated rotations) while whole-head fNIRS is recorded.
• Data Acquisition: The experimental session is video-recorded using a standard camera.
• Motion Data Extraction: Use a deep neural network (e.g., SynergyNet) to analyze the video footage frame-by-frame to compute precise head orientation angles. This provides ground-truth movement data.
• Artifact Identification: In the synchronized fNIRS data, identify motion artifacts (spikes, baseline shifts) using standard algorithms.
• Correlation Analysis: Statistically correlate the extracted head movement metrics (amplitude, speed) with the identified fNIRS artifacts to characterize which movements most severely compromise signal quality in different head regions [46].

The Scientist's Toolkit: Essential Research Reagents & Materials

This table details key materials and tools essential for conducting fNIRS experiments with high signal quality, particularly in the context of optode placement and motion artifact management.

Table: Essential Materials for fNIRS Signal Quality

Item	Function & Rationale
3D-Printed Optode Holders	Ensures precise and reproducible optode array geometry relative to individual brain anatomy, critically improving inter-session reliability and signal quality [47].
Spring-Loaded Optode Holders	Maintains consistent and optimal pressure between the optode and scalp, automatically compensating for minor movements or variations in hair thickness [45].
fOLD Toolbox	A software toolbox that uses photon migration simulations on head atlases to guide the placement of optodes to maximize sensitivity to specific brain Regions-of-Interest (ROIs) [5].
Auxiliary Accelerometer	A hardware sensor integrated into the fNIRS system that provides a direct measure of head motion, which is used as a reference signal for many real-time motion artifact correction algorithms [44].
Digitization Kit	A tool (e.g., 3D stylus) to record the precise 3D locations of optodes relative to cranial fiducial points. This information significantly improves the accuracy of source localization [4] [1].
Hair Parting Tool	A simple non-magnetic tool to part hair underneath each optode, reducing a major source of signal attenuation and improving light-scalp coupling [45].
MNE-Python Software	An open-source Python package that provides a complete and standardized pipeline for fNIRS data preprocessing, including SCI calculation, filtering, and epoching [48].

FAQs: Core Concepts and Problem Identification

What is systemic noise in fNIRS and why is it a problem? Systemic noise, or extracerebral interference, refers to physiological signals originating from the scalp (skin, skull, muscles) that contaminate fNIRS measurements of cerebral brain activity. These signals are not related to neuronal activation but can obscure or mimic true brain responses, leading to false positives or false negatives in your data [41]. This contamination reduces the validity and reliability of fNIRS findings, posing a significant challenge for spatial resolution studies where precise localization of brain activity is critical [1].

How can I visually identify systemic noise in my fNIRS signal? Systemic noise often manifests as low-frequency drifts or high-frequency oscillations in the hemodynamic response that are not time-locked to your experimental task. A key indicator is a highly correlated signal between long-separation channels (which capture both brain and scalp signals) and nearby short-separation channels (which primarily capture scalp signals) [41]. If the signal pattern in your long channels mirrors that of your short channels, it is likely contaminated by systemic noise.

Does systemic noise affect HbO and HbR differently? Yes, research indicates that oxygenated hemoglobin (HbO) is generally more reproducible and less susceptible to certain noise artifacts than deoxygenated hemoglobin (HbR) [4]. This is a critical consideration for your analysis, as focusing on HbO might provide more reliable results, though both signals should be examined.

How does optode placement relate to noise and spatial accuracy? Increased shifts in optode position between sessions directly correlate with reduced spatial overlap in measured brain activity [4]. Inconsistent placement not only reduces your ability to reliably target the same cortical region across sessions but can also vary the degree of extracerebral contamination, compromising the validity of spatial resolution comparisons.

Troubleshooting Guides: Mitigation Strategies

Guide 1: Implementing Short-Channel Regression (SCR)

Short-channel regression is a powerful method to subtract scalp-level interference from your fNIRS signals [41].

• Objective: To remove the influence of systemic physiological noise from the scalp from long-channel fNIRS data.
• Rationale: Short-separation channels (typically 8 mm apart) are predominantly sensitive to hemodynamic changes in the scalp. By using them as a regressor, their contribution can be statistically removed from nearby long-separation channels (typically 30 mm apart), which capture both cerebral and scalp signals [41].
• Experimental Protocol:
  • Setup: Use an fNIRS system and cap that supports both long-separation (~30 mm) and short-separation (~8 mm) channels. Each short channel should be placed near the long channels it will regress [41].
  • Data Collection: Collect data simultaneously from all channels during your experimental task.
  • Analysis:
    • For each long channel, identify the nearest short-separation channel.
    • Use a general linear model (GLM) or similar regression technique. Enter the short-channel signal as a nuisance regressor to model and remove the scalp-based hemodynamic component from the long-channel signal [41].
• Expected Outcome: SCR enhances the statistical effects of your task conditions on the measured hemodynamic responses. It improves contrast-to-noise ratio and increases the number of significant channels, even in low-motion cognitive tasks [41].

Guide 2: Optimizing Hardware and Probe Design

The physical setup of your fNIRS system is a first line of defense against noise.

• Strategy: Utilize High-Density (HD) fNIRS arrays.
• Rationale: Compared to traditional sparse arrays, HD arrays with overlapping, multi-distance channels offer superior depth sensitivity and spatial resolution [3]. This improved configuration provides a more accurate separation of cerebral signals from superficial artifacts.
• Evidence: A direct comparison showed that while sparse arrays (30 mm spacing) could detect activation during high cognitive load tasks, HD arrays outperformed them in precisely localizing brain activity, especially during tasks with lower cognitive loads [3].
• Consideration: Weigh the benefits of improved localization against the increased setup time and computational cost of HD-DOT [3]. For studies where broad detection is sufficient, a sparse array with SCR may be adequate.

Guide 3: Standardizing Data Acquisition and Analysis

Inconsistent methodology is a major source of unreliable data and increased vulnerability to noise.

• Problem: Variability in how researchers handle poor-quality data, model responses, and conduct statistical analyses is a primary driver of non-reproducible fNIRS results [30].
• Solution:
  • Standardize Optode Placement: Use digitized optode positions for anatomically accurate source localization across sessions [4]. This reduces spatial inaccuracy introduced by cap shifts.
  • Adopt Standardized Pipelines: Use the NIRS-BIDS standard to organize your data [49]. This promotes reproducibility and makes data sharing and re-analysis more reliable.
  • Apply Artifact Correction: Consistently use methods for systemic and extracerebral artifact correction. A systematic review of occupational workload studies found that only 17 out of 41 studies applied such corrections, highlighting a common gap in methodology [50] [51].

Data and Methodologies

Table 1: Quantitative Comparison of fNIRS Array Performance

The following table summarizes key findings from a direct comparison of sparse and high-density (HD) fNIRS arrays, informing hardware selection for your research [3].

Metric	Sparse Array (30mm grid)	High-Density (HD) Array	Implication for Spatial Resolution Research
Spatial Localization	Limited, poor specificity	Superior, precise localization	HD arrays are essential for mapping precise regions of activation.
Sensitivity to Activation	Suitable for high cognitive load (e.g., incongruent Stroop)	Strong for both high and low cognitive load tasks	HD arrays provide more robust detection across varying experimental conditions.
Inter-subject Consistency	Reduced reproducibility	Improved localization consistency	HD arrays yield more reliable group-level results and cross-session comparisons.
Depth Sensitivity	Poor without short channels	Improved with multi-distance channels	HD arrays better separate superficial (noise) from deep (neural) signals.
Practical Setup	Faster setup, less complex	Increased setup time and cost	Choice involves a trade-off between data quality and practical resources.

Table 2: Essential Research Reagent Solutions

This table lists key methodological "reagents" for effective noise mitigation in fNIRS experiments.

Solution Item	Function / Purpose	Key Technical Details
Short-Separation Channels (SSCs)	Directly measures hemodynamic signals from the scalp for use as a nuisance regressor [41].	Typical separation: ~8 mm. Should be placed near long-separation channels they are regressing.
Short-Channel Regression (SCR)	Statistical removal of scalp-derived physiological noise from long-channel fNIRS data [41].	Implement via Generalized Linear Model (GLM). Enhances t-values and contrast-to-noise ratio.
Digitized Optode Localization	Records exact optode positions on the scalp for improved anatomical accuracy and cross-session alignment [4].	Used with subject-specific MRI or a standard head model for source reconstruction.
High-Density (HD) DOT Arrays	Improves spatial resolution, depth sensitivity, and signal localization through overlapping source-detector pairs [3].	Uses multiple source-detector distances (e.g., 10-45 mm). Outperforms sparse arrays in image space.
NIRS-BIDS Standard	Provides a standardized framework for organizing and sharing fNIRS data and metadata, promoting reproducibility [49].	File format: SNIRF. Includes mandatory files for data, channels, optodes, and coordinate systems.

Visual Guide: Signal Contamination and Mitigation Pathway

The following diagram illustrates the pathway of systemic noise contamination and the primary methods to mitigate it, integrating the concepts discussed in this guide.

Individual anatomical differences are a critical confound in functional near-infrared spectroscopy (fNIRS) studies. Variations in head size, cortical folding, and scalp characteristics (including hair thickness and density) significantly influence how near-infrared light propagates through tissues, affecting the sensitivity and accuracy of your measurements [52] [53]. Since fNIRS does not provide intrinsic anatomical information, failing to account for these factors can introduce systematic biases, especially when comparing groups that differ in age, sex, or other characteristics [52]. This guide provides targeted troubleshooting advice to help you identify and mitigate these challenges in your experiments.

---

Troubleshooting Guides & FAQs

FAQ: How does head size and scalp-cortex distance affect my fNIRS signal, and how can I account for it?

The Problem: The distance light must travel from a source on the scalp to the cortical brain tissue and back to a detector is not consistent across participants. This distance is influenced by head circumference and the thickness of extracerebral layers (scalp, skull, CSF). A greater scalp-cortex distance results in a weaker and more attenuated brain signal [52] [53].

Troubleshooting Steps:

• Estimate the Depth: For standard 3 cm source-detector separations, you can use the Differential Pathlength Factor (DPF) and Partial Pathlength Factor (PPF) to estimate the effective light pathlength through the brain. Be aware that these factors are wavelength-specific and vary with age and individual anatomy [52].
• Incorporate Structural Data: Where resources allow, use subject-specific anatomical MRI data to create individual head models. Software like AtlasViewer and Nirstorm can use these models to calculate a sensitivity profile for each channel, quantifying how much of the signal originates from the brain versus extracerebral tissues [54] [9].
• Use Population Averages: If subject-specific MRI is not available, use probabilistic atlases (e.g., Colin27, ICBM152) to estimate typical scalp-cortex correlations for your optode placement [5] [9]. Be aware that this method will not capture individual variability.

Experimental Protocol: Quantifying Scalp-Cortex Correlation

• Objective: To establish a mapping between fNIRS optode positions and the underlying brain regions for a specific subject, accounting for their unique anatomy.
• Materials: Subject-specific structural MRI (T1-weighted), fNIRS cap with digitized optode positions, software for light modeling (e.g., AtlasViewer, Nirstorm).
• Method:
  • Acquire a high-resolution structural MRI scan with fiducial markers (e.g., MRI-visible markers at nasion, inion, pre-auricular points) or a subsequent digitization of optode locations.
  • Segment the MRI data into different tissue types (scalp, skull, CSF, gray matter, white matter).
  • Register the fNIRS optode positions to the segmented head model.
  • Perform Monte Carlo simulations of light transport to compute a spatial sensitivity profile (SSP) for each source-detector channel on the gray matter surface [53] [54].
  • The resulting sensitivity map shows which cortical areas each channel is most sensitive to, moving beyond a simple geometrical projection.

FAQ: My signals are noisy and variable across subjects. Could cortical folding be the cause?

The Problem: The fNIRS signal is highly sensitive to the gyral-sulcal pattern beneath the optodes. Channels positioned over sulci (the folds of the brain) have a significantly longer path to the cortical surface and thus a much weaker sensitivity to brain activation compared to those over gyri [54]. Individual differences in cortical folding mean that the same optode placement on two different subjects may be measuring from entirely different functional brain areas.

Troubleshooting Steps:

• Adopt Sensitivity-Based Matching (SBM): Move beyond simple geometrical matching (GM), which assumes the measured brain region is directly below the channel midpoint. Implement SBM, which uses light propagation models to determine the precise brain regions a channel is sensitive to, given the individual's cortical structure [53].
• Use High-Density (HD) Arrays: Sparse optode layouts have poor spatial resolution and cannot distinguish between signals from gyri and sulci. HD-fNIRS arrays with overlapping, multi-distance channels improve spatial resolution and depth discrimination, leading to better localization of brain activity and more consistent results across subjects [3] [54].
• Leverage Probabilistic Atlases: Tools like the fNIRS Optodes' Location Decider (fOLD) toolbox use pre-computed photon transport simulations on head atlases to recommend optode placements that maximize anatomical specificity to your desired brain regions-of-interest (ROIs) [5].

Table 1: Impact of Cortical Folding on fNIRS Sensitivity

Cortical Feature	Impact on fNIRS Signal	Recommended Mitigation Strategy
Gyrus (ridge)	Shorter path from scalp; higher sensitivity and signal strength [54]	Use sensitivity-based mapping to confirm channel sensitivity to target ROI.
Sulcus (groove)	Longer path from scalp; lower sensitivity and attenuated signal [54]	Implement high-density (HD-DOT) arrays to improve depth resolution [3].
Inter-subject Variability in Folding	The same scalp position can correspond to different brain areas across subjects [53]	Guide optode placement using individual fMRI or probabilistic functional maps (fOLD) [2] [5].

FAQ: What is the best way to deal with hair to ensure good signal quality?

The Problem: Hair is a major obstacle for fNIRS, as it creates a barrier between the optode and the scalp, reducing light coupling and increasing signal loss. Dark, thick hair absorbs more light, further degrading the signal-to-noise ratio (SNR).

Troubleshooting Steps:

• Part the Hair: Meticulously part the hair using a blunt tool to create a clear path for each optode to make direct contact with the scalp.
• Use Abundant Gel: Optical gel not only improves optical coupling but also helps to hold the parted hair away from the optode.
• Select Appropriate Optode Holders: Choose holder designs and spring tensions that are robust enough to maintain good contact pressure through the hair.
• Implement Real-Time Quality Control: Use the instrument's data quality check features (e.g., signal strength, coefficient of variation) during setup to identify and fix problematic channels before data collection begins.

---

Table 2: Key Resources for Anatomically-Informed fNIRS Research

Resource / Tool	Function	Example Use Case
AtlasViewer	Open-source software for probe design, registration to anatomical atlases, and visualization of sensitivity profiles [9].	Designing an optimal probe layout and projecting measured data onto a cortical surface for group-level analysis.
fOLD Toolbox	Automatically recommends optode locations from a predefined set (10-10/10-5 system) to maximize specificity to user-defined brain regions-of-interest [5].	Determining the best scalp positions to target the Dorsolateral Prefrontal Cortex (dlPFC) before fabricating a cap.
Nirstorm / Brainstorm	Open-source software for M/EEG and fNIRS analysis; includes pipelines for processing fNIRS data and integrating with subject-specific anatomy (SSA) [54].	Solving the forward problem (computing sensitivity matrices) and the inverse problem (mapping signals to the cortex) using individual MRI data.
Subject-Specific Anatomy (SSA)	Anatomical MRI from an individual participant. Used to create accurate head models for light transport simulation [54].	Accounting for an individual's unique scalp-skull-brain geometry to accurately interpret channel data.
Population-Averaged Brain Atlases	Digital head models (e.g., Colin27, ICBM152) representing a standardized anatomy.	Used when subject-specific MRI is unavailable, for probe design and initial anatomical guidance [5] [9].

---

Experimental Workflow: From Probe Design to Cortical Mapping

The following diagram summarizes the key steps for an fNIRS experiment that robustly accounts for individual anatomical variability.

Fundamental Principles and Key Definitions

What is the relationship between interoptode distance, penetration depth, and signal quality? The interoptode distance (IOD), which is the physical separation between a light source and a detector on the scalp, is a primary factor determining the sensitivity profile of an fNIRS measurement. The light emitted from a source propagates through head tissues in a banana-shaped path before reaching the detector [55]. The IOD directly influences two critical and competing parameters:

• Penetration Depth: The maximum depth of light penetration into the cortex is approximately half the source–detector distance [56]. Therefore, a larger IOD is required to probe deeper cortical structures.
• Signal-to-Noise Ratio (SNR): The light intensity decreases exponentially with distance due to scattering and absorption in tissue. As the IOD increases, the number of photons reaching the detector drops drastically, leading to a lower SNR [2] [1]. A good optical coupling between the optodes and the scalp is essential to maintain SNR, as poor contact can lead to light loss and motion artifacts [57] [43].

