Bridging Neuroimaging Modalities: Integrating fMRI and fNIRS for Advanced Motor Paradigms in Research and Drug Development

Abstract

This article synthesizes current evidence and methodologies for integrating functional magnetic resonance imaging (fMRI) and functional near-infrared spectroscopy (fNIRS) in motor task paradigms. Aimed at researchers, scientists, and drug development professionals, it explores the foundational principles of hemodynamic correlation, presents robust methodological frameworks for simultaneous and asynchronous data acquisition, and addresses key technical challenges in spatial co-registration and signal quality. Furthermore, it examines the validation of fNIRS against the fMRI gold standard and discusses the transformative applications of this multimodal approach in clinical trials, neurofeedback, and real-world motor assessment, providing a comprehensive roadmap for its implementation in neuroscience and therapeutic development.

The Hemodynamic Bridge: Understanding the Physiological Basis for fMRI-fNIRS Integration in Motor Neuroscience

Understanding neurovascular coupling— the critical relationship between neuronal activity, cerebral blood flow, and subsequent metabolic changes—is fundamental to interpreting functional neuroimaging data. Two primary non-invasive imaging techniques used to study this phenomenon are functional magnetic resonance imaging (fMRI) and functional near-infrared spectroscopy (fNIRS). fMRI measures the Blood Oxygen Level Dependent (BOLD) signal, which is primarily sensitive to changes in deoxygenated hemoglobin (HbR) [1] [2]. In contrast, fNIRS directly measures concentration changes in both oxygenated (HbO) and deoxygenated hemoglobin (HbR) in the cortical microvasculature [3]. Framed within a broader thesis on integrating fMRI and fNIRS for motor task paradigms, this Application Note provides a detailed comparative analysis of the hemodynamic signals measured by these techniques. We present standardized protocols for simultaneous data acquisition, analysis frameworks for comparing BOLD and HbO/HbR dynamics, and practical tools for researchers investigating neurovascular coupling in both basic science and clinical drug development contexts.

Theoretical Foundations: BOLD, HbO, and HbR

The BOLD Signal in fMRI

The fMRI BOLD signal is an indirect and complex measure of neuronal activity. It relies on detecting localized changes in the magnetic properties of blood, specifically the concentration of paramagnetic deoxyhemoglobin (HbR) [1]. During neural activation, a cascade of events known as the hemodynamic response leads to an increase in cerebral blood flow (CBF) that exceeds the local brain tissue's oxygen consumption. This results in a net decrease in HbR concentration in the venous capillaries and surrounding tissue, which in turn increases the T2* relaxation time measurable with MRI, producing the positive BOLD signal [1] [2]. The BOLD signal is therefore not a direct measure of blood flow or oxygenation, but rather a composite signal influenced by changes in cerebral blood flow, blood volume, and the cerebral metabolic rate of oxygen (CMRO2) [2]. Its temporal resolution is constrained by the sluggishness of the hemodynamic response, which typically peaks 3-5 seconds after stimulus onset [1].

Hemodynamic Signals in fNIRS

fNIRS utilizes near-infrared light (650-950 nm) to measure changes in HbO and HbR concentrations based on the modified Beer-Lambert law [3] [4]. Unlike fMRI, fNIRS provides direct, quantitative measurements of both hemoglobin species, offering a more straightforward interpretation of the hemodynamic response. During neural activation, the typical fNIRS response shows a characteristic increase in HbO and a concurrent decrease in HbR, reflecting the neurovascularly coupled increase in blood flow and oxygen delivery [3]. fNIRS boasts superior temporal resolution (often millisecond-level precision) compared to fMRI, allowing it to capture rapid hemodynamic dynamics [3]. However, it is limited to monitoring superficial cortical regions due to the limited penetration depth of light and offers lower spatial resolution than fMRI [3].

The Neurovascular Coupling Link

The physiological connection between the BOLD signal and fNIRS measurements lies in the shared hemodynamic origin. The positive BOLD signal is predominantly determined by the change in deoxygenated hemoglobin (ΔHbR) [1] [2]. Consequently, the fNIRS-measured ΔHbR timecourse is theoretically the most direct fNIRS correlate of the BOLD signal. However, the relationship is not one-to-one, as the BOLD signal is also influenced by blood volume and flow changes in larger draining veins, which can spatially blur the underlying neural activity [1]. The integrated HbO response from fNIRS often provides a robust and sensitive measure of the focal hemodynamic change, though it may not correlate with the BOLD signal as directly as HbR [3].

The diagram below illustrates the shared neurovascular coupling pathway that links neural activity to the measurable signals in fMRI and fNIRS.

Diagram 1: The neurovascular coupling pathway demonstrates how neural activity is translated into measurable signals for fNIRS and fMRI. The BOLD signal is most directly influenced by the change in HbR concentration (highlighted in red).

Quantitative Comparison of fMRI and fNIRS Signals

The table below summarizes the fundamental characteristics of the hemodynamic signals measured by fMRI and fNIRS, highlighting their complementary nature.

Table 1: Quantitative and Qualitative Comparison of fMRI-BOLD and fNIRS Signals

Feature	fMRI-BOLD Signal	fNIRS Signals (HbO/HbR)
Primary Physiological Basis	Change in deoxyhemoglobin (HbR) concentration affecting T2* relaxation [1]	Direct concentration changes of oxyhemoglobin (HbO) and deoxyhemoglobin (HbR) [3]
Spatial Resolution	High (millimeter-level), whole-brain coverage including subcortical structures [3]	Lower (1-3 cm), restricted to superficial cortical regions [3]
Temporal Resolution	Low (0.33-2 Hz), limited by hemodynamic lag (4-6 s) [3]	High (up to millisecond precision) [3]
Portability & Environment	Low (requires immobile, shielded scanner); unsuitable for naturalistic settings [3]	High (portable systems); suitable for bedside, clinic, and real-world environments [3] [5]
Primary Signal Correlate	Negative correlation with HbR [1] [2]	HbO: Generally increases with activation.HbR: Generally decreases with activation [3]
Sensitivity to Confounds	Sensitive to motion, magnetic susceptibility artifacts [3]	Sensitive to scalp blood flow, hair color, and motion (though more resilient) [3] [6]

Application Notes & Protocols for Motor Paradigms

Integrating fMRI and fNIRS is particularly valuable in motor research, where understanding the spatiotemporal dynamics of activation is key. The following protocol provides a framework for simultaneous data acquisition during a motor task.

Synchronized fMRI-fNIRS Data Acquisition Protocol

Objective: To acquire co-registered BOLD and HbO/HbR signals during a motor execution task for the study of neurovascular coupling. Primary Application: Validation of fNIRS against fMRI, high-resolution spatiotemporal mapping of motor cortex hemodynamics. Experimental Setup: The subject lies in the MRI scanner. The fNIRS optodes are integrated into an MRI-compatible cap and positioned over the primary motor cortex (C3/C4 locations of the 10-20 system). The fNIRS system must be MR-compatible to avoid interference and ensure subject safety [3].

Procedure:

• Subject Preparation: Secure the integrated fNIRS-fMRI cap on the subject's head. Ensure firm optode-scalp contact. Use a plastic mirror attached to the head coil to allow the subject to view visual cues.
• Hardware Synchronization: Use a transistor-transistor logic (TTL) pulse from the MRI scanner to trigger the onset of the fNIRS recording, ensuring temporal alignment of both data streams.
• Task Paradigm (Block Design):
  • Baseline (30 s): The subject remains at rest, fixating on a cross.
  • Activation (30 s): The subject performs a motor task (e.g., sequential finger-thumb opposition at 2 Hz) with their right hand, cued by a visual stimulus.
  • Repeat: Alternate between baseline and activation blocks for a total of 5-6 cycles.
• Data Acquisition:
  • fMRI: Acquire whole-brain BOLD-weighted EPI images (TR = 2000 ms, TE = 30 ms, voxel size = 3x3x3 mm³).
  • fNIRS: Continuously record light intensity at dual wavelengths (e.g., 760 nm and 850 nm) from all channels over the motor cortex at a high sampling rate (≥ 10 Hz).

The workflow for this integrated experiment is visualized below.

Diagram 2: Workflow for a synchronized fMRI-fNIRS motor paradigm experiment, showing the integration of setup, task design, and simultaneous data acquisition.

Protocol Variations and Considerations

• Cerebellar Motor Tasks: For investigating cerebellar motor circuits, a validated bedside task like alternating pronation-supination of the hand (to assess diadochokinesia) can be used [4]. fNIRS optode placement should target the cerebellar region (below the inion), acknowledging the depth-related sensitivity challenges. Protocol 2 from [4], which utilized more detectors and closer optode spacing (2 cm), demonstrated superior feasibility (100% stable recordings) over simpler configurations.
• Data Quality Control: fNIRS data quality is paramount for reproducibility. The scalp-coupling index (SCI) should be calculated to identify and exclude channels with poor contact [7] [6]. Note that data quality can be affected by task type (e.g., overt speech like Picture Naming can produce lower quality signals) and participant characteristics (e.g., hair color, skin pigmentation) [6].
• Clinical Populations: When studying populations with potential neurovascular pathology (e.g., stroke, Type 2 Diabetes), the Hemodynamic Response Function (HRF) may be altered (e.g., more sluggish peak, altered initial dip/undershoot) [8]. In such cases, using a deconvolution approach to model the HRF, rather than assuming a canonical shape, is critical for accurate analysis [8].

Data Analysis and Integration Workflow

The analysis of simultaneous fMRI-fNIRS data involves parallel processing streams followed by a multimodal integration stage, as outlined in the table below.

Table 2: Data Analysis Pipeline for Simultaneous fMRI-fNIRS Data

Analysis Stage	fMRI (BOLD) Processing	fNIRS (HbO/HbR) Processing	Integration & Comparison
Preprocessing	Slice-time correction, motion realignment, spatial normalization, smoothing [3]	Conversion of raw light intensity to optical density, filtering of cardiac/pulse (0.5-2 Hz) and respiratory (0.1-0.5 Hz) noise, motion artifact correction [3] [7]	Temporal down-sampling of fNIRS data to match fMRI TRs.
First-Level Modeling	General Linear Model (GLM) analysis with a canonical HRF convolved with the task paradigm to generate statistical parametric maps (e.g., T-maps) [8]	GLM analysis using the same task regressor to estimate beta coefficients for HbO and HbR changes for each channel. Alternatively, block-average the response.	Spatial correlation of fNIRS channel locations with underlying fMRI activation foci.
Signal Comparison	Extract the mean BOLD timecourse from a cluster of activated voxels corresponding to the fNIRS measurement location.		Quantitative Correlation: Calculate the correlation coefficient between the preprocessed BOLD signal and the fNIRS-derived HbR timecourse. Theoretically, a strong negative correlation is expected [3].
Advanced Modeling	-	-	Joint HRF Modeling: Fit a physiological model (e.g., within a Dynamic Causal Modeling framework) to the combined BOLD and HbO/HbR data to infer underlying neural activity and neurovascular coupling parameters [2].

The Scientist's Toolkit: Research Reagent Solutions

The table below lists key materials and tools essential for conducting integrated fMRI-fNIRS studies on neurovascular coupling.

Table 3: Essential Research Tools for Integrated fMRI-fNIRS Studies

Item / Solution	Function & Application Note
MR-Compatible fNIRS System	A specialized fNIRS device with fiber-optic cables and optodes made from non-magnetic materials to operate safely inside the MRI scanner without causing interference or artifacts [3].
Integrated fNIRS-fMRI Caps	Head caps with pre-configured holders that securely position fNIRS optodes over cortical regions of interest (e.g., motor cortex) while being compatible with the MRI head coil.
TTL Pulse Generator	A critical synchronization tool that sends a transistor-transistor logic pulse from the MRI scanner to the fNIRS system at the start of the scan, aligning both data streams in time [3].
3D Digitizer	A device (e.g., electromagnetic or optical) used to record the precise 3D spatial coordinates of fNIRS optodes relative to cranial landmarks. This enables accurate co-registration with the high-resolution anatomical MRI scan.
Quality Testing Toolboxes (e.g., QT-NIRS)	Software toolboxes that calculate metrics like the Scalp-Coupling Index (SCI) to automatically identify and flag poor-quality fNIRS channels for exclusion or further processing, improving data reliability [6].
Physiological Informed Dynamic Causal Modeling (P-DCM)	An advanced computational framework for analyzing effective connectivity between brain regions. It uses a generative physiological model of the BOLD signal, which can be informed and constrained by simultaneous fNIRS measurements of HbO/HbR [2].
Validated Motor Paradigms	Standardized task scripts (e.g., finger-thumb opposition, alternating pronation-supination) that reliably activate the targeted motor circuitry (corticospinal vs. cerebellar) and are suitable for use in both fMRI and fNIRS environments [4].

The combined use of fMRI and fNIRS provides a powerful, multi-faceted lens through which to study neurovascular coupling. By leveraging fMRI's high spatial resolution and whole-brain coverage alongside fNIRS's direct measurement of hemoglobin dynamics, superior temporal resolution, and portability, researchers can construct a more complete and quantifiable picture of the brain's hemodynamic response. The protocols and analyses detailed in this Application Note provide a concrete framework for designing and executing studies that bridge these two complementary modalities, with specific relevance to motor system research. This integrated approach is poised to advance our fundamental understanding of neurovascular physiology and enhance the evaluation of diagnostic and therapeutic strategies in neurological disorders and drug development.

The integration of functional near-infrared spectroscopy (fNIRS) and functional magnetic resonance imaging (fMRI) presents a powerful multimodal approach for investigating motor network functions. While fMRI provides high spatial resolution for deep brain structures, fNIRS offers superior temporal resolution, portability, and tolerance for motion artifacts, making it suitable for dynamic motor tasks and naturalistic environments [9]. This application note details protocols and analytical frameworks for establishing spatial correspondence between fNIRS channels and fMRI activation clusters, specifically within motor paradigms. Such correspondence is crucial for translating well-established fMRI paradigms to more flexible fNIRS setups, thereby advancing cognitive and clinical neuroscience research [10].

Quantitative Benchmarking of Spatial Correspondence

Empirical studies directly comparing fNIRS and fMRI during motor tasks demonstrate a significant spatial overlap, validating fNIRS as a reliable tool for mapping cortical motor activity. The following table summarizes key quantitative findings from recent studies.

Table 1: Spatial Correspondence Metrics between fNIRS and fMRI in Motor Tasks

Study Reference	Participant Count	Motor Task Paradigm	Spatial Overlap (True Positive Rate)	Positive Predictive Value (PPV)	Key Findings
Zinos et al. (2024) [11] [12]	22	Finger tapping, Visual checkerboard	Up to 68% (group analysis); 47.25% average (within-subject)	51% (group level); 41.5% (within-subject)	Good spatial correspondence, supporting clinical use for superficial cortex.
Multimodal Assessment (2023) [10]	9	Motor imagery and execution	Significant peak activation overlapping individually-defined M1 and PMC	No significant difference between HbO, HbR, and HbT	Validated translation of neuronal information from fMRI to fNIRS setup.

These studies confirm that fNIRS can reliably detect hemodynamic activity in primary motor (M1) and premotor cortices (PMC) that corresponds to fMRI activation clusters [11] [10]. The within-subject analysis shows moderate spatial overlap and PPV, highlighting the importance of individualized assessment for clinical applications. The PPV, which was lower for within-subject analyses, indicates the presence of fNIRS activity in regions without corresponding fMRI signals, potentially due to physiological noise or differing sensitivities of the modalities to hemodynamic changes [11].

Detailed Experimental Protocols

This section provides standardized methodologies for conducting simultaneous and asynchronous fMRI-fNIRS studies on motor networks.

Protocol 1: Synchronous fMRI-fNIRS Acquisition for Validation

This protocol is designed for the direct spatial comparison of fNIRS channels and fMRI activation clusters.

• Participant Preparation: Recruit healthy adult participants with no history of neurological disorders. Obtain informed consent. For simultaneous acquisition, use MRI-compatible fNIRS optodes to ensure safety and data integrity [9].
• Equipment and Setup:
  • fMRI: 3T Siemens Magnetom TimTrio scanner with a 12-channel head coil. Acquire high-resolution anatomical images (e.g., MPRAGE: 176 slices, 1x1x1 mm voxels) [10].
  • fNIRS: Use a whole-head fNIRS system (e.g., NIRSport2) with optodes placed over motor regions. A typical setup includes 16 sources and 15 detectors (54 channels) with an inter-optode distance of 30 mm to ensure sufficient penetration depth [11] [10]. Integrate short-distance detectors (e.g., 8 mm separation) to regress out superficial physiological noise [10].
• Motor Task Paradigm: Employ a block design. A sample paradigm includes:
  • Rest Baseline: 30 seconds.
  • Activation Block: 30 seconds of bilateral finger tapping (e.g., a sequence like 1-2-1-4-3-4, where each number represents a different finger) [10] or a validated cerebellar task (e.g., alternating pronation/supination) [4].
  • Repeat for 4-5 cycles per run. Total run duration: ~8.5 minutes [10].
• Data Acquisition:
  • fMRI: Acquire functional images with a T2*-weighted echo-planar imaging (EPI) sequence (e.g., TR/TE = 1500/30 ms, 26 slices, 3x3 mm in-plane resolution) [10].
  • fNIRS: Record continuous-wave light intensity at two wavelengths (e.g., 760 and 850 nm) at a sampling rate of 5.08 Hz or higher [10] [4].

Protocol 2: Asynchronous fNIRS-fMRI for Protocol Translation

This protocol is used when direct simultaneous acquisition is not feasible, leveraging subject-specific fNIRS signals to model fMRI data.

• Participant Preparation: Same as Protocol 1.
• Experimental Procedure: Conduct the fMRI and fNIRS scanning sessions on the same day but separately. Ensure the participant performs the identical motor task (e.g., motor execution and imagery) in both scanners to maintain consistency [10].
• Data Processing and Modeling:
  • fNIRS Processing: Preprocess signals (pruning bad channels, converting to optical density, filtering) and calculate concentration changes for HbO, HbR, and total hemoglobin (HbT) using the modified Beer-Lambert law [10] [4].
  • fMRI Modeling: Instead of a standard GLM with task predictors, use the preprocessed, subject-specific fNIRS time-series (e.g., HbO from a key channel over M1) as a regressor of interest in the fMRI General Linear Model (GLM). This tests the ability of the fNIRS signal to predict fMRI activation in motor networks [10].

Visualization and Workflow

The following diagram illustrates the core data processing workflow for establishing spatial correspondence, applicable to both synchronous and asynchronous protocols.

The Scientist's Toolkit: Essential Research Reagents and Materials

Table 2: Key Equipment and Software for fMRI-fNIRS Motor Studies

Item Name	Function/Application	Example Specifications/Models
High-Field MRI Scanner	Provides high-spatial-resolution whole-brain BOLD signals and anatomical reference.	3T Siemens Magnetom TimTrio with 12-channel head coil [10].
Portable fNIRS System	Measures cortical hemodynamics (HbO, HbR) during motor tasks outside or inside the scanner.	NIRSport2 (NIRx) continuous-wave system [10] [13].
MRI-Compatible fNIRS Optodes	Enables safe, simultaneous data acquisition inside the MRI bore without causing interference.	Fiber-optic bundles with non-magnetic components [9].
fNIRS Cap with Short-Distance Detectors	Standardizes probe placement and helps separate cerebral from extracerebral physiological noise.	Caps based on the 10-20 EEG system; integrated SDDs at 8 mm [10].
Data Processing Software	For preprocessing, analyzing, and co-registering multimodal neuroimaging data.	BrainVoyager QX (fMRI); Homer3 (fNIRS); Custom scripts in MATLAB [10].
Validated Motor Task Protocols	Ensures robust and reproducible activation of targeted motor networks (M1, PMC, Cerebellum).	Bilateral finger tapping; Motor Imagery; Single-leg squat; Diadochokinesia tasks [10] [13] [4].

This application note synthesizes current evidence and methodologies for establishing robust spatial correspondence between fNIRS channels and fMRI activation clusters in motor networks. The provided quantitative benchmarks, detailed protocols, and essential toolkit offer researchers a foundational framework for designing and executing studies that leverage the complementary strengths of fMRI and fNIRS. This multimodal approach is poised to advance our understanding of motor control in both healthy and clinical populations, such as stroke survivors [6] and patients with ACL injuries [13], by enabling flexible and reliable functional brain imaging.

Understanding the temporal dynamics of the hemodynamic response function (HRF) is fundamental to the accurate interpretation of non-invasive neuroimaging data. In the context of motor task paradigms, analyzing the concordance of HRFs measured by functional magnetic resonance imaging (fMRI) and functional near-infrared spectroscopy (fNIRS) provides a critical foundation for robust multimodal research. Both techniques rely on neurovascular coupling, where neuronal activity triggers a hemodynamic response, but they measure related yet distinct aspects of this complex physiological process [14] [15]. fMRI detects the blood oxygenation level-dependent (BOLD) signal, which primarily reflects changes in deoxygenated hemoglobin [14], while fNIRS directly measures concentration changes in both oxygenated (Δ[HbO]) and deoxygenated (Δ[HbR]) hemoglobin in cortical blood vessels [16] [15]. This application note examines the temporal concordance between these hemodynamic measures and provides detailed protocols for their analysis in motor task research, supporting the broader thesis that integrated fMRI-fNIRS approaches yield more comprehensive insights into brain function.

Theoretical Background: HRF Fundamentals

The hemodynamic response function characterizes the temporal evolution of blood flow and oxygenation changes following neural activation. The canonical HRF exhibits a characteristic shape: an initial dip, a main peak occurring 4-6 seconds after stimulus onset, a return to baseline, and sometimes a post-stimulus undershoot [14] [15]. However, substantial evidence indicates that HRF shape varies significantly across brain regions, individuals, and brain states [17]. For instance, recent investigations have revealed that white matter HRFs demonstrate reduced peak amplitudes, delayed onset times, and prolonged initial dips compared to gray matter responses [18] [19]. These variations have profound implications for data analysis and interpretation in both unimodal and multimodal neuroimaging studies.

Table 1: Key Characteristics of Hemodynamic Response Functions Across Modalities and Tissue Types

Characteristic	fMRI (GM)	fNIRS (Δ[HbO])	fNIRS (Δ[HbR])	fMRI (WM)
Primary Signal	BOLD (mainly HbR)	Δ[HbO] concentration	Δ[HbR] concentration	BOLD (mainly HbR)
Typical Peak Time	4-6 seconds [14]	5-7 seconds [15]	5-8 seconds [15]	8-10 seconds [18]
Initial Dip	Sometimes present	Variable	More pronounced	Prolonged [18]
Spatial Specificity	High (1-3mm) [14]	Moderate (1-3cm) [14]	Higher than HbO [16]	Variable
Temporal Resolution	0.3-2 Hz [14]	5-10 Hz [15]	5-10 Hz [15]	0.3-2 Hz

Quantitative Data Synthesis

Empirical studies directly comparing fMRI and fNIRS hemodynamic responses during motor tasks provide valuable insights into their concordance. A validation study focusing on the supplementary motor area (SMA) during motor execution and motor imagery found that fNIRS reliably captured SMA activation patterns corresponding to fMRI BOLD responses [16]. Notably, the study revealed subtle differences between motor tasks, indicating that for whole-body motor imagery as well as for motor imagery of hand movements, Δ[HbR] provided a more specific signal than Δ[HbO] [16]. This finding is particularly relevant for designing neurofeedback protocols where signal specificity is crucial.

Reproducibility investigations examining fNIRS for motor and visual tasks across multiple sessions have demonstrated that Δ[HbO] is significantly more reproducible across sessions than Δ[HbR] (F(1, 66) = 5.03, p < 0.05) [20]. This enhanced reproducibility, coupled with typically larger amplitude changes, explains why many neurofeedback applications preferentially utilize Δ[HbO] signals despite the potentially superior specificity of Δ[HbR] in certain paradigms [16] [20].

Table 2: HRF Parameter Comparisons Between GM and WM During Task Performance

HRF Parameter	Gray Matter (Mean ± Variance)	White Matter (Mean ± Variance)	Statistical Significance
Time to Peak (TTP)	6.14 ± 0.27 seconds [18]	8.58-10.00 seconds [18]	P < 0.05 for 9 of 11 tracts [18]
Peak Magnitude	5.3x higher than WM [18]	~19% of GM response [18]	P < 0.05 for all tracts [18]
Area Under Curve (AUC)	Significantly larger [18]	Reduced [18]	P < 0.05 for all tracts [18]
Initial Dip Duration	Standard	Prolonged [18]	Region-dependent

Experimental Protocols

Protocol 1: Concurrent fMRI-fNIRS Validation for Motor Paradigms

Purpose: To validate fNIRS measurements against the fMRI gold standard for motor execution and imagery tasks targeting the supplementary motor area and primary motor cortex [16].

Materials:

• 3T MRI scanner with compatible fNIRS system
• 16-source, 16-detector fNIRS optode array configured for motor cortex coverage
• Electromyography (EMG) system for monitoring inadvertent movements
• Response recording device for task performance metrics

Procedure:

• Participant Screening: Recruit right-handed healthy older participants (56-71 years); exclude for cognitive decline (MoCA), neurological conditions, or contraindications for MRI [16].
• Optode Placement: Position fNIRS optodes over SMA and bilateral M1 using international 10-20 system guidance; digitize positions for coregistration with anatomical MRI.
• Task Design: Implement block design (30s task/30s rest) for: (a) motor execution of left-hand movements, (b) motor execution of right-hand movements, (c) motor imagery of left-hand movements, (d) motor imagery of right-hand movements, (e) motor imagery of whole-body movements [16].
• Data Acquisition: Collect simultaneous fMRI (BOLD contrast) and fNIRS (Δ[HbO], Δ[HbR]) data during task performance; monitor EMG to ensure compliance during imagery conditions.
• Preprocessing: Process fNIRS data to convert raw intensities to concentration changes; reconstruct fMRI data with standard pipelines including motion correction and spatial normalization.
• Analysis: Extract HRF time courses from both modalities; compute spatial correlations between fMRI beta maps and fNIRS topographic maps; assess task sensitivity and spatial specificity for each hemoglobin species.

Protocol 2: Interactive Motor-Cognitive Dual-Task Assessment

Purpose: To investigate how varying levels of interactive motor-cognitive dual-task difficulty affect brain activation, functional connectivity, and behavioral performance using fNIRS [21].

Materials:

• Wireless continuous-wave fNIRS system with 102 channels spanning entire head
• Inertial Measurement Unit (IMU) sensors for gait analysis
• Stroop task presentation system
• Cognitive performance recording apparatus

Procedure:

• Participant Preparation: Recruit 28 healthy right-handed adults; screen for normal cognitive function (MMSE ≥ 24); obtain informed consent.
• Optode Montage: Implement comprehensive optode placement covering prefrontal, dorsolateral prefrontal, premotor, sensorimotor, and motor cortices bilaterally.
• Task Conditions: Implement three difficulty levels of interactive motor-cognitive dual task: (a) Easy Task (ET), (b) Medium Task (MT), (c) Difficult Task (DT) combining walking with Stroop color-word interference tasks [21].
• Data Collection: Acquire fNIRS data (Δ[HbO], Δ[HbR]) throughout task performance; simultaneously record gait parameters (speed, stride length) and cognitive performance metrics.
• Signal Processing: Apply motion correction, bandpass filtering, and general linear model (GLM) approaches to fNIRS data; compute functional connectivity matrices between regions of interest.
• Statistical Analysis: Perform repeated measures ANOVA to examine differences in activation, connectivity, and behavioral measures across difficulty conditions; compute lateralization indices to assess hemispheric dominance.

Visualization and Workflow Diagrams

Multimodal HRF Analysis Workflow - This diagram illustrates the integrated experimental and analytical pipeline for assessing HRF concordance between fMRI and fNIRS modalities.

HRF Variation Between Tissues - This diagram visualizes the key temporal differences between gray matter and white matter hemodynamic response functions.

