Optimizing EEG-fNIRS Sensor Placement: A Comprehensive Guide for Simultaneous Acquisition in Clinical Research

Abstract

This article provides a detailed guide for researchers and drug development professionals on achieving optimal sensor placement for simultaneous EEG-fNIRS acquisition. It covers the foundational principles of both modalities, exploring their synergistic potential and the core challenge of designing compatible montages. The content delves into practical methodologies, including the use of integrated caps and 3D coordinate systems, and addresses critical troubleshooting steps for mitigating common hardware and signal artifacts. Furthermore, it reviews validation frameworks and comparative analyses that demonstrate the enhanced spatiotemporal resolution and clinical utility of hybrid systems in areas such as Brain-Computer Interfaces (BCIs) and neurorehabilitation. The goal is to equip scientists with the knowledge to design robust, reliable, and effective multimodal neuroimaging studies.

The Synergistic Principles of EEG and fNIRS: Why Combine Them?

Understanding the brain's complex activity requires tools that can capture its rapid electrical fluctuations and the metabolic changes that support them. Electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) are two non-invasive neuroimaging techniques that provide complementary insights into these processes by measuring fundamentally different physiological phenomena [1] [2]. EEG captures the brain's direct electrical activity with millisecond temporal precision, while fNIRS measures the slower hemodynamic responses coupled to neural activity through neurovascular coupling [2]. This application note explores the core biophysical principles of these modalities and provides detailed protocols for their integrated use in simultaneous research, with particular emphasis on sensor placement compatibility—a critical consideration for obtaining clean, artifact-free data from both systems simultaneously.

Core Principles and Comparative Analysis

The Biophysical Basis of EEG

EEG measures the electrical potential generated by the synchronized firing of populations of cortical pyramidal neurons [2]. When these neurons fire synchronously, their post-synaptic potentials summate sufficiently to propagate through the skull and scalp, where they can be detected as tiny voltage fluctuations (typically in the microvolt range) by electrodes placed on the scalp surface [1] [2]. The resulting signal represents the macroscopic electrical activity of the brain, which can be divided into various oscillatory rhythms (e.g., theta: 4-7 Hz, alpha: 8-14 Hz, beta: 15-25 Hz, gamma: >25 Hz) that correlate with different cognitive states and functions [2].

The Biophysical Basis of fNIRS

fNIRS is an optical neuroimaging technique that leverages the relative transparency of biological tissue to near-infrared light (wavelengths 600-1000 nm) to measure hemodynamic changes in the cerebral cortex [2] [3]. It detects changes in the concentration of oxygenated hemoglobin (HbO) and deoxygenated hemoglobin (HbR)—the primary light-absorbing chromophores in this wavelength range—which fluctuate in response to neural activity via neurovascular coupling [1] [2]. When a brain region becomes active, a complex physiological process directs increased blood flow to that region, typically resulting in increased HbO and decreased HbR, which fNIRS detects by measuring how much near-infrared light is absorbed as it passes through the tissue [2] [3].

Comparative Analysis of Technical Specifications

Table 1: Comparative analysis of EEG and fNIRS technical specifications and performance characteristics.

Feature	EEG (Electroencephalography)	fNIRS (Functional Near-Infrared Spectroscopy)
What It Measures	Electrical activity of neurons [1] [2]	Hemodynamic response (blood oxygenation levels) [1] [2]
Signal Source	Postsynaptic potentials in cortical neurons [1]	Changes in oxygenated and deoxygenated hemoglobin [1]
Temporal Resolution	High (milliseconds) [1] [2]	Low (seconds) [1] [2]
Spatial Resolution	Low (centimeter-level) [1] [2]	Moderate (better than EEG, but limited to cortex) [1] [2]
Depth of Measurement	Cortical surface [1]	Outer cortex (~1–2.5 cm deep) [1] [3]
Sensitivity to Motion	High – susceptible to movement artifacts [1]	Low – more tolerant to subject movement [1]

The Rationale for Integration and Neurovascular Coupling

The integration of EEG and fNIRS is powerfully motivated by the physiological phenomenon of neurovascular coupling [2] [3]. This mechanism describes the tight relationship between neural electrical activity and subsequent hemodynamic changes in the brain [2]. When neurons fire, they create an immediate electrical signature detectable by EEG. This activity increases the metabolic demand for oxygen and glucose, triggering a delayed hemodynamic response (over 2-6 seconds) that delivers oxygenated blood to the active region—a response measurable by fNIRS [1] [2].

This complementary relationship enables a more complete investigation of brain function than either modality can provide alone. EEG offers unparalleled temporal resolution to track rapid neural dynamics, while fNIRS provides superior spatial localization and is less susceptible to motion artifacts [2]. Together, they enable researchers to study both the initial electrical events and their metabolic consequences, providing a more comprehensive picture of brain activity across different temporal and spatial scales [4] [2].

Experimental Protocols for Simultaneous EEG-fNIRS Recording

Protocol 1: Semantic Decoding During Mental Imagery Tasks

This protocol is adapted from a study investigating semantic neural decoding to differentiate between imagined categories (animals vs. tools) during various mental imagery tasks [5].

Research Objective: To determine whether simultaneous EEG-fNIRS can distinguish between semantic categories (animals vs. tools) during silent naming and sensory-based imagery tasks.

Participant Preparation:

• Screen participants for eligibility (e.g., right-handed, native English speakers for language tasks, normal or corrected-to-normal vision) [5].
• Obtain informed consent according to institutional ethical guidelines.
• Measure head circumference and select appropriate cap size.

Equipment and Reagent Setup:

• Table 2 lists the essential materials and their functions for this protocol.

Table 2: Research reagents and essential materials for simultaneous EEG-fNIRS recording.

Item	Function/Application	Specifications/Notes
Integrated EEG-fNIRS Cap	Provides stable platform for co-located sensor placement [4].	Uses international 10-20 system for placement; ensures optodes do not obscure electrodes.
EEG Amplifier	Amplifies microvolt-scale electrical signals from the scalp.	Multi-channel system; sync capability with fNIRS hardware.
fNIRS Continuous Wave System	Emits NIR light and detects attenuation to calculate HbO/HbR [2] [3].	Typically uses two wavelengths (~690nm, ~830nm); laser/LED sources.
Electrolyte Gel	Ensures conductive connection between scalp and EEG electrodes.	Apply carefully to avoid short-circuiting; use minimal gel near fNIRS optodes.
fNIRS Optode Holders	Maintains optical contact with scalp for consistent light transmission.	Ensures consistent source-detector distance; typically 3 cm for cortical sensitivity [3].
Stimulus Presentation Software	Presents visual cues and records event markers.	Sends synchronization pulses (TTL) to both EEG and fNIRS systems.

Simultaneous Sensor Placement Procedure:

• Fit the integrated cap on the participant, aligning the Cz electrode with the vertex of the head [4].
• For EEG: Carefully abrade the scalp at electrode sites and inject conductive gel to achieve impedances below 10 kΩ [1].
• For fNIRS: Position optodes to ensure firm but comfortable contact with the scalp. Verify that no fNIRS optode holders are placing pressure on EEG electrodes [4].
• Verify that all fNIRS source-detector pairs are functional and check signal quality before proceeding.

Experimental Task:

• Present participants with images of animals and tools in a randomized order.
• For each trial, instruct participants to perform one of four mental tasks (randomized across blocks):
  • Silent Naming: Silently name the object in their mind.
  • Visual Imagery: Visualize the object in their mind.
  • Auditory Imagery: Imagine the sounds the object makes.
  • Tactile Imagery: Imagine the feeling of touching the object [5].
• Each mental task period should last 3-5 seconds, during which participants must minimize movement [5].

Data Acquisition Parameters:

• EEG: Sampling rate ≥ 500 Hz, appropriate online filters (e.g., 0.1-100 Hz).
• fNIRS: Sampling rate ≥ 10 Hz, record changes in HbO and HbR concentrations.
• Synchronization: Use hardware TTL pulses or shared clock system to synchronize EEG and fNIRS data streams with stimulus events [1] [4].

Protocol 2: Assessing Implicit Learning During Cognitive Tasks

This protocol is based on a study using simultaneous EEG-fNIRS to identify neural correlates of implicit learning during a serial reaction time task [6].

Research Objective: To identify electrophysiological and hemodynamic biomarkers of implicit learning in healthy adults.

Participant Preparation: Follow the same preliminary steps as Protocol 1.

Equipment Setup: Utilize the same core equipment as listed in Table 2.

Experimental Task (Serial Reaction Time Task):

• Participants respond to a sequence of visual cues (e.g., colored boxes) by pressing corresponding keys.
• The relationship between cues and correct responses is governed by a set of rules, one of which is hidden and designed to be learned implicitly through repetition.
• The task consists of multiple blocks with embedded structured sequences and random sequences to test for learning.
• Record behavioral data (reaction time and accuracy) simultaneously with neural data.