Table 1: Key Technical Definitions

Term	Definition	Relevance to IOD Optimization
Penetration Depth	The maximum depth in tissue from which usable signal can be obtained.	Roughly 1/2 the IOD. A 30 mm distance probes ~15 mm deep [56].
Signal-to-Noise Ratio (SNR)	The ratio of the power of the brain signal of interest to the power of background noise.	Decreases with increasing IOD due to exponential light attenuation [1].
Sensitivity Profile	The spatial distribution of tissue regions that contribute to the measured fNIRS signal.	The region of highest sensitivity is typically cortical gray matter beneath the channel midpoint [5].
Phonon Migration	The path of near-infrared light through highly scattering biological tissues.	Follows a "banana-shaped" path between source and detector [55].

Troubleshooting Common Experimental Challenges

FAQ 1: My fNIRS signals are consistently weak across all channels. What could be the cause? Weak signals are often related to insufficient light reaching the detectors. Please check the following:

• Optode Coupling: Ensure all optodes have firm and stable contact with the scalp. Hair, dry skin, or poor cap fit can block or scatter light [57] [1].
• Interoptode Distance: Verify that your IOD is within the recommended range. Distances that are too large will lead to severe attenuation, while distances that are too short may only sample the scalp [2] [56].
• Equipment Function: Check light source intensity and detector sensitivity. Ensure fiber optics are not damaged or overly bent.

FAQ 2: How can I confirm that my optodes are targeting the correct region of the cortex? Accurate spatial targeting is a common challenge. The literature-based approach using the 10-20 system is a good start but can be suboptimal due to individual anatomical differences [2].

• Use Neuronavigation: For critical applications, use a neuronavigation system with individual MRIs to co-register optode positions with underlying anatomy [2] [1].
• Leverage Probabilistic Tools: Use software toolboxes like the fNIRS Optodes' Location Decider (fOLD), which uses photon transport simulations on head atlases to guide optode placement for maximum sensitivity to specific brain regions-of-interest [5].

FAQ 3: My data is contaminated by strong physiological noise (e.g., heart rate, blood pressure waves). How can I mitigate this? fNIRS signals are inherently contaminated by physiological noises.

• Apply Digital Filtering: Use a band-pass finite impulse response (FIR) filter to remove noise outside the frequency range of the hemodynamic response. A common practice is to use a 1000th order band-pass FIR filter to preserve the task-evoked response while removing cardiac (~1 Hz) and respiratory (~0.3 Hz) oscillations [56].
• Short-Separation Channels: Incorporate short-separation channels (e.g., IOD < 10 mm). These channels are predominantly sensitive to systemic artifacts in the scalp and can be used as regressors to clean the signals from standard channels [56] [43] [1].

Experimental Protocols for IOD Validation

Protocol: Systematically Evaluating the IOD-SNR Relationship

Aim: To empirically determine the optimal IOD for a specific fNIRS system and experimental setup by quantifying the relationship between distance and signal quality.

Materials:

• fNIRS system with multiple source-detector pairs.
• Optode holder cap capable of precise distance adjustments.
• A healthy, consenting participant.

Methodology:

• Setup: Place a single source and a detector on the prefrontal cortex. Ensure excellent optode-scalp coupling.
• Data Acquisition: Acquire resting-state fNIRS data at a minimum of five different IODs (e.g., 15 mm, 20 mm, 25 mm, 30 mm, 35 mm). Record each distance for 3-5 minutes.
• Signal Quality Metric: For each IOD, calculate the Relative Light Intensity at the detector. This is a direct proxy for SNR, as higher received intensity indicates a stronger signal against the system's noise floor.

Table 2: Example Data Sheet for IOD-SNR Validation

Interoptode Distance (mm)	Theoretical Penetration Depth (mm)	Mean Relative Light Intensity (a.u.)	Standard Deviation	Recommended for Cortical Studies?
15	7.5	1,500,000	15,000	No (Too Superficial)
25	12.5	250,000	8,000	Yes (Good Balance)
30	15.0	80,000	5,000	Yes (Optimal)
35	17.5	15,000	2,000	Marginal (Low SNR)
40	20.0	2,000	500	No (Poor SNR)

Analysis:

• Plot the Mean Relative Light Intensity against the IOD. An exponential decay is expected.
• Identify the "knee" of the curve—the point beyond which intensity drops precipitously. The IOD just before this drop often represents the best trade-off between penetration depth and SNR for your system.

Essential Research Reagents and Tools

Table 3: The Scientist's Toolkit for fNIRS Optode Optimization

Tool / Reagent	Category	Function & Application
fOLD Toolbox [5]	Software	Automates optode placement on the 10-5/10-10 systems using photon Monte Carlo simulations on head atlases to maximize sensitivity to target ROIs.
Montreal Neurological Institute (MNI) Atlas	Template	Standard stereotaxic space for reporting and simulating optode locations and their sensitivity profiles [2] [5].
3D-Printed Custom Helmets [57] [43]	Hardware	Enables precise, stable, and subject-specific co-registration of fNIRS optodes (and EEG electrodes), improving spatial specificity and reproducibility.
Short-Separation Detectors	Hardware	Specialized detectors placed 8-10 mm from a source to measure systemic physiological noise from the scalp, used as a regressor to enhance cortical signal quality [56] [1].
Neuronavigation System	Hardware	Uses individual MRI data to track and confirm the real-time position of fNIRS optodes on the scalp relative to the participant's underlying cortical anatomy [2].

Signaling Pathways and Experimental Workflows

fNIRS Experiment Workflow

Signal & Artifact Separation

Frequently Asked Questions (FAQs)

1. What are the most critical factors for achieving good signal quality in real-time fNIRS? Maintaining high signal quality in real-time fNIRS depends on two pillars: ensuring good spatial specificity (accurately targeting brain regions) and effectively managing signal contaminants like motion artifacts and physiological noise. Unlike offline analysis, real-time processing cannot correct data after acquisition, making robust preprocessing techniques essential to prevent the system from operating on noise rather than brain activity [12] [1].

2. How can I minimize false positives in my fNIRS-BCI experiment? False positives—where non-neuronal hemodynamic changes are mistaken for brain activity—can be minimized through several strategies [58]:

• Careful Experimental Design: Avoid tasks that provoke strong systemic responses (e.g., significant changes in heart rate, blood pressure, or respiration).
• Depth-Resolved Techniques: Use multidistance measurements to help separate cerebral signals from extracerebral confounders.
• Signal Processing: Employ methods like adaptive filtering or principal component analysis (PCA) to remove systemic physiological noise.

3. Is there a minimum number of channels required for effective neurofeedback? No, effective neurofeedback does not strictly require a high number of channels. While some approaches like 19-channel quantitative EEG (QEEG) exist, many neurofeedback protocols can function successfully with only 2 or 4 channels. The choice depends on the specific clinical or experimental goals [59].

4. Why is optode placement so crucial, and how can I optimize it? Optode placement directly determines the quality of the measured signal and the system's sensitivity to your target brain region. Incorrect placement can result in missing the region of interest entirely or capturing excessive noise [11] [60]. Optimization can be achieved by using subject-specific anatomical data (from MRI) to guide placement, ensuring the optodes are positioned over the cortical areas you intend to monitor [11].

Troubleshooting Guides

Problem: Low Classification Accuracy in BCI

Possible Causes and Solutions:

• User-Related Factors

  • Cause: The user may be inexperienced, unmotivated, fatigued, or using an incorrect mental strategy. Inherent physiological differences can also make the brain signals less distinguishable [61] [62].
  • Solution: Ensure proper user training and motivation. The BCI paradigm should be well-explained, and sessions should be kept engaging and not overly long. For some paradigms like motor imagery, extensive training over multiple days may be necessary [61] [62].

• Hardware and Acquisition Issues

  • Cause: Poor optode-scalp coupling, which leads to a low signal-to-noise ratio [61].
  • Solution: For prolonged acquisitions, consider using a clinical adhesive like collodion to secure optodes and maintain excellent optical contact for several hours. This is particularly useful for subjects with challenging hair types [11].
  • Cause: Suboptimal optode placement, missing the target region of interest [60].
  • Solution: Utilize neuronavigation systems guided by individual MRI data to place optodes with high precision over the target brain areas [11].

• Software and Processing Issues

  • Cause: The signal processing pipeline is not adequately removing motion artifacts (MAs) [63] [56].
  • Solution: Implement a real-time, deep-learning-based denoising autoencoder (DAE) model with a sliding window strategy. This has been shown to outperform traditional MA correction methods in terms of mean squared error and correlation with clean data while maintaining the low latency required for real-time applications [63].
  • Cause: Unoptimized parameters in the classification algorithm [61] [62].
  • Solution: Re-calibrate and tune the classifier parameters for the individual user, as brain signals can vary significantly between sessions and users [61].

Problem: Poor Signal-to-Noise Ratio (SNR) and Excessive Motion Artifacts

Recommended Protocol: Real-Time Motion Artifact Correction [63]

This protocol leverages a deep-learning model for superior motion artifact correction.

• 1. Baseline Calibration: Establish a baseline for the real-time measurement at the start of the experiment.
• 2. Sliding Window Application: Continuously feed data to the processing system using a short, sliding time window.
• 3. DAE Processing: Process the data within the window through a pre-trained Denoising Autoencoder (DAE) model. This model automatically corrects for motion artifacts across all channels simultaneously.
• 4. Real-Time Output: Output the cleaned, motion-corrected hemodynamic data for neurofeedback or BCI classification.

This system has demonstrated the capability to process up to 750 fNIRS/DOT channels in real-time, making it suitable for high-density setups.

Problem: Inconsistent Spatial Targeting Across Sessions

Recommended Protocol: Personalized fNIRS Montage [11]

This methodology ensures you are consistently and accurately measuring from the same target brain area in every session.

• 1. Define Target Region: Identify the specific brain Region of Interest (ROI) using the subject's anatomical MRI.
• 2. Compute Optimal Montage: Using the subject's head model and light sensitivity profiles, formulate and solve an optimization problem to determine the set of optode positions on the scalp that provides the best sensitivity to the target ROI.
• 3. Guided Optode Placement: Use a 3D neuronavigation device to place each optode at its computed coordinate on the subject's scalp with high precision.
• 4. Secure Attachment: Attach optodes using collodion or another clinical adhesive to ensure stable optical coupling for the duration of the experiment.

The following table summarizes key quantitative findings from recent research on signal processing and optode layout design.

Table 1: Signal Processing Performance in Real-Time fNIRS

Metric	Traditional MA Methods	Deep-Learning (DAE) Method	Notes
Motion Artifact Correction	Variable performance, often requires manual parameter tuning [63]	Outperforms traditional methods in MSE and correlation with ground truth [63]	DAE uses automated feature extraction and is robust to non-stationary noise [63]
System Latency	Must be maintained low for real-time use [63]	Maintains low latency critical for BCI/NFB [63]	Achieved via a sliding window strategy [63]
Processing Scale	Varies by method	Capable of real-time processing for ~750 channels [63]	Enables high-density DOT applications [63]
Optimal Filtering (Offline)	Various filters used with large heterogeneity [56]	N/A	A 1000th order band-pass Finite Impulse Response (FIR) filter is identified as optimal for recovering the hemodynamic response in a GLM framework [56]

Table 2: Impact of Optode Layout Design on Signal Quality

Design Approach	Description	Key Finding	Practical Implication
Literature-Based (LIT)	Optodes placed based on standard coordinates or literature review [60]	Serves as a baseline, but can be suboptimal for individual users [60]	Most accessible method when subject-specific MRI is unavailable [60]
Probabilistic (PROB)	Uses individual anatomy and probabilistic fMRI maps from independent datasets [60]	Outperforms LIT in signal quality and sensitivity; results similar to iFMRI [60]	A robust compromise when subject-specific fMRI is not available [60]
Individual fMRI (iFMRI)	Uses individual anatomical and functional data [60]	Superior to LIT; results similar to PROB and fVASC [60]	Leverages individual brain function for high spatial specificity [60]
Vascular (fVASC)	Uses individual anatomical, functional, and vascular data [60]	Superior to LIT; results similar to PROB and iFMRI [60]	Incorporates vascular information, but may not be necessary for all applications [60]

Experimental Workflow and Signaling Pathways

Workflow for a Personalized fNIRS Investigation

The following diagram illustrates the end-to-end workflow for setting up a personalized fNIRS study, from defining the target region to analyzing the reconstructed data.

Signaling Pathway: From Neural Activity to fNIRS Measurement

This diagram outlines the physiological and technical pathway from increased neural activity to the measured fNIRS signal, including key sources of false positives.

The Scientist's Toolkit: Key Research Reagents and Materials

Table 3: Essential Materials for Advanced fNIRS Investigations

Item	Function in the Experiment
3T MRI Scanner	Acquires high-resolution anatomical T1-weighted images, functional MRI (fMRI) data, and vascular information to create a personalized head model and define functional Regions of Interest (ROIs) [11] [60].
3D Neuronavigation System	Precisely guides the placement of fNIRS optodes on the subject's scalp according to the coordinates calculated by the optimal montage algorithm, ensuring accurate targeting of the brain region [11].
Collodion Adhesive	A water-resistant clinical adhesive used to glue optodes directly to the scalp. It ensures excellent and stable optical coupling for prolonged acquisitions (6+ hours), which is critical for signal quality and studying tasks under realistic conditions [11].
High-Density DOT System	A high-density diffuse optical tomography system featuring a dense array of sources and detectors. It enables overlapping spatial sampling and the reconstruction of 3D images of cortical hemodynamic activity with enhanced spatial resolution [63].
Denoising Autoencoder (DAE) Model	A deep learning model trained to remove motion artifacts from fNIRS data in real-time. It is a key software component for maintaining signal quality in movement-intensive scenarios like motor rehabilitation BCIs [63].
Physiological Monitoring	Additional devices (e.g., heart rate monitor, blood pressure cuff, capnometer) are used to measure systemic physiological variables. This data helps identify and separate non-neuronal confounders from the desired brain activity [58].

Quantitative Performance Data of Collodion Fixation

The following table summarizes key quantitative findings on the performance of collodion-fixed optical fibers compared to standard Velcro-based probes.

Table 1: Performance Comparison of Collodion vs. Velcro-Based fNIRS Probes

Performance Metric	Collodion-Fixed Probe	Standard Velcro Probe	Improvement/Reduction
Motion Artifact Signal Change	9% [64]	103% [64]	90% reduction [65] [64]
SNR at 690 nm	Increased [64]	Baseline	6-fold increase [65] [64]
SNR at 830 nm	Increased [64]	Baseline	3-fold increase [65] [64]
SNR at Rest (no motion)	Increased [64]	Baseline	2-fold increase [64]
Typical Signal Quality Duration	≥ 6 hours [11]	Varies	Prolonged stability for long-term monitoring

Detailed Experimental Protocol: Collodion Application for fNIRS

This protocol details the methodology for affixing fNIRS optodes using collodion, as validated in clinical studies [65] [64] [11].

Materials and Equipment

• fNIRS Optodes: Miniaturized optical fiber tips are recommended. A typical design includes a glass prism, a mirrored surface, and a prism-housing [65] [64].
• Clinical Adhesive: Collodion (e.g., Mavidon, FL) [65] [64].
• Collodion-Impregnated Gauze: Squares of approximately 2–3 cm [65].
• Personal Protective Equipment (PPE): Gloves, lab coat.
• Other Materials: Cotton-tipped applicators, compressed air canister, towel to protect patient clothing [65].

Step-by-Step Procedure

• Subject Preparation: Place a towel around the subject's shoulders to protect clothing from adhesive. Ensure the subject is in a comfortable and stable position [65].
• Scalp Preparation: Part the hair at the desired optode location using a cotton-tipped applicator to expose the scalp [65].
• Gauze Application: Place a square of collodion-impregnated gauze on the exposed scalp. This gauze will serve as the anchor point for the optode [65].
• Optode Positioning: Place the tip of the miniaturized optical fiber onto the gauze at the desired location.
• Adhesive Drying: Use compressed air to dry the collodion, firmly adhering the gauze and optode to the scalp [65].
• Curing: Allow a few moments for the collodion to fully set and create a secure, water-resistant bond [11].

Safety and Environmental Notes

• The procedure should be performed in a well-ventilated room to dissipate the fumes emitted by collodion during installation [11].
• Use only clinical-grade collodion intended for this purpose.

Troubleshooting Guide & FAQs

Q1: We are getting poor signal quality even with collodion-fixed probes. What could be the issue?

• A: Ensure that the scalp is properly prepared before application. Part the hair thoroughly to expose the skin and ensure the collodion-soaked gauze makes direct contact with the scalp. Inadequate scalp contact is the most common cause of poor signal [65] [11].

Q2: How long does the collodion fixation typically last, and how is it removed?

• A: A properly applied collodion fixation can maintain excellent signal quality for at least 6 hours [11]. For removal, specific collodion removers (solvents) are available that safely dissolve the adhesive without damaging the optodes or causing discomfort to the subject.

Q3: Are there any subject populations for which collodion fixation is particularly advantageous?