The Scientist's Toolkit: Essential Research Reagents and Materials

Table 3: Essential Materials for Multimodal HRF Research

Item	Specification	Function/Purpose
fNIRS System	Continuous-wave, 16+ sources, 16+ detectors [16]	Measures cortical Δ[HbO] and Δ[HbR] concentrations with 1-3cm spatial resolution
MRI Scanner	3T with BOLD capability, head coil [16]	Provides high-resolution spatial localization of neural activity (1-3mm)
Optode Digitizer	3D position tracking system [20]	Records precise optode placement for coregistration with anatomical MRI
EMG System	Surface electrodes, multi-channel [16]	Monitors muscle activity to ensure compliance during motor imagery tasks
Stimulus Presentation	Paradigm software with timing precision	Controls task timing and records behavioral responses
Motion Tracking	Inertial Measurement Units (IMUs) [21]	Quantifies head movement and gait parameters during tasks
Analysis Software	GLM-capable packages (SPM, NIRS-based) [15]	Implements statistical models for HRF estimation and concordance testing

The concordance between fMRI and fNIRS hemodynamic response functions provides a robust foundation for multimodal investigations of motor task paradigms. Empirical evidence confirms that fNIRS reliably captures task-related activation in motor regions, with Δ[HbR] often demonstrating superior spatial specificity while Δ[HbO] offers better reproducibility [16] [20]. Critically, researchers must account for substantial HRF variations across different tissue types, with white matter responses showing characteristically delayed and attenuated profiles compared to gray matter [18] [19]. The protocols and analytical frameworks presented herein enable comprehensive assessment of temporal dynamics in hemodynamic responses, facilitating more accurate interpretation of neural activity across diverse populations and experimental conditions. Future methodological advances should focus on optimizing integrated analysis pipelines that leverage the complementary strengths of these hemodynamic modalities while accounting for their inherent physiological differences.

The integration of functional Magnetic Resonance Imaging (fMRI) and functional Near-Infrared Spectroscopy (fNIRS) represents a powerful multimodal approach for investigating brain function, particularly within motor task paradigms. The efficacy of this integration hinges on a deep understanding of the common physiological origin of their signals: the hemodynamic response mediated by neurovascular coupling (NVC). This is formally described by the Balloon Model, a theoretical framework that mathematically relates changes in blood flow and oxygen metabolism to the measured signals [22] [23]. This article details the theoretical underpinnings of these signals and provides application notes and protocols for researchers aiming to employ these modalities in tandem, especially for motor task research in drug development and cognitive neuroscience.

Theoretical Foundations: The Balloon Model and Neurovascular Coupling

The Balloon Model

The Balloon Model provides a biophysically grounded description of the hemodynamic changes that occur in response to neuronal activity. It conceptualizes a venular compartment as a "balloon" that inflates with blood and deflates as blood drains away [22] [23].

• Core Premise: The model posits that an increase in neuronal activity triggers a localized increase in Cerebral Blood Flow (CBF). This influx of oxygenated blood initially exceeds the oxygen consumption of the tissue, leading to a washout of deoxygenated hemoglobin (HbR).
• Signal Generation: In fMRI, this decrease in HbR, which is paramagnetic, leads to an increase in the Blood Oxygen Level-Dependent (BOLD) signal [24]. For fNIRS, the same physiological process results in a measured increase in oxygenated hemoglobin (HbO) and a decrease in HbR [25] [10].
• Dynamic Mass Balance: The model is governed by equations of mass conservation for blood volume and deoxyhemoglobin content, effectively linking the dynamics of blood inflow, outflow, and oxygen extraction to the observed signals [23].

The following diagram illustrates the core logic of the Balloon Model and its relationship to the measured signals in fMRI and fNIRS.

Neurovascular Coupling (NVC)

NVC is the biological process that the Balloon Model describes mathematically. It involves a coordinated response within the Neurovascular Unit (NVU), which includes neurons, astrocytes, vascular smooth muscle cells, and pericytes [26].

• Sequence of Events: Active neurons release signaling molecules (e.g., glutamate). This leads to the generation of vasoactive agents such as nitric oxide (NO) and prostaglandins (PG) from neurons and astrocytes.
• Vasodilation: These agents cause the dilation of arterioles, increasing local CBF—a process known as functional hyperemia [26] [22].
• Pathological Implications: Dysregulation of NVC is a key pathophysiological feature in conditions like stroke, cerebral small vessel disease, and Alzheimer's disease, making it a critical target for drug development [26].

The diagram below maps the key cellular interactions within the neurovascular unit that underpin NVC.

Quantitative Comparison of fMRI and fNIRS Signals

A solid understanding of the quantitative relationship between fMRI and fNIRS signals is essential for designing multimodal studies. The table below summarizes the key comparative characteristics of these two modalities, synthesized from empirical studies.

Table 1: Quantitative and Qualitative Comparison of fMRI and fNIRS Hemodynamic Signals

Feature	fMRI (BOLD Signal)	fNIRS	Supporting Evidence
Primary Signal Source	Changes in deoxygenated hemoglobin (HbR) concentration [24].	Changes in oxygenated (HbO) and deoxygenated hemoglobin (HbR) concentration [25] [10].	[25] [24]
Temporal Correlation	Gold standard reference.	HbO often shows the strongest correlation with BOLD; HbR is inversely correlated but may have lower SNR [25] [27].	[25] [10] [27]
Spatial Correspondence	High resolution (~1-2 mm), whole-brain coverage [24].	Lower resolution, superficial cortex only; activation foci show good spatial overlap with fMRI in motor areas [25] [10].	[25] [10]
Signal-to-Noise Ratio (SNR)	Generally high.	Weaker SNR, influenced by scalp-skull distance and probe placement [25].	[25]
Key Advantages	High spatial resolution, whole-brain capability.	Portable, tolerant of movement, direct measure of HbO/HbR, quieter environment [10] [24].	[10] [24]

Application Note: A Protocol for Multimodal Motor Task Research

This protocol outlines a procedure for conducting asynchronous fMRI and fNIRS recordings during motor execution and imagery tasks, adapted from validated experimental designs [10].

Experimental Design and Paradigm

• Participants: Recruit healthy adults with no neurological history. Secure informed consent and ethical approval.
• Task Design: Employ a block design. The paradigm should include:
  • Motor Action (MA) Blocks: Participants execute a bilateral finger-tapping sequence (e.g., 1-2-1-4-3-4, assigned to different fingers).
  • Motor Imagery (MI) Blocks: Participants imagine performing the same sequence without physical movement.
  • Baseline Blocks: Participants remain at rest, often fixating on a crosshair.
• Block Structure: Each block typically lasts 30 seconds, with condition cues presented for 2 seconds at the block start. The total run may comprise 17 blocks (e.g., 9 Baseline, 4 MA, 4 MI) for a total duration of 8.5 minutes [10].

The workflow for data acquisition and analysis in a multimodal study is summarized below.

Data Acquisition Parameters

Table 2: Acquisition Parameters for Simultaneous or Asynchronous fMRI-fNIRS Studies

Modality	Key Parameters	Recommended Setting
fMRI	Magnetic Field Strength	3 Tesla
	Sequence	Echo-Planar Imaging (EPI)
	Repetition Time (TR)	1500-2000 ms
	Echo Time (TE)	~30 ms
	Voxel Size	3 × 3 × 3.5 mm
	Slices	Cover motor and premotor cortex
fNIRS	System Type	Continuous-Wave (CW)
	Sources/Detectors	16 sources, 15 detectors (example)
	Wavelengths	760 nm, 850 nm
	Sampling Rate	> 5 Hz
	Optode Distance	30 mm (long), 8 mm (short-distance for extracerebral signal regression)

Data Preprocessing and Analysis

• fMRI Preprocessing: Conducted using standard software (e.g., SPM, FSL, BrainVoyager). Steps include slice-time correction, motion correction, temporal high-pass filtering, co-registration to structural images, and spatial normalization [10].
• fNIRS Preprocessing: Performed using tools like Homer3 or NIRS-SPM. Steps involve:
  • Channel Pruning: Remove channels with a low signal-to-noise ratio (SNR < 15 dB).
  • Conversion: Convert raw light intensity to optical density.
  • Hemodynamic Conversion: Use the Modified Beer-Lambert Law to calculate concentration changes in HbO and HbR.
  • Short-Channel Regression: Use signals from short-distance detectors to regress out systemic physiological noise [10].
• Modeling and Analysis:
  • General Linear Model (GLM): Use both fMRI and fNIRS data to identify activation clusters for contrasts like MA > Baseline and MI > Baseline. Define ROIs such as primary motor cortex (M1) and premotor cortex (PMC) [10].
  • Dynamic Causal Modelling (DCM): Apply DCM to fNIRS data to infer effective connectivity (directed influences) between motor regions at the neuronal level. For example, test how connectivity from the supplementary motor area to M1 is modulated by motor imagery [28].

The Scientist's Toolkit: Research Reagent Solutions

Table 3: Essential Materials and Analytical Tools for Multimodal fMRI-fNIRS Research

Item / Reagent	Function / Application	Example / Note
fNIRS System	Portable measurement of HbO and HbR concentration changes.	NIRSport2 (NIRx) or similar CW systems.
MRI Scanner	High-resolution structural and functional (BOLD) imaging.	3T MRI scanner with a head coil.
Stimulus Presentation Software	Precise delivery of task paradigms in the scanner and fNIRS settings.	E-Prime, PsychoPy, Presentation.
fMRI Analysis Suite	Preprocessing and statistical analysis of BOLD data.	SPM, FSL, AFNI, BrainVoyager.
fNIRS Analysis Package	Preprocessing, visualization, and statistical analysis of optical data.	Homer3, NIRS-SPM, FieldTrip.
Dynamic Causal Modelling (DCM)	Toolbox for model-based inference of effective connectivity from data.	Available within SPM software [28].
Motor Task Paradigm	Standardized protocol to elicit hemodynamic responses in motor networks.	Bilateral finger-tapping sequence [10].

The integration of functional Magnetic Resonance Imaging (fMRI) and functional Near-Infrared Spectroscopy (fNIRS) represents a paradigm shift in neuroimaging, particularly for motor task paradigms. While fMRI is renowned for its high spatial resolution and capacity for whole-brain coverage, fNIRS offers superior portability, temporal resolution, and resilience to motion artifacts [3] [29]. This combination is especially powerful in motor research, where it enables the study of brain activity from precise localization in controlled settings to dynamic movement in ecologically valid environments. This article delineates the inherent trade-offs between spatial resolution, portability, and depth penetration of these modalities and provides detailed application notes and experimental protocols for their integrated use in motor task research, framed within a broader thesis on multimodal neuroimaging.

Comparative Analysis of fMRI and fNIRS

Technical and Performance Trade-offs

The core trade-offs between fMRI and fNIRS stem from their fundamental physical principles. fMRI measures the Blood-Oxygen-Level-Dependent (BOLD) signal, which is influenced by the magnetic properties of deoxygenated hemoglobin [30]. fNIRS, in contrast, uses near-infrared light to measure changes in both oxygenated (HbO) and deoxygenated (HbR) hemoglobin concentrations based on their distinct absorption spectra [31] [32]. Table 1 summarizes the inherent compromises between these two technologies.

Table 1: Inherent Trade-offs Between fMRI and fNIRS

Feature	fMRI	fNIRS
Spatial Resolution	High (millimeter-level) [3]	Low (1-3 cm) [3] [33]
Temporal Resolution	Slow (0.5-2 Hz, limited by hemodynamics) [3]	Superior (up to 100 Hz, millisecond-level precision possible) [3] [32]
Portability	Very low (immobile scanner, restrictive environment) [3]	High (portable/wireless systems available) [3] [29]
Depth Penetration	Whole-brain (cortical and subcortical) [3]	Superficial cortical regions only (limited to 1-1.5 cm in adults) [3] [34] [29]
Tolerance to Motion	Low (highly sensitive to motion artifacts) [3]	High (robust against movement) [33] [29]
Participant Limitations	Contraindicated for individuals with metal implants, claustrophobia [29]	Few limitations; suitable for infants, children, and patients with implants [29]
Operational Environment	Dedicated, shielded room [35]	Almost any environment (bedside, laboratory, real-world) [3] [34]
Key Measured Signal	BOLD signal (primarily reflects HbR) [30]	Direct concentration changes of HbO and HbR [3] [32]

Quantitative Data for Motor Paradigms

Empirical studies directly comparing fMRI and fNIRS during motor tasks provide critical validation data. A 2022 study focusing on the Supplementary Motor Area (SMA) during motor execution and imagery provides key quantitative insights, as summarized in Table 2 [16].

Table 2: fMRI and fNIRS Performance in Motor Task Validation Study

Parameter	fMRI Findings	fNIRS Findings
SMA Activation (Motor Execution)	Robust activation detected with high spatial specificity.	Reliably detected SMA activation. Δ[HbR] showed higher spatial specificity [16].
SMA Activation (Motor Imagery)	Activation patterns observed.	SMA activation detected for both hand and whole-body motor imagery. Δ[HbR] was the more specific signal [16].
Temporal Correlation	BOLD signal time course served as the reference.	fNIRS Δ[HbO] and Δ[HbR] signals showed a time course matching the fMRI BOLD signal [16].
Key Outcome	Gold standard for localization.	Confirmed CW-fNIRS can reliably measure SMA activation for neurofeedback and BCI applications [16].

Application Notes for Motor Task Research

Synergistic Integration Strategies

The combination of fMRI and fNIRS is not merely sequential but synergistic, allowing researchers to bridge the spatial-temporal gap in neuroimaging.

• Synchronous Data Acquisition: Simultaneous collection of fMRI and fNIRS data capitalizes on their complementary strengths. fMRI provides the anatomical framework and high-resolution spatial map, validating fNIRS channel placement and confirming that fNIRS signals originate from targeted regions like the SMA or primary motor cortex (M1) [3] [16]. fNIRS can then provide a cleaner, higher-temporal-resolution account of the hemodynamic response, useful for tracking rapid changes in brain state or connectivity during task performance [3].
• Asynchronous Data Acquisition: In this paradigm, an initial fMRI session is used for precise task localization and individual anatomical mapping. This information then guides the optode placement and channel selection for subsequent fNIRS sessions, which can be conducted in more naturalistic or clinical settings, such as during rehabilitation exercises or at the bedside [3] [16]. This approach is highly efficient for longitudinal studies where frequent fMRI scanning is impractical.
• Hyperscanning for Interactive Motor Paradigms: The portability of fNIRS enables "hyperscanning," where multiple brains are measured simultaneously during social or collaborative motor tasks. This opens new avenues for researching brain-to-brain coupling in activities requiring coordinated movement, which is impossible with traditional fMRI [3].

Protocol 1: Validation of fNIRS for Supplementary Motor Area (SMA) Activation

This protocol is designed to validate fNIRS measurements of the SMA against the gold standard of fMRI, a critical step before deploying fNIRS in standalone applications [16].

1. Objective: To establish the spatial specificity and task sensitivity of continuous-wave (CW) fNIRS for detecting SMA activation during motor execution (ME) and motor imagery (MI).

2. Experimental Design:

• Participants: 16 healthy older adults (e.g., mean age 64 ± 5 years) to enhance relevance for neurorehabilitation populations [16].
• Tasks: A block design is recommended for optimal detection power.
  • Motor Execution (ME): Repetitive squeezing of a ball with the left or right hand.
  • Motor Imagery (MI): Kinesthetic imagination of the same hand movements without physical motion.
  • Control Condition: Rest.
• Data Acquisition:
  • fMRI Session: Conducted on a 3T scanner. Acquire high-resolution T1-weighted anatomical images. For functional scans, use a T2*-weighted EPI sequence. Monitor for head motion and use EMG to ensure absence of muscle activation during MI blocks [16].
  • fNIRS Session: Use a continuous-wave fNIRS system. Place optodes over the SMA based on the international 10-20 system (e.g., between Fz and Cz). A source-detector separation of 3 cm is standard for adults. Record both HbO and HbR concentrations at a sampling rate ≥ 10 Hz [16].

3. Data Analysis:

• fMRI Processing: Preprocessing (motion correction, slice timing, spatial smoothing). Statistical analysis using a General Linear Model (GLM) to generate activation maps for ME and MI. Define the SMA region of interest (ROI) individually for each participant [16] [30].
• fNIRS Processing: Convert raw light intensity to HbO and HbR concentrations using the Modified Beer-Lambert Law. Apply band-pass filtering to remove physiological noise. Use a GLM with a canonical hemodynamic response function to assess task-related activation [16].
• Validation Analysis:
  • Spatial Specificity: Correlate the topographical activation maps from fNIRS (for both HbO and HbR) with the fMRI activation map from the cortical region corresponding to the fNIRS channels.
  • Task Sensitivity: Compare the ability of both modalities to detect the expected stronger activation for ME versus MI and to show lateralized activation for hand tasks in M1.

Protocol 2: Integrated fMRI-fNIRS for a Motor Rehabilitation Paradigm

This protocol leverages the strengths of both modalities to track cortical reorganization in patients recovering from stroke.

1. Objective: To use fMRI for baseline mapping of the motor network and fNIRS for longitudinal, bedside monitoring of therapy-induced neuroplastic changes.

2. Experimental Design:

• Participants: Patients with motor impairments following stroke.
• Tasks:
  • Simple hand motor tasks (execution and imagery).
  • Rehabilitation-specific tasks (e.g., reaching, grasping) that can be performed in a MRI simulator and later at the bedside.
• Data Acquisition:
  • Baseline (fMRI + fNIRS): Conduct a synchronous session. Use fMRI to map the entire motor network, including primary motor cortex, premotor areas, and SMA. Simultaneous fNIRS data is used to establish a cross-modal correlation.
  • Longitudinal Monitoring (fNIRS only): Conduct weekly fNIRS sessions at the bedside during physical or mental practice therapy. The fNIRS optode placement is guided by the baseline fMRI map to ensure coverage of key motor areas.

3. Data Analysis:

• Use the baseline fMRI to create subject-specific models for fNIRS source reconstruction, improving the spatial accuracy of fNIRS [3].
• Analyze longitudinal fNIRS data for changes in HbO/HbR activation strength, lateralization index (shift of activity from contralesional to ipsilesional hemisphere), and functional connectivity between motor areas.
• Correlate fNIRS-derived biomarkers with clinical measures of motor recovery.

The Scientist's Toolkit: Research Reagent Solutions

Table 3: Essential Materials for Integrated fMRI-fNIRS Motor Research

Item	Function/Application
High-Density fNIRS System (>32 channels)	Provides greater cortical coverage and improved spatial resolution for mapping motor areas like SMA and M1 [32].
MRI-Compatible fNIRS Optodes and Cabling	Essential for conducting simultaneous fMRI-fNIRS recordings without causing artifacts or safety hazards [3].
3D Digitizer	Precisely records the 3D locations of fNIRS optodes on the subject's head relative to anatomical landmarks (e.g., nasion, inion). This allows for co-registration with the subject's anatomical MRI scan [29].
Electromyography (EMG) System	Critical for monitoring muscle activity during motor imagery tasks to ensure the absence of overt movement, which is a common confound [16].
AtlasViewer or fOLD Software	Brain mapping tools used for optode placement planning and for projecting fNIRS data onto anatomical images, addressing fNIRS's lack of inherent anatomical information [16] [29].
Riemannian Geometry Classifier	An advanced machine learning tool for fNIRS brain-state classification that leverages spatial co-activation patterns of HbO and HbR, significantly improving classification accuracy for motor imagery tasks [33].

Visualizing the Integrated Workflow

The following diagram illustrates the logical workflow and synergistic relationship between fMRI and fNIRS in a motor task research paradigm.

Integrated fMRI-fNIRS Workflow for Motor Research

The integration of fMRI and fNIRS effectively navigates the inherent trade-offs in neuroimaging, creating a powerful framework for motor task research. While fMRI provides the essential structural and high-resolution functional blueprint, fNIRS offers a flexible and practical tool for longitudinal monitoring and studying brain function in real-world contexts. The protocols and application notes detailed herein provide a roadmap for researchers to validate fNIRS signals and implement this multimodal approach. Future advancements in low-field, portable MRI [35], sophisticated machine learning algorithms for fNIRS [33], and standardized integration protocols [3] [32] will further solidify this synergy, ultimately accelerating discovery in basic neuroscience and improving patient outcomes in clinical neuromotor rehabilitation.

From Theory to Practice: Methodological Frameworks and Translational Applications

The study of motor execution (ME) and motor imagery (MI) is a cornerstone of cognitive neuroscience, with profound implications for brain-computer interfaces (BCIs) and neurorehabilitation. Functional magnetic resonance imaging (fMRI) and functional near-infrared spectroscopy (fNIRS) are two pivotal neuroimaging techniques that enable non-invasive investigation of the hemodynamic responses underlying these motor processes. The choice of experimental design—blocked or event-related—fundamentally shapes the quality, interpretability, and practical application of the acquired data. This article details the application notes and protocols for these paradigms, framed within a broader research initiative on the integration of fMRI and fNIRS for motor task research. The complementary nature of these modalities is clear: fMRI provides high spatial resolution for localizing deep and superficial brain activity, while fNIRS offers superior temporal resolution and portability for more naturalistic or longitudinal studies [3]. This integration is particularly valuable for translating laboratory findings into clinical rehabilitation settings.

Core Paradigm Designs: Blocked and Event-Related

Blocked Designs

In a blocked design, stimuli or tasks of the same condition are grouped together in extended periods (blocks), which are alternated with blocks of a control condition or rest.

• Principles and Advantages: This design maximizes the amplitude of the hemodynamic response by allowing the blood-oxygen-level-dependent (BOLD) or hemoglobin concentration signals to accumulate over a sustained period. It offers superior statistical power for detecting subtle activation differences and is simpler to implement, as it does not require careful randomization of trials [36]. The predictable signal pattern also facilitates the visual identification of artifacts.
• Typical Parameters: Block durations commonly range from 20 to 30 seconds [37]. For example, a typical protocol may consist of 4-5 active blocks alternating with rest blocks of similar or slightly longer duration [37].
• Protocol: Standard Blocked Design for Motor Execution/Imagery:
  • Participant Preparation: Instruct the participant on the task (e.g., finger tapping, kinesthetic motor imagery) and the meaning of visual cues.
  • Baseline Acquisition: Begin with a 30-second rest period to establish a stable hemodynamic baseline.
  • Task Block: Present a visual or auditory cue for 20-30 seconds during which the participant continuously performs the ME or MI task.
  • Rest Block: The cue disappears, and the participant rests for 20-30 seconds, refraining from movement or focused mental imagery.
  • Repetition: Repeat steps 3 and 4 for 4-8 cycles per run to ensure robust signal averaging.
  • Control Condition: For stronger inference, alternate task blocks with a different motor task or a cognitively engaging control task (e.g., size judgment of strings of "O's") instead of simple rest [36].

Event-Related Designs

In an event-related design, discrete trials of different conditions are presented in a randomized order, with varying inter-trial intervals.

• Principles and Advantages: This design avoids the stimulus-order predictability inherent in block designs, thereby reducing potential confounds like participant anticipation or task-switching effects [36]. Its chief advantage is the ability to sort trials post-hoc based on behavioral outcomes (e.g., response accuracy, latency) or to model the hemodynamic response function for each trial type separately [36].
• Typical Parameters: Individual trials are typically short (~2 seconds). The critical parameter is the inter-trial interval (ITI), which is jittered (e.g., 4-10 seconds) to allow the hemodynamic response to return to baseline and to improve the estimation efficiency of the response model [36]. The mean ITI is often around 18 seconds for each task type [36].

Hybrid Designs

To harness the strengths of both approaches, hybrid blocked fast-event-related designs have been introduced, particularly for MVPA-based BCI applications.

• Principles and Advantages: This design combines rest periods similar to block designs with shorter, randomly alternating trials from a rapid event-related design [38]. It maintains high decoding accuracy comparable to block designs while allowing for random trial presentation and providing clear intervals for BCI feedback processing.
• Protocol: Hybrid Design for BCI Classification:
  • Structure: Organize the experiment into blocks, but within each block, present multiple rapid, randomized trials of different conditions (e.g., imagery of different finger movements).
  • Trial Presentation: Present each trial for 1-2 seconds with a short, jittered ISI within the block.
  • Inter-Block Rest: Separate each block of trials with a prolonged rest period (e.g., 15-20 seconds) to allow the participant to reset and to process potential neurofeedback [38].

Table 1: Quantitative Comparison of fMRI Design Performance in Motor Tasks [36] [38]

Design Feature	Block Design	Slow Event-Related Design	Hybrid Blocked Fast-Event-Related
Relative Statistical Power	Highest	Lower	High (Close to block design)
Stimulus Order Predictability	High	Low	Low
Post-hoc Trial Sorting	Not possible	Possible	Possible
Suitability for BCI Feedback	Moderate	Low	High
Decoding Accuracy (Example)	Highest	Worst Performance	Similar to Block Design
Incremental Decoding Stability	Lower	Lower	Most Stable

The following workflow diagram illustrates the decision process for selecting an appropriate experimental design based on research goals.

The Scientist's Toolkit: Essential Research Reagents and Materials

Successful execution of motor paradigm studies requires specific hardware, software, and methodological components. The following table details the essential "research reagent solutions" for this field.

Table 2: Key Research Reagents and Materials for Motor Paradigm Studies [3] [39] [37]

Item Category	Specific Examples / Properties	Primary Function in Research
fMRI Scanner	3T, 1.5T systems; Gradient-echo EPI sequence	High-spatial-resolution whole-brain imaging; detects BOLD signal changes associated with neural activity.
fNIRS System	Continuous-wave (CW) systems; 650-950 nm wavelengths	Portable, tolerant cortical monitoring of HbO/HbR concentration changes during movement or in naturalistic settings.
Optode Configurations	3 cm source-detector distance (adults); 10-100+ channels	Measures cortical hemodynamics; configuration impacts spatial coverage and resolution.
Task Presentation Software	Visual cueing systems (e.g., PsychToolbox, E-Prime)	Presents standardized visual/auditory stimuli and records participant responses or performance.
Physiological Monitors	Heart rate, breathing rate, skin conductance, blood pressure	Monitors autonomic nervous system activity; can be used as covariates or for hybrid BCI classification.
Data Analysis Suites	BrainVoyager, SPM, NIRS-SPM, HomER2	Pre-processing, statistical analysis (GLM), and visualization of fMRI and fNIRS data.
Digital Filters	High-pass (e.g., 0.01 Hz), Low-pass (e.g., 0.2 Hz), Band-pass FIR filters	Removes physiological noise (heart rate, respiration) and low-frequency signal drift from fNIRS data.

Integration of fMRI and fNIRS in Motor Research

The combination of fMRI and fNIRS is a powerful multimodal approach that leverages their complementary strengths.

• Spatiotemporal Synergy: Integrating fMRI's high spatial resolution with fNIRS's superior temporal resolution enables robust mapping of neural activity across both cortical and subcortical structures with fine temporal detail [3]. This is crucial for capturing the rapid dynamics of motor planning and execution.
• Validation and Application: Synchronous fMRI-fNIRS measurements are often used to validate fNIRS signals against the well-established fMRI BOLD response [3]. This integrated approach has advanced research in neurological disorders (e.g., stroke, Alzheimer's), social cognition, and neuroplasticity.
• Protocol: Synchronous fMRI-fNIRS Data Acquisition:
  • Hardware Setup: Use MRI-compatible fNIRS probes and equipment to prevent electromagnetic interference and ensure patient safety [3].
  • Probe Placement: Prior to the scan, place fNIRS optodes on the scalp over the motor cortex regions of interest (e.g., primary motor cortex, premotor cortex). Use a 3D digitizer to record precise optode locations relative to cranial reference points (e.g., nasion, inion) for co-registration with anatomical MRI data [40].
  • Synchronous Data Acquisition: Start both fMRI and fNIRS recordings simultaneously. Use a shared trigger pulse from the stimulus presentation computer to synchronize the timing of task events with data acquisition from both modalities.
  • Data Fusion: During analysis, align fNIRS channels to their corresponding cortical locations using the individual's structural MRI. The fMRI data can provide a spatial prior for source localization in fNIRS analysis, improving the interpretation of the optical signals [3].

The following diagram outlines the workflow for a synchronized multimodal experiment.

Detailed Experimental Protocols

Protocol: fNIRS for Motor Execution and Imagery

This protocol is adapted from studies investigating the differential cortical activation between actual and imagined movement [40].

• Participant Preparation: Recruit right-handed participants with no neurological history. Obtain informed consent. Place the fNIRS cap or probe set over the motor cortex. A configuration with 10 sources and 10 detectors (forming 31 channels) covering the primary motor, premotor, and supplementary motor areas is typical [40].
• Experimental Design: Use a block design. Each session should contain four conditions: ME with the right hand, ME with the left hand, MI with the right hand, and MI with the left hand. The order of sessions should be counter-balanced across participants.
• Task Structure: Each condition should consist of 20 trials. A single trial comprises:
  • Rest Period: 15 seconds. A fixation cross is displayed, and the participant is instructed to relax.
  • Task Period: 10 seconds. A visual cue (e.g., the word "SQUEEZE" or "IMAGINE") is displayed. During ME, the participant performs an isometric finger tapping task at ~1 Hz. During MI, the participant engages in kinesthetic motor imagery of the same movement without any physical motion [40].
• Data Acquisition: Collect data at a sampling rate of ≥ 10 Hz using a continuous-wave fNIRS system with multiple wavelengths (e.g., 780, 805, 830 nm).
• Data Pre-processing:
  • Convert raw light intensity into optical density and then into concentration changes of HbO and HbR using the Modified Beer-Lambert Law.
  • Apply a band-pass filter (e.g., 0.01 - 0.2 Hz Finite Impulse Response filter) to remove physiological noise (heart rate, respiration) and slow drifts [41].
  • Identify and correct for motion artifacts using validated algorithms (e.g., spline interpolation, wavelet-based methods) [41].
• Statistical Analysis: Use a General Linear Model (GLM) framework for statistical inference. Contrast the task periods against the rest periods to generate activation maps for HbO and HbR for each condition.