Post-Experiment Procedure:

• Conduct a structured interview to assess participants' explicit awareness of the hidden rule.
• Classify participants into "Implicit Learning" or "No Implicit Learning" groups based on their verbal reports [6].

Data Analysis Workflow:

• Preprocess EEG and fNIRS data through separate pipelines.
• For EEG: Extract event-related potentials (ERPs) and time-frequency components (e.g., theta, alpha power).
• For fNIRS: Analyze HbO and HbR concentration changes during task periods compared to baseline.
• Apply data fusion techniques (e.g., jICA, canonical correlation analysis) to identify coupled electrical-hemodynamic features that distinguish the two groups [1] [2].

Sensor Placement Compatibility: Critical Considerations and Protocols

Successful simultaneous EEG-fNIRS recording hinges on resolving the technical and physical challenges of sensor co-location. The primary goal is to achieve optimal signal quality from both modalities without cross-interference [4].

Physical Integration Strategies

Several helmet fusion designs have been developed for simultaneous operation:

• Integrated Caps with Pre-Defined Openings: High-density EEG caps with pre-defined fNIRS-compatible openings allow for optimal placement of both sensor types [1] [4]. This approach ensures consistent inter-sensor distances and proper scalp contact.

• 3D-Printed Custom Helmets: For precise requirements, 3D-printed custom helmets tailored to individual head morphology provide the most accurate and stable sensor placement, though at a higher cost [4].

• Thermoplastic Custom Helmets: Composite polymer cryogenic thermoplastic sheets offer a cost-effective alternative that can be molded to fit individual head shapes when heated (~60°C), providing good stability upon cooling [4].

Spatial Configuration Guidelines

• Reference System: Both systems typically use the international 10-20 system for standardized sensor placement, facilitating data co-registration and interpretation [1] [4].
• Optode-Electrode Priority: When conflict arises, prioritize EEG electrode placement for electrical signal integrity, as fNIRS optodes are more flexible in exact positioning within a general region.
• Contact Pressure: Ensure firm but comfortable contact for both electrode and optodes. Excessive pressure can cause discomfort and physiological artifacts, while insufficient contact degrades signal quality [4].

Motion Artifact Mitigation

• Secure Fitting: Use tight but comfortable cap fittings to minimize relative movement between sensors and scalp [1].
• Task Design: For tasks involving inevitable movement, consider simplifying the design or accepting that EEG data quality may be compromised while fNIRS remains viable [1].
• Post-Hoc Correction: Apply motion correction algorithms during preprocessing for both modalities, using accelerometer data if available [1].

EEG and fNIRS provide complementary, non-invasive windows into brain function by measuring its electrical and hemodynamic activity, respectively. Their successful integration in simultaneous recordings offers a powerful approach to studying brain dynamics across multiple spatiotemporal scales. The protocols and considerations outlined herein provide a framework for designing and conducting simultaneous EEG-fNIRS studies, with particular attention to the critical aspect of sensor placement compatibility. As this multimodal approach continues to evolve, it holds significant promise for advancing our understanding of brain function in both research and clinical applications, from brain-computer interfaces and cognitive neuroscience to drug development and clinical monitoring [5] [4] [7].

The integration of Electroencephalography (EEG) and functional Near-Infrared Spectroscopy (fNIRS) represents a transformative approach in neuroimaging, creating a system where the whole is significantly greater than the sum of its parts. These two non-invasive technologies capture complementary aspects of brain activity: EEG measures electrical potentials from synchronized neuronal firing with millisecond temporal resolution, while fNIRS detects hemodynamic responses correlated with neural activity through optical measurements, offering superior spatial localization [8] [3]. This complementary relationship is fundamentally rooted in their measurement of linked physiological processes—electrical neuronal activity and the subsequent hemodynamic response—connected through neurovascular coupling mechanisms [8] [9].

The fusion of these modalities is particularly valuable for advancing brain-computer interfaces (BCIs), cognitive neuroscience research, and clinical applications ranging from stroke rehabilitation to monitoring neurological disorders [4] [10]. Unlike more restrictive neuroimaging methods like fMRI or MEG, combined EEG-fNIRS systems offer portability, relatively low cost, and significantly reduced physical constraints, making them ideal for studying brain function in naturalistic environments and across diverse populations [4] [3]. This integration enables researchers to capture both the rapid neural dynamics detectable through electrical signals and the more localized cortical activation patterns revealed by hemodynamic responses, providing a more complete picture of brain function than either modality could deliver independently.

Technical Foundations: How EEG and fNIRS Complement Each Other

Fundamental Principles and Characteristics

Electroencephalography (EEG) operates by measuring electrical potentials generated by synchronized neuronal activity through electrodes placed on the scalp surface. This neuroelectrical activity manifests with exceptional temporal resolution at the millisecond level, allowing researchers to capture rapid neural dynamics including event-related potentials, oscillatory patterns, and coherence across neural networks [8] [4]. However, as electrical signals pass through multiple layers of non-neural tissue including cerebrospinal fluid, skull, and scalp, they become significantly attenuated and spatially blurred, resulting in limited spatial resolution of approximately 2 centimeters [5] [4]. This fundamental limitation makes precise localization of neural generators challenging from scalp recordings alone.

Functional Near-Infrared Spectroscopy (fNIRS) employs optical principles to measure hemodynamic responses correlated with neural activity. By emitting near-infrared light (typically at wavelengths between 650-950 nm) through the scalp and detecting the backscattered light, fNIRS can quantify concentration changes in oxygenated hemoglobin (Δ[HbO]) and deoxygenated hemoglobin (Δ[HbR]) in cortical tissues [11] [3]. This optical measurement technique provides several advantages: better spatial localization (5-10 mm resolution) than EEG, direct measurement of both hemoglobin species, and relative robustness to motion artifacts [10] [3]. However, fNIRS tracks the slower hemodynamic response to neural activity, which unfolds over seconds rather than milliseconds, resulting in fundamentally inferior temporal resolution compared to EEG [8] [3].

Table 1: Comparative Technical Specifications of EEG and fNIRS

Parameter	EEG	fNIRS
Temporal Resolution	Millisecond level (~1-1000 Hz) [5] [4]	~0.1-10 Hz, limited by hemodynamic response [8] [3]
Spatial Resolution	~2 cm, limited by volume conduction [5] [4]	5-10 mm, determined by source-detector separation [10] [3]
Measured Physiological Signals	Electrical potentials from synchronized neuronal firing [4]	Hemodynamic changes (Δ[HbO] and Δ[HbR]) [11] [3]
Depth Sensitivity	Primarily cortical, with sensitivity to deeper sources limited by spatial spread [4]	Superficial cortex (up to ~2-3 cm depth) [11] [3]
Primary Artifact Sources	Ocular, muscle, cardiac, and movement artifacts [8]	Scalp hemodynamics, motion, and systemic physiology [8] [11]
Typical Sampling Rates	100-2000 Hz [10] [9]	2-100 Hz, commonly ~10 Hz [10] [9]

Neurovascular Coupling: The Biological Bridge

The complementary nature of EEG and fNIRS measurements is fundamentally grounded in the biological process of neurovascular coupling—the mechanism that links neural activity to subsequent changes in cerebral blood flow and oxygenation [8] [9]. When a brain region becomes active, the increased metabolic demands of firing neurons trigger a complex signaling cascade that ultimately leads to increased blood flow to that region, delivering oxygen and nutrients. This hemodynamic response manifests with a characteristic time delay of 1-6 seconds after the initial neural activation [8] [3]. This temporal relationship means that EEG captures the initial electrical neural events, while fNIRS detects the subsequent hemodynamic consequences, making these modalities inherently complementary rather than redundant.

Recent studies investigating structure-function relationships in brain networks using simultaneous EEG-fNIRS recordings have demonstrated that the functional information captured by both modalities shows consistency in early sensory cortical hierarchy but may diverge in higher-order association areas [9]. This suggests that the coupling between electrical and hemodynamic activities varies across different brain regions, following the unimodal to transmodal organizational gradient of the cortex [9]. Understanding these regional variations in neurovascular coupling is essential for proper interpretation of multimodal data and represents an active area of research in neuroimaging.

Practical Integration: Hardware and Signal Acquisition

Sensor Placement and Hardware Integration

Successful simultaneous EEG-fNIRS recording requires careful consideration of sensor placement compatibility to ensure both modalities can acquire quality data from the regions of interest without interference. The fundamental challenge lies in the limited scalp real estate that must be shared between EEG electrodes and fNIRS optodes while maintaining proper coverage of target brain regions [12] [13]. The most common approach involves integrating both sensor types into a single headcap, with EEG electrodes typically placed according to international 10-20 or 10-5 systems, and fNIRS optodes positioned in between at appropriate source-detector distances [12] [13].