• A: Yes. This method is highly beneficial for vulnerable populations who are likely to move frequently, such as infants, children, and patients with neurological conditions like epilepsy. It is especially critical for studying events intrinsically linked to movement, such as epileptic seizures, where it has enabled the collection of good-quality data despite excessive motion [65] [64].

Q4: How does collodion fixation compare to post-processing algorithms for motion artifact correction?

• A: While algorithms like wavelet filtering, spline interpolation (MARA), and PCA are effective at correcting motion artifacts in post-processing, they operate on data that is already contaminated [65] [64] [66]. Collodion fixation is a prospective approach that minimizes the introduction of motion artifacts at the source by robustly coupling the optode to the scalp. It is considered more effective to avoid the artifact altogether than to try to remove it later [65] [64].

Workflow Diagram: Implementing Collodion-Fixed fNIRS

The following diagram illustrates the key stages of implementing a collodion-fixed fNIRS setup, from preparation to data acquisition.

The Scientist's Toolkit: Essential Research Reagents & Materials

Table 2: Essential Materials for Collodion-Fixed fNIRS Experiments

Item	Function/Description	Example/Note
Collodion Adhesive	Fast-drying, water-resistant clinical glue that forms a flexible film to fix optodes to the scalp.	Mavidon brand Collodion [65] [64].
Miniaturized Fiber Tips	Low-profile optodes designed specifically for secure fixation with adhesive.	Includes a glass prism and mirrored housing [65] [64].
Collodion-Impregnated Gauze	Small squares of gauze that act as a physical interface and anchor point between the optode and scalp.	~2-3 cm squares [65].
Compressed Air	Used to rapidly dry the collodion adhesive, securing the optode in place.	Standard medical-grade canister [65].
3D Neuronavigation System	(Optional but recommended) Guides precise placement of optodes over subject-specific target brain regions.	Used for "optimal montage" methodologies [2] [11].
fNIRS Processing Software	For data analysis and application of motion artifact correction algorithms if needed.	e.g., HOMER2, which includes tools like MARA [66].

Functional near-infrared spectroscopy (fNIRS) is a non-invasive neuroimaging technique that measures hemodynamic changes associated with brain activity by detecting near-infrared light passed through the scalp and skull [11]. Its clinical application has been limited by challenges in maintaining good optical coupling and achieving optimized optode coverage of specific brain regions, particularly for prolonged monitoring [11]. The precision of optode placement is critical for data quality and reproducibility, as traditional placement methods using external landmarks or standardized caps do not account for individual neuroanatomical variations [67].

3D neuronavigation integrates individual MRI data with real-time optode tracking to provide precise guidance for optode positioning, bridging the gap between individual anatomy and measurement setup [11] [67]. This technical support document provides comprehensive troubleshooting guidance and experimental protocols for researchers integrating 3D neuronavigation into fNIRS studies, particularly within the context of optimizing optode placement for spatial resolution research.

Essential Research Toolkit

Table 1: Essential Components for fNIRS Neuronavigation Research

Component	Function & Specification	Research Considerations
Neuronavigation System (e.g., Brainsight TMS)	Visualizes individual's brain based on MRI data to navigate and position optodes over specified cortical targets [68].	Compatible with various MRI formats (DICOM, NIfTI); allows target definition via anatomy, coordinates, or functional overlays (fMRI, EEG, fNIRS) [68].
Individual Anatomical MRI	Provides the structural dataset for creating a 3D head model and calculating light sensitivity profiles for optimal montage planning [11] [15].	Essential for personalized montages; template anatomy (e.g., Colin27) can be used for protocol development but lacks individual specificity [15].
fNIRS Optodes & Hardware	Sources emit and detectors receive NIR light (e.g., 690-850 nm) to measure concentration changes in HbO and HbR [11].	System should be compatible with the navigation environment; consider electromagnetic interference in MRI settings [69] [19].
Optical Probe/Holder Design	Secures optodes in the navigated positions on the scalp.	Design affects practicality and stability; flexible setups allow for optimal montages beyond fixed grid caps [11].
Clinical Adhesive (e.g., Collodion)	Glues optodes to the scalp to ensure stable optical coupling for prolonged acquisitions (≥6 hours) [11].	Requires a properly ventilated room; enables studies on subjects with any hair color [11].
Software for Optimal Montage (e.g., NIRSTORM/Brainstorm)	Computes the set of optode positions that maximize sensitivity to target Regions of Interest (ROIs) using the individual's head model [15].	Requires solving a mixed linear integer programming problem (e.g., using IBM CPLEX); depends on pre-computed light propagation (fluence) maps [15].

Frequently Asked Questions (FAQs)

Q1: What are the primary documented benefits of using neuronavigation for fNIRS optode placement? Using a real-time neuronavigation protocol to guide optode placement significantly increases the within-subject reproducibility of fNIRS data [67]. Unlike traditional cap-based approaches, which often show poor intra-subject reproducibility, neuronavigation enables consistent and robust activation detection across multiple sessions on the same and different days [67]. Furthermore, it is a core component of a procedure for personalized fNIRS investigations, allowing for accurate targeting of specific brain regions with fewer optodes, thereby reducing setup time and improving subject comfort [11].

Q2: My lab has access to an MRI but not a commercial neuronavigation system. Can I still implement a form of 3D-guided optode placement? Yes, a feasible alternative is to use software like NIRSTORM within Brainstorm to compute an optimal, personalized montage based on the individual's MRI [15]. This method involves:

• Segmenting the scalp and cortical surfaces from the T1-MRI.
• Pre-computing light propagation (fluence) patterns for the scalp vertices.
• Defining your target cortical region (e.g., manually or via an atlas).
• Running an optimization algorithm to find the best set of source and detector positions on the scalp that maximize sensitivity to your target [15]. The calculated optode positions, defined in the individual's anatomical space, can then be measured and marked on the scalp using a physical measurement device, providing a customized guidance approach.

Q3: How does the use of collodion improve fNIRS investigations, and when is it necessary? Using a water-resistant adhesive like collodion to glue optodes onto the scalp is particularly beneficial for prolonged acquisitions (e.g., several hours) and in populations with dense or dark hair that can interfere with optical contact [11]. Based on research experience, collodion can maintain optical signals of excellent quality for at least 6 hours. Its key advantage is that it enables stable investigations in almost any subject, as hair can be moved aside during the gluing process. This method requires a properly ventilated room to dissipate fumes [11].

Q4: What are the key trade-offs between using a high-density (HD) fNIRS array versus a sparse, optimized array planned with neuronavigation? High-density arrays with overlapping, multi-distance channels provide superior sensitivity and localization, especially for detecting activity during lower cognitive load tasks [3]. However, they come with increased costs, longer setup times, and greater computational demands for data processing [3]. In contrast, a sparse but optimal montage derived via neuronavigation is designed to maximize sensitivity to a specific, pre-defined brain region while significantly reducing the number of optodes [11]. Simulations show that maps from optimal montages can achieve spatial resolution only slightly lower than ultra-high-density montages while being more practical for clinical or realistic lifestyle settings [11]. The choice depends on whether the research goal requires broad, high-fidelity mapping (favors HD) or efficient, targeted investigation (favors optimized sparse montage).

Troubleshooting Common Issues

Problem: Poor or Inconsistent Signal Quality After Navigated Placement

• Potential Cause 1: Inadequate Optical Coupling. Even with perfect placement, signal quality depends on a clear path between the optode and scalp.
• Solution: Ensure the scalp is clean at the target positions. Part the hair thoroughly and use enough conduction gel. For long studies, strongly consider using collodion to secure the optode and maintain coupling [11].
• Potential Cause 2: Systemic Physiological Interference. The fNIRS signal is susceptible to contamination from systemic factors like heart rate, blood pressure, and scalp blood flow [67].
• Solution: Integrate short-separation channels into your montage. These channels are primarily sensitive to superficial, extracerebral layers and their signals can be regressed out from the standard long-channel measurements to improve the specificity of the brain signal [3]. Monitor heart rate and blood pressure across sessions to account for this variability [67].

Problem: Low Reproducibility of Activation Across Sessions

• Potential Cause: Inconsistent Optode Positioning. Without navigation, the primary cause of low intra-subject reproducibility is the inability to place optodes in the same scalp location relative to the underlying cortex across different sessions [67].
• Solution: Implement a neuronavigation protocol for all sessions. Research has demonstrated that using real-time neuronavigation to guide probe positioning directly increases within-subject reproducibility by ensuring the montage is sensitive to the same cortical region each time [67].

Problem: Difficulty Integrating fNIRS with an MRI Environment for Synchronous Data Collection

• Potential Cause: Hardware Incompatibility and Electromagnetic Interference. Standard fNIRS equipment can interfere with, or be interfered by, the MRI scanner [69] [19].
• Solution: Use an MRI-compatible fNIRS system with components (optodes, cables) specifically designed to operate safely inside the scanner without causing artifacts or being damaged. These systems use fiber-optic cables to separate the control unit (outside the scanner room) from the subject interface (inside the bore) [69] [19].

Problem: The Optimal Montage Algorithm Fails to Find a Valid Solution

• Potential Cause 1: Overly Restrictive Constraints. The optimization problem may be unsolvable if the constraints (e.g., number of sources/detectors, minimum source-detector distance, adjacency number) are too strict for the given target and search space [15].
• Solution: Relax the constraints. Start with a higher number of allowed optodes and a larger minimum source-detector distance, then iteratively refine the parameters. The "adjacency number" (minimum number of detectors each source must connect with) is particularly important for ensuring overlapping measurements [15].
• Potential Cause 2: Incorrect or Missing Anatomical Data. The algorithm requires a proper head model and pre-computed fluence maps for the sensitivity calculations [15].
• Solution: Verify that the subject's MRI has been properly processed (segmentation of scalp, skull, CSF, and gray/white matter) and that the fluence maps have been generated for the correct wavelengths and cover the intended search space on the scalp [15].

Experimental Protocols & Data Presentation

Protocol: Validating Neuronavigation for Intra-Subject Reproducibility

This protocol is based on a study that tested the hypothesis that precise anatomical information increases fNIRS reproducibility [67].

1. Subjects & Sessions:

• Recruit healthy volunteers.
• Plan multiple sessions (e.g., 3-5) per subject, conducted at the same time of day to minimize circadian effects, with additional sessions at different times if investigating physiological variability [67].

2. Anatomical Data & Montage Design:

• Acquire a T1-weighted MRI for each subject.
• For a within-subject comparison, define a target region (e.g., the hand knob area of the primary motor cortex). Design a single montage optimized for this target [67].

3. fNIRS Data Acquisition:

• Task: Use a block-designed protocol, such as a contralateral hand finger-tapping task (e.g., 30 blocks of 2-second stimulation interleaved with 10-20 second random rest periods) [67].
• Setup: Use a continuous-wave fNIRS system. The montage should include both standard long-distance channels (~3 cm) and short-separation channels (~0.8 cm) for superficial signal regression [67].
• Procedure: In each session, use the neuronavigation system to place the optodes in the pre-defined positions. Record physiological parameters (heart rate, blood pressure) before and after the task [67].

4. Data Analysis:

• Pre-process the data (filtering, motion artifact correction) and convert optical density changes to HbO and HbR concentrations.
• Use the General Linear Model (GLM) to assess activation for each channel and session.
• Calculate reproducibility metrics (e.g., intra-class correlation coefficient) for the activation strength (beta values) across sessions, particularly for the channel over the target motor cortex.

Table 2: Key Findings from a Neuronavigation Reproducibility Study [67]

Metric	Traditional Cap-Based Placement	Neuronavigation-Guided Placement
Within-Subject Reproducibility	Poor to acceptable	Consistent and robust
Activation in Target ROI	Variable across sessions	Highly consistent across sessions
Impact of Systemic Physiology	Significant confound	Still present, but less impact on core activation pattern
Conclusion	High variability not solely due to physiology; lack of spatial precision is a major factor.	Significantly increases intra-subject reproducibility.

Protocol: Implementing a Personalized Optimal fNIRS Montage

This protocol outlines the steps for a personalized investigation using the methodology detailed in the NIRSTORM tutorial and related research [11] [15].

1. Subject-Specific Anatomical Modeling:

• Acquire an individual T1-MRI.
• Process the MRI using segmentation software (e.g., Brainvisa, Freesurfer) to generate meshes of the scalp, skull, and cortical surface [11] [15].

2. Light Propagation (Fluence) Calculation:

• Use Monte Carlo simulations (e.g., with MCXlab) to compute the light sensitivity profiles for every possible source and detector position on the scalp mesh. This simulates how photons travel through the head tissues [15].

3. Define the Target and Constraints:

• Target Region of Interest (ROI): Define on the cortical surface. This can be done manually, using an anatomical atlas (e.g., Destrieux, Desikan-Killiany), or based on functional localizer data from fMRI or EEG [15].
• Optimization Constraints: Set the functional parameters for the montage:
  • Number of sources and detectors.
  • Minimum and maximum source-detector distance (e.g., 26-34 mm).
  • Adjacency number: The minimum number of channels each source must form, ensuring spatial overlap for better reconstruction [15].

4. Compute the Optimal Montage:

• Run the optimal montage algorithm (e.g., in NIRSTORM) which solves a mixed linear integer programming problem to select the optode set that maximizes sensitivity to the target ROI under the given constraints [11] [15].

5. Navigated Optode Placement and Data Collection:

• Use a 3D neuronavigation system to place the optodes at the computed coordinates on the subject's scalp [11].
• Secure the optodes, preferably with collodion for longer studies [11].
• Conduct the fNIRS experiment (e.g., a motor or cognitive task).

6. Data Reconstruction and Validation:

• Use Diffuse Optical Tomography (DOT) or similar inverse modeling techniques to reconstruct the local hemodynamic activity on the cortical surface, which improves quantitative accuracy compared to raw channel data [11].
• Validate the results by checking if the peak activation is located within the pre-defined target ROI.

Workflow for Personalized fNIRS with Neuronavigation

Table 3: Example Optimization Constraints and Outcomes for Different Montage Types [11] [3]

Montage Type	Typical Optode Count	Spatial Resolution	Key Advantage	Key Disadvantage
Sparse Grid Array	~Varies (Non-overlapping)	Low (1-3 cm)	Simple, fast setup with commercial systems.	Poor localization, misses focal activity, lower reproducibility [3].
High-Density (HD-DOT)	Many (Overlapping)	High (Approaches fMRI)	Superior localization and sensitivity, especially for low cognitive load [3].	High cost, long setup time, computationally intensive [3].
Optimized Sparse Montage	Low (e.g., 3S, 7D)	Moderate to High (in target region)	Maximizes target sensitivity with minimal hardware; ideal for clinical/longitudinal studies [11].	Coverage limited to pre-defined target region [11].

Frequently Asked Questions: fNIRS Data Quality

Q1: What are the primary indicators of a good quality fNIRS signal? A strong indicator of good signal quality is the presence of a clear cardiac rhythm in the raw light intensity data, particularly in the oxygenated hemoglobin (O2Hb) signal. This pulsation demonstrates effective coupling between the optodes and the scalp. Conversely, signals that are unstable (showing high variability from motion) or flat (showing low variability) are typically classified as poor quality [70].

Q2: Which automated methods can I use to assess signal quality on my fNIRS channels? Several automated algorithms exist to evaluate signal quality. The table below summarizes four common methods, their operating principles, and key considerations for their use [70].

Table 1: Automated Algorithms for fNIRS Signal Quality Assessment

Algorithm	Full Name & Principle	Pros	Cons / Key Considerations
CV	Coefficient of Variation: Measures relative variability in raw light intensity (Standard Deviation / Mean) * 100% [70].	Simple to calculate and automate [70].	Cannot distinguish physiological signals from motion noise; may falsely reject good data; requires careful threshold tuning [70].
SCI	Scalp Coupling Index: Assesses optode-scalp contact by correlating the two wavelength signals (e.g., 760 & 850 nm), as physiological signals affect both similarly [70].	Simple principle; a common threshold for a good channel is >0.75 or >0.80 [70].	Susceptible to motion artifacts that affect both wavelengths equally, potentially leading to false positives [70].
PHOEBE	Placing Headgear Optodes Efficiently Before Experimentation: Combines SCI with spectral analysis of the cross-correlation between wavelengths, focusing on the power of the cardiac peak [70].	Improved sensitivity by specifically isolating the heartbeat component, even in the presence of motion [70].	More complex than SCI [70].
SQI	Signal Quality Index: A multi-stage algorithm that detects the cardiac component to provide a numeric quality rating on a scale from 1 (poor) to 5 (excellent) [70].	Provides a more nuanced, flexible, and informative assessment than binary good/bad classification; demonstrated higher performance [70].	Algorithm is more sophisticated [70].