Protocol: fMRI Neurofeedback for Motor Recovery Post-Stroke

This protocol outlines a graded neurofeedback training regimen for stroke patients, targeting the supplementary motor area (SMA) [42].

• Participant Preparation: Recruit patients with first-time middle cerebral artery stroke and residual upper limb impairment. Screen for MRI contraindications. Obtain informed consent.
• Localizer Scan: Conduct an initial fMRI scan where the patient performs kinesthetic motor imagery of the affected limb. Use this data to localize the SMA and define it as the region of interest (ROI) for neurofeedback.
• Neurofeedback Training Structure: Employ a block design. In each training run, patients perform multiple trials of motor imagery.
  • Task: Patients engage in kinesthetic motor imagery of moving their affected limb.
  • Feedback: The real-time BOLD signal from the SMA ROI is processed and displayed to the patient as a visual gauge (e.g., a thermometer) that moves between two discrete target levels (low and high).
  • Instruction: Patients are instructed to use their mental imagery to raise the feedback signal to the designated target level for each trial [42].
• Data Analysis: Monitor the SMA-ROI activation over time to assess the patient's ability to voluntarily regulate this brain region. The success of self-regulation is a key feasibility metric for this therapeutic approach.

Blocked and event-related designs each offer distinct advantages for probing the neural correlates of motor execution and imagery. The choice of paradigm must be guided by the specific research question, whether it demands the high statistical power of a block design or the trial-by-trial analytical flexibility of an event-related design. The emerging trend of hybrid designs and the synergistic integration of fMRI and fNIRS are pushing the boundaries of what is possible, both in fundamental neuroscience and in translational applications like BCIs and stroke rehabilitation. By providing standardized, detailed protocols and highlighting essential methodological tools, this article aims to facilitate rigorous and reproducible research in this dynamic field.

The integration of functional magnetic resonance imaging (fMRI) and functional near-infrared spectroscopy (fNIRS) presents a powerful multimodal approach for brain research, particularly within motor task paradigms. This integration capitalizes on their complementary strengths: fMRI provides high spatial resolution and whole-brain coverage, including subcortical structures, while fNIRS offers superior temporal resolution, portability, and greater resilience to motion artifacts [14]. This application note details standardized protocols for simultaneous and asynchronous data acquisition, tailored for research in motor neuroscience. The guidance provided herein is framed within the context of a broader thesis on leveraging multimodal neuroimaging to achieve a more comprehensive characterization of the neural underpinnings of motor execution and imagery.

Table 1: Fundamental Characteristics of fNIRS and fMRI

Feature	Functional Near-Infrared Spectroscopy (fNIRS)	Functional Magnetic Resonance Imaging (fMRI)
Spatial Resolution	1-3 cm; limited to cortical surfaces [14]	Millimeter-level; whole-brain including subcortical structures [14]
Temporal Resolution	High (up to millisecond precision) [14]	Low (limited by hemodynamic response; typically 0.33-2 Hz) [14]
Measured Parameters	Concentration changes in oxygenated (HbO) and deoxygenated hemoglobin (HbR) [25]	Blood Oxygen Level-Dependent (BOLD) signal, sensitive to deoxyhemoglobin [14]
Key Advantages	Portable, tolerant of movement, suitable for naturalistic settings and bedside monitoring [14]	Gold standard for spatial localization, comprehensive brain coverage [14] [25]
Primary Limitations	Superficial penetration, lower spatial resolution, sensitive to extracerebral hemodynamics [14] [43]	Expensive, immobile, sensitive to motion artifacts, noisy and restrictive environment [14]

Synchronous Data Acquisition Protocol

Simultaneous fNIRS-fMRI recording allows for the direct temporal correlation of signals from both modalities, enabling the validation of fNIRS signals against the fMRI gold standard and providing a rich dataset for advanced multimodal fusion [14] [44].

Equipment and Setup

• fMRI Scanner: A 3T MRI scanner is standard. A 32-channel head coil is recommended for improved signal-to-noise ratio [43].
• fNIRS System: A continuous-wave fNIRS system (e.g., NIRSport2, NIRScout) with dual wavelengths (e.g., 760 and 850 nm) is used. The system must be MR-compatible to function safely and effectively within the high-field environment [10] [43].
• Optode Probe Design: A cap-based probe set should be used to ensure stable optode placement. For motor paradigms, optodes must cover the primary motor (M1) and premotor cortices (PMC) bilaterally [10]. A typical setup may include 16 sources and 15 detectors arranged with an inter-optode distance of 30 mm, generating ~50 channels over the motor network [10].
• Critical Synchronization: The fNIRS and fMRI systems must be synchronized at the start of data acquisition using a shared TTL pulse or trigger from the MRI scanner to align the data streams temporally [44].

Procedural Workflow

The experimental workflow for a simultaneous recording session, from preparation to data acquisition, is outlined below.

Key Technical Considerations

• Hardware Integration: The fNIRS system must be placed outside the MRI scanner room, with fiber optic cables passing through a waveguide in the RF shield. All fNIRS components inside the scanner room (optodes, cables, cap) must be non-magnetic and non-conductive to ensure safety and prevent artifacts [14].
• Artifact Mitigation: The fNIRS signal is susceptible to MR-induced artifacts, primarily from gradient switching and radiofrequency pulses. These can be mitigated by recording the MR sequence timing (gradient triggers) for post-hoc correction and using robust motion artifact correction algorithms (e.g., spline interpolation with wavelet decomposition) during fNIRS preprocessing [43].
• Co-registration: Precise spatial mapping of fNIRS channels to brain anatomy is critical. This is achieved by digitizing the 3D positions of fNIRS optodes and key anatomical landmarks (e.g., Nz, Iz, Cz, A1, A2) relative to the participant's MRI-compatible head coil using an MR-tracked sensor [43]. These digitized positions are then co-registered to the participant's structural T1-weighted MRI scan.

Asynchronous Data Acquisition Protocol

Asynchronous acquisition involves collecting fNIRS and fMRI data in separate sessions, often to translate a paradigm from the fMRI setting to more naturalistic fNIRS environments or to leverage the respective strengths of each modality for different parts of a study [10].

Core Strategy and Paradigm Matching

• Paradigm Replication: The experimental task (e.g., motor execution/imagery), block/event timing, and instructions must be replicated as precisely as possible across the fMRI and fNIRS sessions [10].
• Probe Placement Fidelity: The fNIRS probe layout should be designed to cover the brain regions identified as active in the group-level or individual fMRI analysis. Using individual anatomical scans to guide optode placement maximizes spatial correspondence [10].

Data Integration and Analysis

In asynchronous studies, integration occurs at the analysis level. Subject-specific fNIRS signals (e.g., HbO and HbR time series from a motor cortex channel) can be used as regressors of interest in a General Linear Model (GLM) applied to the fMRI data. This tests the ability of the fNIRS-derived cortical signal to predict activation in the spatially detailed fMRI data, validating the spatial localization of the fNIRS measurement [10].

Performance and Validation Metrics

Table 2: Quantitative Comparison from Simultaneous Recordings

Metric	Findings	Experimental Context
Temporal Correlation	HbO often shows higher correlation with BOLD; wide variance (r = 0.2 to 0.8) reported [10]. HbR is theoretically linked to BOLD via balloon model [25].	Motor, visual, and cognitive tasks during simultaneous fNIRS-fMRI [25] [10].
Spatial Correspondence	fNIRS-based signals can model fMRI activation in primary and premotor cortices. No statistically significant difference between HbO and HbR in spatial correspondence with BOLD [10].	Asynchronous fMRI modeled by fNIRS data during motor imagery and execution [10].
Brain Fingerprinting	fNIRS classification accuracy: 75% to 98%. fMRI accuracy: ~99.9%. Accuracy depends on number of runs and spatial coverage [43].	Subject identification based on resting-state functional connectivity patterns [43].
Reproducibility	HbO is significantly more reproducible across sessions than HbR. Source localization improves reliability [20].	Test-retest fNIRS across multiple sessions for motor and visual tasks [20].

The Scientist's Toolkit

Table 3: Essential Research Reagents and Materials

Item	Function in Protocol	Specification Examples
MR-Compatible fNIRS System	Measures hemodynamic activity safely inside the MRI scanner environment.	Continuous-wave systems (e.g., NIRScout, NIRSport2); must have MRI certification [43] [10].
fNIRS Optode Cap	Holds sources and detectors in stable, pre-defined positions on the scalp.	Dense caps based on the 10-20 system; materials must be non-metallic [43] [45].
Digitization System	Records 3D positions of fNIRS optodes for co-registration with anatomical MRI.	MR-tracked sensor (e.g., Polhemus Fastrak) [43].
Short-Distance Detectors	Measures and enables removal of signals originating from superficial tissues (scalp, skull).	Detectors placed 8 mm from a source [10].
Synchronization Hardware	Generates a shared timing pulse to align fNIRS and fMRI data streams.	TTL pulse generator or cable from MRI scanner to fNIRS system [44].

The strategic integration of fNIRS and fMRI, through either simultaneous or asynchronous protocols, offers a robust framework for advancing motor task research. Simultaneous acquisition provides the highest level of temporal correspondence for signal validation and complex model testing, while asynchronous acquisition allows for greater flexibility and ecological validity. Adherence to the detailed protocols for hardware setup, synchronization, artifact correction, and spatial co-registration outlined in this document is critical for generating high-quality, reliable multimodal data. This approach effectively bridges the spatial-temporal resolution gap, paving the way for more nuanced investigations into brain function in both controlled and naturalistic settings.

The integration of functional magnetic resonance imaging (fMRI) and functional near-infrared spectroscopy (fNIRS) presents a powerful multimodal approach for brain research, particularly in motor task paradigms. By combining fMRI's high spatial resolution with fNIRS's superior temporal resolution and portability, researchers can achieve comprehensive spatiotemporal mapping of neural activity [3] [14]. However, the effectiveness of this integrated approach heavily depends on robust preprocessing pipelines that address the unique characteristics and artifacts inherent in each modality. This application note details best practices for preprocessing steps, with a specific focus on motion correction, filtering, and signal enhancement techniques tailored for fMRI-fNIRS studies of motor function.

The complementary nature of fMRI and fNIRS is particularly advantageous for studying complex motor functions. fMRI provides whole-brain coverage, including subcortical structures, with millimeter-level spatial precision, making it indispensable for localizing activity in deep brain regions such as the basal ganglia and thalamus [3]. Conversely, fNIRS offers greater flexibility for studying naturalistic behaviors. Its tolerance to motion artifacts and portability enables brain imaging during active motor tasks, rehabilitation exercises, and in real-world environments [14] [46]. Both modalities measure hemodynamic responses related to neural activity, enabling direct comparison of signals, though they differ fundamentally in their physical principles and specific artifact profiles.

Developing standardized preprocessing protocols is essential for ensuring data quality, reproducibility, and valid cross-modal comparisons. A recent multi-lab collaboration, the fNIRS Reproducibility Study Hub (FRESH), highlighted that while analytical flexibility is valuable, variability in preprocessing choices—particularly in handling poor-quality data—significantly impacts research outcomes [47]. The following sections provide detailed methodologies for preprocessing both fMRI and fNIRS data, with specialized considerations for their integrated use in motor research.

fMRI Preprocessing with fMRIPrep

fMRIPrep is a robust preprocessing pipeline for fMRI data that requires minimal user input while providing comprehensive error and output reporting [48]. This tool is particularly valuable for standardizing the often complex and variable preprocessing steps across studies. Built on a "glass box" philosophy, fMRIPrep provides visual reports for each subject, enabling researchers to understand the process and assess the accuracy of critical processing steps [48].

The pipeline performs minimal preprocessing, defined as motion correction, field unwarping, normalization, bias field correction, and brain extraction [48]. It utilizes a combination of tools from well-known software packages including FSL, ANTs, FreeSurfer, and AFNI, selecting what the developers consider the best software implementation for each preprocessing stage.

Structured Protocol for fMRI Preprocessing

Table 1: Key Steps in fMRI Preprocessing Using fMRIPrep

Processing Stage	Description	Tools Used	Key Parameters
Head Motion Correction	Estimates and corrects for head movement using rigid-body transformation	FSL (MCFLIRT)	6 degrees of freedom, normalized correlation
Slice Timing Correction	Corrects for acquisition time differences between slices	AFNI (3dTshift)	Fourier interpolation, slice order specification
Susceptibility Distortion Correction	Corrects for field inhomogeneities using fieldmaps	FSL (TOPUP)	Phase encoding direction, echo spacing
Anatomical Coregistration	Aligns functional and structural images	FSL (BBR), FreeSurfer	Boundary-based registration, 6 DOF
Spatial Normalization	Warps images to standard space (MNI)	ANTs	SyN transformation, CC optimization
Brain Extraction	Removes non-brain tissue	ANTs (atropos)	N4 bias field correction, Otsu thresholding
Spatial Smoothing	Increases signal-to-noise ratio (optional)	FSL (sus)	Gaussian kernel, FWHM 5-8mm

Table 2: fMRI Quality Control Metrics in Preprocessing

Quality Metric	Acceptance Threshold	Assessment Method
Head Motion	< 2mm translation, < 2° rotation	Framewise displacement
Signal-to-Noise Ratio	> 100	Mean signal intensity / standard deviation
Temporal Signal-to-Noise	> 50 for cortex	Mean over time / standard deviation over time
Coregistration Accuracy	< 3mm error	Visual inspection, boundary alignment
Normalization Accuracy	> 0.8 correlation	Cross-correlation with template

For motor task paradigms, special attention should be paid to motion correction, as motor execution can induce significant head movement. fMRIPrep's robust motion correction using FSL's MCFLIRT algorithm is particularly valuable here [48]. The pipeline also handles susceptibility distortion correction, which is crucial for accurate localization of activity in motor regions near tissue boundaries.

The visual reports generated by fMRIPrep are essential for quality control, allowing researchers to identify outliers and make informed decisions about data inclusion [48]. These reports include sections on anatomical processing, functional preprocessing, and the coregistration between functional and anatomical spaces.

fNIRS Preprocessing Methodologies

Core Preprocessing Steps

fNIRS preprocessing faces unique challenges, particularly regarding motion artifacts and physiological confounds. Unlike fMRI, fNIRS signals are contaminated by various disturbance factors including heartbeats, breathing, shivering, and instrumental noises [49]. The FRESH initiative revealed substantial variability in how researchers handle these challenges, with pruning choices, hemodynamic response function models, and statistical analysis space being key sources of variability across research teams [47].

Table 3: fNIRS Preprocessing Pipeline Components

Processing Step	Purpose	Common Methods	Motor Task Considerations
Channel Pruning	Remove poor-quality channels	Signal-to-Noise Ratio (SNR < 15 dB)	Preserve motor cortex coverage
Motion Artifact Correction	Reduce movement-induced noise	PCA, wavelet, spline interpolation	Critical for active movement tasks
Physiological Filtering	Remove cardiac/respiratory signals	Bandpass filtering (0.01-0.5 Hz)	Heart rate may elevate with movement
Hemodynamic Conversion	Convert light intensity to HbO/HbR	Modified Beer-Lambert Law	DPF may vary by brain region
Temporal Preprocessing	Enhance signal quality	Detrending, Gaussian smoothing	Align with task timing

Advanced Signal Enhancement Techniques

For motor paradigms, where motion artifacts are prevalent, advanced processing methods are often necessary. A maximum likelihood generalized extended stochastic gradient (ML-GESG) estimation method has been proposed as an alternative filtering approach designed to reduce multiple disturbances originating from heartbeats, breathing, shivering, and instrumental noises as multivariate parameters [49]. This method has demonstrated superior performance compared to conventional filtering when applied to auditory-motor integration tasks.

In pain assessment studies involving motor responses, novel feature extraction methods such as Empirically Transformed Energy Patterns (ETEPs) have been developed to capture fNIRS signal dynamics more effectively [50]. These patterns retain short-term fluctuations and sustained hemodynamic changes, improving the ability to identify task-related neural dynamics with greater precision.

For motor imagery and execution tasks, research indicates that both oxygenated (HbO) and deoxygenated hemoglobin (HbR) provide valuable information. A multimodal assessment study found no statistically significant differences in spatial correspondence with fMRI between HbO, HbR, and total hemoglobin (HbT) for motor tasks [10]. This suggests flexibility in chromophore selection for motor paradigms.

Integrated fMRI-fNIRS Preprocessing

Spatial Coregistration and Alignment

For effective integration of fMRI and fNIRS data, precise spatial coregistration is essential. This process involves mapping fNIRS channels to corresponding cortical locations and aligning them with fMRI activation maps. The procedure typically involves:

• 3D Digitization: Recording the positions of fNIRS optodes relative to cranial landmarks (nasion, inion, preauricular points) using a 3D digitizer.

• MNI Coordinate Transformation: Using the structural T1-weighted image from fMRI to transform optode locations to standard Montreal Neurological Institute (MNI) space.

• Channel Projection: Projecting fNIRS channels onto the cortical surface using photon migration models.

Research has demonstrated the ability to identify motor-related activation clusters in fMRI data using subject-specific fNIRS-based cortical signals as predictors of interest, with significant peak activation found overlapping the individually-defined primary and premotor motor cortices [10].

Temporal Integration Approaches

Temporal integration of fMRI and fNIRS data presents challenges due to their different sampling rates (typically 0.3-2 Hz for fMRI vs. 5-100 Hz for fNIRS) and hemodynamic response characteristics. Two primary approaches exist:

• Synchronous Acquisition: Simultaneous data collection requires careful hardware synchronization and addressing potential electromagnetic interference between systems [3].

• Asynchronous Acquisition: Separate data collection sessions, as used in motor imagery studies, require careful paradigm matching and normalization of hemodynamic responses [10].

For asynchronous designs, studies have successfully modeled fMRI data using corresponding fNIRS measurements as predictors, demonstrating significant spatial correspondence in motor-network regions [10].

Experimental Protocols for Motor Task Paradigms

Protocol for Multimodal Motor Execution and Imagery

This protocol is adapted from established methods for investigating motor function with integrated fMRI-fNIRS [10]:

Participant Preparation:

• Screen for MRI contraindications and obtain informed consent
• Measure head circumference and mark fiducial points (nasion, inion, preauricular)
• Position fNIRS cap, ensuring optode contact quality while allowing for fMRI head coil placement

Data Acquisition Parameters:

• fMRI: 3T scanner, gradient-echo EPI sequence, TR=1500ms, TE=30ms, 3×3mm in-plane resolution, 26 slices covering motor areas
• fNIRS: NIRSport2 system, 16 sources (760/850nm), 15 detectors, 5.08Hz sampling, 30mm optode distance, 54 channels covering bilateral motor areas

Task Paradigm:

• Block design with 17 blocks (8min 30sec total)
• Conditions: Motor Action (MA), Motor Imagery (MI), and Baseline
• Block duration: 30sec each (9 Baseline, 4 MA, 4 MI blocks)
• During MA: Execute bilateral finger tapping sequence (1-2-1-4-3-4) at 2Hz
• During MI: Imagine the same sequence without movement

Data Processing:

• Preprocess fMRI and fNIRS data separately using pipelines described above
• Coregister fNIRS channels to fMRI space using digitized positions
• Extract fNIRS signals from motor regions (M1, PMC) for fMRI modeling

Protocol for Naturalistic Motor Task Validation

This protocol validates fNIRS against fMRI for naturalistic motor tasks, adapted from dance video game paradigms [51]:

Setup Modification:

• Use modified Dance Dance Revolution (DDR) game with StepMania open-source software
• Implement block design (30sec play/30sec rest) within game configuration
• For fMRI: Use foot buttons with limited movement (left/right taps only)
• For fNIRS: Allow full-body movement in naturalistic setting

Data Collection:

• Collect fMRI data in scanner with restricted movement
• Collect fNIRS data in open environment with full movement
• Maintain identical task structure and timing between modalities

Analysis:

• Define ROI in temporal and motor areas based on fMRI data
• Compare fNIRS activity patterns in same ROIs during naturalistic performance
• Correlate activation strength and timing between modalities

Visualization of Preprocessing Workflows

Diagram 1: Comprehensive preprocessing workflow for integrated fMRI-fNIRS studies, showing parallel processing streams that converge during multimodal integration. ML-GESG = Maximum Likelihood Generalized Extended Stochastic Gradient; mBLL = modified Beer-Lambert Law; HRF = Hemodynamic Response Function.

Diagram 2: Motor task paradigm design for multimodal fMRI-fNIRS studies, showing the block structure and modality-specific implementation considerations.

The Scientist's Toolkit

Table 4: Essential Research Reagents and Materials for fMRI-fNIRS Motor Studies

Item	Specification	Function/Purpose
fMRIPrep Software	Version 21.0.0 or later	Automated, robust fMRI preprocessing pipeline
Homer3 Software	MATLAB-based	fNIRS data processing and visualization
NIRSport2 System	NIRx Medical Technologies	Portable fNIRS acquisition with 16 sources, 15 detectors
3D Digitizer	Polhemus Patriot or similar	Precise optode localization for spatial coregistration
MRI-Compatible Response Devices	fMRI-compatible button boxes, foot pedals	Motor response collection in scanner environment
Short-Distance Detectors	8mm source-detector separation	Extracerebral signal regression in fNIRS
StepMania Software	Open-source DDR clone	Customizable motor task paradigm implementation
BrainVoyager QX	Commercial fMRI analysis software	Additional fMRI preprocessing and ROI analysis

Effective preprocessing of fMRI and fNIRS data is fundamental to successful multimodal studies of motor function. The pipelines and protocols detailed in this application note provide a framework for addressing the unique challenges presented by each modality while facilitating their integration. Key considerations include rigorous motion correction—particularly critical for active motor paradigms—appropriate physiological filtering, and careful spatial coregistration.

The FRESH initiative findings highlight that while analytical flexibility is valuable, standardization in critical preprocessing steps—especially handling poor-quality data—can significantly enhance reproducibility [47]. The developing field of data-driven fusion methods, including symmetric multimodal techniques that jointly analyze fMRI and fNIRS signals, shows promise for revealing more complex latent neurovascular coupling processes [52].

As hardware innovations continue, including MRI-compatible fNIRS probes and more portable systems, the potential for naturalistic motor studies will expand accordingly. By implementing robust, standardized preprocessing pipelines, researchers can maximize the complementary strengths of fMRI and fNIRS, advancing our understanding of motor control and its disruption in neurological disorders.

The integration of functional Magnetic Resonance Imaging (fMRI) and functional Near-Infrared Spectroscopy (fNIRS) provides a powerful multimodal framework for investigating brain function. fMRI offers high spatial resolution and whole-brain coverage, including deep structures, while fNIRS provides superior temporal resolution, portability, and tolerance to motion artifacts [14]. This combination is particularly advantageous in motor task paradigms, where it enables comprehensive investigation of the spatiotemporal dynamics of motor execution and imagery. The General Linear Model (GLM) serves as a foundational statistical framework for analyzing evoked hemodynamic responses in both modalities, while connectivity analyses, including effective and functional connectivity, reveal the complex network interactions underlying motor function [10] [53]. This Application Note provides detailed protocols for implementing these analysis techniques within an integrated fMRI-fNIRS research paradigm.

Theoretical Foundations of GLM in Hemodynamic Data Analysis

The General Linear Model is a multivariate framework that defines the relationship between a set of explanatory variables (experimental design) and observed neuroimaging data. For both fMRI and fNIRS, the GLM expresses measured signals as a linear combination of predicted hemodynamic responses and error terms [54]. The model is formulated as:

Y = Xβ + ε

Where Y is the matrix of observed data, X is the design matrix containing hypothesized hemodynamic response predictors, β represents the unknown parameters (weights) to be estimated, and ε is the error term assumed to be normally distributed [54].

In fMRI analysis, the GLM is applied to Blood Oxygen Level Dependent (BOLD) signals, which primarily reflect changes in deoxygenated hemoglobin [14]. In fNIRS, the model can be simultaneously applied to both oxygenated (HbO) and deoxygenated (HbR) hemoglobin concentrations [55]. A key advantage of the GLM approach is its ability to incorporate nuisance regressors (e.g., physiological noise, motion artifacts) directly into the design matrix, allowing for simultaneous estimation of task-evoked responses and confound suppression [55].

For multimodal integration, the GLM facilitates cross-validation of findings between fMRI and fNIRS. The spatial specificity of fMRI can validate fNIRS channel placement and activation patterns, while the temporal resolution of fNIRS can inform the modeling of hemodynamic response functions in both modalities [16] [10].

Quantitative Comparison of fMRI and fNIRS Characteristics

Table 1: Technical specifications and performance metrics of fMRI and fNIRS for motor task paradigms

Parameter	fMRI	fNIRS
Spatial Resolution	1-3 mm (whole-brain including subcortical) [14]	1-3 cm (superficial cortex only) [14]
Temporal Resolution	0.3-2 Hz (limited by hemodynamic response) [14]	Up to 100 Hz (typically 5-10 Hz) [14]
Measured Signal	BOLD (primarily reflects HbR changes) [14]	HbO and HbR concentration changes [55]
Portability	Low (requires fixed scanner environment) [14]	High (wearable systems available) [14]
Spatial Correlation	Reference modality	HbO: 0.65; HbR: -0.76 with fMRI BOLD [10]
Optimal Signal for Motor Tasks	BOLD signal	HbR shows higher specificity for SMA activation [16]
Tolerability to Motion	Low (susceptible to artifacts) [14]	High (suitable for naturalistic movements) [14]

Table 2: Performance comparison of GLM analysis applied to motor task paradigms

Analysis Metric	fNIRS with GLM+SS	fNIRS Conventional	fMRI with GLM
Binary Classification Accuracy	+7.4% improvement [55]	Baseline	Reference standard
Contrast-to-Noise Ratio	Significantly enhanced [55]	Moderate	High
Single-Trial Estimation	Improved [55]	Limited	Good
Feature Separability	Enhanced [55]	Moderate	High
Reproducibility (HbO)	High (across sessions) [20]	Moderate	High
Reproducibility (HbR)	Lower than HbO [20]	Lower than HbO	Reference standard
Effective Connectivity Estimation	Good (with complementary EEG) [56]	Limited	Excellent (DCM) [53]

Experimental Protocols

GLM Analysis Protocol for Integrated fMRI-fNIRS Motor Paradigms

Objective: To implement a synchronized GLM analysis pipeline for fMRI and fNIRS data acquired during motor execution and imagery tasks.

Materials and Equipment:

• 3T fMRI scanner with head coil
• Continuous-wave fNIRS system (e.g., NIRSport2) with 760nm and 850nm wavelengths
• 16 fNIRS sources and 15 detectors (54 channels) with 8 short-distance detectors
• Stimulus presentation system
• Eye tracking and motion capture systems
• Software: SPM12, Homer3, NIRS-KIT, BrainVoyager QX

Step-by-Step Procedure:

• Experimental Design

  • Implement block design with 30-second blocks for Motor Execution (ME), Motor Imagery (MI), and Baseline conditions
  • Counterbalance conditions across participants
  • For MI: Instruct participants to imagine bilateral finger tapping sequence (1-2-1-4-3-4) without overt movement [16]
  • For ME: Same sequence with physical execution at 2Hz frequency
  • Total duration: 8.5 minutes per run [10]

• Data Acquisition Parameters

  • fMRI: TR=1500ms, TE=30ms, 26 slices, voxel size 3×3×3.5mm, FOV=210×210mm [10]
  • fNIRS: Sampling rate=5.08Hz, source-detector distance=30mm, short-distance detectors=8mm [10]

• fMRI Preprocessing (using BrainVoyager QX)

  • Slice timing correction
  • 3D motion correction (6 parameters)
  • Spatial smoothing (Gaussian filter, FWHM=6mm)
  • High-pass filtering (GLM-Fourier, 2 cycles)
  • Co-registration to anatomical and normalization to Talairach space [10]

• fNIRS Preprocessing (using Homer3)

  • Channel pruning (SNR threshold: 15dB)
  • Conversion to optical density
  • Motion artifact correction (e.g., wavelet-based)
  • Bandpass filtering (0.01-0.5Hz)
  • Conversion to hemoglobin concentration changes [10]

• GLM Specification for fMRI

  • Design matrix: ME, MI, and Baseline conditions as predictors
  • Include 6 motion parameters as nuisance regressors
  • Hemodynamic Response Function (HRF) convolution
  • Contrasts: ME > Baseline, MI > Baseline, ME > MI [10]

• GLM Specification for fNIRS

  • Design matrix identical to fMRI
  • Incorporate short-separation channels as nuisance regressors for systemic artifact removal [55]
  • Estimate subject-specific and channel-specific HRF parameters [55]
  • Simultaneous estimation of HbO and HbR responses

• Cross-Modal Validation

  • Extract fMRI BOLD response from cortical regions corresponding to fNIRS channels
  • Compute spatial correlation between fMRI and fNIRS activation maps [10]
  • Compare temporal dynamics of HbO/HbR with BOLD signal

Troubleshooting Tips:

• Poor fNIRS signal quality: Check optode-scalp coupling and hair obstruction
• Low fMRI-fNIRS spatial correlation: Verify accurate co-registration using digitized optode positions
• Motion artifacts: Implement robust motion correction algorithms and exclude excessive movement trials

Dynamic Causal Modeling (DCM) Protocol for Effective Connectivity

Objective: To investigate directed influences between motor regions during task performance using DCM.