A critical technical consideration for fNIRS is maintaining proper inter-optode distance (IOD), typically around 30 mm for adults, which determines the depth sensitivity and penetration of the optical measurement [3] [13]. Shorter distances (~15 mm) primarily sensitivity to extracerebral layers (skin, skull), while longer distances (30-40 mm) enable sampling of cerebral cortical tissues [3]. Modern integration approaches include using customized headcaps with combined holders that accommodate both EEG electrodes and fNIRS optodes in precise spatial configurations [4] [13]. For high-density measurements, 3D-printed customized helmets or cryogenic thermoplastic sheets can provide optimal positioning and stability, though at increased cost and complexity [4].

Table 2: Equipment Specifications for Simultaneous EEG-fNIRS Recording

Component	Specifications	Purpose/Function
EEG Amplifier	g.HIamp, g.Nautilus, g.USBamp (g.tec); Sampling: 256-1000 Hz [12] [10]	Signal acquisition and amplification of electrical potentials
fNIRS System	Continuous wave systems (NirScan, NIRSport2); Sampling: 11-12.5 Hz; Wavelengths: 760 & 850 nm [10] [9]	Measurement of hemodynamic changes via light absorption
EEG Electrodes	Active wet or hybrid dry electrodes (g.SCARABEO); 16-64 channels [12]	Scalp contact for electrical signal acquisition
fNIRS Optodes	Sources (LEDs/lasers) & detectors; Typical power: ~1mW [3]	Light emission and detection through scalp and brain tissue
Headcap	Neoprene fabric with optode/electrode holders; Light-absorbing properties [12] [13]	Secure sensor placement, stability, and light blocking
Synchronization	Event markers from E-Prime, LabStreamingLayer [10] [9]	Temporal alignment of multimodal data streams

Signal Acquisition and Synchronization

Precise temporal synchronization between EEG and fNIRS data streams is essential for meaningful multimodal analysis. There are two primary approaches to synchronization: (1) separate systems synchronized via host computer with external triggers, and (2) integrated systems with unified processors that simultaneously handle both data streams [4]. While the first approach offers flexibility in using existing equipment, the second provides more precise synchronization and streamlined data handling [4]. Modern systems typically use shared event markers from stimulus presentation software (e.g., E-Prime) that simultaneously trigger both recording systems during experimental paradigms [10] [9].

The significant disparity in typical sampling rates between EEG (often 256-1000 Hz) and fNIRS (typically 10-12.5 Hz) necessitates careful consideration in data analysis [10] [9]. In practice, the EEG system often acts as the "master" device in synchronized setups due to its higher sampling frequency requirements [12]. Successful integration requires ensuring that the electronic systems do not interfere with each other—thankfully, fNIRS as an optical technique typically doesn't electrically interfere with EEG, though proper grounding and maintaining low skin-electrode impedance can minimize potential artifacts [13].

Experimental Design and Protocol Implementation

Protocol for Motor Imagery Paradigms

Motor imagery (MI) represents one of the most established and clinically relevant paradigms for combined EEG-fNIRS research, particularly in brain-computer interfaces and neurorehabilitation [10]. The following protocol outlines a standardized approach for upper limb motor imagery experiments:

Participant Preparation and Setup: Begin with explaining the motor imagery concept to participants, emphasizing that it involves mentally simulating movements without physical execution. For enhanced task engagement, incorporate a grip strength calibration procedure using a dynamometer or stress ball to reinforce tactile and force-related aspects of the movement [10]. Proper cap placement is critical—use the international 10-20 system for EEG electrode placement and position fNIRS optodes over the primary sensorimotor cortex (C3, Cz, C4 regions) with appropriate inter-optode distances (typically 30mm for adults) [10] [13]. Ensure proper light coupling for fNIRS and electrode impedances below 10 kΩ for EEG quality.

Experimental Paradigm Structure: Implement a block design with the following structure for each trial: (1) Visual cue presentation (2 seconds) displaying a directional arrow (left/right) indicating the required MI; (2) Execution phase (10 seconds) where participants perform kinesthetic MI of the corresponding hand grasping movement at approximately one imagined grasp per second while fixating on a central cross; (3) Inter-trial interval (15 seconds) with blank screen for rest [10]. Include at least 30 trials per session (15 left/right each), with multiple sessions separated by sufficient rest intervals to mitigate fatigue. Begin with baseline recordings: 1-minute eyes-closed followed by 1-minute eyes-open states before task initiation [10].

Data Acquisition Parameters: Set EEG sampling rate to 256 Hz or higher with appropriate bandpass filtering (e.g., 0.5-40 Hz). Configure fNIRS sampling at 10 Hz or higher using dual wavelengths (typically 760 and 850 nm) to compute hemoglobin concentration changes [10]. Ensure precise synchronization between modalities using shared event markers from stimulus presentation software.

Protocol for Semantic Decoding Paradigms

Semantic decoding using combined EEG-fNIRS offers promising avenues for developing more intuitive brain-computer interfaces for communication [5]. The following protocol enables investigation of neural representations during semantic category processing:

Stimuli and Task Design: Select appropriate visual stimuli representing distinct semantic categories (e.g., animals vs. tools) [5]. Present images against a neutral background with standardized size and luminance. Implement multiple mental tasks: (1) Silent naming—participants silently name the displayed object; (2) Visual imagery—participants visualize the object without focusing on the specific image presented; (3) Auditory imagery—participants imagine sounds associated with the object; (4) Tactile imagery—participants imagine the feeling of touching the object [5].

Trial Structure and Timing: Structure each trial as follows: (1) Stimulus presentation (3-5 seconds) showing the category image; (2) Mental task period (3-5 seconds) where participants perform the cued imagery task; (3) Inter-trial interval (15-20 seconds) for baseline recovery [5]. Use a block design with randomized task order across participants. Include sufficient trials per condition (typically 15-20) to ensure adequate statistical power for decoding analyses.

Data Acquisition Considerations: Position fNIRS optodes over language-related regions (inferior frontal gyrus, temporal cortex) alongside standard EEG placements [5]. For semantic tasks requiring higher cognitive processing, extend mental task periods to 5 seconds to capture the slower hemodynamic response more completely [5]. Implement appropriate artifact removal strategies, particularly for EEG during visual tasks where ocular artifacts are prominent.

Data Processing and Analytical Approaches

Preprocessing Pipelines

Robust preprocessing is essential for extracting meaningful signals from both modalities while addressing their unique artifact profiles:

EEG Preprocessing: Apply bandpass filtering (0.5-40 Hz) to remove slow drifts and high-frequency noise. Identify and remove artifacts using independent component analysis (ICA) with a focus on ocular, cardiac, and muscle artifacts [8]. Implement artifact subspace reconstruction (ASR) for continuous data cleaning, particularly important for real-time applications [11]. For event-related analyses, epoch data around stimulus events and apply baseline correction.

fNIRS Preprocessing: Convert raw light intensity to optical density, then to hemoglobin concentration changes using the modified Beer-Lambert law [9]. Apply bandpass filtering (0.01-0.2 Hz) to isolate hemodynamic responses from physiological noise (cardiac ~1 Hz, respiratory ~0.3 Hz) [11] [9]. Implement motion artifact correction using wavelet-based methods or robust regression [11]. Incorporate short-separation regression to remove superficial scalp hemodynamics when short-distance channels are available [8].

Multimodal Quality Metrics: Calculate scalp-coupling index (SCI) for fNIRS to identify channels with poor optode-scalp contact [9]. For EEG, monitor electrode impedances throughout recording. Reject channels with excessive noise in either modality before further analysis.

Multimodal Fusion Strategies

Data fusion represents the core analytical challenge and opportunity in combined EEG-fNIRS research:

Data-Level Fusion: Concatenate features from both modalities into a unified feature space for machine learning applications [8] [10]. This approach requires careful normalization to address the different scales and dimensionalities of EEG and fNIRS data. For classification tasks, this strategy has demonstrated 5-10% improvement in accuracy compared to unimodal systems [10].

Model-Based Fusion: Implement joint generative models that incorporate neurovascular coupling principles to estimate underlying neural activity [8]. These approaches can include dynamic causal modeling or state-space models that formally represent the relationship between electrical neural activity and hemodynamic responses.

Decision-Level Fusion: Process each modality independently through separate pipelines before combining results at the decision stage through voting schemes or confidence-weighted integration [8]. This approach offers robustness to modality-specific failures but may fail to capture more subtle cross-modal interactions.

Asymmetric Fusion: Use one modality to inform processing of the other—for instance, using EEG-derived neural events to inform fNIRS general linear model analysis, or using fNIRS spatial information to constrain EEG source localization [8].