Q3: How do analysis choices impact the reproducibility of my fNIRS results? The reproducibility of fNIRS findings can vary significantly with data quality, analysis pipeline choices, and researcher experience. Key sources of variability across research teams include how poor-quality data is handled, how the hemodynamic response is modeled, and the specifics of statistical analysis. Reproducibility is higher for strong, literature-supported hypotheses and in teams with greater fNIRS experience [30]. Furthermore, using source localization (image space) instead of just channel data improves the reliability of capturing brain activity across sessions [4].

Q4: What are the most impactful steps I can take during data acquisition to ensure signal reliability? To ensure reliable signals, focus on optode placement and stability. Even small shifts in optode position between sessions can reduce the spatial overlap of measured brain activity [4]. Furthermore, the adherence and fit of the cap are critical, as specific head movements (like upward/downward motions) can introduce motion artifacts, with certain brain regions being more susceptible to particular types of movement [46].

Q5: How can my experimental design choices affect data quality? Using a block design paradigm, where task periods alternate with rest periods, can help achieve a high signal-to-noise ratio and robust measurements. However, this design is sensitive to periodic physiological confounds like heart and respiration rates. To mitigate this, it is recommended to "jitter" or slightly randomize the duration of the rest periods (e.g., 28-32 seconds instead of a fixed 30 seconds) to prevent these confounds from aligning with your task rhythm [71].

Experimental Protocols for Key Assessments

Protocol 1: Establishing Signal Quality with the SQI Algorithm

The Signal Quality Index (SQI) provides a robust, automated method for channel assessment [70].

• Data Input: Use the raw light intensity signals from your fNIRS system.
• Processing Stages:
  • Stage 1 - Identify Low Quality: The algorithm first identifies very low-quality signals.
  • Stage 2 - Identify High Quality: It then identifies very high-quality signals.
  • Stage 3 - Signal Rating: The remaining signals undergo a final rating process.
• Output: Each channel is assigned a numeric score from 1 (poor) to 5 (excellent), allowing you to make informed decisions about which channels to include in your analysis [70].

Protocol 2: Quantifying Within-Subject Reproducibility

This protocol outlines a method to assess the reproducibility of fNIRS signals across multiple sessions in the same individual [4].

• Participant & Sessions: Recruit participants to complete at least ten separate testing sessions on different days.
• Task Paradigm: Employ well-defined block-design tasks (e.g., motor tasks like finger tapping, visual stimuli) to evoke a consistent hemodynamic response.
• Data Acquisition: Use a high-density cap (e.g., 102 channels) covering the entire head. For highest accuracy, digitize the optode positions in each session.
• Data Analysis:
  • Preprocessing: Apply standard filtering and motion correction to your raw data.
  • Level of Analysis: Analyze data at both the channel level and the source level using anatomical head models from the digitized optode positions.
  • Quantification: For each session, identify channels or source vertices showing statistically significant task-related activity. Calculate reproducibility as the percentage of sessions in which a given channel/vertex shows significant activation.

Visual overview of the multi-session reproducibility assessment protocol.

The Scientist's Toolkit: Key Reagents & Materials

Table 2: Essential Materials for fNIRS Data Quality Assessment

Item / Solution	Function & Application in Quality Control
High-Density fNIRS System	Systems with multiple overlapping source-detector pairs enable high-density diffuse optical tomography (HD-DOT), which provides superior spatial resolution, sensitivity, and localization of brain activity compared to sparse arrays, directly impacting data quality [3].
Digitization Equipment	A 3D digitizer (e.g., Polhemus) is used to record the precise spatial coordinates of optodes on the participant's head. This is critical for accurate source localization, which improves reproducibility, and for tracking placement consistency across sessions [4].
Computer Vision Software	Software like SynergyNet can analyze video recordings of sessions to compute head orientation angles frame-by-frame. This provides ground-truth movement data to characterize and validate motion artifact correction algorithms [46].
Signal Quality Index (SQI)	An algorithm that acts as a quantitative reagent for data validation. It automatically grades channel quality from 1-5 based on cardiac component presence, allowing researchers to objectively include or exclude data segments [70].
Motion Correction Algorithms	Computational methods (e.g., targeted PCA, wavelet-based filters, or novel neural network frameworks) are applied to raw fNIRS data to identify and remove signal components caused by head movement, thereby improving signal clarity [72] [46].

Evaluating Method Performance, Reproducibility, and Comparative Efficacy

Frequently Asked Questions (FAQs)

Q1: Why does my fNIRS signal reliability drop significantly between experimental sessions?

The most common cause is inconsistent optode placement after cap removal. One study found that test-retest reliability without cap removal was excellent (ICC ≥ 0.78) for several cortical regions but deteriorated substantially after cap removal and repositioning (ICC as low as 0.00 for some regions) [73]. Additional factors include systemic physiological noise, motion artifacts, and insufficient signal preprocessing to separate neuronal signals from confounding physiological components [1] [74].

Q2: Which fNIRS hemoglobin parameter provides more reproducible results?

Oxyhemoglobin (HbO) is generally more reproducible than deoxyhemoglobin (HbR) [4]. Studies across various tasks, including visual and motor paradigms, have consistently demonstrated that task-related changes in HbO show significantly higher reproducibility across multiple sessions [4].

Q3: How can I improve the spatial specificity and reliability of my fNIRS measurements for a region of interest?

To improve spatial specificity and reliability:

• Use Individual Anatomical Data: Guide optode layout design using subject-specific MRI data. Approaches using individual anatomical information (probabilistic or individual fMRI) outperform literature-based approaches in signal quality and sensitivity [60].
• Utilize Specialized Toolboxes: Employ toolboxes like the fNIRS Optodes' Location Decider (fOLD), which uses photon transport simulations on head atlases to automatically decide optode locations that maximize anatomical specificity to your target brain regions [5].
• Ensure Consistent Cap Placement: Implement precise and reproducible cap positioning protocols using anatomical landmarks (e.g., the 10-20 system) [73].

Q4: What is a minimally acceptable scanning duration for achieving reliable resting-state fNIRS metrics?

For resting-state measurements in patient populations, a scanning duration of more than 4 minutes is recommended. Research in stroke patients has shown that most fNIRS metrics, particularly in the low-frequency band, achieve higher reliability (ICC > 0.5) when the scan time exceeds 4 minutes [75].

Troubleshooting Guides

Problem: Poor Test-Retest Reliability at the Individual Subject Level

Potential Causes and Solutions:

• Cause 1: Inadequate Correction for Systemic Physiology

  • Solution: Implement Systemic Physiology Augmented fNIRS (SPA-fNIRS). Simultaneously measure systemic physiological fluctuations (e.g., heart rate, respiration) and employ processing techniques that regress these nuisance variables from your fNIRS signal [74]. One study showed that physiological correction yielded the highest test-retest reliability score for single-subject auditory data [74].

• Cause 2: Suboptimal Signal Preprocessing

  • Solution: Incorporate short-separation channels and advanced processing algorithms. Short channels help separate superficial, extracerebral signals from cortical signals. For data without short channels, consider using the Hemodynamic Modal Separation (HMS) algorithm, which has been shown to improve the reliability of low-frequency band signals [75].

• Cause 3: High Intra-Individual Variability in Activation Patterns

  • Solution: Focus on group-level analyses when individual reliability is low. Be cautious when interpreting single-subject fNIRS data, as studies on executive functions have found good group-level reliability but considerably lower reliability and strong variability at the individual level [76].

Problem: Low Spatial Overlap of Brain Activation Across Sessions

Potential Causes and Solutions:

• Cause: Shifts in Optode Position Between Sessions
  • Solution: Use source localization techniques and digitized optode positions. Research shows that increased shifts in optode position correlate with less spatial overlap across sessions. Using digitized optode locations from each session with anatomy-specific source localization significantly improves the reliability of capturing brain activity [4].

Problem: Unreliable fNIRS Metrics in Clinical Populations

Potential Causes and Solutions:

• Cause: Inappropriate Preprocessing or Analysis Frequency Bands
  • Solution: For resting-state studies in clinical populations (e.g., stroke), prioritize metrics and frequency bands with known high reliability. One study found that local efficiency and global network metrics reached high and excellent reliability after 4 minutes of scanning, while degree and betweenness showed only moderate or poor reliability [75]. It is recommended to use a global correction method like HMS and to be cautious when reporting single-channel level data [75].

Quantitative Data on fNIRS Reliability

The table below summarizes key test-retest reliability findings from recent fNIRS studies, quantified using the Intraclass Correlation Coefficient (ICC).

Table 1: Test-Retest Reliability (ICC) Across Different fNIRS Paradigms and Conditions

Study Paradigm / Condition	Brain Region(s)	Key Reliability Finding (ICC)	Notes
Postural & Finger-Tapping Task [73]	Prefrontal Cortex (PFC), Premotor Cortex (PMC), Somatosensory Cortex (SSC)	Excellent (ICC ≥ 0.78) without cap removal	Reliability reduced significantly after cap removal (e.g., PMC ICC=0.00) [73]
Postural & Finger-Tapping Task [73]	Hand Motor Region	Good (ICC=0.66) without cap removal	Reliability deteriorated (ICC=0.38) after cap removal [73]
Inhibitory Control & Working Memory [76]	Prefrontal Network	Lower at individual level	Good group-level reliability, but strong intra-individual variability hampers single-subject interpretation [76]
Visual & Motor Tasks [4]	Visual & Motor Cortex	HbO more reproducible than HbR	Source localization improved reliability; optode shifts reduced spatial overlap [4]
Resting-State in Stroke [75]	Whole Cortex	High reliability (ICC > 0.5) for most metrics with >4 min scan time	Local efficiency & global metrics were most reliable; degree & betweenness were less reliable [75]
Auditory Task (Single-Subject) [74]	Auditory Cortex	Physiology correction improved single-subject reliability	Short-channel correction alone reduced reliability by removing global systemic artifacts [74]

Detailed Experimental Protocols

Protocol 1: Assessing Reliability with and without Cap Removal

This protocol is derived from a study investigating test-retest reliability during postural and finger-tapping tasks in older adults [73].

• Participants: Recruit healthy older adults (e.g., >65 years). Exclude for major neurological, cardiovascular, or musculoskeletal disorders.
• Equipment: Portable continuous-wave fNIRS system (e.g., NIRSport2). Use a block design.
• Procedure:
  • Day 1, Test 1: Perform the motor tasks (e.g., postural weight-shifting and finger-tapping) with fNIRS recording.
  • Day 1, Test 2: After a 30-minute rest period with the cap kept in place, repeat the tasks.
  • Day 2, Test 3: After complete cap removal, re-attach the cap the next day at the same time. Standardize the cap position using anatomical landmarks (e.g., Cz) and head circumference measurements. Repeat the tasks.
• Data Analysis: Calculate ICC for oxyhemoglobin (HbO) signals in Regions of Interest (ROIs) between test sessions (Test1-Test2 for no removal; Test1-Test3 for with removal).

Diagram 1: Cap Removal Reliability Workflow

Protocol 2: Optimizing Single-Subject Reliability with Physiological Correction

This protocol is based on a repeated-measures study on a single subject for auditory-evoked responses [74].

• Participants: Any cohort, designed for deep individual-level analysis.
• Equipment: fNIRS system capable of simultaneous physiology monitoring (e.g., heart rate, respiration) and incorporating short-separation channels.
• Stimuli & Design: Use a passive block-design paradigm (e.g., auditory sentences vs. silence).
• Procedure:
  • Conduct multiple recording sessions (e.g., 10 sessions over 5 days).
  • In each session, simultaneously record fNIRS signals, short-channel signals, and systemic physiological signals.
• Data Analysis: Compare the test-retest reliability (using ICC) of the HbO signal after applying different preprocessing pipelines:
  • No correction
  • Physiology correction only
  • Short-channel correction only
  • Combined short-channel and physiology correction

Diagram 2: Signal Processing for Single-Subject Reliability

Table 2: Key Research Reagents and Tools for fNIRS Reliability Research

Tool / Solution	Function / Purpose	Relevance to Reliability
fOLD Toolbox [5]	Informs optode placement on the scalp based on photon transport simulations to maximize sensitivity to specific brain regions.	Directly improves spatial specificity and consistency of measurement target across sessions.
Short-Separation Channels [74] [75]	Measure signals predominantly from extracerebral tissues (scalp, skull).	Used as regressors in data processing to remove systemic physiological noise, improving signal quality and potentially reliability.
SPA-fNIRS Framework [74]	A methodological framework for the simultaneous acquisition of fNIRS and systemic physiological data (e.g., heart rate, blood pressure).	Enables more precise isolation of the neuronal signal of interest, enhancing single-subject reliability.
Hemodynamic Modal Separation (HMS) Algorithm [75]	A signal preprocessing algorithm designed to separate different components of the hemodynamic signal.	Has been shown to improve the test-retest reliability of fNIRS metrics, particularly in resting-state studies.
Kernel Flow2 (TD-fNIRS) [77]	A time-domain fNIRS system that provides improved, depth-resolved estimates of hemoglobin concentrations.	The system's design aims for reliable replacement after removal and offers metrics that have demonstrated high test-retest reliability.
Individual MRI Data [60]	Subject-specific anatomical (and optionally functional) images.	Informs subject-specific optode layout design, accounting for individual anatomical differences, which boosts signal quality and sensitivity.

Functional near-infrared spectroscopy (fNIRS) has emerged as a powerful neuroimaging tool for assessing brain function in naturalistic settings. However, as data analysis pipelines grow more complex, understanding how methodological choices affect results has become essential for ensuring reproducibility and transparency. The fNIRS Reproducibility Study Hub (FRESH) initiative, which involved 38 research teams independently analyzing the same datasets, revealed that while nearly 80% of teams agreed on group-level results for strongly supported hypotheses, significant variability emerged at the individual level [30]. This variability stems from multiple sources, including how poor-quality data are handled, how responses are modeled, and how statistical analyses are conducted [78]. This technical support guide addresses these challenges within the specific context of optimizing optode placement for spatial resolution research, providing troubleshooting guidance and standardized protocols to enhance reproducibility.

Quantitative Data on Reproducibility and Performance

Table 1: Key Factors Affecting fNIRS Reproducibility

Factor	Impact on Reproducibility	Evidence
Hemoglobin Species	Oxyhemoglobin (HbO) is significantly more reproducible than deoxyhemoglobin (HbR) [4].	F(1, 66) = 5.03, p < 0.05 [4].
Analytical Flexibility	Different analysis pipelines can lead to markedly divergent results and interpretations [30].	Nearly 80% agreement on group-level results, but lower at individual level [30] [78].
Researcher Experience	Higher self-reported analysis confidence (correlated with fNIRS experience) leads to greater inter-team agreement [30] [78].	Identified as a key variable in the FRESH initiative [30].
Data Quality	Better data quality improves agreement across analysis pipelines, especially for individual-level results [30].	A primary source of analytical variability [78].
Optode Placement	Increased shifts in optode position correlate with reduced spatial overlap across sessions [4].	Impacts consistency of targeting specific regions of interest (ROIs) [1].

Table 2: Performance Comparison of fNIRS Array Designs

Parameter	Sparse Array (e.g., 30mm grid)	High-Density (HD) Array
Spatial Resolution	Limited, poor ability to differentiate nearby regions [3].	Superior, improved anatomical specificity [3].
Sensitivity	Lower sensitivity to brain activity [3].	Higher sensitivity, captures stronger signal amplitude [3].
Localization	Poor functional localization, especially for low-load tasks [3].	Excellent and consistent localization of brain activity [3].
Reproducibility	Poor signal reproducibility due to nonuniform spatial sensitivity [3].	Improved inter-subject consistency [3].
Setup Practicality	Faster setup time, lower cost [3].	Increased setup time, higher cost and processing requirements [3].

Troubleshooting Guides & FAQs

Frequently Asked Questions (FAQs)

Q1: Why is my group-level HRF signal weak or unexpected, showing multiple dips and negative HbO averages?

This is a common preprocessing issue. Potential causes and solutions include:

• Cause 1: Non-robust brain activation. The experimental design or analysis might not be capturing robust activity. Check if individual subjects or blocks show expected signals [79].
• Cause 2: Suboptimal analysis pipeline. The preprocessing parameters (e.g., bandpass filter ranges, time range for block averaging) may be inappropriate. Re-evaluate parameter choices against established literature and ensure they fit your stimulus design (e.g., 30-second blocks) [79].
• Cause 3: Incorrect target region. The fNIRS channels might not be covering the brain region activated by the task. Re-assess your optode placement based on anatomical landmarks or neuroanatomy knowledge [79] [1].

Q2: How much variability in signals across subjects and sessions is considered normal?

Some variability is inevitable. Acceptable ranges depend on your specific paradigm and hardware. Key principles are:

• Focus on Reproducible Metrics: Prioritize HbO over HbR, as it has been shown to be more reproducible across sessions [4].
• Control What You Can: Minimize variability from methodological sources. Use consistent optode placement across sessions, as "increased shifts in optode position correlate with less spatial overlap" [4]. Employ high-density arrays where possible for better inter-subject consistency [3].
• Assess Data Quality: The FRESH initiative found that agreement between analysis pipelines improves with better data quality [30] [78]. Systematically quantify data quality (e.g., signal-to-noise ratio) to contextualize subject-level variability.