Materials and Equipment:

• Processed fMRI data from GLM analysis
• Software: SPM12 with DCM toolbox

Step-by-Step Procedure:

• Region of Interest (ROI) Selection

  • Identify ROIs from GLM analysis: Primary Motor Cortex (M1), Supplementary Motor Area (SMA), Premotor Cortex (PMC)
  • Use individual anatomical images for precise ROI definition [16]
  • Extract BOLD time series using spheres (8mm radius) around peak activation coordinates [53]

• DCM Model Specification

  • Specify full model with driving inputs (entering via SMA) and modulatory inputs (task conditions)
  • Use bilinear one-state DCM for fMRI [53]
  • Include parametric modulations (e.g., learning effects) as appropriate [57]

• Model Estimation and Comparison

  • Perform Bayesian model inversion using variational Laplace [53]
  • Compare evidence for different connectivity architectures
  • Use Parametric Empirical Bayes (PEB) for group-level analysis [53]

• Connectivity Analysis for fNIRS

  • Implement functional connectivity using Pearson correlation between fNIRS channels
  • Construct brain networks for graph theory analysis [58]
  • For effective connectivity, combine with EEG using multilayer network models [56]

Application Notes:

• DCM is hypothesis-driven; clearly articulate connectivity hypotheses before analysis
• For fNIRS, effective connectivity is best estimated when combined with EEG [56]
• In motor paradigms, focus on connections between SMA, M1, and PMC

Diagrammatic Representations

GLM Workflow for Multimodal fMRI-fNIRS Analysis

Connectivity Analysis Framework

The Scientist's Toolkit: Essential Research Reagents and Materials

Table 3: Essential research tools and solutions for multimodal fMRI-fNIRS studies

Tool/Reagent	Specification	Application Purpose	Example Vendor/Software
fNIRS System	Continuous-wave, 16 sources, 15 detectors, 760/850nm	Hemodynamic response measurement	NIRSport2 (NIRx)
fMRI Scanner	3T with head coil, EPI sequence	BOLD signal acquisition	Siemens Magnetom TimTrio
Short-Distance Detectors	8mm source-detector distance	Superficial signal regression for improved CNR [55]	Custom fNIRS setups
Analysis Software	SPM12, Homer3, NIRS-KIT, BrainVoyager	GLM implementation, preprocessing, connectivity analysis	Open source/Commercial
Digitization System	3D position digitizer	Accurate optode localization and co-registration	Polhemus Patriot
Motor Task Interface	Response box, tactile stimulator	Precise task timing and participant response	Current Design Inc.
Physiological Monitoring	Pulse oximeter, respiration belt	Physiological noise modeling	BIOPAC Systems
DCM Toolbox	SPM12 integration	Effective connectivity analysis	SPM Software Package [53]

The emerging field of neurobiomechanics provides an integrative framework for understanding human movement by combining insights from functional anatomy, musculoskeletal physiology, central nervous system function, and physics [59] [60]. This approach is particularly valuable for translating laboratory-based motor task paradigms into real-world gait and balance assessment, addressing a critical gap in neurological rehabilitation and drug development. Traditional diagnostic approaches often overlook the intricate interplay between neural signals and mechanical forces that characterize motor function in pathological conditions [59]. The integration of functional magnetic resonance imaging (fMRI) and functional near-infrared spectroscopy (fNIRS) offers a powerful multimodal approach to bridge this gap, leveraging their complementary strengths for comprehensive motor assessment across environments [3].

Human movement results from highly coordinated mechanical interactions between bones, muscles, ligaments, and joints, regulated by the nervous system [59] [60]. This complex process involves transformations between neural inputs, mechanical forces, and sensory feedback, creating a continuous loop between central nervous system commands and peripheral execution. In neurological disorders, this delicate interplay becomes disrupted, leading to altered motor patterns and functional impairments that require integrated assessment approaches [59]. The neurobiomechanics framework enables researchers and clinicians to decompose these disruptions by simultaneously evaluating neurophysiological and biomechanical parameters, providing a more complete picture of motor dysfunction and recovery trajectories [59] [60].

Comparative Technical Specifications of fMRI and fNIRS

Methodological Synergies and Complementarity

The combined use of fMRI and fNIRS capitalizes on their complementary capabilities for motor assessment. fMRI provides high spatial resolution (millimeter-level) and whole-brain coverage, including deep brain structures, making it ideal for localizing specific brain regions involved in motor tasks [3]. However, its temporal resolution is constrained by the hemodynamic response (typically lagging 4-6 seconds behind neural activity), and it requires immobile participants in a restrictive scanner environment [3]. Conversely, fNIRS offers superior temporal resolution (millisecond-level), portability, and resilience to motion artifacts, enabling brain imaging during active motor tasks such as walking and balance activities [3] [61]. These complementary features create an ideal methodological synergy for translating laboratory findings to real-world assessment.

Table 1: Technical Comparison of fMRI and fNIRS for Motor Assessment

Parameter	fMRI	fNIRS
Spatial Resolution	High (1-3 mm)	Moderate (1-3 cm)
Temporal Resolution	Low (0.33-2 Hz)	High (up to 100 Hz)
Penetration Depth	Whole-brain (cortical & subcortical)	Superficial cortical (2-3 cm)
Portability	No (fixed scanner)	Yes (wearable systems)
Motion Tolerance	Low	High
Primary Signal	BOLD response	HbO, HbR concentration changes
Motor Task Environment	Restricted to laboratory	Naturalistic settings
Cost & Accessibility	High, limited	Moderate, increasing

Spatial and Temporal Correspondence Validation

Multimodal validation studies have demonstrated significant spatial correspondence between fMRI and fNIRS hemodynamic responses in motor-network regions. Research investigating motor imagery and execution tasks found that group-level activation identified in fMRI data could be modeled using subject-specific fNIRS signals, with significant peak activation overlapping individually-defined primary and premotor cortices [10]. No statistically significant differences were observed in multimodal spatial correspondence between oxygenated hemoglobin (HbO), deoxygenated hemoglobin (HbR), and total hemoglobin (HbT) for both tasks, suggesting that both oxy- and deoxyhemoglobin data can effectively translate neuronal information from fMRI to fNIRS setups [10].

The temporal correlation between modalities varies across studies, with HbO generally showing higher correlation with the fMRI BOLD signal, though reported values range from 0 to 0.8 [10]. This variability highlights the importance of standardized protocols while confirming the fundamental relationship between the hemodynamic measurements captured by each modality.

Application Protocols for Motor Task Paradigms

Protocol 1: Upper Limb Motor Execution and Imagery

Objective: To assess cortical activation patterns during prepared and executed upper limb movements for quantifying motor planning and execution deficits in neurological disorders.

Experimental Design:

• Participants: Patients with mild cognitive impairment, Parkinson's disease, stroke, or healthy controls matched for age and gender [62]
• Task Paradigm: Block design alternating between motor action (MA) and motor imagery (MI) conditions with baseline periods [10]
• MA Blocks: Participants execute bilateral finger tapping sequences (e.g., 1-2-1-4-3-4 where 1=left middle finger, 2=left index, 3=right index, 4=right middle finger) at 2 Hz frequency [10]
• MI Blocks: Participants imagine the same sequence without overt movement [10]
• Block Structure: 17 blocks of 30-second duration (9 baseline, 4 MA, 4 MI) for total 8.5-minute protocol [10]

Data Acquisition:

• fMRI Parameters: 3T scanner, TR=1500ms, TE=30ms, 26 slices, 3×3mm in-plane resolution, 3.5mm slice thickness [10]
• fNIRS Setup: 16 sources (760nm & 850nm), 15 detectors, 8 short-distance detectors (8mm), 54 channels covering bilateral motor areas, sampling at 5.08Hz [10]
• Concurrent Measures: EMG of forearm flexor/extensor muscles to monitor unintended movement during imagery tasks

Data Analysis:

• fMRI Preprocessing: Slice timing correction, motion correction, spatial smoothing (FWHM=6mm), normalization to standard space [10]
• fNIRS Preprocessing: Conversion to optical density, motion artifact correction, bandpass filtering (0.01-0.2Hz), conversion to hemoglobin concentrations [10]
• Statistical Analysis: General Linear Model (GLM) with predictors for conditions, motion parameters, and short-distance channels as nuisance regressors [63] [10]

Protocol 2: Interactive Motor-Cognitive Dual-Task Assessment

Objective: To evaluate the impact of increasing cognitive load on motor performance and cortical activation during walking tasks.

Experimental Design:

• Participants: 28 healthy adults or neurological patients with no cognitive impairment (MMSE ≥24) and ability to walk independently [61]
• Task Paradigm: Interactive motor-cognitive dual-task with three difficulty levels performed during walking:
  • Easy Task (ET): Simple color-word Stroop task
  • Medium Task (MT): Increased Stroop complexity
  • Difficult Task (DT): High-conflict Stroop trials [61]
• Task Structure: Each difficulty level performed for 5-minute walks with 3-minute rest between conditions

Data Acquisition:

• fNIRS Configuration: 36-channel system with 12 short channels covering 10 ROIs: left/right prefrontal cortex (LPFC/RPFC), left/right dorsolateral prefrontal cortex (DLPFC/DRPFC), left/right premotor cortex (LPMC/RPMC), left/right sensorimotor cortex (LSC/RSC), and left/right motor cortex (LMC/RMC) [61]
• Kinematic Measures: Inertial Measurement Unit (IMU) sensors on feet, trunk, and head to capture gait parameters (speed, stride length, variability) [61]
• Cognitive Performance: Accuracy and reaction time recorded for Stroop tasks [61]

Data Analysis:

• Hemodynamic Response: Mean HbO changes calculated for each ROI across difficulty levels
• Functional Connectivity (FC): ROI-to-ROI connectivity analysis using correlation methods
• Lateralization Index (LI): Calculation of hemispheric dominance (LI range: 0.10-0.35 indicates right-brain lateralization) [61]
• Gait Parameter Correlation: Regression analysis between cortical activation and gait metrics

Table 2: Key Findings from Dual-Task Motor-Cognitive Assessment [61]

Measurement	Easy Task	Medium Task	Difficult Task	Statistical Significance
HbO in RPMC	0.12 ± 0.08 μM	0.18 ± 0.09 μM	0.27 ± 0.11 μM	p < 0.05
HbO in LMC	0.09 ± 0.07 μM	0.14 ± 0.08 μM	0.22 ± 0.10 μM	p < 0.05
Functional Connectivity	0.46 ± 0.21	0.54 ± 0.19	0.61 ± 0.21	p = 0.023 (DT vs ET)
Gait Speed	1.32 ± 0.15 m/s	1.28 ± 0.16 m/s	1.15 ± 0.18 m/s	p < 0.05
Cognitive Accuracy	95.2 ± 3.1%	89.7 ± 4.5%	78.3 ± 6.8%	p < 0.01
Lateralization Index	0.10 ± 0.08	0.22 ± 0.11	0.35 ± 0.14	p < 0.05

Protocol 3: Prepared versus Single Walking Task Assessment

Objective: To investigate how motor preparation influences brain activation and gait performance in healthy and neurologically impaired populations.

Experimental Design:

• Participants: 57 individuals with Mild Cognitive Impairment (MCI) and 67 healthy controls (HCs) matched for age and gender [62]
• Task Paradigm: Two walking conditions:
  • Prepared Walking (PW): "Ready" verbal cue provided 3 seconds before initiation of walking
  • Single Walking (SW): Immediate walking without preparatory cue [62]
• Task Structure: 10 trials per condition, 20-second walking bouts with 30-second rest between trials

Data Acquisition:

• fNIRS Configuration: Portable system covering 4 ROIs: prefrontal cortex (PFC), primary motor cortex (M1), secondary motor cortex (M2), and parietal lobe [62]
• Gait Analysis: Timed walk, gait speed measurement, and stride characteristic assessment

Data Analysis:

• Activation Differences: Contrast between PW and SW conditions across groups
• Correlation Analysis: Montreal Cognitive Assessment (MoCA) scores correlated with cortical activation
• Gait-Brain Relationship: Regression between gait speed and oxygenation changes

Computational Integration and Analytical Approaches

Data Fusion Methodologies

The integration of fMRI and fNIRS data can be implemented through synchronous or asynchronous detection modes [3]. Synchronous acquisition provides direct temporal correspondence but presents technical challenges regarding electromagnetic compatibility in the MRI environment. Asynchronous approaches involve separate sessions with careful temporal alignment during analysis, enabling the translation of fMRI-localized regions to fNIRS montages for naturalistic assessment [3] [10].

Advanced computational platforms facilitate the integration of neural and biomechanical data:

• OpenSim: Enables estimation of joint forces and muscle activations based on kinematic and EMG data [59] [60]
• MOtoNMS: MATLAB-based toolbox for standardized preprocessing of motion capture and EMG data [59] [60]
• NEUROiD and NfMBS: Multiscale environments integrating neural, muscular, and skeletal models for simulating pathological movement [59] [60]

Reproducibility and Quality Control Considerations

fNIRS reproducibility varies significantly with data quality, analysis pipelines, and researcher experience [7]. Key factors affecting reproducibility include:

• Data Quality: Better quality data improves individual-level agreement [7]
• Analysis Confidence: Teams with higher self-reported analysis confidence (correlated with fNIRS experience) showed greater agreement [7]
• Optode Placement: Increased shifts in optode position reduce spatial overlap across sessions [20]
• Chromophore Choice: HbO demonstrates higher reproducibility than HbR across sessions [20]
• Source Localization: Anatomically specific head models improve reliability of capturing brain activity [20]

The FRESH (fNIRS Reproducibility Study Hub) initiative, involving 38 research teams analyzing identical datasets, found that nearly 80% agreed on group-level results when hypotheses were strongly supported by literature, highlighting the importance of standardized protocols and analytical transparency [7].

The Scientist's Toolkit: Essential Research Reagent Solutions

Table 3: Essential Materials and Analytical Tools for fMRI-fNIRS Motor Studies

Item	Specification	Function/Application
fNIRS System	NIRSport2 (NIRx) or comparable; 16+ sources, 15+ detectors, 8+ short-distance detectors	Portable cortical hemodynamic monitoring during natural movement [10]
fMRI Scanner	3T with 12-channel head coil, EPI sequence capability	High-spatial resolution localization of motor network activation [10]
Motion Capture	Inertial Measurement Units (IMU), optical systems (Vicon)	Kinematic analysis of gait parameters and movement quality [61]
EMG System	Wireless surface electrodes, >1000Hz sampling	Muscle activation timing and coordination assessment [59]
Analysis Software	Homer3, BrainVoyager, FieldTrip, OpenSim	Data preprocessing, statistical analysis, biomechanical modeling [59] [63] [10]
Montage Design Tools	AtlasViewer, fOLD toolbox	Optode placement optimization for target brain regions [20]
Short-Distance Channels	8mm source-detector separation	Superficial signal regression for improved brain specificity [63] [10]

Implementation Workflow and Translational Pathway

The following diagram illustrates the integrated workflow for translating laboratory-based motor assessments to real-world applications:

The translational application of integrated fMRI-fNIRS protocols for motor assessment represents a significant advancement in neurorehabilitation and therapeutic development. By leveraging the spatial precision of fMRI with the ecological validity of fNIRS, researchers can now bridge the critical gap between laboratory findings and real-world motor function. The structured protocols outlined provide methodological rigor for assessing motor imagery, execution, and complex motor-cognitive interactions across neurological populations.

Future directions should focus on standardizing analytical pipelines across research sites, developing age- and pathology-specific normative databases, and advancing artifact correction algorithms for increasingly naturalistic movement assessment. As fNIRS technology continues to evolve with improved spatial resolution and whole-head coverage, its integration with fMRI will become increasingly seamless, ultimately enabling comprehensive neuromotor assessment from laboratory to home and community environments. For drug development professionals, these approaches offer quantitative biomarkers for evaluating therapeutic efficacy on functional outcomes that matter to patients' daily lives.

The integration of real-time functional magnetic resonance imaging (rt-fMRI) and functional near-infrared spectroscopy (fNIRS) represents a transformative approach in translational neuroscience, particularly for developing novel neurorehabilitation strategies and quantifying target engagement in clinical trials. This application note details structured protocols and empirical data for employing these multimodal neurofeedback paradigms in stroke and Parkinson's disease (PD) motor rehabilitation, and for validating biomarker engagement in therapeutic development. The content is framed within a broader research thesis on integrated fMRI-fNIRS motor task paradigms, providing methodologies for leveraging their complementary strengths: fMRI's high spatial resolution for precise localization and fNIRS's portability for real-world training and longitudinal monitoring [14].

Neurofeedback for Motor Rehabilitation: Stroke and Parkinson's Disease

Neurofeedback (NFB) operates as a closed-loop system that provides real-time information on a participant's brain activity, enabling the development of self-learning strategies to modulate these signals via operant conditioning. This approach can induce neural plasticity and promote functional recovery [64] [65].

Quantitative Evidence for Neurofeedback Efficacy

Table 1: Neurofeedback Efficacy in Motor Rehabilitation

Condition	Neurofeedback Approach	Key Neural Targets	Reported Clinical Outcomes	Evidence Source
Stroke	rt-fMRI & rt-fNIRS (sequential)	Supplementary Motor Area (SMA), Premotor Cortex, Primary Motor Cortex (M1)	Clinically significant recovery of arm coordination and active wrist extension; up to 71% brain signal accuracy during rt-fNIRS [46].	Combined Protocol Feasibility Study [46]
Parkinson's Disease (PD)	rt-fMRI-guided Motor Imagery	Supplementary Motor Area (SMA)	Average improvement of 4.5 points on the MDS-UPDRS motor scale ("off-medication"), meeting the minimal clinically important difference [66].	Randomized Controlled Trial [66]

Experimental Protocol: Combined Sequential fMRI-fNIRS for Stroke Motor Rehabilitation

This protocol outlines a method to leverage the spatial precision of fMRI initially, followed by multiple practical fNIRS sessions to consolidate learning.

Protocol 1: Sequential fMRI-fNIRS Neurofeedback for Chronic Stroke

• Objective: To promote recovery of upper limb coordination and wrist extension in chronic stroke survivors with moderate/severe motor impairment.
• Patient Population: Adults >6 months post-stroke with persistent upper extremity motor impairment [46].
• Equipment:
  • MRI scanner with real-time fMRI processing capabilities.
  • fNIRS system with optodes placed over the contralesional motor cortex (e.g., SMA, M1).
  • Custom software (e.g., Matlab/Python) for real-time signal processing and feedback.
  • Functional Electrical Stimulation (FES) device for assisted movement.
• Procedure:
  • Localizer Session (fMRI): Conduct an initial fMRI scan with a hand grasp task (or motor imagery if movement is limited) to identify the target Region of Interest (ROI) for neurofeedback, typically in the ipsilesional motor network [46].
  • rt-fMRI Neurofeedback Phase (Sessions 1-4):
    • Patients perform kinesthetic motor imagery of the affected hand.
    • Provide real-time visual feedback based on the BOLD signal from the pre-defined ROI (e.g., a thermometer display whose height varies with activation level).
    • The goal is to train the patient to volitionally upregulate activity in the target area [46] [65].
  • rt-fNIRS Neurofeedback Phase (Sessions 5-15+):
    • Transition to an upright, sitting position using fNIRS.
    • Optodes are placed based on the fMRI localizer data or using standard positions over the motor cortex.
    • Patients again perform motor imagery, now receiving feedback based on fNIRS hemodynamic signals (typically oxygenated hemoglobin, HbO).
    • FES is applied to the affected wrist extensor muscles concurrently with successful neurofeedback trials, providing afferent feedback [46].
  • Motor Learning Sessions: Intersperse neurofeedback sessions with physical practice sessions without neural feedback to consolidate re-learned coordination into functional tasks [46].
• Key Outcome Measures:
  • Primary Clinical: Fugl-Meyer Assessment for upper extremity, active range of motion for wrist extension.
  • Neurophysiological: Self-regulation accuracy (% of run above threshold), amplitude of brain signal during wrist extension, and focality of activation on fMRI [46].

Experimental Protocol: fMRI-Neurofeedback for Parkinson's Disease

This protocol targets the underactive Supplementary Motor Area (SMA) in PD to improve motor symptoms.

Protocol 2: rt-fMRI Neurofeedback for Parkinson's Disease

• Objective: To improve motor symptoms in PD patients (Hoehn & Yahr I-III) by upregulating SMA activity.
• Study Design: Randomized, controlled trial [66].
• Intervention Group (NF):
  • rt-fMRI Neurofeedback: Three rt-fMRI sessions (weeks 2, 6, 12). Patients use kinesthetic motor imagery to upregulate the SMA, guided by real-time visual feedback of their BOLD signal.
  • Motor Imagery Homework: Regular practice of motor imagery between scanner sessions.
  • Motor Training (Weeks 5-10): Supervised exercise using a virtual reality gaming device (e.g., Nintendo Wii Fit) [66].
• Control Group (MOT): Receives only supervised motor training on the gaming device throughout the 10-week intervention [66].
• Primary Outcome: Change in the Movement Disorder Society-Unified PD Rating Scale-Motor scale (MDS-UPDRS-MS) in the "off-medication" state [66].

Diagram 1: Experimental workflow for PD neurofeedback training.

Target Engagement Biomarkers in Drug Development

Target engagement biomarkers are critical in early-phase clinical trials for confirming that a drug candidate interacts with its intended biological target and connects this interaction to a physiological effect [67].

Conceptual Framework for Biomarker Validation

Table 2: Framework for Target Engagement Biomarker Development

Stage	Action	Purpose	Example (MetAP2 Inhibitors)
Identification	Discover a proximal, quantifiable marker linked to the target.	To have a direct measure of drug-target interaction.	Identification of NMet14-3-3γ, a MetAP2 substrate that accumulates upon inhibition [67].
Preclinical Validation	Correlate biomarker changes with efficacy in disease models.	To establish a PK-PD-E (Pharmacokinetic-Pharmacodynamic-Efficacy) relationship.	In obese mice, increased NMet14-3-3γ in adipose tissue correlated with dose-dependent body weight loss [67].
Assay Development	Create a robust, translatable assay for clinical use.	To measure target engagement in human trials.	Development of an assay to measure inhibitor-bound MetAP2 levels in human blood [67].
Clinical Translation	Use biomarker to guide dosing and predict efficacy in trials.	To make informed go/no-go decisions and select optimal dosing regimens.	Using the NMet14-3-3γ biomarker and MetAP2 occupancy to predict weight loss efficacy in humans [67].

Experimental Protocol: Validating a Target Engagement Biomarker

This generic protocol can be adapted for validating target engagement biomarkers for novel therapeutics in neurological disorders.

Protocol 3: Validation of a Target Engagement Biomarker

• Objective: To establish a quantitative relationship between drug exposure, target engagement, and a downstream efficacy readout.
• Application: Preclinical phase of drug development for novel neurological targets (e.g., enzymes, receptors implicated in PD or stroke recovery) [67].
• In Vitro Phase:
  • Primary Human Cells: Treat relevant primary human cells (e.g., neurons, glial cells) with a range of drug concentrations.
  • Biomarker Measurement: Quantify the proposed target engagement biomarker (e.g., levels of a phosphorylated protein, accumulated substrate) using ELISA, mass spectrometry, or other specific assays.
  • Dose-Response: Establish a concentration-response curve for target engagement [67].
• In Vivo Phase (Animal Model):
  • Dosing: Administer the drug candidate to a relevant animal disease model (e.g., PD mouse model) at multiple doses and frequencies.
  • Sample Collection: Collect plasma and target tissue (e.g., brain regions) at various time points.
  • Analysis: Measure:
    • PK: Compound concentration in plasma.
    • PD: Biomarker level in the target tissue.
    • Efficacy: Relevant functional or behavioral outcome [67].
• Data Integration and Modeling:
  • Develop a PK-PD model linking plasma concentration to biomarker level.
  • Develop a PK-PD-E model linking biomarker level to the efficacy outcome.
  • Use this model to predict the dose and regimen required for efficacy in humans [67].

Diagram 2: Workflow for target engagement biomarker validation.

The Scientist's Toolkit: Essential Research Reagents and Materials

Table 3: Key Research Reagents and Solutions for Neurofeedback and Biomarker Studies

Item	Function/Application	Examples & Notes
rt-fMRI Software Suite	Real-time processing of BOLD signal, ROI definition, and feedback display generation.	Custom Matlab or Python scripts; Turbo-BrainVoyager; OpenNFT [46].
rt-fNIRS System	Portable measurement of HbO/HbR concentration changes for neurofeedback in naturalistic positions.	Continuous-wave (CW) systems; ensure compatibility with individual anatomy for channel placement [16] [46].
Motor Imagery Paradigms	Standardized tasks to elicit activation in motor networks without physical movement.	Kinesthetic imagery of hand grasping or whole-body movement; validated instructions are critical [16] [66].
Functional Electrical Stimulation (FES)	Provides afferent feedback and assists movement during fNIRS training, engaging Hebbian plasticity.	Used to stimulate peripheral nerves or muscles contingent with successful brain self-regulation [46].
Target Engagement Assay Kits	Quantifying biomarker levels in biological samples (plasma, tissue, CSF).	ELISA kits; reagents for seed amplification assays (e.g., for α-synuclein in PD); mass spectrometry protocols [68] [67].
Validated Clinical Scales	Standardized assessment of clinical efficacy in trials.	Stroke: Fugl-Meyer Assessment (FMA). PD: Movement Disorder Society-Unified PD Rating Scale (MDS-UPDRS) [46] [66].

Navigating Technical Challenges: Optimization Strategies for Robust Data Quality

Functional near-infrared spectroscopy (fNIRS) has emerged as a prominent neuroimaging technique due to its portability, cost-efficiency, and tolerance for motion artifacts. However, a significant challenge persists: fNIRS measures cortical activity from the scalp surface without providing intrinsic anatomical information about the underlying brain structures [69]. This limitation becomes particularly critical in research integrating fNIRS with functional magnetic resonance imaging (fMRI) for motor task paradigms, where precise spatial correspondence between modalities is essential for validating findings and translating fMRI-based paradigms to fNIRS setups [10].

The spatial resolution of fNIRS is fundamentally constrained by how optodes (light sources and detectors) are arranged on the scalp. The quality of the measured signal and sensitivity to cortical regions-of-interest (ROIs) depend heavily on this arrangement [70]. Unlike fMRI, which provides whole-brain coverage with high spatial resolution, fNIRS is limited to superficial cortical regions and suffers from a limited number of measurement channels in typical setups [14]. Overcoming these spatial constraints requires sophisticated methods for optode placement that account for individual neuroanatomy and the physics of light propagation in biological tissues.

This application note details advanced methodologies for probabilistic registration and anatomical guidance to optimize fNIRS optode placement, specifically framed within motor task research integrating fNIRS and fMRI. We provide structured protocols, quantitative comparisons, and practical tools to enhance spatial specificity and improve cross-modal validation in neuroscientific investigations and clinical applications.

Background and Significance

The Spatial Challenge in fNIRS

fNIRS operates by emitting near-infrared light through the scalp and detecting photons that have traveled through cerebral tissues. Each source-detector pair forms a measurement channel sensitive to hemodynamic changes in the underlying cortex. However, the technique faces two primary spatial limitations: (1) an inherent ambiguity in localizing the precise origin of measured signals, and (2) significant inter-individual anatomical variability that affects scalp-cortex correspondence [71].