The Scientist's Toolkit: Essential Research Reagents and Materials

Table 3: Essential Equipment and Software for EEG-fNIRS Research

Category	Item	Specification/Purpose	Example Brands/Platforms
Hardware	EEG Amplifier	32-64 channels; Sampling ≥256 Hz; Referenced recording	g.HIamp, g.Nautilus, BrainAmp [12] [10]
	fNIRS System	Continuous wave; 2+ wavelengths (760, 850 nm); Sampling ≥10 Hz	NirScan, NIRSport2, Artinis systems [10] [13]
	Integrated Caps	Neoprene fabric; Combined optode/electrode holders; Light-blocking	Custom designs with 10-20 system markers [12] [13]
Software	Stimulus Presentation	Precision timing, trigger generation	E-Prime, PsychToolbox, Presentation [10] [9]
	Data Acquisition	Synchronized multi-modal recording	LabStreamingLayer, BrainVision Recorder [10]
	Analysis Platforms	Preprocessing, fusion, visualization	MNE-Python, Brainstorm, Homer2, NIRS-KIT [9]
Accessories	EEG Electrodes	Active/passive wet electrodes or hybrid dry systems	g.SCARABEO, sintered Ag/AgCl [12]
	fNIRS Optodes	Source-detector pairs with spring-loaded holders	Custom designs for specific systems [13]
	3D Digitization	Optode/electrode position registration	Polhemus, Structure Sensor, photogrammetry [9]

Applications and Future Directions

The integration of EEG and fNIRS has demonstrated particular value across several research and clinical domains:

Brain-Computer Interfaces and Neurorehabilitation: Hybrid EEG-fNIRS BCIs have shown significantly improved classification accuracy compared to unimodal systems, particularly for motor imagery paradigms [10]. In clinical populations such as intracerebral hemorrhage patients, this multimodal approach can track both electrical abnormalities and hemodynamic impairments, providing a more complete assessment for neurorehabilitation [10].

Cognitive Neuroscience Research: The combined temporal and spatial resolution enables investigation of complex cognitive processes including semantic decoding, working memory, and attention [5] [9]. Studies examining structure-function relationships using simultaneous recordings have revealed how electrical and hemodynamic networks align with underlying structural connectivity across different brain states [9].

Clinical Monitoring and Assessment: The portability and relatively low cost of combined systems make them ideal for bedside monitoring in various clinical conditions including ADHD, epilepsy, disorders of consciousness, and stroke recovery [4]. The ability to capture both rapid electrical events (e.g., seizures) and slower hemodynamic changes provides complementary information for diagnosis and treatment monitoring.

Future advancements in EEG-fNIRS integration will likely focus on improving hardware miniaturization and wireless capabilities, developing more sophisticated real-time processing algorithms, and establishing standardized analytical frameworks for multimodal data fusion [4] [11]. As these technologies continue to evolve, they hold the promise of making high-quality brain imaging more accessible, portable, and applicable to real-world environments beyond traditional laboratory settings.

Neurovascular coupling (NVC) describes the fundamental physiological process that links neural activity with subsequent changes in cerebral blood flow and hemodynamics [14] [15]. This connection forms the foundational basis for several functional neuroimaging techniques, including functional magnetic resonance imaging (fMRI) and functional near-infrared spectroscopy (fNIRS), which rely on hemodynamic changes as a proxy for neural activity [14]. The adult brain, while constituting only approximately 2% of total body weight, accounts for about 20% of the body's total energy consumption [15]. To meet this high metabolic demand, the brain requires a continuous supply of oxygen and glucose, delivered via cerebral blood flow [15]. The NVC mechanism mediates this delivery through a complex biological signaling cascade wherein synaptic activity triggers the release of vasoactive molecules that act on vascular smooth muscle cells and pericytes, ultimately resulting in changes to cerebral blood flow (CBF), cerebral blood volume (CBV), and the cerebral metabolic rate of oxygen (CMRO₂) [14] [15].

Table 1: Key Components of Neurovascular Coupling

Component	Description	Role in NVC
Neural Activity	Electrical signals from neuronal firing, particularly from pyramidal cells	Initiates the coupling process by triggering vasoactive signaling
Vasoactive Molecules	Nitric oxide (NO), prostaglandins, neuropeptides	Act as messengers that communicate between neurons and blood vessels
Vascular Response	Dilation/constriction of arterioles and capillaries	Directly alters blood flow and volume to active brain regions
Hemodynamic Response	Changes in oxygenated/deoxygenated hemoglobin	Measurable outcome used in fNIRS and fMRI imaging

The intricate cellular pathways underlying NVC involve multiple cell types and signaling molecules. As illustrated in Figure 1, neuronal signaling activates GABAergic interneurons, pyramidal neurons, and astrocytes by stimulating calcium (Ca²⁺) influx [15]. This calcium facilitates distinct signaling pathways in different cells: in GABAergic interneurons, it promotes nitric oxide (NO) synthesis; in pyramidal neurons, it facilitates arachidonic acid metabolism to prostaglandin E₂ (PGE₂); and in astrocytes, it triggers production of additional vasoactive molecules including PGE₂, EET, and 20-HETE [15]. These pathways collectively regulate vascular tone in response to neural activity.

EEG-fNIRS Integration Fundamentals

Complementary Modality Characteristics

The integration of electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) provides a powerful multimodal approach for studying neurovascular coupling by capturing complementary aspects of brain activity [16]. EEG measures electrical activity resulting from synchronized firing of cortical neurons, primarily pyramidal cells, providing millisecond-level temporal resolution ideal for tracking rapid neural dynamics [16] [17]. In contrast, fNIRS monitors cerebral hemodynamic responses by measuring changes in oxygenated (HbO) and deoxygenated hemoglobin (HbR) using near-infrared light, offering better spatial resolution for surface cortical areas but slower temporal resolution due to the inherent delay of the hemodynamic response (2-6 seconds) [16].

Table 2: Comparative Characteristics of EEG and fNIRS

Feature	EEG (Electroencephalography)	fNIRS (Functional Near-Infrared Spectroscopy)
What It Measures	Electrical activity of neurons	Hemodynamic response (blood oxygenation levels)
Signal Source	Postsynaptic potentials in cortical neurons	Changes in oxygenated and deoxygenated hemoglobin
Temporal Resolution	High (milliseconds)	Low (seconds)
Spatial Resolution	Low (centimeter-level)	Moderate (better than EEG, limited to cortex)
Depth of Measurement	Cortical surface	Outer cortex (~1-2.5 cm deep)
Sensitivity to Motion	High - susceptible to movement artifacts	Low - more tolerant to subject movement
Best Use Cases	Fast cognitive tasks, ERP studies, sleep research	Naturalistic studies, child development, motor rehab

Sensor Placement Compatibility

For simultaneous EEG-fNIRS recordings, proper sensor placement is crucial for effective data collection and interpretation. Both systems typically use the international 10-20 system for electrode and optode placement, ensuring standardized positioning across subjects and studies [16]. Technical considerations for compatible sensor placement include:

• High-density EEG caps with pre-defined fNIRS-compatible openings can accommodate both modalities simultaneously [16]
• Some fNIRS systems are designed to be embedded within EEG caps or mounted using optode holders that avoid electrode contact points [16]
• Careful cap fitting is essential to minimize motion artifacts while ensuring participant comfort [16]
• Avoiding sensor overlap prevents interference between EEG electrodes and fNIRS optodes [16]

Modern experimental setups have successfully implemented integrated systems, such as a custom-designed hybrid EEG-fNIRS cap with 32 EEG electrodes and 62 fNIRS optodes (32 sources, 30 detectors) arranged to achieve 90 fNIRS measurement channels through source-detector pairing [10]. The systematic expansion of the international 10-20 system for EEG electrode placement ensures comprehensive coverage of major functional cortical areas, while fNIRS optodes follow an anatomically-guided configuration aligned with functional neuroanatomical parcellations [10].