Q3: What is the single most impactful step I can take to improve the spatial reproducibility of my fNIRS measurements?

The most impactful step is to implement a protocol for highly consistent and anatomically accurate optode placement. This directly addresses the finding that "increased shifts in optode position correlated with less spatial overlap across sessions for each participant" [4]. Use digitized optode positions for anatomy-specific source localization to improve the reliability of capturing brain activity [4].

Troubleshooting Guide: Common fNIRS Analysis Issues

Table 3: Troubleshooting Analysis and Reproducibility Problems

Problem	Potential Causes	Solutions & Best Practices
Low Spatial Overlap/Specificity	Shifts in optode placement between sessions [4]. Using a sparse optode array [3]. Limited anatomical information for localization [1].	Use digitized optode positioning for source localization [4]. Upgrade to a high-density (HD) array for better sensitivity and localization [3]. Use individual MRIs or standardized atlases for accurate co-registration [1].
Poor Signal Quality (Noise/Artifacts)	Motion artifacts [1]. Systemic physiological noise (cardiac, respiratory) [1]. Poor probe-scalp contact [1].	Apply real-time motion correction algorithms and signal quality metrics [1]. Use short-separation channels to regress out superficial noise [3] [1]. Implement robust preprocessing pipelines with artifact rejection [30].
Low Analytical Reproducibility	High flexibility in analysis pipelines [30]. Inconsistent handling of poor-quality data [30] [78]. Variability in statistical modeling [30].	Pre-register analysis plans to reduce researcher degrees of freedom [30]. Adopt and report standardized processing pipelines where possible. Foster expertise, as experienced researchers showed higher agreement [30] [78].

Experimental Protocols for Enhancing Reproducibility

Protocol 1: Standardized Optode Placement and Coregistration

Purpose: To ensure consistent spatial targeting of Regions of Interest (ROIs) across subjects and sessions, which is critical for reproducibility in spatial resolution studies [4] [1].

Detailed Methodology:

• Cap Placement: Use a reliable EEG-style cap system with pre-defined coordinates based on the international 10-10 or 10-5 systems for consistent positioning.
• Digitization: Use a 3D digitizer (e.g., Polhemus Patriot) to record the 3D coordinates of all optodes (sources and detectors), as well as key anatomical landmarks (nasion, inion, left/right pre-auricular points). This step is crucial for anatomically specific source localization [4].
• Co-registration and MRI Mapping: Coregister the digitized optode positions with the subject's individual structural MRI or a standard brain template (e.g., MNI). This allows for projecting fNIRS data onto an anatomical image and defining ROIs based on anatomy, not just channel numbers [1].
• Source Reconstruction: Use the digitized positions and anatomical data to perform source localization, which moves the analysis from channel-space to image-space (e.g., using AtlasViewer or Homer3). This improves the reliability of capturing brain activity [4].

Protocol 2: High-Density fNIRS for Superior Spatial Resolution

Purpose: To statistically compare and leverage the superior localization and sensitivity of HD arrays over traditional sparse arrays for detecting brain activity [3].

Detailed Methodology:

• Array Design: Employ a high-density, multidistance array with a hexagonal pattern of overlapping channels, including short-separation channels (e.g., < 15 mm). This layout is necessary for improved depth sensitivity and spatial resolution [3].
• Data Acquisition: Collect data during a well-established paradigm (e.g., the Word-Color Stroop task) known to activate the target region (e.g., dorsolateral prefrontal cortex) [3].
• Image Reconstruction: Process the HD data using Diffuse Optical Tomography (DOT) algorithms to reconstruct 3D images of brain activation, rather than analyzing single channels. This provides a direct statistical comparison of activation maps between sparse and HD layouts [3].
• Validation: The protocol should confirm that the HD array provides better localization and sensitivity, particularly for tasks with lower cognitive load, making it more suitable for neuroimaging applications requiring high spatial specificity [3].

Workflow Visualization for fNIRS Reproducibility

fNIRS Reproducibility Workflow

Optode Optimization Pathway

The Scientist's Toolkit: Essential Research Reagents & Materials

Table 4: Key Materials and Tools for fNIRS Spatial Resolution Research

Item	Function/Application	Key Consideration
High-Density (HD) fNIRS System	Enables overlapping, multidistance channel measurements for superior spatial resolution and localization of brain activity [3].	Prefer systems with integrated short-separation channels for more effective superficial noise regression [3].
3D Digitizer	Records precise 3D locations of optodes and anatomical landmarks for co-registration with anatomical images [4].	Essential for moving from channel-space to anatomically accurate image-space analysis, improving reliability [4] [1].
Structural MRI	Provides individual anatomical data for accurate co-registration of fNIRS data and source reconstruction [1].	If unavailable, a standardized atlas (e.g., MNI) can be used, though with reduced individual accuracy [1].
Source Localization Software	Reconstructs hemodynamic activity in the brain in 3D (e.g., using AtlasViewer, Homer3) rather than showing data only from surface channels [4].	Critical for improving the spatial specificity and interpretation of fNIRS signals [4].
Short-Separation Channels	Specialized optode pairs (< 15 mm) used to measure and regress out systemic physiological noise originating from the scalp [3] [1].	Their use is a best practice for improving signal quality by separating cerebral and extracerebral signals [3].

This guide provides technical support for researchers optimizing optode placement to enhance the spatial specificity of functional Near-Infrared Spectroscopy (fNIRS) experiments.

Frequently Asked Questions (FAQs)

1. What is the core challenge in fNIRS optode placement that these methods aim to solve? Ensuring that the optodes (sources and detectors) placed on the scalp are positioned to maximize the signal quality and sensitivity to the specific cortical brain regions of interest (ROIs) involved in your experiment. Inaccurate placement can result in weak signals, poor sensitivity to brain activation, and data that does not reliably reflect activity in the target area [12] [60].

2. How do I choose between a Literature-Based, Probabilistic, or Individual fMRI approach? The choice often depends on the resources available (access to MRI/fMRI scanners, computational tools) and the specific requirements of your study concerning spatial precision and individual accuracy [60]. Please refer to the comparison table in the next section for a detailed breakdown.

3. What are "light-sensitivity profiles" and how are they used? Light-sensitivity profiles are probabilistic models, often computed using Monte Carlo simulations, that predict how near-infrared light propagates and is absorbed through the different tissues of the head (scalp, skull, brain). These profiles are used to optimize optode layouts by estimating the sensitivity of each potential source-detector channel to the cortical region you wish to study [11] [60].

4. My research involves patient populations or children where acquiring individual fMRI data is not feasible. What is the recommended approach? The Probabilistic (PROB) approach is highly suitable in these scenarios. It leverages individual anatomical data (if available) or an atlas, combined with probabilistic functional activation maps from independent group-level studies, to inform the optode layout. This method has been shown to outperform the simple literature-based approach and can yield results comparable to using individual fMRI data [60].

5. What is the recommended source-detector distance for a good signal-to-noise ratio? A distance between 25 and 40 mm is typically recommended. This range provides a reasonable trade-off, ensuring sufficient light penetration to the cerebral cortex while maintaining an acceptable signal-to-noise ratio [60]. Distances shorter than this are often used as "short-separation channels" to measure and remove signals originating from the scalp [80].

Methodology and Experimental Protocols

Literature-Based (LIT) Approach

This method uses existing scientific literature to determine where to place optodes.

• Procedure: Based on a review of prior fNIRS and fMRI studies, identify standard brain regions (e.g., the Dorsolateral Prefrontal Cortex) and their corresponding scalp coordinates according to the international 10-20 or 10-5 system for EEG electrode placement.
• Implementation: Place your fNIRS optodes in a grid-like fashion over these standardized scalp locations [60].

Probabilistic (PROB) Approach

This method enhances the LIT approach by incorporating anatomical and group-level functional data.

• Procedure:
  • Acquire an individual T1-weighted anatomical MRI scan for each participant.
  • Use software (e.g., NIRS-SPM, AtlasViewer) to coregister the participant's scalp surface and anatomy with standard brain atlases (e.g., Montreal Neurological Institute - MNI).
  • Incorporate probabilistic maps of functional MRI (fMRI) activation from an independent dataset of participants who performed a similar task. These maps indicate the likelihood of activation in a given brain voxel.
  • Use light-modeling software to compute sensitivity profiles and algorithmically determine the optode positions that maximize sensitivity to the probabilistically defined ROIs [60].

Individual fMRI (iFMRI) Approach

This is the most tailored method, using the participant's own functional and anatomical data.

• Procedure:
  • In a separate session, acquire the participant's anatomical (T1-weighted) and functional (fMRI) data while they perform the same (or a very similar) experimental task that will be used in the fNIRS study.
  • Analyze the fMRI data to generate a statistical map of individual brain activation, identifying the precise, subject-specific ROI.
  • Coregister the individual fMRI activation map with the anatomical scan and the digitized scalp surface.
  • Use computational tools to run optimization algorithms that determine the optode layout providing the highest sensitivity to the individual's specific activation focus [11] [60].

The following workflow illustrates the decision process and increasing level of personalization involved in selecting an optode placement strategy:

Comparative Performance Data

The table below summarizes the key characteristics and performance outcomes of the three optode placement approaches based on a controlled study [60].

Table 1: Quantitative Comparison of Optode Placement Approaches

Feature	Literature-Based (LIT)	Probabilistic (PROB)	Individual fMRI (iFMRI)
Core Data Source	Published studies & standard head atlases [60]	Individual anatomy + group fMRI maps [60]	Individual anatomy & individual task-fMRI [11] [60]
Required Resources	Low (no neuroimaging needed)	Medium (requires individual MRI)	High (requires individual MRI & fMRI)
Spatial Specificity	Low (assumes group-level anatomy/function)	Medium-High (individual anatomy, group function)	High (individual anatomy & function)
Signal Quality (SNR)	Lower	Higher	Higher [60]
Sensitivity to Brain Activation	Lower	Higher	Higher [60]
Best Use Cases	Exploratory studies, limited resources, large groups	Robust setups for clinical/patient populations, when individual fMRI is unavailable [60]	Studies requiring highest spatial precision, fundamental research

The Scientist's Toolkit

Table 2: Essential Research Reagents and Solutions

Item	Function/Description
fNIRS System	A continuous-wave, frequency-domain, or time-domain system to emit near-infrared light and detect its attenuation after passing through tissue [80].
Optodes & Cap	The optical components (sources and detectors) and a head cap to hold them in place. Commercial caps often follow the 10-20 system [80].
3D Digitizer	A device to record the precise 3D locations of optodes on the scalp relative to anatomical landmarks, crucial for accurate coregistration [11].
Collodion Adhesive	A clinical adhesive used to fix optodes directly to the scalp, ensuring excellent optical coupling and signal stability for prolonged recordings [11].
Neuronavigation System	A device that uses the participant's MRI to guide and verify the real-time placement of optodes on the exact target scalp locations [11] [60].
Software for Coregistration & Light Modeling	Tools like AtlasViewer, NIRS-SPM, and PHOEBE are used to coregister optode positions with anatomical images and compute light sensitivity profiles [80] [11] [60].

Troubleshooting Guide

Table 3: Common Issues and Solutions

Problem	Potential Cause	Solution Steps
Poor Signal-to-Noise Ratio	Optodes too far from target cortex; poor optical contact; excessive hair under optodes.	1. Verify optode placement with a digitizer and coregistration software.2. Ensure firm optode-scalp contact; consider using collodion for long studies [11].3. Part hair and use a generous amount of optode gel.
Weak or No Task-Related Hemodynamic Response	Optodes are not sensitive to the true, individual-specific functional area.	1. For future studies, adopt a Probabilistic or iFMRI approach for better targeting [60].2. In existing data, use short-separation channels to confirm the signal is of cerebral origin [80].
Inconsistent Results Across Participants	High inter-subject variability in anatomy and functional localization not accounted for.	Move from a Literature-Based to a Probabilistic approach, which uses individual anatomy to improve consistency and robustness across participants [60].
Difficulty Targeting Specific Deep Brain Regions	fNIRS has limited penetration depth (typically <½ of source-detector separation) and cannot directly measure subcortical areas [81] [82].	Focus on cortical regions. Ensure source-detector distance is sufficient (25-40 mm) but not so large that the signal is lost [60]. Acknowledge the inherent limitation of fNIRS for deep structures.

Frequently Asked Questions

Q1: What are the most common factors that reduce the anatomical targeting accuracy of my fNIRS setup?

The primary factors affecting targeting accuracy are inter-subject anatomical variability, inconsistent optode placement across sessions, and the use of generic head models instead of subject-specific anatomy [1] [13]. Even with careful cap placement, studies show high dispersion in sensitivity profiles between different individuals targeting the same cortical region [13]. Furthermore, increased shifts in optode positioning directly correlate with reduced spatial overlap of measured brain activity across repeated sessions [4].

Q2: How much does subject-specific anatomy actually impact sensitivity measurements?

Subject-specific anatomy significantly alters sensitivity profiles compared to atlas-based models. Research quantifying this effect found that approximately 70% of the fNIRS signal originates from gyri, with sensitivity profiles showing broad patterns in the source-detector direction (20.953 ± 5.379 mm FWHM) and steeper drops in the transversal direction (6.082 ± 2.086 mm) [13]. The table below summarizes key quantitative differences:

Table 1: Sensitivity Profile Characteristics Based on Depth Quartiles

Depth Quartile	Median Depth (mm)	Signal Coverage (%)
First	< 11.8	0.391 (0.087)
Second	< 13.6	0.292 (0.009)
Third	< 15.7	0.185 (0.011)
Fourth	≥ 15.7	0.132 (0.082)

Q3: What evidence supports investing in high-density fNIRS arrays for improved spatial specificity?

High-density (HD) multidistance arrays demonstrate superior localization capabilities compared to traditional sparse arrays. Statistical comparisons show HD arrays provide significantly better detection and localization of brain activity in image space, particularly during lower cognitive load tasks [3]. While sparse arrays (typically with 30mm channel spacing) may suitably detect activation during cognitively demanding tasks, HD arrays with overlapping channels improve spatial resolution, depth sensitivity, and inter-subject consistency [3].

Q4: Which hemoglobin species provides more reproducible measurements across multiple sessions?

Oxyhemoglobin (Δ[HbO]) demonstrates significantly higher reproducibility over multiple sessions compared to deoxygenated hemoglobin (Δ[HbR]) [4]. Quantitative analyses show task-related changes in HbO are more consistently detected across repeated measurements, making it a more reliable metric for longitudinal studies [4].

Q5: What tools are available to help optimize my optode placement for specific regions of interest?

The fNIRS Optodes' Location Decider (fOLD) toolbox provides data-driven guidance for probe arrangement [5]. This approach uses photon transport simulations on head atlases to automatically determine optode locations that maximize anatomical specificity to predefined brain regions-of-interest, bringing parcellation methods from fMRI to fNIRS experimental design [5].

Troubleshooting Guides

Issue: Poor Reproducibility Across Multiple Measurement Sessions

Problem: Inconsistent spatial overlap and activation patterns when repeating the same experiment across different sessions.

Solution:

• Implement precise optode localization: Use digitized optode positions for each session co-registered with individual anatomical MRI when possible [4]. This controls for placement variability.
• Apply source localization techniques: Transform channel-based data to source-reconstructed images using anatomical constraints. Research shows this improves the reliability of capturing brain activity compared to channel-level analysis alone [4].
• Establish standardized cap placement protocols: Develop detailed procedures for identifying fiducial points and positioning measurement caps to minimize placement variations between sessions [1].

Issue: Inadequate Sensitivity to Target Cortical Regions

Problem: Uncertainty whether fNIRS channels adequately sample from intended brain regions-of-interest.

Solution:

• Utilize photon migration modeling: Conduct Monte Carlo simulations on appropriate head models to estimate sensitivity profiles for your specific optode configuration [13] [5].
• Consider high-density arrays: Implement overlapping, multidistance channel configurations where possible. Studies demonstrate these provide improved spatial resolution and depth discrimination compared to sparse arrays [3].
• Incorporate subject-specific anatomy: When feasible, use individual MRI scans to generate personalized sensitivity profiles, as atlas-based evaluations show substantial differences from subject-specific calculations [13].

Table 2: Comparison of fNIRS Array Configurations for Spatial Specificity

Array Type	Channel Spacing	Spatial Resolution	Depth Sensitivity	Localization Accuracy
Sparse	~30 mm (non-overlapping)	Limited	Poor without short-separation	Low
High-Density (HD)	Multiple distances (overlapping)	Improved	Good with multidistance	High (approaching fMRI)
HD-DOT	Multiple distances (highly overlapping)	High	Excellent	Superior, consistent across subjects

Issue: Contamination by Extracerebral Signals and Systemic Physiology

Problem: Inability to distinguish cerebral hemodynamic responses from confounding systemic signals.