The spatial sensitivity profile of an fNIRS channel is not confined to a single point beneath the optode midpoint but extends to a broad area due to strong light scattering in biological tissues. This scattering means that each channel captures a weighted sensitivity profile across multiple cortical regions, necessitating computational approaches to estimate the specific cortical areas being measured [71]. Furthermore, individual differences in cortical folding, head size, and skull thickness dramatically influence how optodes on the scalp relate to underlying functional areas, particularly in motor regions where precise localization is crucial for paradigm validation [71].

fMRI-fNIRS Integration in Motor Task Paradigms

Motor tasks—including execution, imagery, and learning—have become benchmark paradigms for evaluating fMRI-fNIRS integration due to their well-characterized cortical representations [10]. The primary motor cortex (M1) and premotor areas (PMA) exhibit robust, reproducible hemodynamic responses during motor activities, making them ideal targets for cross-modal validation studies.

Research demonstrates that fNIRS can reliably detect motor-related activation clusters identified through fMRI when optodes are properly placed. Studies investigating spatial correspondence have found that fNIRS signals from motor regions can successfully predict fMRI activation patterns, with no statistically significant differences observed between oxyhemoglobin (HbO), deoxyhemoglobin (HbR), and total hemoglobin (HbT) in their ability to identify motor cortex activation in corresponding fMRI data [10]. This correspondence provides a solid foundation for translating sophisticated fMRI motor paradigms to more flexible fNIRS setups, with applications ranging from basic neuroscience to clinical rehabilitation and drug development.

Table 1: Comparison of fMRI and fNIRS Characteristics for Motor Task Research

Characteristic	fMRI	fNIRS
Spatial Resolution	High (1-3 mm)	Moderate (1-3 cm)
Temporal Resolution	Low (0.3-2 Hz)	High (up to 100 Hz)
Depth Penetration	Whole-brain	Superficial cortex (2-3 cm)
Portability	Low	High
Sensitivity to Motion	High	Moderate
Target Populations	Limited due to constraints	Broad (including pediatric and patient populations)
Primary Signal	BOLD (Δ[HbR])	Δ[HbO], Δ[HbR], Δ[HbT]

Methodological Approaches

Probabilistic Registration Methods

Probabilistic registration approaches enable researchers to estimate the relationship between scalp positions and cortical areas without requiring individual MRI data for each subject. These methods leverage reference MRI databases and standardized coordinate systems to compute the most likely cortical projection points for fNIRS channels.

The international 10-20, 10-10, and 10-5 systems provide standardized frameworks for describing scalp landmarks based on anatomical reference points (nasion, inion, and preauricular points) [69]. These systems enable consistent positioning of optodes across subjects and studies. By correlating these scalp positions with cortical anatomy through probabilistic atlases, researchers can estimate which brain regions are measured by each fNIRS channel.

Advanced probabilistic methods incorporate light propagation models to account for the scattering effect of near-infrared light in head tissues. The Sensitivity-Based Matching (SBM) method has demonstrated superior performance compared to conventional geometrical matching approaches by incorporating the broad spatial sensitivity of probe pairs due to light scattering [71]. This method computes the sensitivity profile of each channel through photon migration simulations, providing a more accurate mapping between scalp positions and cortical regions than point-to-point geometrical projection.

Table 2: Quantitative Comparison of Optode Placement Approaches for Motor Cortex Targeting

Approach	Required Resources	Spatial Accuracy	Signal Quality (SNR)	Setup Robustness	Best Use Cases
Literature-Based (LIT)	Literature review only	Low	Reference level	Low	Preliminary studies, limited resources
Probabilistic (PROB)	Individual anatomy + probabilistic fMRI maps	Moderate-High	Comparable to iFMRI/fVASC	High	Most studies without subject fMRI
Individual fMRI (iFMRI)	Individual anatomical + fMRI data	High	High	High	Studies requiring maximal precision
Vascular (fVASC)	Individual anatomical, functional, and vascular data	High	High	High	Specialized applications

Anatomical Guidance with Subject-Specific MRI

When subject-specific MRI data is available, anatomical guidance significantly enhances optode placement precision. This process involves co-registering fNIRS probe locations with the individual's structural MRI to determine the exact sensitivity profile for each channel [69].

The process begins by mapping the 3D positions of optodes on the subject's scalp using digitization techniques. These positions are then co-registered with the individual's structural MRI through alignment of common fiducial points (nasion, inion, preauricular). Once co-registered, Monte Carlo simulations or simplified light transport models compute the sensitivity profile for each source-detector pair, revealing which cortical areas contribute most significantly to the measured signals [72].

Research demonstrates that approaches incorporating individual functional MRI data (iFMRI) outperform literature-based methods in both signal quality and sensitivity to task-related activation [70]. Interestingly, studies comparing progressively individualized approaches found that probabilistic methods (PROB) incorporating individual anatomical data with probabilistic fMRI maps performed nearly as well as fully individualized methods (iFMRI) that used subject-specific fMRI data, suggesting that probabilistic approaches represent a favorable balance between practicality and performance [70].

Experimental Protocols

Protocol 1: Probabilistic Registration for Group Studies

Purpose: To implement a standardized probabilistic registration procedure for fNIRS group studies targeting motor regions without requiring individual MRI data.

Materials and Equipment:

• fNIRS system with compatible optodes and cap
• 3D digitizer (e.g., Polhemus Fastrak)
• fNIRS Optodes' Location Decider (fOLD) toolbox [73]
• AtlasViewer software [72]
• Reference MRI database (e.g., Colin27, MNI152)

Procedure:

• Cap Placement: Secure an fNIRS cap incorporating the 10-5 system on the participant's head. Identify and mark fiducial points (nasion, inion, left/right preauricular).
• Optode Configuration: Arrange sources and detectors over motor regions (C3, C4, FC3, FC4, CP3, CP4 positions) with inter-optode distances of 25-35 mm.
• Digitization: Using a 3D digitizer, record the precise spatial coordinates of all optodes and fiducial points relative to the head coordinate system.
• Spatial Registration: Input digitized positions into AtlasViewer. Use the automated processing pipeline to co-register optode positions with the standard brain atlas through non-linear transformation.
• Sensitivity Analysis: Generate spatial sensitivity profiles for each channel using Monte Carlo simulations based on the atlas head model.
• Region-of-Interest Assignment: Map sensitivity profiles to anatomical atlases (e.g., AAL) to determine the primary cortical target for each channel, focusing on precentral gyrus (M1) and premotor areas.

Validation: Conduct a motor execution task (e.g., finger tapping) to verify expected HbO increases and HbR decreases in primary motor regions.

Protocol 2: Subject-Specific MRI Co-registration

Purpose: To achieve maximal spatial precision in optode placement using individual MRI data for targeting motor regions.

Materials and Equipment:

• fNIRS system with compatible optodes
• MRI scanner
• MRI-visible fiducial markers
• Neuronavigation system or 3D digitizer
• AtlasViewer or NIRSTORM software [74]

Procedure:

• Marker Placement: Prior to MRI scanning, attach MRI-visible fiducial markers (e.g., vitamin E capsules) at standard 10-5 positions covering motor areas.
• Structural MRI Acquisition: Acquire a high-resolution T1-weighted structural scan with the following typical parameters: 1×1×1 mm voxels, 176 sagittal slices, TE/TR = 3.42/2530 ms, flip angle = 7° [10].
• fNIRS Setup: Following scanning, place fNIRS optodes on the participant using the same marked positions from the MRI session.
• Optode Digitization: Digitize the 3D positions of all optodes and fiducial markers using a neuronavigation system or 3D digitizer.
• Co-registration: Import structural MRI and digitized optode positions into AtlasViewer. Manually co-register by aligning digitized fiducials with corresponding markers visible in the MRI.
• Head Model Creation: Segment head tissues (scalp, skull, CSF, gray matter, white matter) and generate a volume mesh for light transport simulation.
• Sensitivity Profile Calculation: Run Monte Carlo simulations to compute the sensitivity profile for each channel, identifying the cortical regions with highest sensitivity.
• Probe Optimization: Adjust optode placement iteratively to maximize sensitivity to target motor regions while maintaining practical constraints.

Validation: Compare fNIRS activation patterns with subject-specific fMRI data from a matching motor task to quantify spatial correspondence.

Diagram 1: Probabilistic Registration Workflow for fNIRS Motor Studies

The Scientist's Toolkit

Table 3: Essential Research Reagents and Tools for fNIRS-fMRI Integration in Motor Research

Tool/Resource	Type	Primary Function	Application in Motor Paradigms
AtlasViewer [72]	Software	Spatial registration and probe design	Visualize probe placement on standard brain; compute sensitivity profiles
fOLD Toolbox [73]	Software	Optode placement decision	Determine optimal positions for targeting motor regions
NIRSTORM [74]	Software	Optode montage optimization	Personalized fNIRS investigations with EEG integration
Array Designer [74]	Software	Automated array design	Generate optimized probe layouts for motor cortex coverage
3D Digitizer	Hardware	Spatial localization	Record precise 3D coordinates of optodes on scalp
MRI-Visible Markers	Material	Cross-modal registration	Create common reference points between MRI and fNIRS
Colin27 Atlas	Reference Data	Standard brain model	Template for probabilistic registration
SPM12 Tissue Probability Maps [73]	Reference Data	Tissue segmentation	Priors for head model creation in photon migration simulations

Data Analysis and Interpretation

Quantitative Assessment of Spatial Correspondence

Evaluating the success of optode placement strategies requires quantitative metrics of spatial correspondence between fNIRS channels and target brain regions. The sensitivity-based matching (SBM) method provides a superior approach compared to geometrical methods by accounting for light scattering in head tissues [71].

To calculate spatial correspondence for motor tasks:

• Generate Sensitivity Profiles: Compute the normalized sensitivity for each channel to every cortical region in the anatomical atlas.
• Define Primary Target: Identify the cortical region with the highest sensitivity value for each channel.
• Calculate Specificity Metrics: Determine the percentage of total sensitivity captured by the target region (e.g., precentral gyrus) versus neighboring areas.
• Cross-Modal Validation: For studies with concurrent fMRI, compute the spatial correlation between fNIRS activation maps and fMRI BOLD signals in motor regions.

Research shows that inter-individual anatomical variability significantly affects scalp-cortex correlation, with the SBM method achieving more consistent targeting of motor regions across subjects compared to geometrical methods [71]. Studies report that probabilistic approaches can achieve approximately 70-80% of the signal quality obtained with full individual fMRI data when targeting motor regions [70].

Signal Quality Metrics

The effectiveness of optode placement directly impacts signal quality and the ability to detect task-related activation. Key metrics include:

• Signal-to-Noise Ratio (SNR): Calculated as the ratio of task-related signal power to noise power
• Contrast-to-Noise Ratio: Measure of the strength of hemodynamic response relative to background fluctuations
• Reproducibility: Consistency of activation patterns across repeated measurements

Studies comparing placement approaches found that methods incorporating individual anatomical information (PROB, iFMRI, fVASC) significantly outperformed literature-based approaches in signal quality, with oxyhemoglobin (HbO) demonstrating higher reproducibility across sessions compared to deoxyhemoglobin (HbR) [20]. For motor tasks, source localization techniques that incorporate spatial sensitivity profiles improve the reliability of capturing brain activity compared to channel-based analyses [20].

Diagram 2: Subject-Specific MRI Co-registration Workflow

Implementation Considerations

Practical Recommendations for Motor Paradigms

Based on comparative studies, we recommend the following approach for motor task research:

• Resource Allocation: When subject-specific fMRI is unavailable, probabilistic approaches (PROB) using individual anatomical data with probabilistic fMRI maps provide favorable performance without the cost and complexity of acquiring individual functional scans [70].

• Optode Configuration: For motor cortex targeting, implement a minimum of 2-3 channels per hemisphere with source-detector distances of 25-35 mm to balance sensitivity to cerebral signals and adequate signal-to-noise ratio [70].

• Validation Procedure: Include a simple motor execution task (e.g., finger tapping) in initial sessions to verify expected activation patterns in primary motor regions before proceeding to more complex motor imagery or learning paradigms.

• Consistent Placement: Use customized caps with predefined optode positions based on 10-5 landmarks to ensure consistent placement across multiple sessions, as increased shifts in optode position correlate with reduced spatial overlap across sessions [20].

Advancements and Future Directions

Emerging methodologies continue to enhance spatial precision in fNIRS optode placement:

• Hybrid Approaches: Combining probabilistic registration with subject-specific vascular information (fVASC) may further improve sensitivity estimates, particularly in populations with atypical vasculature [70].

• Machine Learning Integration: Automated algorithms for optimizing probe layouts using constraints from both anatomical and functional priors are showing promise for maximizing sensitivity to target networks.

• Multimodal Integration: Simultaneous EEG-fNIRS setups benefit from integrated source localization approaches that leverage electrical and hemodynamic information for improved spatial specificity [75].

• Real-Time Applications: For neurofeedback and brain-computer interface applications, maintaining consistent spatial targeting across sessions is crucial, necessitating careful attention to probe placement reproducibility [75].

In conclusion, overcoming the spatial constraints of fNIRS through probabilistic registration and anatomical guidance enables more precise targeting of motor regions, enhancing the validity and reproducibility of motor task paradigms. By implementing these methodologies, researchers can strengthen the integration between fNIRS and fMRI, leveraging the complementary strengths of each modality to advance our understanding of motor function in both healthy and clinical populations.

The integration of functional magnetic resonance imaging (fMRI) and functional near-infrared spectroscopy (fNIRS) offers a powerful, multimodal approach for investigating brain function during motor task paradigms. This synergy capitalizes on fMRI's high spatial resolution for deep brain structures and fNIRS's superior temporal resolution and portability for cortical monitoring [14]. However, a significant challenge confronting both techniques, particularly in the context of naturalistic motor tasks, is the contamination of neural signals by physiological noise and systemic confounds. These nuisance signals originate from systemic physiology—including cardiac pulsation, respiration, and blood pressure oscillations (e.g., Mayer waves)—and from motion, which can be especially pronounced in movement-based studies [76]. For the robust estimation of evoked brain activity, it is crucial to reduce these confounding signals to isolate the hemodynamic response specific to neural activity effectively [76].

Physiological noise can manifest with several challenging characteristics: non-instantaneous and non-constant coupling between fNIRS channels and other modalities, pronounced correlation of physiological nuisance signals across measurement channels, and statistical dependencies of the underlying physiological processes regulated by the autonomous nervous system [76]. If unaddressed, these confounds can lead to inaccurate biomarkers and misinterpretations of brain-behavior relationships, as predictive models may capture these spurious effects instead of the neural features genuinely linked to the motor outcome of interest [77]. This application note provides detailed protocols and advanced algorithmic solutions for mitigating these confounds, framed within a multimodal fMRI-fNIRS research context.

Advanced Algorithmic Approaches for Deconfounding

The General Linear Model (GLM) with Short-Separation Regression

The current best practice for analyzing fNIRS signals involves using a General Linear Model (GLM) with short-separation (SS) regression [76]. This supervised approach simultaneously extracts the hemodynamic response function (HRF) while filtering confounding signals using nuisance regressors. Short-separation detectors (typically placed ~10 mm from a source) are used to measure scalp hemodynamics, which predominantly capture systemic physiological fluctuations. This scalp-only measurement is incorporated as a regressor in the GLM to remove the superficial contamination from the brain signal measured with standard-separation source-detector pairs.

Experimental Protocol: GLM with SS Regression

• Data Acquisition: Collect fNIRS data using a system equipped with both standard-length (e.g., 30 mm) and short-separation (8-12 mm) channels. Record an accelerometer synchronized with the fNIRS data to capture head motion.
• Preprocessing: Apply standard fNIRS preprocessing steps: converting raw light intensity to optical density, detecting and correcting motion artifacts (e.g., using wavelet-based or moving average techniques), and band-pass filtering (e.g., 0.01 - 0.5 Hz) to remove drift and high-frequency noise.
• Regressor Construction: Extract the data from the short-separation channels. This signal, SS(t), serves as a nuisance regressor to account for systemic scalp hemodynamics.
• GLM Design: Construct the design matrix for the GLM. The model for a single fNIRS channel is: Y(t) = β₀ + β₁*HRF(t) + β₂*SS(t) + ε(t) Where:
  • Y(t) is the preprocessed fNIRS signal (e.g., HbO concentration).
  • HRF(t) is the canonical hemodynamic response function convolved with the task paradigm (e.g., block or event-related design for motor tasks).
  • SS(t) is the short-separation regressor.
  • β₁ is the parameter of interest, representing the magnitude of the task-evoked HRF.
  • ε(t) is the error term.
• Model Estimation and Inference: Estimate the β parameters using ordinary least squares. Statistical significance of β₁ can be assessed via t-tests or F-tests, corrected for multiple comparisons across channels.

GLM with Temporally Embedded Canonical Correlation Analysis (GLM with tCCA)

Building upon the GLM with SS, a more advanced method incorporates Blind Source Separation (BSS) principles. The GLM with temporally embedded Canonical Correlation Analysis (tCCA) integrates the advantages of multimodality and temporal embedding into the conventional supervised GLM [76]. This method flexibly combines any number of auxiliary signals (short-separation fNIRS, accelerometers, physiological monitors) into optimal nuisance regressors by identifying underlying components that are maximally correlated with the physiological noise.

Experimental Protocol: GLM with tCCA

• Auxiliary Signal Acquisition: Collect a comprehensive set of auxiliary signals in addition to standard and short-separation fNIRS:
  • Accelerometer Data: 3-axis accelerometer data synchronized with fNIRS to capture motion.
  • Physiological Monitoring: Heart rate (via PPG), respiration (via belt), and blood pressure (if available) to capture systemic physiology.
• Temporal Embedding: For each auxiliary signal (e.g., accelerometer x-axis, short-separation signal), create a temporally embedded matrix by stacking time-lagged versions of the signal. The embedding dimension L (e.g., 5-10) must be optimized for the specific dataset and represents the number of time lags.
• tCCA Execution: Perform temporally embedded CCA between the temporally embedded auxiliary signal matrices and the standard fNIRS channel data. This identifies common temporal patterns that are maximally correlated between the nuisance sources and the fNIRS data.
• Nuisance Regressor Construction: The resulting canonical variates from the auxiliary signals serve as the optimized nuisance regressors. These regressors inherently account for non-instantaneous and non-constant coupling between the confounds and the brain signal.
• GLM Design and Estimation: Construct the GLM design matrix as: Y(t) = β₀ + β₁*HRF(t) + Σ(β_i * NV_i(t)) + ε(t) Where NV_i(t) are the nuisance regressors derived from the tCCA. Proceed with model estimation and inference as in the standard GLM.

The performance of GLM with tCCA has been quantitatively shown to significantly improve upon the GLM with SS, yielding markedly better results in the recovery of evoked HRFs across metrics like correlation, root mean squared error, and statistical power [76].

Diagram 1: GLM with tCCA workflow for advanced physiological noise regression.

Quantitative Performance Comparison of Deconfounding Algorithms

The efficacy of deconfounding algorithms must be quantitatively evaluated against a known ground truth. This is typically done by adding a synthetic hemodynamic response function (HRF) to resting-state fNIRS data and then assessing how accurately each algorithm can recover it in the presence of physiological noise.

Table 1: Performance Metrics for HRF Recovery Using Different Algorithms

Algorithm	Correlation (HbO)	Root Mean Squared Error (HbO)	F-Score (HbO)	Key Advantages	Key Limitations
GLM with Short-Separation Regression [76]	Baseline	Baseline	Baseline	Simple to implement; current best practice.	Less effective for non-instantaneous/correlated noise.
GLM with tCCA [76]	Max. +45%	Max. -55%	Up to 3.25x	Handles complex, multimodal noise; flexible auxiliary signal use.	More complex; requires parameter optimization (e.g., embedding dimension).

Experimental Protocol for Algorithm Validation

• Synthetic HRF Addition: Use a canonical HRF model (e.g., double-gamma function) convolved with a boxcar function representing the motor task (e.g., 20s movement, 40s rest). Add this synthetic HRF at a known amplitude to resting-state fNIRS data, which contains the true, inherent physiological noise.
• Algorithm Application: Process the combined (HRF + resting noise) data with the standard GLM+SS and the advanced GLM+tCCA algorithms.
• Performance Calculation: Compare the estimated HRF from each algorithm to the known, synthetic HRF using the following metrics calculated over the estimated HRF time course:
  • Correlation (Corr): Pearson's correlation coefficient between the estimated and true HRF.
  • Root Mean Squared Error (RMSE): Square root of the average squared differences between estimated and true HRF.
  • F-Score: Combines precision and recall in detecting significant activation.
• Statistical Comparison: Use paired statistical tests (e.g., paired t-test across multiple trials or participants) to determine if the performance improvements of one algorithm over another are significant [78].

The Scientist's Toolkit: Research Reagent Solutions

Table 2: Essential Materials for fNIRS-fMRI Motor Paradigm Studies

Item	Function & Specification	Application in Deconfounding
fNIRS System with Short-Separation Capability	Measures cortical hemodynamics (HbO, HbR). Must support integration of additional short-separation detector channels (~8-12 mm).	Essential for acquiring the primary brain signal and the key nuisance regressor (scalp hemodynamics) for GLM+SS and GLM+tCCA.
MRI-Compatible fNIRS Setup	fNIRS probes and fibers designed to operate safely inside the MRI scanner bore, resistant to electromagnetic interference [14].	Enables simultaneous fMRI-fNIRS data acquisition, crucial for validating fNIRS source localization with fMRI and developing multimodal biomarkers.
3-Axis Accelerometer	A small, lightweight sensor synchronized with the fNIRS system to record head motion acceleration in three dimensions.	Provides critical data for motion artifact detection and serves as an essential auxiliary input for the GLM+tCCA algorithm to model motion-induced physiological noise.
Physiological Monitoring System	Includes photoplethysmography (PPG) for heart rate, respiratory belt for respiration, and continuous blood pressure monitoring if available.	Provides direct measurements of systemic physiological processes that are major sources of confounding noise; used as auxiliary inputs in GLM+tCCA.
Digitization System	A 3D digitizer (e.g., Polhemus) or photogrammetry system to record the precise locations of fNIRS optodes on the scalp.	Critical for accurate source localization and co-registration with anatomical MRI. Improves reproducibility and spatial accuracy of fNIRS findings [20].

Integrated Experimental Protocol for a Motor Task Study

This protocol outlines the steps for a combined fMRI-fNIRS study investigating hand motor function, incorporating advanced deconfounding.

• Participant Setup & Digitization:

  • Place the MRI-compatible fNIRS cap on the participant, ensuring coverage of the primary motor cortex.
  • Attach all optodes (sources, standard detectors, and short-separation detectors).
  • Securely attach the accelerometer to the fNIRS cap.
  • Attach physiological monitors (PPG, respiration belt).
  • Using a 3D digitizer, record the positions of fNIRS optodes and fiducial landmarks (nasion, left/right pre-auricular points) relative to the participant's head.

• Simultaneous Data Acquisition:

  • Position the participant in the MRI scanner.
  • fMRI Parameters: Acquire T1-weighted anatomical scan. For functional scans, use a standard BOLD fMRI sequence (e.g., gradient-echo EPI, TR=2s, voxel size=3x3x3 mm³).
  • fNIRS Parameters: Start fNIRS recording with a sampling rate ≥ 10 Hz. Ensure synchronization pulses are sent from the MRI scanner to the fNIRS system to mark the onset of each volume acquisition and any gradient artifacts for post-hoc correction.
  • Task Paradigm: Execute a block-design motor task (e.g., 30s rest, 30s repeated finger tapping, 5 cycles). Use a visual cue to instruct the participant when to move and rest.

• Data Preprocessing:

  • fNIRS Preprocessing: Convert light intensity to optical density, then to HbO/HbR concentrations using the Modified Beer-Lambert Law. Perform motion artifact correction. Apply a band-pass filter (e.g., 0.01 - 0.5 Hz).
  • fMRI Preprocessing: Perform standard preprocessing steps including slice-time correction, motion realignment, co-registration to the anatomical scan, normalization to standard space (e.g., MNI), and spatial smoothing.

• Deconfounding Analysis (fNIRS):

  • Implement GLM with tCCA: Follow the protocol in Section 2.2, using the short-separation channels, accelerometer data, and physiological recordings (PPG, respiration) as auxiliary inputs to the tCCA.
  • Generate Statistical Maps: For each fNIRS channel, obtain the beta value and p-value for the task HRF regressor. Create statistical parametric maps of brain activation.

• Data Fusion & Validation:

  • Co-registration: Use the digitized optode positions to project the fNIRS statistical maps onto the individual's anatomical MRI or a standard brain template.
  • Comparison: Compare the spatial location and extent of activation in the deconfounded fNIRS map with the activation map derived from the simultaneously acquired BOLD fMRI data. This serves as a robust validation of the fNIRS deconfounding pipeline.

Diagram 2: Integrated experimental protocol for a motor task study.

Functional neuroimaging techniques, particularly functional Magnetic Resonance Imaging (fMRI) and functional Near-Infrared Spectroscopy (fNIRS), are indispensable for investigating brain function during motor tasks. However, their application in paradigms involving naturalistic movement is severely constrained by motion artifacts, which introduce noise that can obscure genuine neural signals and lead to spurious conclusions [79] [80]. This challenge is especially pertinent in the context of integrating fMRI and fNIRS data, where consistent artifact mitigation across modalities is crucial for valid data fusion and interpretation. Motion artifacts manifest differently across modalities; in fNIRS, they often appear as high-frequency spikes and baseline shifts due to optode-scalp decoupling [79] [81], while in fMRI, they cause image misalignment and spin-history effects [82] [83]. This application note details advanced hybrid correction techniques, providing a structured framework to overcome these barriers and enable robust neuroimaging in ecologically valid movement paradigms.

Core Principles of Motion Artifacts and Hybrid Correction

Characterization of Motion Artifacts

A practical understanding of motion artifact typology is the first step toward effective correction. The following table classifies common artifacts and their features.

Table 1: Classification and Features of Motion Artifacts in Neuroimaging

Modality	Artifact Type	Temporal/ Spatial Profile	Primary Cause
fNIRS	High-Frequency Spike	Short-duration, large amplitude	Rapid head movement, poor optode contact [79]
fNIRS	Baseline Shift (BS)	Sustained signal drift	Slow head rotation, optode resettling [79] [81]
fNIRS	Slow Oscillation	Low-frequency signal variation	Body movements, speaking [80]
fMRI	Spin History Effect	Signal loss/gain in slices	Through-slice movement during acquisition [82]
fMRI	Image Misalignment	Volumetric displacement between scans	Head translation and rotation [83]

The hybrid correction philosophy is predicated on the recognition that no single algorithm can optimally address all artifact types. Each method has inherent strengths and weaknesses; for instance, wavelet-based methods excel at suppressing spikes but are less effective against baseline shifts, while spline interpolation effectively models and removes baseline shifts but may leave high-frequency spikes untouched [79] [81]. Therefore, a sequential, hybrid approach that combines complementary techniques yields superior outcomes compared to any method used in isolation [84].

The Hybrid Workflow Logic

The following diagram illustrates the logical decision-making workflow for applying a hybrid motion artifact correction strategy, from data input to final output.

Quantitative Comparison of Motion Correction Techniques

Evaluating the performance of different algorithms is essential for selecting an appropriate method. The following tables summarize key findings from comparative studies.

Table 2: Performance Summary of fNIRS Motion Correction Techniques on Pediatric Data [85]

Correction Method	Key Performance Insight	Notable Advantage	Notable Limitation
Moving Average (MA)	Ranked among the best outcomes for pediatric data	Effective simplicity	May not handle complex artifacts [85]
Wavelet Filtering	Ranked among the best outcomes for pediatric data	Effective for spike removal [79]	Can exacerbate baseline shifts [79]
Spline Interpolation	Produced greatest improvement in mean-squared error [79]	Excellent for baseline shift correction [81]	Poor performance on high-frequency spikes [79]
Spline + Wavelet (Hybrid)	Outperformed individual use on infant data [84]	Combats complex artifact profiles; saves corrupted trials [84]	Increased computational complexity

Table 3: Quantitative Metrics for Hybrid Method Performance Evaluation [79] [81]

Evaluation Metric	Definition and Purpose	Hybrid Method Performance
Signal-to-Noise Ratio (SNR)	Measures the level of desired signal relative to noise.	Shows significant improvements post-correction [79]
Pearson's Correlation (R)	Quantifies similarity to a ground-truth signal.	Strong stability and high correlation values [79]
Mean-Squared Error (MSE)	Measures the average squared difference between estimated and true values.	Spline interpolation provides greatest MSE improvement [79]
Hemodynamic Response Function (HRF) Recovery	Ability to accurately recover the shape and amplitude of the HRF.	Hybrid Spline-SG provides reasonable improvements [81]

Detailed Experimental Protocols

Protocol A: Hybrid Motion Artifact Correction for fNIRS Data

This protocol outlines the application of a proven hybrid method combining spline interpolation and wavelet filtering, which has demonstrated efficacy on real and semi-simulated data, including from infants [84].