Experimental Protocols for Simultaneous EEG-fNIRS

Visual Cognitive Motivation Study Protocol

A comprehensive simultaneous EEG-fNIRS protocol for investigating visual cognitive processing was established by Phukhachee and colleagues [18]. This protocol examines the neural correlates of intentional memory formation using a multimodal approach:

Participant Preparation and Setup

• Recruit participants without history of visual perception or memory disorders (sample size determined using Lemeshow method) [18]
• Apply simultaneous EEG and fNIRS sensors according to the international 10-20 system, ensuring proper contact and signal quality
• Position participants for optimal viewing of visual stimuli at appropriate distance

Experimental Paradigm

• The experiment comprises two parts: a cognitive motivation task followed by a recognition test [18]
• During the cognitive task, present random visual stimuli of indoor or outdoor scenes from the Scene UNderstanding database [18]
• Display each stimulus for 3 seconds (attention span), followed by a fixation cross for 9 seconds (decision period) [18]
• Instruct participants to freely decide whether to remember each scene during the decision period [18]

Data Collection Parameters

• Record EEG and fNIRS signals simultaneously throughout the cognitive task [18]
• For EEG analysis, focus on event-related potentials (ERPs) during the first second following stimulus presentation [18]
• For fNIRS analysis, examine hemodynamic responses during the subsequent 9-second decision period [18]
• Categorize trials based on motivation and subsequent recognition: Want to Remember and Remembered (RR), Want to Remember but Forgot (RF), Did Not Want to Remember but Remembered (FR), and Did Not Want to Remember and Forgot (FF) [18]

Motor Execution, Observation, and Imagery Protocol

Another established protocol examines neural activity during motor execution, observation, and imagery using simultaneous EEG-fNIRS recordings [19]:

Participant Preparation

• Recruit healthy adult participants (18-65 years) without recent concussion history [19]
• Use a 24-channel fNIRS system embedded within a 128-electrode EEG cap [19]
• Digitize fNIRS optode positions relative to anatomical landmarks (nasion, inion, preauricular points) using a 3D magnetic space digitizer [19]

Experimental Conditions

• Motor Execution (ME): Participants grasp, lift, and move a cup approximately two feet toward themselves using their right hand upon audio command [19]
• Motor Observation (MO): Participants observe an experimenter performing the cup-moving task [19]
• Motor Imagery (MI): Participants mentally rehearse the cup-moving task without physical movement [19]

Data Acquisition and Analysis

• Collect simultaneous EEG and fNIRS data throughout all conditions
• Use structured sparse multiset Canonical Correlation Analysis (ssmCCA) to fuse fNIRS and EEG data and identify brain regions consistently detected by both modalities [19]
• Compare unimodal and multimodal results to validate findings across modalities

Figure 1: Experimental workflow for simultaneous EEG-fNIRS studies, illustrating the sequential steps from participant preparation to data analysis.

Signaling Pathways in Neurovascular Coupling

The neurovascular coupling process involves sophisticated signaling pathways between neurons and blood vessels. Research using optogenetics and microscopy in mice has revealed cell-specific contributions to the vascular response [14] [15]:

• The first rapid dilation in the vascular response is caused by NO-interneurons [14]
• The main dilation during longer stimuli is primarily mediated by pyramidal neurons [14]
• The post-peak undershoot characteristic of the hemodynamic response is caused by NPY-interneurons [14]

These specific cellular mechanisms are conserved across species and form the biological basis for interpreting combined EEG-fNIRS measurements [14]. The interplay between different neuronal subpopulations produces the characteristic biphasic hemodynamic response function (HRF) observed in neuroimaging, which features an initial peak at approximately 3-6 seconds after a brief stimulus, followed by a post-peak undershoot, with the entire response typically lasting 15-20 seconds [14].

Figure 2: Neurovascular coupling signaling pathways showing the progression from neural activity to hemodynamic response through cell-type specific mechanisms.

The Scientist's Toolkit: Research Reagent Solutions

Table 3: Essential Research Materials for EEG-fNIRS Neurovascular Coupling Studies

Research Tool	Function/Application	Examples/Specifications
Hybrid EEG-fNIRS Caps	Simultaneous sensor placement	Custom designs with 32 EEG electrodes + 62 fNIRS optodes; International 10-20 system compatibility [10]
fNIRS Systems	Hemodynamic response measurement	Continuous-wave systems (e.g., NirScan); 695 nm & 830 nm wavelengths; ~10 Hz sampling rate [10] [19]
EEG Amplifiers	Neural electrical activity recording	g.HIamp amplifier; 256 Hz sampling rate; 32+ channels [10]
3D Digitizers	Precise optode localization	Fastrak Polhemus system; records coordinates relative to nasion, inion, preauricular landmarks [19]
Stimulus Presentation	Experimental paradigm delivery	E-Prime 3.0; synchronized event markers for both EEG and fNIRS [10]
Data Fusion Algorithms	Multimodal data integration	Structured sparse multiset CCA (ssmCCA); joint Independent Component Analysis (jICA) [20] [19]
Motion Correction Tools	Artifact reduction	Preprocessing algorithms for minimizing movement artifacts in both modalities [16]

Advanced Data Analysis and Fusion Techniques

The integration of EEG and fNIRS data requires sophisticated analytical approaches that accommodate their different temporal characteristics and physiological origins. Several methodological frameworks have been developed specifically for multimodal neurovascular data:

Neurovascular Coupling Analysis Method

Lin et al. (2023) developed a dedicated EEG-fNIRS analysis framework to investigate cognitive-motor interference through neurovascular coupling [20]. This approach:

• Extracts task-related components for EEG and fNIRS signals separately before analyzing their correlation [20]
• Employs within-class similarity and between-class distance indicators to validate the analysis framework [20]
• Reveals how extra cognitive interference in dual-tasking decreases neurovascular coupling across theta, alpha, and beta rhythms [20]
• Demonstrates significantly higher classification performance compared to canonical channel-averaged methods [20]

Deep Learning Architectures for Multimodal Classification

Advanced deep learning approaches have shown promising results for classifying simultaneous EEG-fNIRS data. The Multimodal DenseNet Fusion (MDNF) model represents a significant technical advancement by [17]:

• Transforming EEG data into 2D representations using short-time Fourier transform (STFT) for temporal and spectral feature extraction [17]
• Applying transfer learning to extract discriminative features from transformed EEG data [17]
• Integrating EEG features with fNIRS-derived spectral entropy features [17]
• Achieving superior classification accuracy across various cognitive and motor imagery tasks compared to state-of-the-art methods [17]

This approach effectively bridges the feature representation gap between the temporal richness of EEG and spatial specificity of fNIRS, demonstrating the potential for clinical applications in neurodiagnostics and rehabilitation [17].

Simultaneous EEG-fNIRS recording provides a powerful multimodal framework for investigating neurovascular coupling, leveraging the complementary strengths of electrophysiological and hemodynamic measurement techniques. The integration of these modalities requires careful consideration of sensor placement compatibility, experimental design, and advanced data fusion methodologies. The protocols and analytical approaches outlined in this document establish standardized methods for studying the fundamental relationship between electrical and metabolic brain activity, with significant implications for basic cognitive neuroscience, clinical research, and therapeutic development. As multimodal integration methodologies continue to advance, simultaneous EEG-fNIRS is poised to yield increasingly nuanced insights into the neurovascular basis of brain function in both healthy and pathological states.

The prefrontal cortex (PFC) and sensorimotor cortex (SMC) are critical brain regions for understanding higher-order cognitive functions and motor control. Research using multimodal neuroimaging, particularly simultaneous electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS), has advanced our ability to investigate these regions in both healthy and clinical populations. These techniques leverage the complementary information provided by electrical neural activity (EEG) and hemodynamic responses (fNIRS), which are linked through neurovascular coupling—the relationship between neural electrical activity and subsequent changes in cerebral blood flow [2]. This application note details the anatomical significance, provides validated experimental protocols, and outlines technical considerations for studying the PFC and SMC using combined EEG-fNIRS.

Anatomical and Functional Significance

Prefrontal Cortex (PFC)

The PFC is central to executive functions, including cognitive control, working memory, and decision-making [21]. Key subregions include the dorsolateral PFC (DLPFC), involved in planning and regulation; the ventrolateral PFC (VLPFC), and the orbitofrontal cortex (OFC), which is implicated in emotion and reward. In clinical contexts, PFC dysfunction is a hallmark of various neurodevelopmental and neurodegenerative disorders. For instance, in Parkinson's disease (PD), the progression of cognitive impairment is strongly linked to altered activation and functional connectivity within the PFC and related networks [21].

Sensorimotor Cortex (SMC)

The SMC, encompassing the primary motor cortex (M1) and primary somatosensory cortex (S1), is responsible for motor execution and sensory processing. The premotor and supplementary motor areas (PMC/SMA) play a key role in motor planning and coordination [21]. Investigating the SMC is crucial for understanding motor deficits in conditions like PD and for developing rehabilitation strategies for stroke and other motor disorders.

Protocol for Simultaneous EEG-fNIRS Recordings

Combining EEG and fNIRS provides a comprehensive view of brain activity by capturing its electrical and hemodynamic aspects simultaneously [2] [22]. The following protocol ensures high-quality data acquisition.

Equipment and Setup

• EEG System: A high-density amplifier (e.g., g.Nautilus, g.USBamp) with active electrodes is recommended for its high temporal resolution and resistance to artifacts [22].
• fNIRS System: A continuous-wave fNIRS device (e.g., NIRSport2, ETG-ONE) with sources emitting light at 695-830 nm and detectors placed 20-30 mm apart to penetrate the cortex [21] [23] [22].
• Integrated Cap: Use a specialized cap that accommodates both EEG electrodes and fNIRS optodes. The material should be dark to block ambient light and ensure optode-scalp contact [22].
• Data Synchronization: The EEG amplifier, with its higher sampling rate, should act as the "master" device, streaming data in real-time to a computer where fNIRS data is synchronized [22].