Solution:

• Implement short-separation regression: Include source-detector pairs with distances <15 mm to regress out superficial (extracerebral) components from long-channel measurements [3] [83].
• Apply signal processing techniques: Use methods such as principal component analysis (PCA) or global average signal removal to reduce the influence of global systemic physiology [83].
• Utilize additional physiological monitoring: Incorporate measurements of heart rate, respiration, and blood pressure to account for systemic physiological fluctuations in your signal processing models [1].

Experimental Protocols for Validation

Protocol 1: Quantifying Channel-to-Cortex Sensitivity Profiles

Purpose: To determine the precise cortical regions sampled by each fNIRS channel in a specific experimental setup.

Methodology:

• Acquire anatomical data: Obtain high-resolution T1-weighted MRI scans for each subject (or use appropriate head atlas if unavailable) [13].
• Segment head tissues: Process MRI data to discriminate five tissue types: scalp, skull, cerebrospinal fluid (CSF), gray matter, and white matter [5].
• Co-register optode positions: Map precise optode locations to the scalp surface of the anatomical model using digitized positions or standardized coordinate systems (10/5 or 10/10) [13].
• Perform photon migration simulations: Use Monte Carlo simulations to model light transport from each optode [13] [5]. Key parameters:
  • Number of photons: 10⁸ for sufficient statistics
  • Optical properties based on published values for each tissue type
  • Wavelengths appropriate for your system (typically 760-850nm range)
• Calculate sensitivity profiles: Compute normalized sensitivity for each channel as the voxel-wise product of the photon fluence from source and detector [5].
• Quantify brain specificity: Determine the percentage of sensitivity penetrating to cerebral gray matter for each channel [13].

Protocol 2: Evaluating Inter-Subject Variability in Sensitivity

Purpose: To assess how anatomical differences between subjects affect sensitivity to target regions.

Methodology:

• Recruit participant cohort: Include sufficient subjects (e.g., 10+ ) to account for natural anatomical variation [13].
• Acquire individual anatomy: Perform MRI scans for each participant following standardized protocols [13].
• Maintain consistent optode placement: Use the same cap layout across all subjects, co-registered to individual anatomy [13].
• Compute sensitivity matrices: Follow Protocol 1 for each subject to generate subject-specific sensitivity profiles [13].
• Perform region-of-interest analysis: Define specific cortical regions and quantify the coupling between channels and these regions for each subject [13].
• Statistical analysis: Calculate coefficients of variation across subjects for channel-ROI coupling strengths to quantify inter-subject variability [13].

The Scientist's Toolkit

Table 3: Essential Resources for fNIRS Anatomical Targeting Research

Tool/Resource	Function	Application Context
fOLD Toolbox [5]	Automated decision of optode locations	Optimizing probe arrangement for specific ROIs during experimental design
Monte Carlo Extreme (MCX) [5]	GPU-accelerated photon transport simulation	Calculating sensitivity profiles for specific optode configurations
Nirstorm [13]	fNIRS processing integrated with Brainstorm	MRI-fNIRS integration and cortical mapping
HOMER2 [83]	Comprehensive fNIRS data processing	Implementing motion correction and signal processing pipelines
Subject-Specific Anatomy (SSA) [13]	Individual MRI-based head models	Improving accuracy of sensitivity calculations and cortical mapping
High-Density Arrays [3]	Multidistance, overlapping optode configurations	Enhancing spatial resolution and depth discrimination

Methodological Workflows

Workflow for Quantifying fNIRS Targeting Accuracy

Factors Influencing fNIRS Targeting Accuracy

Technical Support Center

Troubleshooting Guide: Common DOT Validation Challenges

Q1: Our reconstructed activation appears diffuse and poorly localized. What could be the cause?

This issue typically stems from insufficient spatial sampling or suboptimal optode placement. Traditional sparse arrays with 30mm channel spacing fundamentally limit spatial resolution and accurate localization [3]. As optodes shift between sessions, spatial overlap decreases, directly reducing reproducibility [4].

• Solution: Implement high-density (HD) arrays with overlapping, multidistance channels. HD-DOT layouts create a "diffuse focus" of approximately 3cm diameter at 1.4cm depth, significantly improving sensitivity and localization compared to sparse arrays [3] [84].
• Verification Protocol:
  • Use a head phantom with a known activation target to quantify your system's point spread function.
  • Ensure you are using digitized optode positions for anatomy-specific source localization, which improves reliability [4].
  • Statistically compare activation maps from your current setup against HD-DOT results, which have been shown to provide superior localization and inter-subject consistency [3].

Q2: How can we distinguish true cerebral activation from systemic physiological noise?

fNIRS signals are notoriously contaminated by physiological confounds from systemic physiology (e.g., blood pressure, heart rate) and extracerebral tissue (skin, bone) [83] [1]. Failure to account for these is a common source of false positives.

• Solution: Integrate multiple strategies for enhancing depth sensitivity and signal specificity.
• Actionable Steps:
  • Implement Short-Separation Channels: Place detectors 8-10mm from sources to predominantly measure extracerebral hemodynamics. Use this signal as a regressor in a general linear model (GLM) to remove the superficial component from long-channel data [85] [83].
  • Incorporate Physiological Monitoring: Record heart rate, blood pressure, and respiration. Use these as additional regressors in your GLM to account for global systemic physiology [85].
  • Apply Advanced Processing: Utilize data-driven approaches like Principal Component Analysis (PCA) to identify and remove global spatial noise components [83].

Q3: Our results are not reproducible across repeated measurement sessions. How can we improve consistency?

Low reproducibility often relates to two factors: variability in optode placement and the choice of hemoglobin species for analysis.

• Solution:
  • Standardize Optode Placement: Use a reliable cap and 3D digitization (e.g., with a Polhemus or similar system) to record the precise location of every optode in each session. Coregister these positions to individual or standard head anatomy [4] [85] [1].
  • Focus on Oxyhemoglobin (HbO): Evidence indicates that task-related changes in HbO are significantly more reproducible over sessions than changes in deoxyhemoglobin (HbR) [4].

Experimental Protocols for Validation

Protocol 1: Phantom-Based System Characterization

This protocol validates the basic spatial performance of your DOT system using a tissue-like phantom.

• Objective: To quantify the spatial resolution and localization accuracy of the imaging system.
• Materials:
  • Double-Layer Phantom: A concentric beaker setup with a PBS/intralipid/blood mixture (e.g., 97/1/2%) to mimic tissue scattering and absorption [84].
  • Absorbing Inclusion: A smaller, oxygenated target embedded within the phantom. This target can be covered with neoprene sleeves with holes of various sizes to estimate the spatial limits of the measurement volume [84].
• Methodology:
  • Place the HD-DOT probe on the surface of the phantom.
  • Collect data with the inclusion present at various known locations.
  • Reconstruct images and calculate the system's Point Spread Function (PSF) by comparing the known location of the inclusion to its reconstructed location.
  • Quantify metrics like full-width-at-half-maximum (FWHM) of the PSF and localization error.

Protocol 2: In-Vivo Task-Based Validation Against Ground Truth

This protocol uses a well-established brain activation paradigm to validate functional results.

• Objective: To compare DOT-reconstructed brain activity with expected activation patterns from the literature.
• Task Paradigm: Word-Color Stroop (WCS) Task. The incongruent condition (e.g., the word "RED" printed in blue ink) robustly activates the dorsolateral prefrontal cortex (dlPFC), providing a reliable ground truth for validation [3].
• Methodology:
  • Participant Setup: Fit participants with an HD-DOT array covering the prefrontal cortex. Coregister optode positions to individual MRI or a standard head model.
  • Data Acquisition: Record fNIRS data during multiple blocks of congruent and incongruent WCS trials.
  • Image Reconstruction: Use a forward model based on the registered probe geometry to reconstruct images of HbO and HbR changes on the cortical surface.
  • Validation Analysis:
    • Confirm that significant activation is detected in the dlPFC specifically during the incongruent condition.
    • Statistically compare the localization and sensitivity of your HD-DOT results with a simulated sparse array extracted from the same data [3].
    • The expected outcome is that the HD array will show superior localization and stronger signal, particularly for lower cognitive load tasks.

The Scientist's Toolkit: Research Reagent Solutions

Table: Essential Materials and Tools for fNIRS/DOT Validation Research

Item	Function & Explanation
High-Density (HD) DOT Probe	A multidistance, overlapping optode array. It improves spatial resolution, depth sensitivity, and inter-subject consistency compared to traditional sparse arrays, which is fundamental for accurate validation [3].
3D Digitizer	A device (e.g., Polhemus Patriot) to record the 3D coordinates of optodes on the head. This enables precise coregistration to individual anatomy, dramatically improving the accuracy of the forward model used in image reconstruction [4] [85].
Short-Separation Channels	Source-detector pairs with a separation of <1 cm. These channels are primarily sensitive to extracerebral layers and provide a reference signal to regress out systemic physiological noise from the long channels, enhancing cerebral specificity [85] [86].
Anatomical Head Model	An MRI-derived model (individual or atlas-based) of head tissue layers (scalp, skull, CSF, brain). It is used to generate a realistic light propagation model (forward model), which is essential for accurate image reconstruction from boundary measurements [87] [85].
Analysis Toolbox (e.g., AnalyzIR, HOMER2)	Open-source software packages providing standardized pipelines for fNIRS data management, preprocessing, statistical analysis, and image reconstruction. Using such tools enhances reproducibility and allows for method comparison [87] [83].

Signaling Pathways and Workflows

The following diagram illustrates the core workflow for validating DOT reconstruction accuracy, integrating both phantom and in-vivo approaches.

Workflow for DOT Accuracy Validation

The diagram below visualizes the light propagation and signal processing steps that underpin DOT, highlighting sources of potential error that validation seeks to minimize.

fNIRS Signal Path and Processing Chain

Frequently Asked Questions (FAQs)

Q1: Why is consistent optode placement across multiple sessions so critical, especially for clinical neurofeedback studies?

A1: In applications like neurofeedback, where patients undergo repeated training sessions, the repeatability of optode placement is critical for reliably targeting and training the same specific brain regions each time. Inconsistent placement can lead to measuring from different brain areas, which undermines the training's validity and effectiveness. Studies have shown that increased shifts in optode position between sessions correlate with reduced spatial overlap in the measured brain activity, directly impacting the reliability of the intervention [1] [4].

Q2: Our fNIRS signals are often contaminated by large, low-frequency drifts. What is a common source of this noise, and how can we address it in real-time?

A2: These drifts are often caused by physiological noise from systemic activities, such as heart rate, blood pressure changes, and respiration, which affect both cerebral and extracerebral (scalp) tissues. A highly effective method to address this is using short-separation channels. These channels are placed too close to the source (typically < 1.5 cm) to detect brain activity, so they primarily record noise from the scalp. This signal can then be used as a regressor to remove similar noise components from the standard long-separation channels that measure brain activity [88] [3].

Q3: We see motion artifacts in our data. What are some robust preprocessing techniques to handle them for real-time analysis?

A3: Motion artifacts are a common challenge. Unlike offline analysis, real-time processing cannot correct data after acquisition, making robust preprocessing vital. While a comprehensive list of all methods is beyond this scope, effective and real-time feasible techniques include:

• Principal Component Analysis (PCA): Can identify and remove globally uniform components, which often represent motion or other global physiological noise [88].
• General Linear Model (GLM): Allows for the modeling and removal of noise components derived from auxiliary measurements (e.g., motion sensors, short-separation channels) [88]. Insufficient real-time preprocessing can cause a system to operate on noise instead of genuine brain activity, which is particularly detrimental in clinical applications like brain-computer interfaces [1].

Q4: What are the practical trade-offs between using a traditional sparse fNIRS array and a high-density (HD) array?

A4: The choice involves a balance between resource investment and data quality. The table below summarizes the key trade-offs:

Feature	Sparse fNIRS Array	High-Density (HD) fNIRS Array
Spatial Resolution & Specificity	Limited; may miss or average signals from small or adjacent regions [3].	Superior; provides improved sensitivity, localization, and depth discrimination [3].
Resource Requirements (Cost, Setup Time)	Lower; fewer optodes and channels make it less expensive and faster to set up [3].	Higher; requires more hardware, leading to greater cost, longer setup times, and more complex data processing [3].
Best Suited For	Detecting the presence of activation from cognitively demanding tasks in a broad field-of-view [3].	Applications requiring precise localization of brain activity, especially during lower cognitive load tasks, or for network-level analysis [88] [3].

Q5: Which hemoglobin species is generally more reproducible for tracking task-related brain activity over multiple sessions?

A5: Research indicates that oxyhemoglobin (HbO) is significantly more reproducible over sessions than deoxyhemoglobin (HbR) [4]. Therefore, for studies focusing on longitudinal tracking or reproducibility, HbO is often the more reliable metric.

The following table summarizes key quantitative findings from recent fNIRS research relevant to addressing variability.

Metric	Finding	Relevance to Variability	Source
HbO vs. HbR Reproducibility	Oxyhemoglobin (HbO) is significantly more reproducible over multiple sessions than deoxyhemoglobin (HbR) (F(1, 66) = 5.03, p < 0.05) [4].	Supports using HbO as a more stable signal for longitudinal clinical studies.	[4]
Impact of Optode Shift	Increased shifts in optode placement between sessions correlate with reduced spatial overlap of measured brain activity [4].	Highlights the critical importance of consistent cap placement for within-subject study designs.	[4]
Source Localization Impact	Using source localization (e.g., with digitized optode positions) improves the reliability of fNIRS for capturing brain activity compared to channel-level analysis [4].	Recommends advanced analysis methods to mitigate spatial inaccuracies.	[4]

Experimental Protocol: Assessing fNIRS Reproducibility

This protocol outlines a methodology for quantifying the within-subject reproducibility of fNIRS signals, a critical step for validating clinical applications.

Objective: To determine the test-retest reliability of fNIRS-measured hemodynamic responses to a controlled motor and visual task across multiple sessions.

Materials:

• fNIRS system with high channel count (e.g., 102 channels).
• Cap or headband with optodes arranged to cover target regions (motor and visual cortices).
• Digitization equipment (e.g., 3D digitizer) to record precise optode positions per session.
• Stimulus presentation software.

Procedure:

• Participant Setup: Recruit participants who will complete at least ten separate testing sessions on different days. During each session, digitize the 3D positions of the optodes relative to anatomical landmarks on the participant's head [4].
• Task Paradigm: Employ a block design. For example:
  • Motor Task: A 10-second period of repetitive finger tapping, followed by a 20-second rest period. Repeat 5-10 times per session.
  • Visual Task: A 10-second period of a reversing checkerboard stimulus, followed by a 20-second rest period. Repeat 5-10 times per session.
• Data Acquisition: Record fNIRS data from all channels throughout the task performance.
• Data Analysis:
  • Preprocessing: Apply standard filters (bandpass, low-pass) and correct for motion artifacts and physiological noise using techniques like short-separation channel regression [88].
  • Channel-Level Analysis: For each channel and session, calculate the average hemodynamic response (HbO and HbR) to the task. Identify channels with statistically significant task-related activity for each session.
  • Source-Level Analysis: Use the digitized optode positions and a head model (can be a generic or individual anatomical MRI) to reconstruct the brain activity from channel data to the cortical surface [4].
  • Quantify Reproducibility:
    • Calculate the percentage of sessions in which a significant activation is detected at each channel or cortical location.
    • Compute the spatial overlap of activated regions across all sessions for each participant.
    • Corrogate the magnitude of optode shift (from digitization data) with the reduction in spatial overlap across sessions [4].

Workflow & Signaling Pathways

Experimental Workflow for Reproducibility Assessment

Decision Process for fNIRS Array Selection

The Scientist's Toolkit: Essential Research Reagents & Materials

This table details key materials and tools essential for conducting rigorous fNIRS studies, particularly those focused on optimizing spatial specificity.

Item	Function & Importance
High-Density (HD) fNIRS Array	An array with overlapping, multidistance channels. It provides superior spatial resolution, sensitivity, and localization of brain activity compared to traditional sparse arrays, making it ideal for isolating specific regions of interest [3].
Short-Separation Channels	Optode pairs placed <1.5 cm apart. They are critical for measuring and subsequently regressing out the confounding physiological noise originating from the scalp and skull, thereby improving the signal quality of cerebral measurements [88] [3].
3D Digitization System	A tool to record the precise three-dimensional locations of fNIRS optodes relative to anatomical landmarks (e.g., nasion, inion). This is essential for accurate co-registration of fNIRS data with anatomical atlases or individual MRI scans, greatly improving spatial specificity [4].
Auxiliary Sensors	Sensors for measuring heart rate (ECG/pulse oximeter), respiration, and head motion. These provide independent recordings of physiological noise, which can be incorporated into processing pipelines (e.g., General Linear Models) to clean the fNIRS signal more effectively [88].
Linear Mixed Models (LMM)	An advanced statistical method. LMMs are crucial for properly analyzing nested fNIRS data (e.g., channels nested within subjects), as they control for within-subject variability and non-independent observations, leading to more accurate and reliable findings [89].