Step 1: Data Preprocessing and Conversion

• Convert raw light intensity signals (ΔODλ1, ΔODλ2) to optical density units [79].
• Extract hemodynamic signals. Calculate concentration changes for oxy-hemoglobin (Δ[HbO2]) and deoxy-hemoglobin (Δ[Hb]) using the Modified Lambert-Beer Law [79]: Δ[HbO2] = (α1 * ΔODλ1 + α2 * ΔODλ2) / L Δ[Hb] = (β1 * ΔODλ1 + β2 * ΔODλ2) / L where L is the photon pathlength, and α1, α2, β1, β2 are wavelength-specific coefficients [79].

Step 2: Automated Motion Artifact Detection

• Calculate the moving standard deviation t(n) of the optical density or hemodynamic signal using a sliding window (e.g., W = 2k+1, where k = 3 * FsNIRS) [79].
• Generate a dynamic threshold by analyzing the distribution of the moving standard deviation to objectively identify outlier segments contaminated by motion [79].

Step 3: Categorization and Targeted Correction

• Categorize detected artifacts into baseline shifts (BS), slight oscillations, and severe oscillations [79].
• Correct Severe Artifacts & BS: Use cubic spline interpolation to model and subtract the motion component from the identified artifact segments [79] [84].
• Correct Slight Oscillations & Residual Spikes: Apply a dual-threshold wavelet-based filtering method to the corrected signal from the previous step to remove high-frequency noise without distorting the baseline [79] [84].

Step 4: Final Filtering

• Apply a high-pass filter (e.g., cut-off at 0.01-0.02 Hz) to remove any remaining very low-frequency drifts and prepare the signal for statistical analysis [79].

Protocol B: Motion Artifact Correction for Multi-Echo fMRI Data

This protocol integrates RETROICOR, a widely used method for physiological noise correction, with multi-echo acquisition to enhance data quality in movement paradigms [82].

Step 1: Data Acquisition

• Acquire fMRI data using a multi-echo (ME) sequence with multiple echo times (TEs), e.g., 17.00 ms, 34.64 ms, and 52.28 ms [82].
• Simultaneously record physiological data (cardiac pulsation and respiration) using pulse oximetry and a respiratory belt.

Step 2: RETROICOR Implementation

• Two viable strategies exist:
  • RTCind: Apply RETROICOR to each individual echo time series before combining them into a final time series [82].
  • RTCcomp: First combine the multi-echo data into a composite time series (e.g., using T2* weighting), then apply RETROICOR to the composite data [82].
• Model the physiological noise in the fMRI signal using retrospective timing from the recorded cardiac and respiratory phases [82].
• Regress out the modeled noise from the fMRI signal.

Step 3: Data Combination and Quality Assessment

• If using the RTC_ind approach, combine the denoised individual echoes.
• Calculate quality metrics such as the temporal Signal-to-Noise Ratio (tSNR) and Signal Fluctuation Sensitivity (SFS) to evaluate the efficacy of the correction. Studies show RETROICOR improves these metrics, particularly in moderately accelerated acquisitions [82].

The Scientist's Toolkit: Research Reagent Solutions

Table 4: Essential Tools for Motion-Resilient Neuroimaging Research

Tool / Solution	Function / Description	Application Note
Homer2 Software Package	A comprehensive MATLAB-based toolbox for fNIRS data processing.	Implements various motion correction algorithms (e.g., spline, wavelet, PCA) and is widely used in the field [85].
Accelerometer / Inertial Measurement Unit (IMU)	A hardware sensor attached to the participant's head or optode holder to measure motion.	Provides a reference signal for methods like Adaptive Filtering or ABAMAR, improving motion artifact identification [80].
Computer Vision (e.g., SynergyNet)	A deep neural network for frame-by-frame analysis of video recordings to compute head orientation.	Provides ground-truth movement data without physical contact, useful for characterizing and validating artifacts [86].
Retrospective Image Correction (RETROICOR)	A model-based algorithm for removing cardiac and respiratory fluctuations from fMRI data.	Requires concurrent physiological recording; effective in multi-echo fMRI at improving tSNR [82].
Collodion-Fixed Optical Fibers	A method to improve optode-scalp coupling using a fast-drying adhesive.	Physically reduces the occurrence of motion artifacts by enhancing stability, but requires careful application [79].

Signaling Pathways and Workflow Integration

For a multi-modal study integrating fNIRS and fMRI, the data streams must be processed through parallel yet complementary artifact correction pipelines before integration. The following diagram outlines this integrated workflow.

The integration of fMRI and fNIRS for motor task paradigms presents a powerful approach to understanding brain function, but its success is contingent on effectively addressing the challenge of motion artifacts. The hybrid techniques detailed in this document—particularly the sequential application of spline interpolation and wavelet filtering for fNIRS, and the use of RETROICOR with multi-echo acquisition for fMRI—provide a robust methodological foundation. By adopting these structured protocols and utilizing the outlined toolkit, researchers and drug development professionals can significantly enhance data quality, thereby unlocking the potential of neuroimaging in naturalistic movement paradigms and yielding more reliable, translatable findings.

Functional Near-Infrared Spectroscopy (fNIRS) presents a promising complement to functional Magnetic Resonance Imaging (fMRI) in multimodal neuroimaging studies, particularly for motor task paradigms. Its portability, cost-efficiency, and tolerance for motion artifacts enable research in naturalistic settings that would be impractical in an MRI scanner [14]. However, the optical nature of fNIRS measurements makes signal quality particularly vulnerable to individual differences in biophysical factors including hair characteristics, skin pigmentation, and skull thickness [87] [88] [89]. If unaddressed, these factors risk biasing research findings by disproportionately affecting data quality across diverse populations [6] [90], ultimately compromising the validity of correlations with fMRI's blood-oxygen-level-dependent (BOLD) signal.

This application note provides evidence-based strategies to mitigate these challenges, ensuring high-quality fNIRS data collection across diverse participants. By implementing these protocols, researchers can enhance the inclusivity and reliability of their fNIRS-fMRI studies, particularly in motor task research where robust signal detection is paramount for cross-modal validation.

Quantitative Impact of Biophysical Factors on fNIRS Signals

Understanding the specific impacts of various biophysical factors is essential for developing effective mitigation strategies. Recent large-scale studies have quantified how these characteristics affect fNIRS signal quality.

Table 1: Quantitative Impact of Biophysical Factors on fNIRS Signal Quality

Factor	Specific Characteristic	Impact on fNIRS Signal	Evidence Source
Hair	Density & Darkness	Increased light absorption, reduced penetrating/reflected light [90]	Yücel et al., 2025 [88]
	Type (curly, kinky)	Interferes with optode-scalp coupling [90]	Yücel et al., 2025 [88]
Skin	Pigmentation (Melanin Index)	Higher absorption of near-infrared light [88] [90]	Yücel et al., 2025 [88]
Head Anatomy	Scalp & Skull Thickness	Reduced sensitivity to cortical brain regions [91]	Cooper et al., 2015 [91]
	Sex & Age	Correlated with structural changes affecting light propagation [88]	Yücel et al., 2025 [88]

A study examining stroke survivors highlighted the real-world consequences of these factors, finding that "fNIRS signals were generally worse for Black women compared to Black men and White individuals regardless of gender" [6]. This underscores the urgent need for standardized protocols that address these intersecting biophysical characteristics to ensure equity in neuroimaging research.

Experimental Protocols for Signal Quality Assurance

Comprehensive Pre-Data Collection Protocol

Participant Characterization and Metadata Collection Standardized characterization of participant factors enables both proactive optimization and post-hoc analysis of data quality. Researchers should collect the following metadata for each participant:

• Skin Pigmentation: Quantify using a melanometer to determine the Melanin Index at standard cranial locations (Fp1, Fp2, C3, C4, etc.) [90]
• Hair Characteristics: Document hair type (straight, wavy, curly, kinky), color, density, and thickness using standardized classification systems and trichoscopy imaging where available [88] [90]
• Cranial Anatomy: Measure head circumference and document age and sex, as these factors correlate with scalp-to-cortex distance and skull thickness [88] [91]

Table 2: Essential Research Reagents and Equipment

Item	Function/Application	Specifications/Notes
Melanometer	Quantifies skin pigmentation (Melanin Index)	Critical for objective measurement beyond visual categorization [90]
3D-Printed Headgear	Holds optodes in place; accommodates varied hairstyles	e.g., NinjaCap; customizable for individual anatomy [90]
Short-Separation Detectors	Measures superficial hemodynamics for signal regression	Optimal distance: 8.4mm for adults [91]
Cotton-Tipped Applicators	Parts hair and applies gel during cap adjustment	For gentle hair management under optodes [90]
Ultrasound Gel	Ensures optical coupling between optode and scalp	Preferred over other gels for stable coupling [90]

Cap Selection and Preparation

• Utilize customizable, flexible headgear such as 3D-printed NinjaCaps that can adapt to various head sizes and shapes [90]
• Select appropriate cap size based on head circumference measurements (e.g., 55cm or 57cm) [90]
• For diverse hair types, consider caps with larger grommets or adaptable designs that accommodate various hair volumes and structures

Cap Placement and Optode Coupling Protocol

The following workflow details the optimized procedure for cap placement and signal optimization, particularly critical for participants with dense or curly hair:

Key Technical Considerations:

• Directionality: Always place the cap from front to back to prevent hair from accumulating under optodes [90]
• Stabilization: Use Velcro attachments and cable management arms to stabilize the weight of optode bundles, reducing pressure on the participant's head [90]
• Hair Management: For dense or curly hair, use cotton-tipped applicators to gently part hair and create clear paths for optodes. Temporarily remove optodes from grommets if necessary to apply ultrasound gel directly to the scalp [90]
• Environmental Optimization: Reduce ambient light interference by turning off pulse-wave modulated LED lights, using incandescent floor lamps instead, and covering the fNIRS cap with an opaque shower cap [90]

Data Acquisition and Quality Control Protocol

Signal Quality Assessment

• Implement the Signal Optimization function in acquisition software (e.g., Aurora for NIRSport2) both after fast capping and after proper capping to quantify improvement [90]
• Calculate scalp-coupling indices and peak spectral power values using standardized toolboxes like QT-NIRS (Quality Testing of Near Infrared Scans) [6]
• Establish minimum signal quality thresholds for inclusion (e.g., signal-to-noise ratio >15 dB) [10]

Short-Separation Channel Configuration

• Incorporate short-separation detectors (SSD) with optimal source-detector distances:
  • Adults: 8.4 mm source-detector distance [91]
  • Infants: 2.15 mm source-detector distance [91]
• Distribute SSDs evenly throughout the optode array, particularly in regions prone to high superficial interference [10]
• Use SSDs to measure and regress superficial hemodynamic components from standard fNIRS signals, isolating cortical functional responses [91]

Integration with fMRI Motor Task Paradigms

Methodological Synergies for Motor Task Research

The combination of fNIRS and fMRI leverages their complementary strengths for comprehensive motor function assessment. fNIRS provides superior temporal resolution and motion tolerance for capturing rapid motor sequences, while fMRI offers high spatial resolution for precise localization of motor network activity [14] [10].

Spatial Correspondence Validation Studies have demonstrated strong spatial correspondence between fNIRS and fMRI hemodynamic responses in motor regions. A multimodal investigation found that "group-level activation was found in fMRI data modeled from corresponding fNIRS measurements, with significant peak activation found overlapping the individually-defined primary and premotor motor cortices" [10]. This validates fNIRS as a reliable tool for motor cortex assessment, particularly when integrated with fMRI.

Protocol Standardization for Cross-Modal Comparison

• Implement asynchronous fMRI-fNIRS data acquisition when simultaneous collection is impractical, using identical motor task paradigms (e.g., bilateral finger tapping sequences) [10]
• Ensure consistent block designs (e.g., 30-second blocks for motor execution, motor imagery, and baseline) across both modalities [10]
• Use individual anatomical landmarks from fMRI to inform fNIRS optode placement, enhancing spatial accuracy [10]

Data Processing and Fusion Strategies

Advanced Signal Processing

• Apply motion correction algorithms tailored to fNIRS data, particularly important for motor tasks involving physical movement [10]
• Implement component-based noise correction methods (e.g., using short-separation channels to regress superficial physiological interference) [91]
• Utilize standardized preprocessing pipelines such as Homer3 for fNIRS data to enhance reproducibility [10]

Multimodal Data Integration

• Coregister fNIRS channels with anatomical MRI data using digitized optode positions [20] [10]
• Adopt the Brain Imaging Data Structure (BIDS) extension for fNIRS (NIRS-BIDS) to standardize data organization and facilitate sharing and replication [88]
• Employ statistical models that account for the differential sensitivities of fNIRS and fMRI to various aspects of the hemodynamic response

Ensuring fNIRS signal quality across diverse populations requires systematic attention to biophysical factors including hair characteristics, skin pigmentation, and cranial anatomy. The protocols outlined in this application note provide actionable strategies to mitigate these influences, thereby enhancing the inclusivity and reliability of fNIRS research, particularly in multimodal studies with fMRI.

Future advancements should focus on developing more inclusive hardware designs, such as optodes that better accommodate various hair types and textures, and algorithmic approaches that automatically compensate for signal quality variations related to skin pigmentation. Furthermore, adopting standardized reporting of participant metadata, as recommended by Yücel et al. [88] [89], will enable meta-analyses that further elucidate the impact of biophysical factors on fNIRS signal quality across larger, more diverse populations.

By implementing these evidence-based protocols, researchers can minimize systematic biases in fNIRS data collection, thereby strengthening the validity and generalizability of findings in motor task paradigms and beyond. This approach ultimately fosters more equitable and inclusive neuroimaging research that better represents the full diversity of human populations.

Functional near-infrared spectroscopy (fNIRS) has emerged as a powerful, non-invasive neuroimaging tool that measures cerebral hemodynamic activity by quantifying changes in oxygenated hemoglobin (HbO), deoxygenated hemoglobin (HbR), and total hemoglobin (HbT). Within the broader thesis of integrating functional magnetic resonance imaging (fMRI) and fNIRS for motor task paradigms, optimizing chromophore selection is paramount for data interpretation validity. While fMRI provides high spatial resolution for deep brain structures, fNIRS offers superior temporal resolution, portability, and motion artifact resistance, making it ideal for naturalistic motor studies [14]. This protocol provides a structured framework for selecting the most informative chromophore (HbO, HbR, or HbT) based on specific experimental contexts, with a focus on motor task research.

The neurophysiological basis of fNIRS centers on neurovascular coupling, where neural activation triggers a hemodynamic response characterized by increased regional cerebral blood flow. This leads to an increase in HbO and a corresponding decrease in HbR at the activation site [92]. HbT, representing the sum of HbO and HbR, reflects total blood volume changes. Each chromophore offers distinct advantages and limitations for detecting brain activity, influenced by factors such as signal-to-noise ratio, sensitivity to systemic artifacts, and correlation with the fMRI blood-oxygen-level-dependent (BOLD) signal [10]. A critical challenge in fNIRS is its contamination by systemic physiological artifacts, making proper correction essential for valid interpretation [93].

Performance Characteristics of fNIRS Chromophores

Table 1: Comparative characteristics of fNIRS chromophores for motor task paradigms.

Metric	HbO	HbR	HbT
Typical Response to Neural Activation	Increase	Decrease	Increase
Amplitude	High	Low	Intermediate
Signal-to-Noise Ratio	Generally higher [20]	Generally lower [20]	Intermediate
Reproducibility	More reproducible over sessions [20]	Less reproducible [20]	Not specified
Sensitivity to Systemic Artifacts	High	High	High
Spatial Specificity	Good	Potentially higher	Good
Correlation with fMRI BOLD	Variable (r = 0 to 0.8) [10]	Negative correlation (theoretical basis) [10]	High correlation in some studies [10]
Best Use Cases	Primary metric for BCI/Neurofeedback [94], General activation detection	Confirming hemodynamic response pattern, Quality check	When a composite measure is desired

Experimental Protocols for Chromophore Assessment

Protocol: Systematic Comparison of Chromophores in a Motor Paradigm

Objective: To empirically determine the optimal chromophore for a specific motor task by simultaneously assessing HbO, HbR, and HbT signals against performance metrics and gold-standard neuroimaging.

Materials:

• fNIRS system (CW, FD, or TD) with full motor cortex coverage
• Short-distance channels (SDCs) for systemic artifact correction [93]
• EEG cap modified for fNIRS optode placement (if applicable)
• Stimulus presentation software
• Response recording device (e.g., button box, force transducer)
• Data processing software (e.g., Homer3, NIRSlab)

Procedure:

• Participant Preparation & Setup:
  • Recruit right-handed participants with no neurological history.
  • Position optodes over bilateral motor cortices (C3/C4 locations per 10-20 system), ensuring source-detector distances of 30 mm for long channels and 8 mm for short-distance channels [93] [10].
  • Use a digitization pen to record precise optode locations relative to cranial landmarks for coregistration with MRI space.

• Experimental Task (Blocked Design):

  • Baseline (30 s): Participants fixate on a cross-hair.
  • Motor Execution (20 s): Perform self-paced sequential finger tapping (left hand, right hand) [10].
  • Motor Imagery (20 s): Imagine the same finger tapping sequence without movement.
  • Repeat for 6-8 blocks per condition, counterbalancing order across participants.

• fNIRS Data Acquisition:

  • Acquire data at a minimum sampling rate of 10 Hz.
  • Record accelerometer data if available to monitor head movements.

• Data Preprocessing (Homer3/NIRSlab):

  • Convert raw intensity to optical density.
  • Identify and correct motion artifacts (e.g., using wavelet-based methods or Savitzky-Golay filtering) [95].
  • Apply band-pass filter (0.01 - 0.2 Hz) to remove cardiac and respiratory oscillations [92].
  • Convert optical density to hemoglobin concentration changes using the Modified Beer-Lambert Law (MBLL) with appropriate differential pathlength factors (DPF) [92].
  • Critical Step: Perform systemic artifact correction using short-distance channel regression [93].

• Data Analysis:

  • General Linear Model (GLM): Model the hemodynamic response for each chromophore and condition. Use canonical HRF and derivatives as regressors.
  • Contrast Analysis: Calculate contrasts (e.g., Motor Execution > Baseline) for each chromophore.
  • Signal Quality Metrics: Calculate contrast-to-noise ratio (CNR) and effect size for each chromophore.
  • Spatial Specificity: Compare activation maps with known motor cortex topography.
  • Test-Retest Reliability: If multiple sessions are available, calculate intraclass correlation coefficients (ICCs) for each chromophore's activation strength [20] [96].

Protocol: Validating fNIRS Chromophores Against fMRI

Objective: To establish the spatial correspondence between fNIRS chromophores and the fMRI BOLD signal in motor tasks.

Materials:

• fNIRS system compatible with MRI environment
• 3T MRI scanner
• Custom fNIRS caps for simultaneous acquisition

Procedure:

• Asynchronous Scanning: Acquire fMRI and fNIRS data in separate sessions using identical motor paradigms [10].
• Synchronous Scanning (if hardware allows): Acquire both modalities simultaneously.
• Coregistration: Map fNIRS channels to cortical surface using individual T1-weighted anatomical scans and digitized optode positions.
• Analysis: Use subject-specific fNIRS signals (HbO, HbR, HbT) as predictors in fMRI GLM analysis. Identify regions where fNIRS signals significantly predict BOLD activity [10].

Table 2: Key reagents and materials for fNIRS-fMRI motor studies.

Category	Specific Item/Technique	Function/Application
fNIRS Hardware	Short-Distance Channels (SDCs)	Measures extracerebral systemic activity for signal correction [93]
	Time-Domain (TD) fNIRS	Provides superior depth resolution and quantification of hemoglobin [96]
Software & Analysis	General Linear Model (GLM) with HRF	Models task-related hemodynamic responses for each chromophore [93] [10]
	Systemic Artifact Correction Algorithms	Removes confounding physiological signals from fNIRS data [93]
	Homer3 / NIRSlab	Open-source software for fNIRS data preprocessing and analysis [95]
Experimental Materials	Motor Paradigm Tasks	Finger tapping, motor imagery to elicit robust cortical activation [10] [94]
	Digitization System	Records precise 3D optode positions for MRI coregistration [10]

Application Notes and Decision Framework

When HbO is Superior:

• Brain-Computer Interfaces (BCIs) and Neurofeedback: HbO's higher amplitude and signal-to-noise ratio make it the most reliable signal for single-trial classification and real-time applications [94].
• Maximizing Reproducibility: For studies requiring high test-retest reliability across multiple sessions (e.g., longitudinal intervention studies), HbO demonstrates significantly higher reproducibility than HbR [20] [96].
• Initial Learning and Cognitive Effort Assessment: In educational or cognitive training contexts, HbO from the prefrontal cortex is effective for estimating cognitive effort and neural efficiency when combined with performance scores [97].

When HbR is Superior:

• Confirming Canonical Hemodynamic Response: A simultaneous HbO increase and HbR decrease provides strong evidence for true neurovascular coupling, reducing false positives from systemic artifacts.
• Theoretical Link to fMRI BOLD: The BOLD signal is more directly influenced by HbR concentrations, making HbR valuable for biophysical models linking fNIRS and fMRI [10].

When HbT is Superior:

• Robustness to Hemodynamic Variations: HbT is less affected by changes in oxygen extraction fraction and may provide a more stable measure of regional cerebral blood volume.
• High Spatial Correspondence with fMRI: Some studies report that HbT shows the highest spatial correlation with fMRI BOLD signals, making it advantageous for multimodal integration [10].

Critical Considerations:

• Systemic Artifact Correction: The superiority of any chromophore is contingent on effective removal of systemic confounds. Short-distance channels are the gold standard for this correction and significantly improve signal quality for all chromophores, though HbO may benefit most [93].
• Optode Stability: Changes in optode placement between sessions significantly reduce spatial overlap and reproducibility. Use consistent mounting and digitization to mitigate this [20].
• Population-Specific Responses: Chromophore performance may vary in clinical populations with altered neurovascular coupling (e.g., stroke, aging). Pilot testing in the target population is recommended.

Establishing Validity and Utility: Comparative Studies and Clinical Validation

Functional magnetic resonance imaging (fMRI) and functional near-infrared spectroscopy (fNIRS) have become cornerstone technologies in cognitive neuroscience and clinical research for mapping brain function. While both modalities measure hemodynamic responses correlated with neural activity, they possess distinct and complementary strengths and limitations [14]. fMRI is celebrated for its high spatial resolution and whole-brain coverage, including deep structures, whereas fNIRS offers superior temporal resolution, portability, and a higher tolerance for motion [98]. This combination is particularly powerful for studying motor paradigms in naturalistic settings and in populations unsuitable for the restrictive fMRI environment [14] [98].

The core of this application note focuses on a critical question for researchers: how do the activation maps generated by fNIRS quantitatively compare to the gold-standard maps from fMRI in terms of spatial specificity (the accuracy in localizing brain activity) and task sensitivity (the ability to detect changes in activity due to a task)? Understanding this relationship is fundamental for designing robust experiments, especially those aimed at translating validated fMRI paradigms to more flexible fNIRS setups, such as in neurorehabilitation or drug development studies [10] [16]. We frame this discussion within the broader thesis that the integration of fMRI and fNIRS is not redundant but synergistic, enabling a more comprehensive understanding of brain function by leveraging the spatial detail of fMRI with the ecological validity of fNIRS [14].

Quantitative Data Comparison

A synthesis of recent comparative studies reveals a consistent and quantifiable relationship between fNIRS and fMRI-derived activation maps. The following tables summarize key quantitative findings on their spatial correspondence and task sensitivity across various motor tasks.

Table 1: Spatial Correspondence Between fNIRS and fMRI

Study & Task	fNIRS Chromophore	Spatial Correlation / Overlap Findings	Notes
Multimodal Assessment (2023) [10]Motor Execution & Imagery	HbO, HbR, HbT	- Group-level fMRI activation identified using fNIRS signals as predictors.- Significant peak activation overlapping individually-defined primary (M1) and premotor (PMC) cortices.	No statistically significant differences in spatial correspondence between HbO, HbR, and HbT.
fMRI-based Validation (2022) [16]Motor Execution	HbO and HbR	- Significant topographical similarity (Spearman correlation) between fNIRS channels and corresponding fMRI voxels for motor execution tasks.	Spatial specificity for Motor Imagery was more variable than for Motor Execution.
Quantitative Comparison (2002) [99] [27]Motor Task	HbO and HbR	- Highly variable initial correlations.- After correcting for systematic errors, strong correlations were found with all optical measures, with HbO providing the strongest correlation with the BOLD signal.	Suggested variability stems from individual differences and systematic errors in NIRS.

Table 2: Task Sensitivity and Chromophore Performance

Metric	Findings	Implications for Protocol Design
Signal-to-Noise Ratio (SNR)	fNIRS signals have a significantly weaker SNR compared to fMRI, which can be a limiting factor in regions with a greater scalp-to-cortex distance [25].	Studies require careful power analysis; tasks with robust hemodynamic responses (e.g., motor execution) are more reliably detected.
Chromophore Sensitivity	In motor imagery tasks, particularly for whole-body movements, HbR may be the more specific signal compared to HbO [16]. Other studies report HbO often has a higher contrast-to-noise ratio [99] [25].	The choice of chromophore (HbO, HbR, or both) for primary analysis should be task-dependent. HbO is often more robust, but HbR may offer superior specificity.
Temporal Correlation	High temporal correlations between fNIRS and fMRI signals are frequently reported, though they vary widely (from	r	~0.2 to >0.8) depending on the task, region, and signal processing pipeline [10] [25].	Ensuring high-quality data acquisition and preprocessing is critical for achieving reliable temporal correspondence.

Experimental Protocols

This section outlines detailed methodologies for a representative experiment that quantitatively compares fNIRS and fMRI activation during a motor task paradigm.

Protocol: Multimodal Assessment of Motor Execution and Imagery

Objective: To validate the spatial correspondence and task sensitivity of fNIRS-derived hemodynamic signals against fMRI BOLD responses in primary motor (M1) and premotor (PMC) cortices [10].

Participant Preparation:

• Recruitment: Recruit healthy adult volunteers with no history of neurological disorders. Obtain informed consent and ethical approval from the local institutional review board.
• Instructions: Train participants on the motor tasks (execution and imagery) outside the scanner to ensure compliance and understanding. For motor imagery, instruct participants to mentally simulate the kinesthetic sensation of the movement without producing any overt motion [16].

Stimulus Presentation:

• Software: Use a stimulus presentation software (e.g., E-Prime, PsychoPy) synchronized with both scanners.
• Paradigm: Employ a block design. A sample structure is provided in the workflow diagram below, consisting of 17 blocks (9 Baseline, 4 Motor Action (MA), 4 Motor Imagery (MI)), each lasting 30 seconds [10].

Data Acquisition:

• fMRI Parameters:
  • Scanner: 3T Siemens Magnetom Trio.
  • Sequence: Gradient-echo EPI.
  • Parameters: TR/TE = 1500/30 ms, voxel size = 3×3×3.5 mm, 26 slices covering motor areas.
  • High-resolution anatomical scan: MPRAGE, 1×1×1 mm voxels [10].
• fNIRS Parameters:
  • System: Portable continuous-wave (CW) system (e.g., NIRSport2).
  • Setup: 16 sources (760 & 850 nm), 15 detectors, 54 channels over bilateral motor areas. Intra-optode distance: 30 mm.
  • Include short-distance detectors (8 mm) to mitigate extracerebral confounds [10].
  • Sampling Rate: 5.08 Hz or higher.