Sensor Placement and Montage Design

The design principle is to interlace EEG electrodes between fNIRS optodes. A sample montage for investigating the PFC and SMC is detailed in Table 1.

Table 1: Example EEG-fNIRS Montage for PFC and SMC Investigation

Anatomical Region	EEG 10-20 Landmarks	Key fNIRS Channels	Targeted Function
Dorsolateral PFC (DLPFC)	F3, F4	Between F3-AF3, F4-AF4	Executive Function, Working Memory [21]
Medial PFC (mPFC)	FPz, AFz	Between FPz-AFz, AF3-AFz	Social/Emotional Processing [21]
Orbitofrontal Cortex (OFC)	FP1, FP2	Inferior to FP1/FP2	Reward, Decision-Making [21]
Primary Motor Cortex (M1)	C3, C4	Between C3-CP3, C4-CP4	Hand Motor Execution [21]
Premotor Cortex (PMC)	FC3, FC4	Between FC3-C3, FC4-C4	Motor Planning [21]
Supplementary Motor Area (SMA)	FCz, Cz	Between FCz-Cz	Motor Sequencing, Bimanual Coordination [21]

Experimental Procedure

• Participant Preparation: The setup for a combined 32-channel EEG and fNIRS system takes approximately 10 minutes with active electrode technology [22]. Ensure proper scalp contact for both EEG electrodes and fNIRS optodes.
• Task Paradigm: Participants perform tasks while seated. The Stroop task is highly effective for probing executive function and PFC activity [21].
  • Stroop Task Protocol: Present congruent (e.g., "BLUE" in blue font) and incongruent (e.g., "BLUE" in red font) stimuli in a randomized order. Each trial should display the word for 1500-2000 ms, followed by an inter-trial interval of 1000-1500 ms. Block designs (e.g., 30-second task blocks alternating with 30-second rest) are well-suited for fNIRS analysis.
• Data Acquisition: Record both signals simultaneously throughout the task and resting-state periods. Instruct participants to minimize head and body movements to reduce motion artifacts.

Data Analysis and Workflow

The analysis of concurrent EEG-fNIRS data can be performed using three primary approaches: EEG-informed fNIRS analysis, fNIRS-informed EEG analysis, or parallel analysis [2]. A typical workflow is outlined below.

Key Quantitative Findings from fNIRS Studies

fNIRS studies have revealed distinct patterns of cortical activation and functional connectivity (FC) across different clinical stages, particularly in Parkinson's disease (PD). These findings can serve as benchmarks for your own research.

Table 2: Stage-Specific fNIRS Findings in Parkinson's Disease During a Stroop Task [21]

Patient Group	Cortical Activation Pattern	Functional Connectivity (FC) Pattern	Association with Cognition
PD with Normal Cognition (PD-NC)	Not specified in results summary	Significantly enhanced interhemispheric connectivity compared to Healthy Controls (HCs)	Suggests early compensatory mechanisms
PD with Mild Cognitive Impairment (PD-MCI)	Significant hypoactivation in DLPFC, M1, and PMC	Extensive and pronounced interhemispheric connectivity	Suggests expanded cortical network as compensation for reduced activation
PD with Dementia (PDD)	Increased activation in mPFC, OFC, and DLPFC	Reduced connectivity among PMC, VLPFC, and OFC	Increased DLPFC activation correlated with poorer executive function

The Scientist's Toolkit: Research Reagents and Materials

This section lists essential materials and their functions for conducting simultaneous EEG-fNIRS studies.

Table 3: Essential Research Materials for EEG-fNIRS Studies

Item	Function / Purpose	Example / Specification
Biosignal Amplifier	Acquires and digitizes high-quality EEG signals with high temporal resolution.	g.Nautilus, g.USBamp [22]
fNIRS Sensor Module	Adds fNIRS measurement capability to the EEG amplifier for simultaneous recording.	g.SENSOR fNIRS [22]
Integrated Head Cap	Holds both EEG electrodes and fNIRS optodes in a stable, predefined configuration.	g.GAMMAcap with electrode and optode holder rings [22]
Active EEG Electrodes	Improve signal quality and reduce preparation time; can be wet or hybrid dry.	g.SCARABEO electrodes [22]
Conductive Electrode Gel	Ensures low impedance between the scalp and EEG electrodes for optimal signal quality.	Saline-based or abrasive gel
fNIRS Source Detectors	Emit near-infrared light and detect the reflected signal from the cortex.	LEDs/Lasers (695/830 nm) [23]
Stimulus Presentation Software	Presents experimental paradigms and sends triggers to synchronize data with task events.	Presentation, E-Prime, PsychoPy
Integrated Analysis Software	Processes and analyzes synchronized EEG and fNIRS data.	g.HIsys, MATLAB toolboxes (e.g., BBCI, NIRS-KIT)

Clinical Applications and Signaling Pathways

The integration of EEG and fNIRS is particularly valuable for clinical translation. The following diagram summarizes the pathway from neural activity to measurable signals and its application in diagnosing brain disorders.

In conditions like Alzheimer's disease and stroke, the integrity of neurovascular coupling is compromised [2]. This impairment can be detected as a discrepancy or abnormal relationship between the EEG and fNIRS signals, offering a potential diagnostic biomarker. Similarly, in autism spectrum disorder (ASD), fNIRS has revealed abnormal prefrontal activation and reduced functional connectivity in children, demonstrating its utility as a potential biomarker for neurodevelopmental disorders [23].

Designing Integrated Montages: Hardware and Configuration Strategies

Simultaneous electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) monitoring provides a comprehensive picture of brain function by capturing complementary aspects of neural activity—the millisecond-scale electrical discharges and the slower hemodynamic responses related to metabolic demand [24] [25]. The physical integration of these sensors into a single headcap is a critical step in the evolution of this multimodal field [24]. A well-designed integrated cap must overcome significant challenges, including mechanical interference between components, electrical crosstalk, and ensuring precise co-registration of the different signal types [24] [26]. This application note details the core design considerations for integrated EEG-fNIRS caps, surveys commercially available solutions, and provides a foundational protocol for researchers in neuroscience and drug development.

Core Design Considerations for Integrated Caps

The development of an integrated EEG-fNIRS cap involves addressing several intertwined technical and practical challenges to ensure data quality and participant comfort.

Mechanical Integration and Spatial Compatibility

The primary mechanical challenge is the competition for space on the scalp between EEG electrodes and fNIRS optodes (sources and detectors). This is particularly acute for studies involving populations with smaller head sizes, such as infants and children [24] [27].

• Cap Fabric and Fit: The cap base should be made of a lightweight, breathable, and durable fabric. A proper fit is crucial for stable sensor placement; some manufacturers offer different cap cuts (e.g., A-Cut for rounder heads, C-Cut for more oval heads) and fabric types (e.g., "High Precision" for reproducible positioning, "High Comfort" for sensitive skin or long-term recordings) to accommodate anatomical diversity [28] [29].
• Modular vs. Integrated Designs: Researchers can choose between two main cap styles:
  • Modular Caps: Feature holders into which loose electrodes and optodes are inserted. This offers flexibility in montage design and can lower the cost of entry for labs needing multiple cap sizes [28].
  • Integrated Caps: Come with a fixed channel layout and sensors fully braided into the fabric. These are always ready for use with minimal setup time and provide built-in cable management [28].
• Co-located Sensor Arrangement: For high spatial-temporal correspondence, a "patch" design can be employed where an EEG electrode is placed directly between a fNIRS source and detector, allowing both modalities to probe the same cortical location [26].

Electrical Integration and Crosstalk Mitigation

fNIRS systems often use rapidly switching currents to drive their light sources (LEDs or laser diodes), which can create electrical noise that interferes with the sensitive analog EEG signals [24].

• Shared Circuit Architecture: A fully integrated system uses a common control module and circuit board, which simplifies synchronization and allows for engineered crosstalk suppression [24] [30] [26].
• Source Switching Frequency: A key strategy is to set the fNIRS source switching frequency above the EEG band of interest (typically > 40 Hz, and often ≥ 100 Hz). This pushes the switching noise outside the relevant EEG spectrum, where it can be filtered out [24] [26].
• Shielding and Grounding: Proper electronic design, including shielding and grounding of the fNIRS driving circuits, is essential to minimize electromagnetic interference.

Signal Synchronization

Precise time-locking of EEG and fNIRS data streams is fundamental for correlating the fast electrical events with the slower hemodynamic changes.