Technical Support & Troubleshooting

Frequently Asked Questions (FAQs)

Q1: How can I quickly check if my fNIRS optode placement is consistent with the intended brain regions of interest? A1: Use automated probe design toolboxes like the fNIRS Optodes' Location Decider (fOLD). The fOLD toolbox uses photon transport simulations on head atlases to determine optode positions that maximize anatomical specificity to your target brain regions. It provides sensitivity profiles for channels and helps select positions from the 10-10 or 10-5 systems that best cover your regions-of-interest [5].

Q2: What is the most effective method to ensure consistent optode placement across multiple sessions or subjects? A2: Implement augmented reality (AR) guidance systems. Software like NeuroNavigatAR (NNAR) uses facial recognition and computer vision to track facial landmarks in real-time (15 frames per second) and displays 10-20 system landmarks directly over a video feed of the subject's head. This approach reduces placement errors to 0.75 cm when using subject-specific head surfaces, significantly improving consistency across sessions [90].

Q3: How can I register my fNIRS data to anatomical brain images without an individual MRI for each subject? A3: Use probabilistic spatial registration methods implemented in software like AtlasViewer or fNIRS-SPM. These tools leverage reference MRI databases to map fNIRS channel positions from scalp-based coordinates (10-5 or 10-20 systems) to standard brain space (MNI coordinates), enabling anatomical interpretation without subject-specific MRIs [9] [20].

Q4: What steps can I take to improve signal quality and reproducibility in my fNIRS experiments? A4: Reproducibility is influenced by multiple factors. Follow these evidence-based practices:

• Ensure High Data Quality: Agreement in individual-level analyses improves significantly with better data quality [30].
• Standardize Analysis Pipelines: Variability in how poor-quality data are handled, how responses are modeled, and how statistical analyses are conducted are major sources of irreproducibility [30].
• Gain Experience: Research teams with higher self-reported analysis confidence (correlated with years of fNIRS experience) showed greater agreement in results [30].

Q5: What methods are available for co-registering fNIRS probe locations with structural MRI scans? A5: For individual analysis with structural MRI, use direct co-registration. Place fiducial markers (e.g., Vitamin E capsules) on key optode locations during MRI scanning. Then apply projection algorithms (e.g., balloon-inflation algorithm) to map scalp positions to the underlying cortical surface, providing coordinates in both MNI and Talairach spaces [91].

Troubleshooting Common Experimental Issues

Issue: Poor spatial specificity in fNIRS measurements during cognitive tasks.

• Potential Cause: Inaccurate optode placement relative to target cortical regions.
• Solution: Prior to experimentation, use the fOLD toolbox [5] to simulate sensitivity profiles for your probe design. This ensures your channel arrangements optimally cover the brain regions relevant to your cognitive paradigm (e.g., prefrontal cortex for executive function tasks).

Issue: Inconsistent findings across repeated measurements in longitudinal studies.

• Potential Cause: Variability in cap placement between sessions.
• Solution: Implement AR-guided placement systems like NeuroNavigatAR [90] to maintain consistent positioning. For manual placement, always use the 10-20 system landmarks and document measured distances from nasion, inion, and preauricular points for verification.

Issue: Difficulty interpreting fNIRS data in anatomical context.

• Potential Cause: Lack of structural co-registration.
• Solution: For group studies without individual MRIs, use probabilistic registration to standard brain space [20]. For maximum anatomical precision in individual subjects, perform MRI co-registration with fiducial markers [91].

Issue: Low signal quality and contamination by motion artifacts or systemic noise.

• Potential Cause: Insufficient real-time preprocessing and poor optode-scalp contact.
• Solution: Ensure good optode-scalp coupling and implement real-time preprocessing techniques to distinguish cerebral signals from extracerebral contaminants [12]. Use secure, comfortable headgear that minimizes movement during tasks.

Quantitative Data Comparison

Table 1: Accuracy Comparison of fNIRS Placement Guidance Methods

Method	Median Position Error	Key Advantages	Limitations	Best Suited Paradigms
AR-Guided Placement (NeuroNavigatAR)	1.52 cm (general atlas)0.75 cm (subject-specific)	Real-time guidance (15 fps), reduced operator dependence, consistent across sessions [90]	Requires camera system, initial setup	Longitudinal studies, multi-site studies, novice operators
Manual 10-20 with Measuring Tape	Not quantified in results	Self-adaptive to head shape, no special equipment needed [90]	Time-consuming, operator-dependent, hair impedes measurement	Single sessions with experienced technicians, limited resources
Prefabricated Caps	Varies with cap fit	Fast application, standardized for group studies [90]	Global offset possible, limited size options, based on atlas not individual anatomy [90]	Large group studies, rapid screening protocols
Digitizer-Based Co-registration	Minimal when properly performed	High precision for individual anatomy, direct MRI correlation [91]	Requires additional equipment, time-consuming, technically complex	Studies requiring maximum anatomical precision, clinical applications

Table 2: fNIRS Co-registration Methods and Their Applications

Co-registration Method	Spatial Precision	Equipment Requirements	Time Investment	Ideal Use Cases
Individual MRI with Fiducial Markers [91]	High (subject-specific)	MRI scanner, fiducial markers, digitizer	High	Studies requiring maximal anatomical precision, clinical populations
Probabilistic Registration (Reference MRI database) [20]	Moderate (group average)	3D digitizer or measurement tools	Moderate	Group studies without individual MRIs, retrospective analysis
Virtual Registration (10-20 to standard space) [20]	Moderate	Basic measurement tools	Low	Standalone fNIRS studies, rapid analysis, large sample sizes
AtlasViewer with Generic Head Model [9]	Moderate	Software only	Low	Probe design, feasibility studies, educational use

Experimental Protocols & Workflows

Protocol 1: AR-Guided Optode Placement for Longitudinal Studies

Purpose: To ensure consistent optode placement across multiple measurement sessions using augmented reality guidance.

Materials:

• NeuroNavigatAR software or similar AR placement system [90]
• Laptop or tablet with camera (minimum 15 fps capability)
• fNIRS headgear with adjustable positioning
• Measurement chair with headrest

Procedure:

• System Setup: Launch the AR guidance software and position the camera to capture a clear view of the subject's head.
• Landmark Detection: The system will automatically detect and track facial landmarks (nasion, preauricular points) using computer vision algorithms.
• Head Model Alignment: Select the appropriate head model (general atlas, age-matched, or subject-specific if available).
• Real-time Guidance: Position the fNIRS headgear while observing the overlaid 10-20 system landmarks on the camera feed.
• Verification: The system provides continuous feedback on placement accuracy. Adjust until target error threshold is achieved (<1.5 cm for general atlas, <0.75 cm for subject-specific).
• Documentation: Save the session data including final placement coordinates for future reference.

Validation: This method has demonstrated consistent head-landmark prediction errors across repeated measurement sessions with no statistically significant difference in accuracy across age groups [90].

Protocol 2: fOLD-Based Probe Design for Region-Specific Studies

Purpose: To optimize probe geometry for specific cognitive paradigms targeting particular brain regions.

Materials:

• fOLD toolbox software [5]
• Head atlases (Colin27 or SPM12 template)
• Computer with MATLAB
• 3D digitizer (optional for verification)

Procedure:

• Region Definition: Input your target brain regions-of-interest based on your cognitive paradigm (e.g., dorsolateral PFC for executive functions, motor cortex for movement tasks).
• Probe Simulation: The toolbox runs photon transport simulations using Monte Carlo methods to model light propagation through head tissues.
• Sensitivity Analysis: Evaluate normalized sensitivity profiles for potential channel configurations.
• Optode Positioning: The algorithm recommends optode locations from 10-10 or 10-5 systems that maximize sensitivity to your target regions.
• Geometry Export: Export the final probe design for cap fabrication or existing system setup.

Validation: The fOLD method has been validated against two head atlases and provides quantitative metrics of anatomical specificity for each channel [5].

Experimental Workflows & Signaling Pathways

Research Reagent Solutions & Essential Materials

Table 3: Essential Tools for fNIRS Optode Placement Research

Tool/Software	Primary Function	Application Context	Access Information
NeuroNavigatAR [90]	AR-guided optode placement	Real-time placement guidance for consistent positioning across sessions	Open-source software
fOLD Toolbox [5]	Probe design optimization	Determining optode locations that maximize sensitivity to target brain regions	Publicly available toolbox
AtlasViewer [9]	Spatial registration & visualization	Mapping fNIRS channels to anatomical locations, probe design evaluation	Part of Homer2 software package
fNIRS-SPM [20]	Statistical parametric mapping	Statistical analysis of fNIRS data with anatomical interpretation	Standalone software package
Vitamin E Capsules [91]	Fiducial markers for MRI	Marking optode positions on scalp for MRI co-registration	Commercial pharmaceutical product
3D Digitizer [20]	Precise coordinate measurement	Capturing exact optode positions on head surface for registration	Commercial hardware (Polhemus, etc.)

Functional near-infrared spectroscopy (fNIRS) and functional magnetic resonance imaging (fMRI) are both hemodynamic-based neuroimaging techniques that measure brain activity through neurovascular coupling. While fNIRS measures concentration changes in oxygenated (Δ[HbO]) and deoxygenated hemoglobin (Δ[HbR]) using near-infrared light, fMRI detects the blood-oxygen-level-dependent (BOLD) signal, which primarily reflects Δ[HbR] changes [92]. Cross-validation between these modalities establishes concurrent validity, ensuring that fNIRS activation patterns accurately reflect genuine cortical activity rather than physiological noise or artifacts [93].

This technical guide addresses the critical challenges in fNIRS-fMRI validation studies, with particular emphasis on optimizing optode placement - a fundamental determinant of spatial accuracy in fNIRS measurements. Proper cross-validation is especially crucial for real-time applications like neurofeedback and brain-computer interfaces where accurate spatial targeting is essential for efficacy [1].

Fundamental Principles & Signaling Pathways

Neurovascular Coupling Pathway

The following diagram illustrates the fundamental signaling pathway that underlies both fNIRS and fMRI measurements, explaining their physiological relationship.

Figure 1: Neurovascular coupling pathway linking neuronal activity to fNIRS and fMRI signals.

This shared physiological foundation enables cross-validation between modalities. However, important differences exist: fNIRS directly measures both HbO and HbR changes separately, while fMRI's BOLD signal primarily reflects HbR changes and is influenced by other physiological factors [92]. Understanding this relationship is crucial for interpreting validation results.

Experimental Design & Methodologies

Concurrent vs. Sequential Validation Designs

Two primary experimental approaches exist for fNIRS-fMRI validation:

Concurrent Measurements: fNIRS and fMRI data are collected simultaneously within the MRI scanner [94]. This approach eliminates temporal variability but presents technical challenges including MR compatibility of fNIRS equipment and potential electromagnetic interference.

Sequential Measurements: fNIRS and fMRI data are collected in separate sessions using identical task paradigms [93]. This requires careful attention to maintaining consistent task conditions and accounting for intersession variability.

Task Paradigms for Validation

The table below summarizes common task paradigms used in fNIRS-fMRI validation studies:

Table 1: Common task paradigms for fNIRS-fMRI validation

Task Category	Specific Tasks	Targeted Brain Regions	Validation Considerations
Motor Tasks	Finger tapping, Hand movements [93]	Primary motor cortex (M1), Supplementary motor area (SMA)	High test-retest reliability, clear lateralization
Cognitive Tasks	Mental calculation, Mental rotation [2]	Prefrontal cortex, Parietal regions	Individual variability in strategy may affect localization
Language Tasks	Verbal fluency, Inner speech [2] [94]	Prefrontal cortex, Language areas	More variable activation patterns between individuals
Working Memory	N-back tasks, Auditory Stroop [94] [95]	Dorsolateral prefrontal cortex (dlPFC)	Sensitivity to executive demand across populations [95]

fNIRS Optode Placement Methodologies

The accuracy of spatial cross-validation critically depends on optode placement strategies. The table below compares different approaches:

Table 2: Optode placement methodologies for improved spatial specificity

Method	Description	Data Requirements	Advantages	Limitations
Literature-Based (LIT)	Placement based on previous studies and standard coordinates [2]	None	Simple, fast implementation	Ignores individual anatomy, lowest sensitivity
Probabilistic (PROB)	Uses individual anatomy + probabilistic fMRI maps from independent datasets [2]	Individual structural MRI, Group fMRI data	Good balance of practicality and accuracy	Limited by quality of probabilistic maps
Individual fMRI (iFMRI)	Uses subject-specific functional and anatomical data [2]	Individual fMRI and structural MRI	High anatomical and functional precision	Requires additional scanning session
fOLD Toolbox	Automated optode positioning based on photon transport simulations [5]	Head atlases (Colin27, SPM12)	Systematic approach, publicly available	Based on atlas rather than individual anatomy
NIRSTORM Optimal Montage	Computational optimization targeting specific ROIs with constraints [15]	Individual or template anatomy	Maximizes sensitivity to target regions	Requires technical expertise with software

Research demonstrates that approaches incorporating individual neuroimaging data (PROB, iFMRI) significantly outperform literature-based approaches in both signal quality and sensitivity to brain activation [2].

Troubleshooting Guides & FAQs

Common Experimental Challenges & Solutions

Q: We're observing poor spatial correlation between fNIRS channels and fMRI activation in the target region. What could be causing this?

A: This frequent issue can stem from several sources:

• Inaccurate optode positioning: Even small displacements (1-2 cm) can significantly reduce sensitivity to target regions [1]. Solution: Use neuronavigation systems for precise optode placement based on individual anatomy.
• Insufficient source-detector distance: Distances less than 2.5-3 cm in adults may primarily sensitive to extracerebral tissues rather than cortical activity [92]. Solution: Ensure appropriate distance (typically 3-5 cm) while maintaining adequate signal-to-noise ratio.
• Individual anatomical variability: Standard positions may not account for sulcal/gyral variability. Solution: Implement individual MRI-guided optode placement [2].

Q: Our fNIRS signals show similar temporal patterns to fMRI but with inconsistent HbO/HbR correlations. Is this normal?

A: The relationship between HbO and HbR can vary based on multiple factors:

• Physiological differences: fNIRS is more sensitive to systemic physiological noise (cardiac, respiratory, blood pressure oscillations) [94] [1]. Solution: Implement short-distance channels to measure and regress out extracerebral contributions.
• Vascular properties: The BOLD signal has complex relationships with both HbR and HbO, while fNIRS measures them directly [92]. Solution: Focus on consistent task-related responses rather than exact waveform matching.
• Signal processing differences: Filtering approaches can differentially affect HbO and HbR. Solution: Ensure comparable filtering parameters between modalities.

Q: How can we optimize fNIRS probe placement for specific regions like the prefrontal cortex when fMRI isn't available?

A: Several approaches can enhance targeting without individual fMRI:

• Use the fOLD toolbox: This publicly available resource provides probe arrangements optimized for specific brain regions based on atlas data [5].
• Implement probabilistic approaches: Combine individual structural MRI with probabilistic functional maps from independent datasets (PROB approach) [2].
• Leverage NIRSTORM: Use the optimal montage tool in NIRSTORM to design probe layouts maximizing sensitivity to target regions [15].

Technical Validation FAQs

Q: Which fNIRS signal (HbO or HbR) correlates better with the fMRI BOLD signal?

A: Research shows complex relationships:

• The BOLD signal has an inverse relationship with HbR concentrations, so increased BOLD typically correlates with decreased HbR [92].
• However, studies have found that HbR often shows better spatial specificity to the activation focus, particularly for motor imagery tasks [93].
• HbO typically has higher amplitude and better signal-to-noise ratio but may be more susceptible to extracerebral contamination [94].
• Recommendation: Analyze both signals initially, as their relationship can provide quality indicators.

Q: What correlation values should we expect between fNIRS and fMRI signals?

A: Reported correlations vary by brain region and task:

• Motor execution tasks: Typically show moderate to strong correlations (r = 0.4-0.8) [93]
• Prefrontal regions: Often show lower correlations due to stronger physiological noise and anatomical variability
• Mental imagery tasks: Generally lower correlations than motor execution due to weaker activation
• Context matters: Focus on statistical significance and consistency across subjects rather than absolute correlation values

Q: How many channels/subjects are typically needed for adequate validation?

A: This depends on your research goals:

• For methodology papers: 12-20 subjects are typical in the literature [94] [93]
• For clinical applications: Larger samples may be needed to account for greater population variability
• Channel count: Dense arrays are preferable, but even 2-well-placed channels can adequately target specific regions like SMA [2] [93]
• Power considerations: Power analysis should account for expected effect sizes in your target region

Experimental Protocols

Comprehensive Cross-Validation Protocol

The following workflow diagram outlines a complete experimental protocol for fNIRS-fMRI cross-validation, emphasizing optode optimization:

Figure 2: Comprehensive experimental workflow for fNIRS-fMRI cross-validation.