The following workflow diagram illustrates the parallel data processing streams for the acquired fMRI and fNIRS data:

Data Preprocessing and Analysis:

• fMRI Preprocessing: Perform using software like BrainVoyager QX or SPM. Steps include slice timing correction, 3D motion correction, temporal high-pass filtering, co-registration to anatomical data, spatial smoothing (e.g., 6mm FWHM Gaussian kernel), and normalization to standard space (e.g., Talairach) [10] [100].
• fNIRS Preprocessing: Perform using tools like Homer3 or NIRS-KIT. Steps include:
  • Channel Pruning: Remove channels with a signal-to-noise ratio (SNR) < 15 dB [10].
  • Conversion: Convert raw intensity to optical density, then to concentration changes of HbO and HbR using the Modified Beer-Lambert Law (mBLL) [101].
  • GLM Analysis: Model the fNIRS data using a GLM with the task conditions (MA, MI, Baseline) as predictors of interest to generate beta maps for each chromophore [10].
• Multimodal Comparison:
  • Coregistration: Coregister fNIRS optode locations to the participant's anatomical MRI or a standard head model to identify the corresponding cortical regions for each channel [101] [16].
  • Spatial Analysis: Extract the fMRI BOLD response from the cortical regions underlying the fNIRS channels. Use Spearman correlation to compare the topographical maps of fNIRS beta values and fMRI t-statistics [16].
  • Statistical Testing: Compare task sensitivity by evaluating the significance of activation in the ROIs for both modalities and for the different chromophores.

The Scientist's Toolkit: Essential Research Reagents & Materials

The following table lists key materials and tools essential for conducting rigorous multimodal fMRI-fNIRS studies.

Table 3: Essential Research Reagents and Solutions

Item Name	Specification / Example	Critical Function in Protocol
High-Density fNIRS System	Continuous-Wave (CW) system, e.g., NIRSport2 (NIRx) with 16+ sources and 15+ detectors.	Provides the hardware platform for measuring cortical hemodynamics with coverage over targeted motor regions.
3T MRI Scanner	Siemens Magnetom Trio, Prisma, or equivalent.	The gold-standard platform for acquiring high-resolution anatomical and functional BOLD data for validation.
Short-Distance Detectors	fNIRS detectors placed at 8-10 mm from a source.	Crucial for measuring and subsequently regressing out systemic physiological noise from scalp and skull, improving brain signal specificity [10].
fNIRS Processing Software	Homer3, NIRS-KIT, or AtlasViewer.	Provides a standardized pipeline for converting raw fNIRS data into analyzed hemodynamic responses, including SNR checking, GLM, and coregistration.
fMRI Processing Software	BrainVoyager QX, SPM, FSL, or AFNI.	Used for preprocessing and statistical analysis of BOLD data, enabling the generation of high-resolution activation maps for comparison.
Coregistration Software	fOLD Toolbox, AtlasViewer, or custom scripts.	Enables precise mapping of fNIRS optode locations onto anatomical MRI data, which is fundamental for accurate spatial comparison with fMRI [16].

Integrated Analysis Workflow

To systematically compare fNIRS and fMRI data, a structured analysis workflow is essential. The following diagram outlines the key steps for integrating data from both modalities to assess spatial specificity and task sensitivity.

This workflow allows researchers to quantitatively answer the core questions of spatial correspondence (Step 3) and task sensitivity (Step 4). The final output (Step 5) is a validated fNIRS channel configuration and analysis protocol that can be confidently used for subsequent studies where fMRI is impractical, thereby operationalizing the integration of these two complementary modalities.

Brain Fingerprinting represents a paradigm shift in neuroimaging, moving from group-level comparisons to the identification of individual neural signatures. This approach relies on the core principle that an individual's functional brain activity and connectivity patterns are unique and stable over time, much like a fingerprint. The assessment of its intra-subject reproducibility and individual identification accuracy is therefore foundational to its clinical and research utility. Framed within a broader thesis on the integration of functional magnetic resonance imaging (fMRI) and functional near-infrared spectroscopy (fNIRS) for motor task paradigms, this document outlines how a multimodal approach can overcome the limitations of either technique used in isolation. By leveraging fMRI's high spatial resolution and fNIRS's portability and tolerance for motion, researchers can create robust, individualized brain maps suitable for applications in personalized medicine, neurological drug development, and advanced cognitive neuroscience [98] [102].

Quantitative Evidence for Reproducibility and Identification

The empirical foundation of brain fingerprinting rests on demonstrating that measured brain patterns are both reproducible within an individual and distinct from those of others. Recent studies provide quantitative evidence supporting this, particularly for fNIRS, which is a key tool for naturalistic data collection.

Table 1: Key Evidence for fNIRS Reproducibility and Individual Identification

Study Focus	Key Quantitative Finding	Implication for Brain Fingerprinting
Hemoglobin Contrast [20]	Oxyhemoglobin (HbO) signals were significantly more reproducible over sessions than deoxyhemoglobin (HbR) signals (F(1, 66) = 5.03, p < 0.05).	HbO should be prioritized as a more reliable metric for individual identification.
Dense-Sampling fNIRS [103]	High test-retest reliability and within-participant consistency were found in functional connectivity across ten self-administered sessions.	Repeated assessments yield robust individual-specific patterns, forming the basis of a reliable "fingerprint."
Analysis Pipelines [7]	Nearly 80% of research teams agreed on group-level fNIRS results when analyzing the same dataset, with higher agreement from experienced teams.	Reproducibility is achievable but is influenced by analytical choices and researcher expertise.
Data Quality [7]	Agreement at the individual level improved with better data quality.	Signal quality is a critical factor for accurate individual identification.

These findings confirm that with appropriate methodologies—such as focusing on HbO, employing dense-sampling strategies, and ensuring high data quality—fNIRS can reliably capture unique and stable neural signatures.

Experimental Protocols for Motor Paradigms

The following protocols are designed for a multimodal fMRI-fNIRS study, focusing on motor tasks to ensure high intra-subject reproducibility and facilitate individual identification.

Protocol 1: Basic Finger-Tapping Task

This classic motor paradigm is ideal for establishing reproducibility and validating the fMRI-fNIRS integration setup.

• Task Design: Employ a block design (e.g., 30s rest vs. 30s tapping). Participants perform self-paced finger opposition with their dominant hand.
• fMRI Acquisition: Use a standard BOLD protocol on a 3T scanner. Acquire whole-brain T1-weighted structural images for anatomical co-registration.
• fNIRS Acquisition: Position optodes over the primary motor cortex (M1) and prefrontal cortex (PFC) contralateral to the tapping hand. Use a system with at least 690 nm and 830 nm wavelengths to measure HbO and HbR changes. Source-detector separation should be 3.0-3.5 cm to ensure sufficient cortical penetration [104].
• Synchronization: Use a transistor-transistor logic (TTL) pulse from the fMRI scanner to synchronize the start of fNIRS recordings and task presentation.
• Analysis:
  • fMRI: Preprocess data (realignment, normalization, smoothing). Perform a general linear model (GLM) analysis to identify active voxels in M1.
  • fNIRS: Convert raw light intensity to optical density, then to HbO and HbR concentrations using the Modified Beer-Lambert Law. Apply a high-pass filter to remove drift and a low-pass filter for physiological noise. Use a GLM with a canonical hemodynamic response function to model task-related responses [105] [104].

Protocol 2: Naturalistic Motor Balancing Task

This protocol leverages fNIRS's strength in measuring brain activity during dynamic movement, which can be correlated with fMRI-derived maps.

• Task Design: Participants perform a balance task, such as using a Nintendo Wii Fit board for a simulated skiing game, with varying difficulty levels [104].
• Data Acquisition:
  • fMRI Session: In a separate session, acquire high-resolution functional and anatomical scans while the participant performs an imagined version of the balance task or simple motor tasks to map the motor and vestibular networks.
  • fNIRS Session: Using a high-density, portable fNIRS system, record from frontal, motor, sensory, and temporal cortices while the participant actively performs the balance task. Digitize optode positions for precise anatomical registration with the fMRI data [20] [104].
• Individual Identification Analysis:
  • Feature Extraction: For each subject and session, extract features such as the average HbO amplitude in the superior temporal gyrus (STG) during the task or whole-brain functional connectivity matrices during a resting-state period.
  • Fingerprint Creation: Generate an individual's functional connectivity profile (a "fingerprint") from a reference fNIRS session.
  • Identification: Use a cross-correlation or machine learning approach to match the fingerprint from a target session to the correct individual from a database of reference fingerprints [103].

Methodologies for fMRI-fNIRS Integration

Combining fMRI and fNIRS capitalizes on their complementary strengths to create a more complete and reliable picture of individual brain function.

Table 2: fMRI-fNIRS Integration Methodologies for Brain Fingerprinting

Integration Mode	Description	Application in Brain Fingerprinting
Synchronous [102]	Concurrent data acquisition during a single session. fNIRS probes are placed inside the MRI scanner, requiring MRI-compatible equipment.	Directly validates fNIRS signals against the fMRI BOLD signal. Provides a ground truth for spatial localization of fNIRS-derived individual fingerprints.
Asynchronous [102] [105]	Data is collected in separate sessions, often with different tasks (e.g., fMRI for precise localization, fNIRS for naturalistic monitoring).	More practical for longitudinal studies. Enables the use of fMRI maps as a spatial prior for analyzing fNIRS data collected over many sessions, improving the reliability of the fingerprint.
Analytical Fusion [102]	Combining datasets post-hoc using advanced computational models (e.g., machine learning).	Creates a unified model of brain function that incorporates both high spatial resolution (fMRI) and high temporal resolution in natural settings (fNIRS), enhancing identification accuracy.

The Scientist's Toolkit

Successful implementation of brain fingerprinting protocols requires specific tools and reagents. The following table details essential components.

Table 3: Research Reagent Solutions for fMRI-fNIRS Studies

Item	Function/Description	Justification
High-Density fNIRS System (>64 channels) [103]	Enables dense spatial sampling of the cortex, crucial for capturing detailed individual connectivity patterns.	Foundational for obtaining the high-quality, comprehensive data needed for reliable individual identification.
MRI-Compatible fNIRS Probes [102]	Optodes constructed from non-magnetic materials (e.g., plastic, fiber optics) for safe and simultaneous data acquisition inside the MRI scanner.	Essential for synchronous fMRI-fNIRS data collection, allowing for direct signal comparison and validation.
Anatomical Registration Kit (e.g., 3D digitizer) [20] [104]	Precisely records the 3D location of fNIRS optodes on the subject's head relative to anatomical landmarks (nasion, inion).	Critical for accurately co-registering fNIRS data with anatomical and functional MRI data, improving spatial accuracy.
Augmented Reality (AR) Guidance System [103]	Software that uses a tablet camera to guide users or technicians for reproducible fNIRS device placement across sessions.	Mitigates the negative impact of optode placement shifts on reproducibility, a key factor in longitudinal fingerprinting.
Standardized Cognitive & Motor Tasks (e.g., N-back, Finger-Tapping) [103] [105]	Well-validated paradigms that reliably activate specific brain networks, providing the behavioral context for measuring neural signatures.	Ensures that the evoked brain activity is consistent and comparable within and across individuals.

Standardization and Reporting Guidelines

To ensure the reproducibility and cross-study comparability of brain fingerprinting findings, adherence to community-driven best practices is imperative.

• Resting-State Protocols: For resting-state fNIRS, which is often used for connectivity fingerprints, a minimum scan length of 12 minutes is recommended to reliably capture slow neural fluctuations. Studies should clearly report whether an eyes-open or eyes-closed condition was used, as this significantly impacts the data [106].
• Analysis Pipeline Transparency: Given that analytical choices significantly impact results [7], researchers must pre-register their analysis plans or provide exhaustive details on their preprocessing steps, statistical models, and quality control thresholds.
• Reporting Optode Placement: The method for optode placement (e.g., based on the 10-20 system) and any session-to-session adjustments must be documented, as increased shifts in optode position reduce spatial overlap and reproducibility [20].

The integration of Functional Near-Infrared Spectroscopy (fNIRS) and functional Magnetic Resonance Imaging (fMRI) represents a powerful multimodal approach for advancing clinical neuroscience research. This framework is particularly valuable for differentiating patient populations where behavioral assessments are insufficient, such as in disorders of consciousness (DOC). fNIRS provides portable, bedside monitoring of cortical hemodynamics with high temporal resolution and tolerance to movement, while fMRI offers high-spatial-resolution whole-brain mapping including subcortical structures [14]. Their combined use enables robust cross-referencing, validating fNIRS findings against the clinical gold-standard of fMRI, and creating a comprehensive picture of brain function for drug development and clinical diagnosis [14] [107].

Clinical Validation: Evidence for Patient Differentiation

Case Study: Identifying Cognitive Motor Dissociation (CMD)

fNIRS has been clinically validated for its ability to identify covert consciousness in patients who are behaviorally unresponsive. A 2025 study with 70 prolonged DOC patients utilized a command-driven hand motor imagery task to detect CMD—a condition where patients have neuroimaging evidence of command-following despite being behaviorally non-responsive [108] [109].

Key Findings:

• Seven CMD patients were identified from vegetative state/unresponsive wakefulness syndrome (VS/UWS) and minimally conscious state minus (MCS–) populations
• CMD patients showed more favorable outcomes at 6-month follow-up (3/4 vs. 1/31, P=0.014) compared to non-CMD patients
• The methodology achieved differentiation through analyzing seven features of hemodynamic responses during task performance versus rest conditions [108] [109]

This demonstrates fNIRS's clinical value in identifying patient subgroups with distinct prognostic outcomes that are indistinguishable through standard behavioral assessment alone.

fNIRS Reproducibility for Patient Monitoring

Longitudinal fNIRS studies establish its reliability for monitoring patient populations over time. Research on visual and motor tasks across multiple sessions found that oxygenated hemoglobin (HbO) signals are significantly more reproducible than deoxygenated hemoglobin (HbR) signals (F(1, 66) = 5.03, p < 0.05) [20]. Source localization techniques further improved reliability, while shifts in optode placement reduced spatial overlap across sessions [20]. This reproducibility is essential for tracking patient progress or therapeutic response in clinical trials.

Comparative Technical Specifications: fNIRS versus fMRI

Table 1: Technical comparison between fNIRS and fMRI for clinical research applications

Parameter	fNIRS	fMRI
Spatial Resolution	1-3 cm [14]	Millimeter-level [14]
Temporal Resolution	Millisecond-level [14]	0.33-2 Hz (limited by hemodynamic response) [14]
Depth Penetration	Superficial cortex (2-3 cm) [14]	Whole brain, including subcortical structures [14]
Portability	High - bedside monitoring capability [14] [108]	Low - requires fixed scanner environment [14]
Motion Tolerance	High - suitable for naturalistic settings [14] [110]	Low - requires head immobilization [14]
Measured Parameters	HbO, HbR, total hemoglobin concentrations [108]	Blood Oxygen Level Dependent (BOLD) signal [14]
Electromagnetic Compatibility	High - operates in various environments [108]	Low - requires specialized MRI-compatible equipment [14]
Patient Population Suitability	Ideal for DOC patients, children, rehabilitation settings [108] [111]	Limited for patients with implants, claustrophobia, or requiring intensive care [14]

Experimental Protocols for Motor Task Paradigms

fNIRS Protocol for Command Following in DOC Patients

Objective: To identify CMD patients among behaviorally unresponsive DOC populations using a motor imagery task [108] [109].

Patient Population:

• Prolonged DOC patients (VS/UWS, MCS–, MCS+)
• Inclusion: DOC >28 days, age 16-80, intact auditory brainstem evoked potentials
• Exclusion: Unstable vital signs, cranial defects, scalp injuries, recent sedatives

Experimental Paradigm:

• Task: Hand-open-close motor imagery (without physical movement)
• Design: Block design with 20-second imagery alternated with 20-second rest
• Repetitions: 5 blocks totaling 300 seconds
• Instruction Delivery: Auditory commands "imagery" and "rest"
• Baseline: 50-second pre- and post-baseline periods for hemodynamic stabilization

fNIRS Data Acquisition:

• System: Continuous-wave fNIRS (e.g., NirScan-6000A)
• Wavelengths: 703 nm, 808 nm, 850 nm
• Sampling Rate: 11 Hz
• Optode Placement: 24 sources, 24 detectors based on international 10-20 system
• Channel Configuration: 63 measurement channels covering frontal, parietal, temporal, occipital regions
• Registration: 3D digitizer for Montreal Neurological Institute (MNI) coordinate conversion

Data Analysis:

• Extract seven hemodynamic features during task versus rest
• Apply support vector machine with genetic algorithm for classification
• Identify CMD based on significant neural response to commands despite behavioral unresponsiveness

Simultaneous fNIRS-fMRI Data Fusion Protocol

Objective: To leverage complementary spatiotemporal information from simultaneous fNIRS-fMRI acquisition for enhanced patient differentiation [14] [107].

Experimental Setup:

• fNIRS in MRI: MRI-compatible fNIRS system with 10m fiber optics
• Task Design: Right finger tapping task (block design: 21s activation/30s rest, 10 repetitions)
• Synchronization: Precise timing synchronization between fNIRS and fMRI acquisitions

Data Fusion Methodology:

• Joint Independent Component Analysis (jICA):
  • Simultaneously decomposes fNIRS and fMRI data into linked components
  • Model: X_fNIRS = AS_fNIRS and X_fMRI = AS_fMRI
  • Shared mixing matrix A identifies coupled spatiotemporal patterns
  • Enables visualization of where (fMRI) and when (fNIRS) hemodynamic signals change

• Spatiotemporal Snapshots:
  • Generate dynamic movies of brain activation
  • fMRI movie: M_fMRI = |T| × S^T (where T is fNIRS time courses, S is fMRI spatial components)
  • fNIRS movie: M_fNIRS = T × |S|^T

Validation Approach:

• Cross-reference fNIRS-identified patient differentiations with fMRI maps
• Use fMRI's spatial precision to validate fNIRS-derived regional activation patterns
• Apply fused data to identify novel biomarkers for patient stratification

Clinical Validation Workflow for Patient Differentiation

The Scientist's Toolkit: Essential Research Reagents & Materials

Table 2: Essential research materials and solutions for fNIRS-fMRI clinical studies

Item	Specification	Research Function
fNIRS System	Continuous-wave, multi-wavelength (e.g., 703, 808, 850 nm) [108]	Measures cortical HbO/HbR concentration changes during tasks
MRI-Compatible fNIRS	Fiber optic cables (10m length), non-magnetic components [107]	Enables simultaneous fNIRS-fMRI data acquisition
3D Digitizer	Electromagnetic (e.g., Patriot, Polhemus) [108]	Records precise optode positions for MNI space registration
Headgear	Flexible cap with adjustable optode holders	Secure sensor placement accommodating various head sizes
Stimulus Delivery	Auditory presentation system with noise cancellation	Presents task instructions consistently across patients
Data Analysis Software	NirSpace, SPM, Homer2, custom MATLAB scripts	Processes raw signals, extracts features, performs statistical analysis
Machine Learning Tools	Support Vector Machine with Genetic Algorithm [108]	Classifies neural responses and identifies patient subgroups
Spatial Registration	MNI standard brain template, Brodmann area atlas [108]	Standardizes anatomical localization across subjects
Joint ICA Algorithms	Custom implementations for multimodal fusion [107]	Identifies coupled spatiotemporal patterns in fNIRS-fMRI data

Signaling Pathways and Analytical Workflows

Multimodal Data Fusion Pathway

The integration of fNIRS and fMRI creates a powerful framework for differentiating patient populations in clinical neuroscience research. fNIRS provides the accessibility, portability, and temporal resolution needed for bedside monitoring of DOC patients and other challenging populations, while fMRI offers the spatial precision for validation and detailed localization. The protocols and methodologies outlined herein enable researchers to cross-reference findings across modalities, validating fNIRS against the established standard of fMRI while leveraging its unique advantages for naturalistic assessment. This multimodal approach accelerates the identification of clinically relevant patient subgroups, such as CMD patients, who may benefit from targeted therapeutic interventions and have distinct prognostic trajectories. As both technologies continue to advance, their synergistic application promises to uncover novel biomarkers for drug development and personalized medicine approaches in neurology and psychiatry.

The integration of functional Magnetic Resonance Imaging (fMRI) and functional near-infrared spectroscopy (fNIRS) represents a transformative approach in neuroimaging, combining excellent spatial resolution with superior temporal resolution and portability [3]. This multimodal framework is particularly powerful for studying motor task paradigms, as it enables comprehensive mapping of both cortical and subcortical brain activities with high spatiotemporal precision. Within this integrated framework, machine learning (ML) classifiers serve as critical analytical tools for translating complex hemodynamic data into clinically actionable diagnostic models. This protocol details the application of Support Vector Machines (SVM) and other ML classifiers for developing validated diagnostic models within fMRI-fNIRS integrated studies, with particular emphasis on motor imagery (MI) tasks relevant to disorders of consciousness and neurological rehabilitation.

Classifier Performance Comparison

Table 1: Performance Comparison of Classifiers on fNIRS Data for Motor Tasks

Classifier	Average Accuracy	Key Strengths	Optimal Use Cases	Citations
Support Vector Machine (SVM)	59.81% (base); 81.63% (with CSP)	Effective in high-dimensional spaces; Robust to overfitting	Small to medium-sized datasets; fNIRS motor imagery classification	[108] [112] [113]
Linear Discriminant Analysis (LDA)	69.00% (base); 84.19% (with CSP)	Computational efficiency; Simple implementation	Real-time BCI applications; Initial model prototyping	[112] [114] [115]
Convolutional Neural Network (CNN)	85.63% (MI); 96.84% (MA)	Automatic feature extraction; Spatial hierarchy learning	Large datasets; Raw signal processing; Spatiotemporal patterns	[116]
Long Short-Term Memory (LSTM)	~83.3% (mental tasks)	Temporal sequence modeling; Long-range dependencies	Time-series hemodynamic data; Complex cognitive tasks	[114] [116]
Temporal Convolutional Network (TCN)	Comparable to LSTM with faster training	Parallel processing; Flexible receptive fields	Motor imagery classification; Real-time implementation	[116]

The performance metrics in Table 1 demonstrate that while traditional classifiers like SVM and LDA provide solid baseline performance, their effectiveness can be significantly enhanced through feature optimization techniques such as Common Spatial Pattern (CSP) algorithms, which improve SVM accuracy from 59.81% to 81.63% and LDA from 69.00% to 84.19% for fNIRS-based motor imagery tasks [112]. Recent advances in deep learning architectures, particularly CNNs with spatiotemporal feature extraction mechanisms, have achieved notable performance (85.63% for motor imagery, 96.84% for mental arithmetic) on publicly available datasets [116]. Benchmarking studies indicate that reported classification accuracies in literature may be overly optimistic, and rigorous validation methodologies are essential for realistic performance assessment [114].

Experimental Protocols

Multimodal Data Acquisition Setup

Participants: For diagnostic model development, recruit 70-100 participants total, including both clinical populations and healthy controls. Specific inclusion criteria should include right-handedness (for motor paradigm consistency), intact auditory brainstem evoked potentials (for command-driven tasks), and age range of 16-80 years [108] [109]. Exclusion criteria should encompass unstable vital signs, cranial defects, history of neurological disorders, recent sedative medication, and scalp injuries preventing proper optode placement [109].

fNIRS Configuration: Utilize a continuous-wave fNIRS system (e.g., NirScan-6000A) with wavelengths of 703, 808, and 850 nm at sampling frequency ≥ 10 Hz [109]. Arrange 24 source optodes and 24 detector optodes in flexible headgear with 3 cm optode spacing, forming 63 measurement channels covering frontal, parietal, temporal, and occipital lobe areas based on the international 10-20 EEG electrode placement system [108] [109]. Perform spatial registration using a 3D electromagnetic digitizer to convert coordinates to Montreal Neurological Institute (MNI) space and project onto standard brain templates [109].

fMRI Configuration: Conduct simultaneous fMRI acquisition using a 3.0T MRI system with echo planar imaging (EPI) sequence parameters: TR/TE = 3000/35 ms, flip angle = 80°, 35 slices, and 4 mm slice thickness [117]. For asynchronous data acquisition, ensure consistent motor task paradigms between fNIRS and fMRI sessions [3].

Motor Imagery Paradigm: Implement a block design consisting of 20-second imagery tasks alternating with 20-second rest periods, repeated 5 times [108] [109]. Provide verbal commands "imagery" and "rest" through MR-compatible headphones, as the auditory pathway is relatively well-preserved in clinical populations [109]. For hand motor imagery, instruct participants to: "Imagine repeatedly opening and closing both hands as quickly and naturally as possible without distinguishing between left and right" [108]. Include 50-second pre-baseline and post-baseline periods to allow cortical hemodynamics to stabilize [109].

Data Preprocessing Pipeline

fNIRS Preprocessing: Convert raw light intensity measurements to oxygenated (HbO) and deoxygenated hemoglobin (HbR) concentrations using the Modified Beer-Lambert Law [116]. Apply a bandpass filter (0.01-0.2 Hz) to remove physiological noise and drift [114]. Perform motion artifact correction using wavelet-based or moving average approaches [114]. Segment data into epochs from -5s pre-stimulus to 25s post-stimulus onset [117].

fMRI Preprocessing: Implement standard preprocessing pipeline including slice timing correction, realignment, coregistration to structural images, normalization to MNI space, and spatial smoothing [117]. For multimodal fusion, extract BOLD time series from regions corresponding to fNIRS channels [3].

Feature Extraction: For traditional ML classifiers, extract seven key features of hemodynamic responses during both task and rest conditions: mean, variance, slope, skewness, kurtosis, peak, and signal slope [108] [112]. For deep learning approaches, use raw preprocessed signals or time-frequency representations as input [116]. Implement Common Spatial Pattern (CSP) algorithm for dimensionality reduction and enhanced feature discriminability [112].

Classifier Training and Validation

Data Partitioning: Employ nested cross-validation with an outer loop (5-fold) for performance estimation and an inner loop (3-fold) for hyperparameter optimization [114]. Ensure data from individual participants is contained within a single fold to prevent leakage and overoptimistic performance [114].

SVM Implementation: For SVM classifier development, utilize linear or radial basis function kernels based on dataset characteristics [113]. Combine with genetic algorithms for feature selection and parameter optimization [108]. Implement cost parameter optimization through grid search (typical range: 0.001 to 1000) [113].

Deep Learning Implementation: For CNN architectures, design layers with 2D time convolution, depth convolution, and separable convolution, followed by batch normalization, ELU activation, average pooling, and dropout layers [116]. Incorporate spatial attention mechanisms and temporal convolutional networks (TCN) for spatiotemporal feature extraction [116].

Validation Metrics: Report accuracy, sensitivity, specificity, F1-score, and area under the receiver operating characteristic curve (AUROC) [114]. Perform statistical significance testing using Fisher's exact test for contingency tables and paired t-tests for classifier comparisons [108] [109].

Signaling Pathways and Workflows

Figure 1: Integrated fMRI-fNIRS diagnostic model development workflow. The pathway begins with motor imagery tasks generating neural activity in motor cortex regions, followed by neurovascular coupling that triggers hemodynamic responses detected simultaneously by fNIRS (HbO/HbR concentrations) and fMRI (BOLD signal). After preprocessing and feature extraction, multiple classifiers (SVM, LDA, CNN) are trained and validated to produce diagnostic models for conditions such as cognitive motor dissociation [108] [117] [3].

Research Reagent Solutions

Table 2: Essential Research Reagents and Equipment for fMRI-fNIRS Integration

Item Category	Specific Examples	Function and Application	Citations
fNIRS Systems	NirScan-6000A; Oxymon MKIII	Continuous-wave fNIRS measurement; HbO/HbR concentration detection	[108] [117]
fMRI Systems	3.0T MRI systems (e.g., ISOL)	BOLD signal acquisition; High-spatial resolution mapping	[117] [3]
Optode Digitization	Patriot 3D electromagnetic digitizer	Spatial registration of fNIRS optodes; MNI coordinate transformation	[109]
Software Platforms	NirSpace; SPM; BenchNIRS	Data processing; Machine learning benchmarking; Statistical analysis	[114] [109]
ML Libraries	Scikit-learn; TensorFlow/PyTorch	SVM/LDA implementation; Deep learning model development	[114] [116]
Experimental Paradigms	Hand-open-close motor imagery; Finger tapping	Motor task execution; Cognitive motor dissociation assessment	[108] [117]

The research reagents and equipment listed in Table 2 represent the essential components for establishing a multimodal fMRI-fNIRS research laboratory. MRI-compatible fNIRS systems are particularly critical for simultaneous data acquisition, requiring specialized fiber optics (10m length) to connect optodes in the MR scanner to the NIRS instrument in the control room [117]. The BenchNIRS framework provides an open-source benchmarking tool for evaluating machine learning models on fNIRS data, implementing robust methodology with nested cross-validation to prevent overoptimistic performance reporting [114].

Methodological Considerations

Technical Integration Challenges

Simultaneous fMRI-fNIRS acquisition presents several technical challenges, including electromagnetic interference in MRI environments, hardware incompatibilities, and restricted motion paradigms that may limit naturalistic movement [3]. Effective data fusion requires addressing the temporal misalignment between fNIRS (high temporal resolution) and fMRI (low temporal resolution) signals, in addition to spatial resolution disparities [117] [3]. Joint Independent Component Analysis (jICA) has been successfully employed to calculate linked temporal fNIRS components and spatial fMRI components, enabling the generation of spatiotemporal "snapshots" of brain activity [117].