• Hardware Synchronization: The most robust method uses a shared clock or a single analog-to-digital converter (ADC) for both modalities, ensuring sample-accurate alignment from the start [24] [30].
• Software Synchronization: When using discrete systems, synchronization can be achieved by sending and recording trigger pulses (e.g., TTL) in both data streams, which are then used for offline alignment [24] [25].

Commercial Availability and System Selection

The market offers solutions ranging from fully integrated wireless headsets to customizable cap systems that accommodate various amplifiers and electrodes.

Table 1: Comparison of Select Commercial Integrated EEG-fNIRS Solutions

Vendor / Product	Key Features	EEG Channels	fNIRS Channels	Notable Integration Aspects
g.tecg.Nautilus with g.GAMMAcap [31]	Wireless headset	8 to 64 (wet or dry)	8	fNIRS sensor easily inserts into cap; pre-defined placement over frontal/sensorimotor cortices.
ArtinisHeadcaps [29]	Compatible with various systems	Customizable	Customizable	Full-head caps with printed fNIRS/10-20 grids; pre-punched or custom hole options.
NIRxNIRScaps [32]	High flexibility	Customizable (active/passive)	Customizable (high-density)	Supports arbitrary layouts; integration with active/passive EEG and other modalities (tDCS).
Easycap [28]	Modular or integrated	19 to 128+	Customizable (via layout)	Broad range of sizes (preterm to adult); options for multimodal bespoke designs (e.g., BrainCaps).
BIOPACMedelOpt+ [33]	Exoskeleton headset	Up to 512 Hz sampling	64 to 128 theoretical	Flexible, adaptable headset fitting ages 4 to adult; unified with AcqKnowledge software.

Experimental Protocol: Prefrontal Cortex Activation During a Cognitive Task

This protocol outlines a standard procedure for conducting a simultaneous EEG-fNIRS study using a Stroop task to elicit cognitive conflict in the prefrontal cortex (PFC) [26].

Materials and Equipment

Table 2: Research Reagent Solutions and Essential Materials

Item	Function / Explanation
Integrated EEG-fNIRS Cap	The central hardware, combining electrodes and optodes in a single headgear for simultaneous data acquisition.
Conductive EEG Gel (for wet electrodes)	Facilitates electrical conduction between the scalp and Ag/AgCl electrodes, crucial for obtaining low impedance.
Abrasive Prep Gel	Gently exfoliates the skin to lower impedance at EEG electrode sites.
Alcohol Wipes & Gauze	For cleaning and drying the scalp before sensor application.
Optode Positioning Tool	Ensures fNIRS sources/detectors are placed at the correct 10-20 locations and maintain proper scalp contact.
Task Presentation Software	Presents the Stroop task stimuli and records behavioral responses (accuracy, reaction time).
Synchronization Trigger Box	Sends precise electronic markers from the stimulus computer to the EEG and fNIRS acquisition systems to align data streams.

Procedure

Step 1: Participant Preparation and Cap Fitting

• Measure Head Circumference: Use a flexible measuring tape to determine the correct cap size [29].
• Prepare Scalp: Part the participant's hair at the intended EEG electrode locations (Fp1, Fp2, etc.). Clean these areas with alcohol wipes and apply a small amount of abrasive gel if using wet electrodes to achieve impedances below 10 kΩ [28]. Thoroughly wipe away any residue.
• Don the Cap: Use the "dunking the head" technique [29]—scoop the forehead into the cap first, then pull the cap down over the rest of the head. Adjust the chin strap for a snug, comfortable fit.
• Position Sensors: For a PFC study, ensure the cap is positioned so that its sensors cover the forehead. Insert EEG electrodes into their designated holders and apply gel. Insert fNIRS optodes into their holders, ensuring they make firm contact with the scalp.

Step 2: System Setup and Signal Quality Check

• Connect Hardware: Link the cap's connectors to the EEG amplifier and fNIRS control unit. Connect the synchronization trigger box.
• Check EEG Quality: Verify that all electrode impedances are within an acceptable range (e.g., < 10 kΩ for active electrodes, < 5 kΩ for passive).
• Check fNIRS Quality: Use the system's software to inspect the signal quality for each fNIRS channel, ensuring light levels are sufficient and stable.

Step 3: Data Acquisition and Task Execution

• Synchronize Systems: Start recording on both the EEG and fNIRS systems. Initiate the stimulus presentation software, which will send synchronization triggers at the beginning of the task and at the onset of each trial.
• Run the Stroop Task:
  • Instruction: Inform the participant that they will see color words (e.g., "BLUE") printed in incongruent inks (e.g., the word "BLUE" printed in red ink) and must name the ink color as quickly and accurately as possible.
  • Task Blocks: The experiment should follow a block design, for example: 30-second baseline (fixation cross) → 3-minute Stroop task block → 30-second rest → 3-minute control task block (e.g., reading color words in black ink) → 30-second rest. Repeat 3-5 times.
  • Behavioral Recording: Log the participant's vocal responses and reaction times.

Step 4: Concluding the Session

• End the recordings in the EEG and fNIRS software.
• Carefully remove the cap from the participant.
• Clean the cap, electrodes, and optodes according to the manufacturer's guidelines (e.g., hand-washing in lukewarm water with mild detergent) [29].

Data Processing and Analysis Workflow

The following workflow outlines the core steps for processing the acquired multimodal data. Separate preprocessing pipelines are required before integration due to the fundamentally different nature of the signals [25].

Diagram 1: EEG-fNIRS Data Processing Workflow. This diagram illustrates the separate preprocessing pipelines for EEG and fNIRS data, followed by joint fusion and analysis after synchronization.

The Scientist's Toolkit: Technical Specifications

Table 3: Key Hardware Components and Their Specifications

Component	Types / Options	Key Considerations for Integration
EEG Electrodes	Wet (Ag/AgCl): Standard, reliable, low impedance. Dry: Faster setup, no gel, but higher impedance. Active: Include a pre-amplifier at the electrode site to reduce noise [24] [28].	Wet electrodes are preferred for high signal quality but are less suitable for long-term monitoring. Dry/active electrodes reduce setup time and cabling. Choice affects cap design and space requirements.
fNIRS Sources	Light-Emitting Diodes (LEDs): Low cost, portable. Laser Diodes (LDs) / Vertical-Cavity Surface-Emitting Lasers (VCSELs): Higher power, better signal quality [30] [26].	Higher power sources (e.g., for sensorimotor cortex) may be needed versus lower power (e.g., for frontal cortex) [31]. Switching frequency must be managed to avoid EEG crosstalk [26].
fNIRS Detectors	Silicon Photodiodes (PDs): Common, good sensitivity. Avalanche Photodiodes (APDs): Higher sensitivity, but more complex and expensive [30].	Sensitivity determines the quality of the detected light signal. The size and profile of the detector can influence mechanical design and comfort.

Integrated EEG-fNIRS caps represent a significant technological advancement, enabling robust and convenient multimodal brain imaging. Successful implementation hinges on carefully balancing mechanical design, electronic integration, and experimental protocol. By understanding the core design principles and leveraging the growing range of commercial and custom solutions, researchers can effectively deploy this powerful technology to uncover new insights into brain function and dysfunction in both laboratory and real-world settings.

The integration of functional near-infrared spectroscopy (fNIRS) and electroencephalography (EEG) provides a powerful multimodal approach for studying brain function by simultaneously capturing hemodynamic and electrical neural activity [2] [4]. A fundamental prerequisite for effective simultaneous recording is the precise spatial co-registration of fNIRS optodes with the International 10-20 EEG system [34] [35]. This co-registration ensures that the brain regions assessed by both modalities are anatomically aligned, enabling valid cross-modal data integration and interpretation. For researchers in neuroscience and drug development, establishing a standardized protocol for sensor placement is crucial for obtaining reliable, reproducible measurements in studies investigating neural correlates of cognition, disease states, or treatment effects [4]. This application note details the methodologies and protocols for achieving precise spatial co-registration, framed within the broader context of sensor placement compatibility for simultaneous EEG-fNIRS research.

The 10-20 System: A Foundational Framework

The International 10-20 System is a globally recognized method for standardizing electrode placement on the scalp based on anatomical landmarks [34]. The system specifies locations using a proportional measurement system that accounts for individual head size and shape.

• Landmarks and Positioning: The system is defined by four primary anatomical reference points: the nasion (Nz), the inion (Iz), and the left and right preauricular points (AL, AR) [34]. The positions of EEG electrodes (e.g., Fp1, Fp2, C3, C4) are determined by measuring arcs along the scalp at 10% or 20% intervals between these reference points.
• Cranio-Cerebral Correlation: A key strength of the 10-20 system is its consistent correspondence with underlying cortical anatomy. Studies using MRI have confirmed that each 10-20 landmark on the scalp reliably corresponds to a specific cortical area, a relationship that generalizes across different subjects [34]. This makes the system an ideal scaffold for positioning other neuroimaging sensors, such as fNIRS optodes.