Step-by-Step Implementation Guide

Phase 1: Pre-Experimental Planning

• Define target regions based on research questions and prior literature
• Select appropriate tasks that reliably activate target regions (see Table 1)
• Determine optode placement strategy based on available resources (see Table 2)
• Plan statistical approach for validation metrics

Phase 2: MRI Data Acquisition

• Acquire high-resolution structural scan (T1-weighted) for individual anatomy
• Collect functional localizer data using predetermined tasks
• Identify individual activation peaks within target regions
• Generate probabilistic maps if using PROB approach [2]

Phase 3: fNIRS Probe Placement

• Coregister optode positions with individual anatomy using neuronavigation
• Select optimal source-detector pairs (typically 2.5-4 cm distance for adults)
• Consider implementing short-distance channels (<1 cm) for signal correction [94]
• Document final positions using photographic or digitization methods

Phase 4: Data Collection

• Implement identical task paradigms across modalities
• Maintain consistent timing and stimulus parameters
• Monitor data quality in real-time (signal-to-noise ratio, motion artifacts)
• Collect sufficient trials for robust statistical power

Phase 5: Analysis & Validation

• Preprocess both datasets with comparable parameters (filtering, artifact removal)
• Co-register fNIRS channels with cortical surfaces
• Extract time courses from corresponding regions
• Compute spatial and temporal correlation metrics
• Assess task-related sensitivity using general linear models

Research Reagent Solutions

Table 3: Essential tools and resources for fNIRS-fMRI validation studies

Tool/Resource	Primary Function	Application in Validation	Availability
fOLD Toolbox [5]	Automated optode placement	Optimizing probe arrangement for specific ROIs	Publicly available
NIRSTORM [15]	fNIRS data analysis & optimal montage	Designing subject-specific optode layouts	Brainstorm plugin
AtlasViewer [5]	Probe design and data visualization	Co-registration of fNIRS channels with anatomy	Publicly available
MCX Simulation [15]	Monte Carlo photon transport	Modeling light propagation in head tissues	Open source
Short-Distance Channels [94]	Superficial signal regression	Removing extracerebral contamination	Hardware implementation
Correlation-based Signal Improvement (CBSI) [95]	Hemodynamic data processing	Validated combined hemoglobin measure for executive demand	Algorithm implementation

Validation Metrics Toolkit

Spatial Specificity Measures:

• Topographical similarity: Spearman correlation between fNIRS and fMRI activation maps [93]
• Center-of-mass distance: Spatial offset between activation peaks across modalities
• Spatial overlap metrics: Dice coefficient or Jaccard index for activated regions

Temporal Validation Measures:

• Time-course correlation: Pearson correlation between hemodynamic responses
• Task sensitivity: Effect sizes for task conditions in both modalities
• Lateralization indices: Consistency in hemispheric dominance measures

Advanced Considerations & Future Directions

Emerging approaches in fNIRS-fMRI cross-validation include:

• Multimodal integration: Combining fNIRS with EEG for improved temporal and spatial resolution [92]
• Advanced signal processing: Machine learning approaches for enhanced signal separation and artifact removal [1]
• Individualized head modeling: Using precise anatomical information to improve light transport modeling [2]
• Real-time applications: Validation approaches optimized for neurofeedback and brain-computer interface systems [1] [93]

Successful cross-validation requires meticulous attention to experimental design, optode placement, and analytical approaches. By implementing the methodologies and troubleshooting guides presented here, researchers can establish robust concurrent validity for their fNIRS activation patterns, strengthening the foundation for subsequent research applications.

Conclusion

Optimizing fNIRS optode placement represents a critical methodology advancement for enhancing spatial resolution and measurement reliability. The integration of computational tools like fOLD, combined with subject-specific anatomical and functional data, significantly improves targeting accuracy beyond traditional literature-based approaches. Future directions should focus on standardizing analysis pipelines, developing more accessible neuronavigation solutions, and establishing robust validation protocols for clinical translation. As fNIRS continues to evolve toward real-world applications and biomarker development, precision optode placement will remain fundamental to unlocking its full potential in both neuroscience research and pharmaceutical development, particularly for longitudinal studies and interventional trials requiring high measurement consistency.

References

• 1. Optimizing spatial specificity and signal quality in fNIRS [https://www.frontiersin.org/journals/neuroergonomics/articles/10.3389/fnrgo.2024.1286586/full]
• 2. Guiding functional near-infrared spectroscopy optode- ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC8211086/]
• 3. High-density multidistance fNIRS enhances detection of ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12412631/]
• 4. Reproducibility of fNIRS within Subject for Visual and Motor ... [https://www.sciencedirect.com/science/article/pii/S1053811925004951]
• 5. fNIRS Optodes' Location Decider (fOLD): a toolbox for ... [https://www.nature.com/articles/s41598-018-21716-z]
• 6. Co-localized optode-electrode design for multimodal ... [https://www.spiedigitallibrary.org/journals/neurophotonics/volume-12/issue-2/025006/Co-localized-optode-electrode-design-for-multimodal-functional-near-infrared/10.1117/1.NPh.12.2.025006.full]
• 7. A Multi-Distance Probe Configuration With PPL Method [https://www.frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2020.569508/full]
• 8. Transcranial brain atlas-based optimization for functional ... [https://pubmed.ncbi.nlm.nih.gov/33332685/]
• 9. Anatomical guidance for functional near-infrared ... [https://www.spiedigitallibrary.org/journals/neurophotonics/volume-2/issue-02/020801/Anatomical-guidance-for-functional-near-infrared-spectroscopy-AtlasViewer-tutorial/10.1117/1.NPh.2.2.020801.full]
• 10. Sensor space group analysis for fNIRS data - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC4840017/]
• 11. Optimal positioning of optodes on the scalp for ... [https://www.sciencedirect.com/science/article/abs/pii/S0165027018302425]
• 12. Optimizing spatial specificity and signal quality in fNIRS [https://pmc.ncbi.nlm.nih.gov/articles/PMC11188482/]
• 13. Impact of Anatomical Variability on Sensitivity Profile in ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC9962997/]
• 14. Monte Carlo-Based Optical Simulation of ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10968032/]
• 15. NIRSTORM: Optimizing fNIRS montage [https://neuroimage.usc.edu/brainstorm/Tutorials/NIRS_Optimal_montage]
• 16. A working hypothesis visualization method for fNIRS ... [https://www.sciencedirect.com/science/article/pii/S2215016123003539]
• 17. Sound intensity-dependent cortical : implications of the... activation [https://pmc.ncbi.nlm.nih.gov/articles/PMC12149058/]
• 18. fNIRS: Advances and Challenges [https://sapienlabs.org/fnirs-advances-and-challenges/]
• 19. Applications and advances of combined fMRI-fNIRs ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11958174/]
• 20. Spatial registration for functional near-infrared ... [https://www.sciencedirect.com/science/article/pii/S1053811913007805]
• 21. Best practices for fNIRS publications [https://www.spiedigitallibrary.org/journals/neurophotonics/volume-8/issue-01/012101/Best-practices-for-fNIRS-publications/10.1117/1.NPh.8.1.012101.full]
• 22. devfOLD: a toolbox for designing age-specific fNIRS ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC8647945/]
• 23. Software [https://fnirs.org/resources/data-analysis/software/]
• 24. 10/20, 10/10, and 10/5 systems revisited: Their validity as relative head-surface-based positioning systems - ScienceDirect [https://www.sciencedirect.com/science/article/pii/S1053811906009724]
• 25. 10/20, 10/10, and 10/5 systems revisited : Neuroimage [https://www.ovid.com/journals/neuima/abstract/10.1016/j.neuroimage.2006.09.024~1020-1010-and-105-systems-revisited-their-validity-as?redirectionsource=fulltextview]
• 26. Understanding the 10-20 System of EEG Electrode Placement – EMOTIV [https://www.emotiv.com/blogs/how-to/understanding-the-10-20-system-of-eeg-electrode-placement]
• 27. Current Status and Issues Regarding Pre-processing of ... [https://www.frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2018.00505/full]
• 28. Probabilistic mapping of human functional brain networks ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC8296467/]
• 29. Probabilistic atlas for the language network based on ... [https://www.nature.com/articles/s41597-022-01645-3]
• 30. fNIRS reproducibility varies with data quality, analysis ... [https://www.nature.com/articles/s42003-025-08412-1]
• 31. Probabilistic Maps [https://www.brainvoyager.com/bvqx/doc/UsersGuide/ProbabilisticMaps/ProbabilisticMaps.html]
• 32. NIRSTORM: a Brainstorm extension dedicated to functional ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12081164/]
• 33. Optimal optode montage on electroencephalography ... [https://www.spiedigitallibrary.org/journals/journal-of-biomedical-optics/volume-19/issue-2/026010/Optimal-optode-montage-on-electroencephalography-functional-near-infrared-spectroscopy-caps/10.1117/1.JBO.19.2.026010.full]
• 34. From lab to life: challenges and perspectives of fNIRS for ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11513164/]
• 35. Multimodal assessment of the spatial correspondence ... [https://www.nature.com/articles/s41598-023-29123-9]
• 36. Photon transport modeling [https://openfnirs.org/2024/01/01/photon-transport-modeling/]
• 37. Modelling sensitivity profile using custom MRIs [https://openfnirs.org/community/homer3-forum/modelling-sensitivity-profile-using-custom-mris/]
• 38. Improving model-based functional near-infrared ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC7035879/]
• 39. A Narrative Review on Clinical Applications of fNIRS - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC7573058/]
• 40. Sensitivity of fNIRS to cognitive state and load - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC3930096/]
• 41. Functional near-infrared spectroscopy short-channel ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12368614/]
• 42. Longitudinal infant fNIRS channel-space analyses are ... [https://www.sciencedirect.com/science/article/pii/S1053811921003451]
• 43. Optimized multimodal functional magnetic resonance ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC4675318/]
• 44. Motion artifacts removal and evaluation techniques for ... [https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2022.878750/full]
• 45. Motion artefacts in (f)NIRS — Not a problem, but not to be ... [https://www.artinis.com/blogpost-all/motion-artefacts-in-fnirs-not-a-problem-but-needs-attention]
• 46. Characterizing motion artifacts in functional near-infrared ... [https://www.sciencedirect.com/science/article/pii/S1746809425007670]
• 47. Combining scalp coupling index with coefficient of variation ... [https://opg.optica.org/abstract.cfm?uri=ECBO-2025-Tu2A.13]
• 48. Preprocessing functional near-infrared spectroscopy ... [https://mne.tools/stable/auto_tutorials/preprocessing/70_fnirs_processing.html]
• 49. NIRS-BIDS: Brain Imaging Data Structure Extended to ... [https://www.nature.com/articles/s41597-024-04136-9]
• 50. The application of fNIRS in studies on occupational workload [https://pubmed.ncbi.nlm.nih.gov/40331113/]
• 51. The application of fNIRS in studies on occupational workload [https://pmc.ncbi.nlm.nih.gov/articles/PMC12053328/]
• 52. Investigation of the sensitivity of functional near-infrared ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC5601503/]
• 53. Targeting brain regions of interest in functional near‐infrared ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC8046049/]
• 54. Impact of Anatomical Variability on Sensitivity Profile in ... [https://www.mdpi.com/1424-8220/23/4/2089]
• 55. Functional near-infrared spectroscopy for the assessment ... [https://www.frontiersin.org/journals/neurology/articles/10.3389/fneur.2025.1524806/full]
• 56. An Investigation of Diverse Signal Filtering Methods Within ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC6336925/]
• 57. A Cross-Disciplinary Review of the fNIRS-EEG Dual ... [https://www.mdpi.com/2076-3425/14/10/1022]
• 58. False positives and false negatives in functional near-infrared ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC4791590/]
• 59. Neurofeedback, OpenBCI Boards, and general getting ... [https://openbci.com/forum/index.php?p=/discussion/1662/neurofeedback-openbci-boards-and-general-getting-started-questions-thoughts]
• 60. Guiding functional near-infrared spectroscopy optode- ... [https://www.spiedigitallibrary.org/journals/neurophotonics/volume-8/issue-2/025012/Guiding-functional-near-infrared-spectroscopy-optode-layout-design-using-individual/10.1117/1.NPh.8.2.025012.full]
• 61. Addressing low BCI accuracy – Error sources & solutions [https://openvibe.inria.fr/addressing-low-bci-accuracy-error-sources-solutions/]
• 62. A Comprehensive Review on Critical Issues and Possible ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC8003721/]
• 63. A Deep-Learning Empowered, Real-Time Processing ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12270413/]
• 64. Reducing motion artifacts for long-term clinical NIRS ... [https://www.sciencedirect.com/science/article/abs/pii/S1053811913006915]
• 65. Reducing motion artifacts for long-term clinical NIRS ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC3849205/]
• 66. Artifact reduction in long-term monitoring of cerebral ... [https://www.spiedigitallibrary.org/journals/neurophotonics/volume-2/issue-2/025004/Artifact-reduction-in-long-term-monitoring-of-cerebral-hemodynamics-using/10.1117/1.NPh.2.2.025004.full]
• 67. Integration of Spatial Information Increases Reproducibility ... [https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2020.00746/full]
• 68. Brainsight TMS Neuronavigation | neurocare technology [https://www.neurocaregroup.com/technology/brainsight-tms]
• 69. Applications and advances of combined fMRI-fNIRs ... [https://www.frontiersin.org/journals/neurology/articles/10.3389/fneur.2025.1542075/full]
• 70. Ensuring good fNIRS data quality - Automatic assessment ... [https://www.artinis.com/blogpost-all/automatic-assessment-of-signal-quality]
• 71. Which experimental designs to use in fNIRS — Block design [https://www.artinis.com/blogpost-all/fnirs-block-design]
• 72. A Novel Optimisation Framework for fNIRS [https://pubmed.ncbi.nlm.nih.gov/40857195/]
• 73. Test-retest reliability of functional near-infrared ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10218660/]
• 74. Intra-subject test-retest reliability for auditory-evoked ... [https://www.spiedigitallibrary.org/journals/neurophotonics/volume-12/issue-1/015015/Intra-subject-test-retest-reliability-for-auditory-evoked-functional-near/10.1117/1.NPh.12.1.015015.full]
• 75. Test-retest reliability of fNIRS in resting-state cortical activity ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10549743/]
• 76. Test–retest reliability of functional near infrared ... [https://www.sciencedirect.com/science/article/pii/S0925492725000484]
• 77. Reliability of brain metrics derived from a Time-Domain ... [https://www.nature.com/articles/s41598-024-68555-9]
• 78. fNIRS reproducibility varies with data quality, analysis ... [https://pubmed.ncbi.nlm.nih.gov/40760004/]
• 79. Preprocessing Troubleshooting - Forum [https://openfnirs.org/community/homer3-forum/preprocessing-troubleshooting/]
• 80. Functional near-infrared spectroscopy [https://en.wikipedia.org/wiki/Functional_near-infrared_spectroscopy]
• 81. Comparison of fNIRS with other neuroimaging modalities [https://www.artinis.com/blogpost-all/comparison-fnirs-vs-fmri]
• 82. Applications of Functional Near-Infrared Spectroscopy (fNIRS ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC6306799/]
• 83. Automated Processing of fNIRS Data—A Visual Guide to ... [https://www.mdpi.com/1999-4893/11/5/67]
• 84. Estimating the spatial of resolution sensors for BCI purposes fNIRS [https://www.academia.edu/78742739/Estimating_the_spatial_resolution_of_fNIRS_sensors_for_BCI_purposes]
• 85. Best practices for fNIRS publications - PMC - PubMed Central [https://pmc.ncbi.nlm.nih.gov/articles/PMC7793571/]
• 86. Functional Near-Infrared Spectroscopy and Its Clinical Application in... [https://pmc.ncbi.nlm.nih.gov/articles/PMC7364176/]
• 87. The NIRS Brain AnalyzIR Toolbox - PMC - PubMed Central [https://pmc.ncbi.nlm.nih.gov/articles/PMC11218834/]
• 88. Correcting physiological noise in whole-head functional ... [https://www.sciencedirect.com/science/article/pii/S0165027021001977]
• 89. Nested and robust modeling techniques for fNIRS data with ... [https://www.sciencedirect.com/science/article/abs/pii/S016801022200267X]
• 90. Improving neuroimaging headgear placement robustness ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12546995/]
• 91. Spatial co-registration of functional near-infrared spectroscopy ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC5581202/]
• 92. Functional Near-Infrared Spectroscopy and Its Clinical ... [https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2020.00724/full]
• 93. fMRI-based validation of continuous-wave fNIRS ... [https://www.nature.com/articles/s41598-022-06519-7]
• 94. Concurrent fNIRS-fMRI measurement to validate a method for ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC4478864/]
• 95. Validation of fNIRS measurement of executive demand ... [https://www.sciencedirect.com/science/article/pii/S2213158224000767]