Clinical Validation Protocols

For diagnostic model validation in clinical populations such as disorders of consciousness, follow-up assessments using standardized outcome measures like the Glasgow Outcome Scale-Extended (GOSE) should be conducted at 6 months post-evaluation to establish prognostic validity [108] [109]. Studies have demonstrated that patients identified as cognitive motor dissociation (CMD) using fNIRS-based classifiers show significantly more favorable outcomes (3/4 vs. 1/31, P = 0.014), supporting the clinical validity of this approach [108] [109].

This application note provides comprehensive protocols for developing and validating SVM and other classifiers within an integrated fMRI-fNIRS framework for motor task paradigms. The systematic comparison of classifier performance, detailed experimental methodologies, and technical considerations outlined herein serve as a foundation for robust diagnostic model development. The integration of multimodal neuroimaging with machine learning classification represents a promising pathway for advancing objective diagnostic biomarkers in neurological and psychiatric disorders, particularly through the detection of covert consciousness in behaviorally non-responsive patients. Future directions should emphasize standardized reporting practices, open-source benchmarking frameworks, and personalized classification approaches to enhance translational impact.

The integration of functional magnetic resonance imaging (fMRI) and functional near-infrared spectroscopy (fNIRS) represents a promising frontier in developing robust biomarkers for neurological drug development. These multimodal neuroimaging approaches leverage complementary strengths: fMRI provides high spatial resolution and whole-brain coverage, including deep structures, while fNIRS offers superior temporal resolution, portability for naturalistic settings, and resilience to motion artifacts [3]. The qualification pathway for such biomarkers is formally established through the FDA's Drug Development Tool (DDT) Qualification Program, created under the 21st Century Cures Act [118] [119]. This program provides a regulatory framework for qualifying tools for specific contexts of use (COU) in drug development, enabling their use across multiple drug development programs without needing re-evaluation for each application [120].

Qualification is particularly valuable for biomarkers intended for repeated use across clinical trials, as it creates a publicly available resource that any sponsor can employ. For multimodal fMRI-fNIRS biomarkers, this pathway offers the opportunity to establish standardized methodologies that leverage the synergistic potential of these technologies, especially in measuring cortical activity during motor tasks with both spatial precision and temporal dynamics [3]. The convergence of technological advances in hardware integration and analytical methods for data fusion now makes this an opportune time to pursue formal qualification of these biomarkers for specific contexts in neurological drug development.

Regulatory Framework and Qualification Process

The Drug Development Tool Qualification Program

The DDT Qualification Program establishes a structured pathway for qualifying biomarkers, clinical outcome assessments, and other drug development tools. According to FDA guidance, the qualification process involves three distinct stages: Letter of Intent (LOI), Qualification Plan (QP), and Full Qualification Package (FQP) [119]. This structured approach allows for early engagement and iterative feedback between biomarker developers and regulatory agencies throughout the development process.

The program operates with a focus on specific Context of Use (COU), which precisely defines how the biomarker will be applied in drug development. The COU statement describes all elements characterizing the purpose and manner of use, establishing boundaries within which available data adequately justify the DDT's application [118]. For a multimodal fMRI-fNIRS biomarker, the COU might specify its use in detecting specific neurophysiological responses to therapeutic interventions in conditions like stroke recovery or Parkinson's disease, particularly for tracking cortical activation patterns during motor task paradigms.

Table 1: Stages in the FDA DDT Qualification Process

Stage	Purpose	FDA Review Timeline	Key Deliverables
Letter of Intent (LOI)	Introduce biomarker concept and proposed COU	3 months	Brief description of biomarker, preliminary COU, rationale for use
Qualification Plan (QP)	Detail development plan and evidence generation strategy	6 months	Complete COU, detailed plans for analytical and clinical validation
Full Qualification Package (FQP)	Submit complete evidence package	10 months	Comprehensive data demonstrating analytical and clinical validity

Evidentiary Standards and Considerations

The level of evidence required for biomarker qualification depends on a benefit-risk analysis tied to the proposed COU [120]. Biomarkers used for high-impact decisions (e.g., clinical trial enrichment or definitive regulatory decisions) require more substantial evidence than those used for exploratory purposes. The FDA's evidentiary framework emphasizes three key components: needs assessment, context of use, and benefit-risk analysis [120].

For multimodal neuroimaging biomarkers, regulators will expect demonstration of analytical validity (reliability and reproducibility of the measurements), clinical validity (association with clinical endpoints or disease states), and clinical utility (value in informing drug development decisions). The reproducibility of fNIRS signals has been demonstrated across multiple sessions, with oxygenated hemoglobin (HbO) showing superior reproducibility compared to deoxygenated hemoglobin (HbR) [20]. This evidence of measurement stability strengthens the case for qualification.

Multimodal fMRI-fNIRS Integration for Motor Task Paradigms

Technical Synergies and Applications

The combined use of fMRI and fNIRS creates a powerful multimodal approach that overcomes the limitations of either technology used independently. fMRI provides high spatial resolution (millimeter-level precision) and comprehensive coverage of both cortical and subcortical structures, enabling localization of brain activity across the entire brain [3]. Meanwhile, fNIRS delivers superior temporal resolution (millisecond-level precision), portability for naturalistic settings, and greater resilience to motion artifacts, making it suitable for studying active motor behaviors and longitudinal monitoring [3].

This complementary relationship is particularly valuable for motor task paradigms, where researchers can use fMRI to precisely localize activity patterns and fNIRS to capture the temporal dynamics of cortical activation. The integration methodologies can be implemented in either synchronous (simultaneous data acquisition) or asynchronous (sequential acquisition) modes, each with distinct advantages for specific research contexts [3]. Synchronous acquisition enables direct correlation of signals, while asynchronous approaches can leverage fMRI for spatial localization to inform fNIRS source reconstruction.

Table 2: Comparative Technical Specifications of fMRI and fNIRS

Parameter	fMRI	fNIRS
Spatial Resolution	1-3 mm	1-3 cm
Temporal Resolution	0.3-2 Hz (limited by hemodynamics)	Up to 100 Hz
Penetration Depth	Whole brain (cortical and subcortical)	Superficial cortex (1.5-2 cm)
Portability	Low (requires fixed scanner)	High (wearable systems available)
Motion Tolerance	Low (requires head immobilization)	Moderate (tolerant of some movement)
Measured Signal	Blood Oxygen Level Dependent (BOLD)	Oxygenated (HbO) and Deoxygenated (HbR) hemoglobin

Experimental Protocol: Motor Task Paradigm with Simultaneous fMRI-fNIRS

Objective: To quantify cortical motor network activation in response to controlled motor tasks for assessing therapeutic interventions in neurodegenerative disorders.

Participants:

• Target population: Adults with diagnosed motor impairments (e.g., Parkinson's disease, stroke recovery)
• Sample size: Determined by power analysis (typically n≥20 per group for pilot studies)
• Control groups: Age-matched healthy controls

Equipment and Materials:

• 3T MRI scanner with compatible fNIRS system
• MRI-compatible fNIRS headset with emitter-detector pairs positioned over motor cortices
• Response recording device (MRI-compatible button box or transducer)
• Stimulus presentation system (visual projection or auditory delivery system)
• Synchronization unit to align fMRI and fNIRS data acquisition

Procedure:

• Participant Preparation:
  • Screen for MRI contraindications
  • Position fNIRS optodes according to international 10-20 system, focusing on primary motor cortex (C3, C4), supplementary motor area, and prefrontal regions
  • Ensure proper light coupling and signal quality before moving into scanner

• Experimental Design:

  • Implement block design with alternating 30-second task and 30-second rest periods
  • Motor tasks: Sequential finger tapping, hand gripping, or foot flexion
  • Include multiple task conditions with varying complexity levels
  • Total paradigm duration: 10-15 minutes

• Data Acquisition:

  • fMRI: Acquire T2*-weighted BOLD images with standard parameters (TR=2000ms, TE=30ms, voxel size=3×3×3mm)
  • fNIRS: Simultaneously record HbO and HbR concentrations at 10Hz sampling rate
  • Synchronize timing pulses between systems

• Data Processing:

  • fMRI: Preprocess with standard pipeline (realignment, normalization, smoothing)
  • fNIRS: Convert raw intensity to optical density, then to hemoglobin concentrations using Modified Beer-Lambert Law
  • Coregister fNIRS channels to anatomical locations using digitization

Biomarker Performance and Validation Data

Establishing Analytical and Clinical Validity

Robust validation is essential for biomarker qualification. For multimodal neuroimaging biomarkers, this requires demonstrating both analytical validity (reliability, reproducibility) and clinical validity (association with clinical endpoints). Recent research provides promising evidence for both aspects.

In fNIRS studies, oxygenated hemoglobin (HbO) has shown significantly better reproducibility across multiple sessions compared to deoxygenated hemoglobin (HbR) [20]. This reproducibility is crucial for longitudinal studies tracking disease progression or treatment response. Source localization techniques and anatomical guidance from concurrent fMRI can further enhance the reliability of fNIRS for capturing brain activity [20].

For clinical validity, studies have demonstrated the capability of fNIRS to differentiate patient populations. In mild cognitive impairment (MCI) research, incorporating neural metrics from time-domain fNIRS significantly improved classification performance (AUC=0.92) compared to using only behavioral or self-report data [121]. This demonstrates the potential added value of neuroimaging biomarkers in diagnostic assessment.

Table 3: Performance Metrics from fNIRS Biomarker Studies

Study Reference	Population	Task Paradigm	Key Performance Metrics
MCI Classification [121]	MCI patients (n=50) vs Healthy Controls (n=51)	Verbal Fluency, N-Back	Neural metrics + behavior: AUC=0.92; Behavior only: AUC=0.79; Self-report only: AUC=0.76
fNIRS Reproducibility [20]	4 participants across ≥10 sessions	Motor, Visual tasks	HbO significantly more reproducible than HbR; Source localization improves reliability
fMRI-fNIRS Integration [3]	Literature review (63 studies)	Motor, Cognitive tasks	Combined approach enables robust spatiotemporal mapping; Synchronous and asynchronous modes available

Advanced Analytical Approaches

The complex, high-dimensional data generated by multimodal neuroimaging requires sophisticated analytical approaches. Machine learning algorithms are increasingly employed for feature selection, classification, and prediction of disease states from neuroimaging data [122]. For multimodal fMRI-fNIRS data, analytical approaches include:

• Multimodal data fusion: Integrating complementary spatial and temporal information to create unified activation maps
• Connectivity analysis: Examining functional connectivity patterns within motor networks
• Multivariate pattern analysis: Identifying distributed activation patterns associated with specific states or responses

Deep learning frameworks have shown particular promise in analyzing complex neuroimaging data. For example, graph-based neural networks that incorporate both brain connectivity and morphological features have achieved high precision in brain age estimation (MAE=2.39 years) and demonstrated strong discriminative capacity between cognitive states (AUC=0.885 for CN vs MCI) [123].

The Scientist's Toolkit: Essential Research Reagent Solutions

Table 4: Essential Materials and Solutions for Multimodal fMRI-fNIRS Research

Item	Function/Application	Specification Considerations
MRI-compatible fNIRS System	Simultaneous data acquisition in scanner environment	Electromagnetic compatibility; Fiber-optic cabling; Minimal metallic components
fNIRS Headcaps	Secure optode positioning on scalp	Material compatibility (MRI safety); Adaptability to international 10-20 system; Various sizes
Source Localization Software	Anatomical registration of fNIRS channels	Integration with MRI anatomical data; Digitization capability; Brain atlas alignment
Multimodal Data Fusion Platform	Integrated analysis of fMRI and fNIRS data	Support for heterogeneous temporal resolutions; Spatial coregistration tools; Statistical parametric mapping
Hemodynamic Response Modeling Tools	Analysis of blood flow dynamics	Physiological noise removal; Hemodynamic response function estimation; Temporal filtering
Motor Task Paradigm Software	Presentation of controlled motor tasks	Precision timing; Synchronization with acquisition systems; Multiple condition support

Path Forward and Implementation Strategy

The qualification pathway for multimodal fMRI-fNIRS biomarkers requires systematic evidence generation across the development pipeline. Based on analysis of the DDT Qualification Program, developers should anticipate an average timeline of 6 years from initial submission to qualification [119]. This extended timeline underscores the importance of early planning and stakeholder engagement.

Successful qualification strategies often involve collaborative consortia that pool resources and data across multiple institutions [120]. For multimodal neuroimaging biomarkers, such consortia could establish standardized acquisition protocols, shared datasets for validation, and unified analytical approaches. The Biomarker Qualification Program encourages the formation of such collaborative groups to increase efficiency and lessen individual resource burdens [118].

To maximize the likelihood of regulatory acceptance, developers should focus on well-defined contexts of use with clear clinical relevance to drug development. Initial targets might include biomarkers for patient stratification in motor disorder trials, pharmacodynamic biomarkers for tracking treatment response, or prognostic biomarkers for predicting disease progression. The evolving landscape of biomarker qualification shows promising growth, with increasing numbers of tools progressing through the regulatory pathway [119].

The integration of artificial intelligence and machine learning approaches will further enhance the value of multimodal biomarkers by enabling the identification of complex patterns in high-dimensional data [122]. As these analytical techniques mature alongside improvements in hardware integration, multimodal fMRI-fNIRS biomarkers are poised to become increasingly valuable tools in the development of therapies for neurological disorders.

Conclusion

The integration of fMRI and fNIRS for motor task paradigms represents a powerful synergy that leverages the high spatial resolution and whole-brain coverage of fMRI with the portability, cost-effectiveness, and ecological validity of fNIRS. Evidence confirms a strong spatial correspondence between hemodynamic signals, enabling the translation of well-established fMRI paradigms to flexible fNIRS setups. Key to success is addressing methodological challenges in co-registration and noise suppression. For the future, this multimodal approach holds immense promise for creating robust biomarkers for drug development, personalizing neurorehabilitation therapies via neurofeedback, and fundamentally advancing our understanding of motor control in real-world environments. The continued development of standardized protocols and analytical tools will be crucial for realizing the full potential of this integration in both research and clinical practice.

References

• 1. Coupling Mechanism and Significance of the BOLD Signal [https://pmc.ncbi.nlm.nih.gov/articles/PMC4147398/]
• 2. Physiological modeling of the BOLD signal and ... [https://www.sciencedirect.com/science/article/pii/S1053811923004007]
• 3. Applications and advances of combined fMRI-fNIRs ... [https://www.frontiersin.org/journals/neurology/articles/10.3389/fneur.2025.1542075/full]
• 4. Near-infrared spectroscopy of the cerebellum in motor ... [https://www.nature.com/articles/s41598-025-04678-x]
• 5. The role of fNIRS in assessing motor function: A bibliometric ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12338178/]
• 6. The effects of protocol factors and participant ... [https://www.sciencedirect.com/science/article/pii/S2666956025000443]
• 7. fNIRS reproducibility varies with data quality, analysis ... [https://www.nature.com/articles/s42003-025-08412-1]
• 8. Changes in hemodynamic response function components ... [https://www.frontiersin.org/journals/physiology/articles/10.3389/fphys.2022.1101470/full]
• 9. Applications and Advances of Combined fMRI-fNIRs ... [https://www.frontiersin.org/journals/neurology/articles/10.3389/fneur.2025.1542075/abstract]
• 10. Multimodal assessment of the spatial correspondence ... [https://www.nature.com/articles/s41598-023-29123-9]
• 11. Spatial correspondence of cortical activity measured with ... [https://www.sciencedirect.com/science/article/pii/S1053811924000648]
• 12. Spatial correspondence of cortical activity measured with ... [https://pubmed.ncbi.nlm.nih.gov/38461959/]
• 13. Requirement of a complex motor task to identify ... [https://www.frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2025.1595284/full]
• 14. Applications and advances of combined fMRI-fNIRs ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11958174/]
• 15. How to design real-world functional near-infrared ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12618016/]
• 16. fMRI-based validation of continuous-wave fNIRS of supplementary motor area activation during motor execution and motor imagery | Scientific Reports [https://www.nature.com/articles/s41598-022-06519-7]
• 17. BOLD Hemodynamic Response Function Varies as a ... [https://archive.ismrm.org/2018/4512.html]
• 18. Characterization of the hemodynamic response function in ... [https://www.nature.com/articles/s41467-019-09076-2]
• 19. Hemodynamic Response Function in Brain White Matter in a ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC7552822/]
• 20. Reproducibility of fNIRS within Subject for Visual and Motor ... [https://www.sciencedirect.com/science/article/pii/S1053811925004951]
• 21. The impact of interactive motor-cognitive dual tasking on ... [https://www.frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2025.1464617/full]
• 22. Investigating Human Neurovascular Coupling Using ... [https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2015.00467/full]
• 23. A dynamical model of the laminar BOLD response [https://www.sciencedirect.com/science/article/pii/S1053811919308006]
• 24. Functional Magnetic Resonance Imaging and ... [https://www.frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2017.00419/full]
• 25. A quantitative comparison of NIRS and fMRI across multiple ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC3021967/]
• 26. From Mechanisms to Medicine: Neurovascular Coupling in the ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11719775/]
• 27. A Quantitative Comparison of Simultaneous BOLD fMRI ... [https://www.sciencedirect.com/science/article/pii/S1053811902912279]
• 28. Dynamic causal modelling for functional near-infrared ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC4401444/]
• 29. Comparison of fNIRS with other neuroimaging modalities [https://www.artinis.com/blogpost-all/comparison-fnirs-vs-fmri]
• 30. Overview of Functional Magnetic Resonance Imaging - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC3073717/]
• 31. Application of Functional Near-Infrared Spectroscopy to the ... [https://www.sciencedirect.com/science/article/pii/S1016847823006428]
• 32. Applications of Functional Near-Infrared Spectroscopy ... [https://www.mdpi.com/2077-0383/14/15/5197]
• 33. Riemannian geometry boosts functional near-infrared ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12523035/]
• 34. Functional near-infrared spectroscopy for the assessment ... [https://www.frontiersin.org/journals/neurology/articles/10.3389/fneur.2025.1524806/full]
• 35. Low-field and portable MRI technology - PubMed Central - NIH [https://pmc.ncbi.nlm.nih.gov/articles/PMC12549457/]
• 36. Comparison of block and event‐related fMRI designs in ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC6871864/]
• 37. Large-scale fMRI dataset for the design of motor-based ... [https://www.nature.com/articles/s41597-025-05134-1]
• 38. Optimizing fMRI experimental design for MVPA-based BCI ... [https://www.sciencedirect.com/science/article/pii/S1053811918320615]
• 39. Detection of motor execution using a hybrid fNIRS-biosignal BCI [https://pmc.ncbi.nlm.nih.gov/articles/PMC3637588/]
• 40. Suppressing Systemic Interference in fNIRS Monitoring of ... [https://www.frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2018.00085/full]
• 41. Current Status and Issues Regarding Pre-processing of ... [https://www.frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2018.00505/full]
• 42. Graded fMRI Neurofeedback Training of Motor Imagery in ... [https://www.frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2020.00226/full]
• 43. a simultaneous fNIRS-fMRI study [https://pmc.ncbi.nlm.nih.gov/articles/PMC9896013/]
• 44. Dynamic causal modelling on infant fNIRS data [https://www.sciencedirect.com/science/article/pii/S1053811918303185]
• 45. Hybrid EEG-fNIRS Asynchronous Brain-Computer Interface for ... [https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0146610]
• 46. Development of a combined, sequential real-time fMRI and ... [https://www.sciencedirect.com/science/article/abs/pii/S0165027020301424]
• 47. fNIRS reproducibility varies with data quality, analysis ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12322209/]
• 48. fMRIPrep: A Robust Preprocessing Pipeline for fMRI Data ... [https://fmriprep.org/]
• 49. Multivariate disturbance filtering in auditory fNIRS signals ... [https://www.sciencedirect.com/science/article/pii/S0010482524009259]
• 50. A novel approach for capturing fNIRS signal dynamics in ... [https://www.sciencedirect.com/science/article/abs/pii/S0010482525006511]
• 51. fMRI Validation of fNIRS Measurements During a ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC4544944/]
• 52. Multimodal fNIRS–EEG sensor fusion: Review of data-driven ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12592382/]
• 53. A guide to group effective connectivity analysis, part 1 [https://pmc.ncbi.nlm.nih.gov/articles/PMC6711459/]
• 54. General Linear Model [https://web.conn-toolbox.org/fmri-methods/general-linear-model]
• 55. Using the General Linear Model to Improve Performance in ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC7040364/]
• 56. Investigating the interaction between EEG and fNIRS [https://www.sciencedirect.com/science/article/pii/S1877750324002096]
• 57. Effective connectivity analysis [https://bio-protocol.org/exchange/minidetail?id=9328208&type=30]
• 58. Functional connectivity in whole-brain and network analysis ... [https://translational-medicine.biomedcentral.com/articles/10.1186/s12967-025-07181-z]
• 59. Neural control meets biomechanics in the motor ... [https://www.frontiersin.org/journals/neural-circuits/articles/10.3389/fncir.2025.1608328/full]
• 60. Neural control meets biomechanics in the motor assessment ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12245779/]
• 61. The impact of interactive motor-cognitive dual tasking on brain ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12237959/]
• 62. Brain Activation During Motor Preparation and Execution in ... [https://www.mdpi.com/2076-3425/15/4/333]
• 63. fNIRS is sensitive to leg activity in the primary motor cortex ... [https://www.sciencedirect.com/science/article/pii/S1053811923000290]
• 64. The potential of real-time fMRI neurofeedback for stroke rehabilitation: A systematic review - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC6182108/]
• 65. Frontiers | Toward an Adapted Neurofeedback for Post-stroke Motor Rehabilitation: State of the Art and Perspectives [https://www.frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2022.917909/full]
• 66. Functional Magnetic Resonance Imaging Neurofeedback ... [https://www.frontiersin.org/journals/behavioral-neuroscience/articles/10.3389/fnbeh.2016.00111/full]
• 67. Using Target Engagement Biomarkers to Predict Clinical ... [https://www.sciencedirect.com/science/article/abs/pii/S0022356524268944]
• 68. Mapping the developmental path for Parkinson's disease ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12594996/]
• 69. Spatial registration for functional near-infrared ... [https://www.sciencedirect.com/science/article/pii/S1053811913007805]
• 70. Guiding functional near-infrared spectroscopy optode- ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC8211086/]
• 71. Targeting brain regions of interest in functional near‐infrared ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC8046049/]
• 72. 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]
• 73. fNIRS Optodes' Location Decider (fOLD): a toolbox for ... [https://www.nature.com/articles/s41598-018-21716-z]
• 74. Probe registration [https://fnirs.org/resources/experiment/probe-registration/]
• 75. Optimizing spatial specificity and signal quality in fNIRS [https://www.frontiersin.org/journals/neuroergonomics/articles/10.3389/fnrgo.2024.1286586/full]
• 76. Improved physiological noise regression in fNIRS [https://pmc.ncbi.nlm.nih.gov/articles/PMC7703677/]
• 77. How to remove or control confounds in predictive models ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC8917515/]
• 78. Paired evaluation of machine-learning models ... [https://www.sciencedirect.com/science/article/pii/S2666389923001460]
• 79. Hybrid motion artifact detection and correction approach for ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC8871689/]
• 80. Motion artifacts removal and evaluation techniques for ... [https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2022.878750/full]
• 81. Motion artifact detection and correction in functional near- ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC5803523/]
• 82. Correction of Physiological Artifacts in Multi‐Echo fMRI Data ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12183604/]
• 83. Iterative Learning for Joint Image Denoising and Motion ... [https://www.sciencedirect.com/science/article/pii/S1361841525004189?dgcid=rss_sd_all]
• 84. Recommendations for motion correction of infant fNIRS ... [https://www.sciencedirect.com/science/article/pii/S1053811919305531]
• 85. Comparison of motion correction techniques applied to ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC9900395/]
• 86. Characterizing motion artifacts in functional near-infrared ... [https://www.sciencedirect.com/science/article/pii/S1746809425007670]
• 87. Quantifying the Impact of Hair and Skin Characteristics on ... [https://pubmed.ncbi.nlm.nih.gov/39554177/]
• 88. Quantifying the impact of hair and skin characteristics on fNIRS ... signal [https://www.nature.com/articles/s41562-025-02274-7?error=cookies_not_supported&code=be9f36c5-6bc3-4cf2-9b72-bacca6a7fbb6]
• 89. Quantifying the impact of hair and skin characteristics on ... [https://pubmed.ncbi.nlm.nih.gov/40897802/]
• 90. Quantifying the Impact of Hair and Skin Characteristics on ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11565806/]
• 91. How short is short? Optimum source–detector distance for... [https://pmc.ncbi.nlm.nih.gov/articles/PMC4478880/]
• 92. Cortical Signal Analysis and Advances in Functional Near ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC4899446/]
• 93. Performance comparison of systemic activity correction in ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC9555616/]
• 94. fNIRS-based brain-computer interfaces: a review [https://www.frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2015.00003/full]
• 95. Sensorimotor adaptation in spatial orientation task: a fNIRS ... [https://www.nature.com/articles/s41598-023-42416-3]
• 96. Reliability of brain metrics derived from a Time-Domain ... [https://www.nature.com/articles/s41598-024-68555-9]
• 97. Hybrid Deep Learning Model to Estimate Cognitive Effort ... [https://arxiv.org/html/2504.13883v3]
• 98. Functional Magnetic Resonance Imaging and Functional Near ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC5563305/]
• 99. A quantitative comparison of simultaneous BOLD fMRI and NIRS recordings during functional brain activation - PubMed [https://pubmed.ncbi.nlm.nih.gov/12377147/]
• 100. A Hitchhiker's Guide to Functional Magnetic Resonance ... [https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2016.00515/full]
• 101. Best practices for fNIRS publications - PMC - PubMed Central [https://pmc.ncbi.nlm.nih.gov/articles/PMC7793571/]
• 102. Applications and advances of combined fMRI-fNIRs ... [https://pubmed.ncbi.nlm.nih.gov/40170894/]
• 103. Wearable fNIRS platform for dense sampling and precision ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12069634/]
• 104. Functional near-infrared spectroscopy (fNIRS) of brain ... [https://www.sciencedirect.com/science/article/abs/pii/S0966636211004115]
• 105. Task-related brain activity and functional connectivity in ... [https://pubmed.ncbi.nlm.nih.gov/33094125/]
• 106. Shedding light on the brain: guidelines to address ... [https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2025.1557471/full]
• 107. Fusion of fNIRS and fMRI data: identifying when and where ... [https://www.frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2013.00676/full]
• 108. Detecting cognitive motor dissociation by functional near- ... [https://www.frontiersin.org/journals/neurology/articles/10.3389/fneur.2025.1532804/full]
• 109. Detecting cognitive motor dissociation by functional near- ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11997382/]
• 110. NIRS-Based Hyperscanning Reveals Inter-brain Neural ... [https://www.frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2016.00082/full]
• 111. The development of functional network organization in ... [https://www.sciencedirect.com/science/article/pii/S1878929317300579]
• 112. Improvement of classification accuracy of functional near- ... [https://www.sciencedirect.com/science/article/pii/S2667102625000245]
• 113. Diagnostic machine learning applications on clinical ... [https://pubmed.ncbi.nlm.nih.gov/38308531/]
• 114. Benchmarking framework for machine learning ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10790918/]
• 115. Feature Extraction and Classification Methods for Hybrid ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC6032997/]
• 116. Classification Algorithm for fNIRS-based Brain Signals ... [https://www.sciencedirect.com/science/article/abs/pii/S0306452224000617]
• 117. Fusion of fNIRS and fMRI data: identifying when and where ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC3797402/]
• 118. Drug Development Tool (DDT) Qualification Programs [https://www.fda.gov/drugs/development-approval-process-drugs/drug-development-tool-ddt-qualification-programs]
• 119. Has FDA's Drug Development Tools Qualification Program ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12181102/]
• 120. Biomarker Regulatory Considerations [https://www.nordicbioscience.com/why-biomarkers/biomarker-regulatory-considerations/]
• 121. A functional neuroimaging biomarker of mild cognitive ... [https://www.nature.com/articles/s44400-025-00018-y]
• 122. Biomarker Discovery: Types, Phases and Applications [https://lifesciences.danaher.com/us/en/library/biomarker-discovery.html]
• 123. Development and Validation of a Brain Aging Biomarker in ... [https://aging.jmir.org/2025/1/e73004]