Co-registration Methodologies

Several methodological approaches exist for co-registering fNIRS optodes to the 10-20 system, ranging from manual measurement to advanced computational and neuroimaging-assisted techniques.

Manual and Semi-Automatic Measurement

• Traditional Manual Method: The conventional approach involves manually identifying and marking the 10-20 landmarks on a participant's scalp using a measuring tape [34]. The fNIRS optodes are then positioned relative to these marks. However, this method is time-consuming (often taking 16 minutes or more) and prone to human error, as the identification of later landmarks depends on the accurate placement of earlier ones [34].
• Semi-Automatic Digitization: To overcome the limitations of manual measurement, semi-automatic methods using 3D magnetic digitizers have been developed [34]. These systems involve:
  • Digitizing the four primary reference points (Nz, Iz, AL, AR).
  • Sampling a cloud of additional points from the subject's head surface.
  • Using software algorithms to reconstruct the head surface geometry and automatically compute the precise 10-20 locations within this virtual space [34]. This approach significantly improves both the reliability and efficiency of landmark identification.

MRI-Assisted and Probabilistic Co-registration

For higher spatial precision, particularly for targeting specific cortical regions, MRI-assisted methods are employed.

• Subject-Specific MRI Co-registration: This "gold-standard" method involves acquiring an individual's structural MRI scan. Vitamin E capsules or other MRI-visible markers are placed on the fNIRS optodes during the scan to visualize their locations relative to cranial anatomy [36]. The balloon-inflation algorithm is then commonly used to project the fNIRS channel locations from the scalp surface to the underlying cortical surface, providing precise anatomical localization [36].
• Probabilistic and Virtual Registration: When subject-specific MRI is unavailable, virtual registration methods offer a practical alternative. These techniques utilize a reference database of MRIs from multiple individuals and established probabilistic maps of the 10-20 system positions to estimate the most likely cortical projection for a given optode placement [36] [37]. Software toolboxes like the fNIRS Optodes' Location Decider (fOLD) leverage photon transport simulations on head atlases to guide optode placement for optimal sensitivity to specific brain regions-of-interest [37].

Experimental Protocols for Co-registration

The following protocols provide a framework for accurate co-registration in a research setting.

Protocol 1: Standardized Manual Co-registration for EEG-fNIRS

This protocol is suitable for studies without access to neuronavigation or individual MRI data.

• Materials: Measuring tape, surgical marker, EEG cap with integrated fNIRS holders (or separate EEG cap and fNIRS headband), 3D digitizer (optional, for improved accuracy).
• Procedure:
  • Landmark Identification: Visually identify and mark the four primary reference points (Nz, Iz, AL, AR) on the participant's scalp.
  • EEG Cap Placement: Position the EEG cap according to the manufacturer's instructions, aligning its pre-marked 10-20 positions with the marked landmarks on the scalp.
  • fNIRS Optode Placement: If using an integrated cap, insert the fNIRS optodes into their designated holders. If using a separate fNIRS headband, position it such that the optodes target the cortical region of interest (e.g., the prefrontal cortex) based on its relationship to the nearest 10-20 positions (e.g., Fp1, Fp2) [36] [35].
  • Verification (Optional): Use a 3D digitizer to record the 3D spatial coordinates of key optodes and EEG electrodes relative to the cranial landmarks. This digital record facilitates more accurate co-registration during data analysis.

Protocol 2: MRI-Guided Co-registration for High-Precision Studies

This protocol is for studies requiring the highest degree of anatomical specificity.

• Materials: Structural MRI scanner, Vitamin E capsules or fiducial markers, fNIRS system with MRI-compatible optodes, neuronavigation system.
• Procedure:
  • Pre-Scan Marker Placement: Prior to the MRI scan, attach Vitamin E capsules to the fNIRS optode holder positions that will be used in the subsequent fNIRS session.
  • MRI Acquisition: Acquire a high-resolution T1-weighted structural MRI scan. The markers will be visible in the resulting images.
  • Co-registration in Software: In a neuroimaging software package (e.g., SPM, AtlasViewer), co-register the marker positions from the MRI to a standard head model or the individual's scalp surface.
  • Cortical Projection: Use an algorithm (e.g., balloon-inflation) to project the fNIRS channel locations (the midpoints between sources and detectors) from the scalp surface onto the cortical surface, obtaining their coordinates in standard space (e.g., MNI or Talairach) [36].

Table 1: Comparison of fNIRS-EEG Co-registration Methods

Method	Key Principle	Accuracy	Time/Cost	Primary Application
Manual Measurement	Proportional measurement from cranial landmarks [34]	Moderate	Low	Standard cognitive studies, field research
Semi-Automatic Digitization	3D head surface reconstruction & virtual measurement [34]	High	Moderate	Studies requiring higher spatial precision
Virtual/Probabilistic Registration	Probabilistic mapping using group-level MRI templates [36] [37]	Good	Low to Moderate	Standalone fNIRS studies without subject-specific MRI
Subject-Specific MRI	Anatomical projection using individual structural MRI [36]	Very High	High	Clinical trials, studies targeting specific brain structures

The Scientist's Toolkit: Essential Research Reagents and Materials

Successful co-registration and simultaneous data acquisition rely on a set of key materials and tools.

Table 2: Essential Materials for fNIRS-EEG Co-registration Research

Item	Function/Description
Integrated EEG-fNIRS Caps	Elastic caps with pre-defined holders for both EEG electrodes and fNIRS optodes, ensuring fixed relative positions [4] [35].
3D Magnetic Digitizer (e.g., Polhemus Fastrak)	Records the 3D spatial coordinates of scalp landmarks, EEG electrodes, and fNIRS optodes for precise digital co-registration [34] [19].
MRI-Visible Fiducial Markers (e.g., Vitamin E Capsules)	Placed on optodes during an MRI scan to make their locations visible on structural images, enabling anatomical co-registration [36].
Neuronavigation System	Uses the participant's MRI and a 3D digitizer to visually guide the experimenter in placing optodes on the scalp over target cortical regions in real-time [38].
Computational Toolboxes (e.g., fOLD, AtlasViewer)	Software that uses head models and photon migration simulations to optimize optode placement for sensitivity to specific regions-of-interest [37].

Workflow and Signaling Pathways

The following diagram illustrates the logical workflow for selecting and implementing a co-registration strategy in simultaneous fNIRS-EEG research.

Co-registration Strategy Workflow

Integration in Combined fNIRS-EEG Setups

The physical integration of both modalities is a critical step. The dominant approach is to use a single, integrated helmet or cap [4] [35].

• Design Configurations: Integration can be achieved by:
  • Modified EEG Caps: Creating punctures in a standard elastic EEG cap to accommodate fNIRS probe fixtures [4].
  • Customized Rigid Helmets: Using 3D printing or thermoplastic sheets to create subject-specific helmets that offer superior stability and consistent optode-scalp coupling pressure, though at a higher cost [4].
• Spatial Arrangement: Typically, the smaller EEG electrodes are placed in between the larger fNIRS optodes [35]. This arrangement allows the EEG electrodes to provide high-density spatial information that aids in the co-registration of the fNIRS channels, leading to a more precise integrated interpretation of the electrophysiological and hemodynamic data [4].

Precise spatial co-registration of fNIRS optodes to the International 10-20 EEG system is not merely a technical preliminary but a foundational step for generating robust and interpretable data in simultaneous multimodal research. The choice of co-registration method—from cost-effective manual measurement to high-precision MRI-guided techniques—should be guided by the study's specific requirements for anatomical accuracy, available resources, and participant population. By adhering to the detailed protocols and leveraging the tools outlined in this document, researchers in both academic and pharmaceutical development settings can ensure sensor placement compatibility, thereby enhancing the validity of their findings on brain function and the effects of neuromodulatory compounds.

Optimizing Source-Detector Distances for fNIRS and Electrode Density for EEG

Simultaneous electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) recording provides a powerful, multimodal approach for studying brain function by capturing complementary electrical and hemodynamic activities [3]. The successful integration of these modalities hinges on optimizing their distinct hardware configurations: electrode density for EEG and source-detector distances for fNIRS. Optimal sensor placement is critical for maximizing signal quality, spatial specificity, and the overall fidelity of the collected data, which is a central theme in advanced neuroimaging research [11]. This protocol details the methodologies for determining these key parameters to ensure high-quality, compatible setups for simultaneous EEG-fNIRS studies.

Comparative Technical Specifications

The fundamental differences between EEG and fNIRS necessitate distinct optimization strategies for sensor placement. The table below summarizes their core technical characteristics.

Table 1: Technical comparison of EEG and fNIRS modalities.

Feature	EEG (Electroencephalography)	fNIRS (Functional Near-Infrared Spectroscopy)
What It Measures	Electrical activity from postsynaptic potentials of cortical neurons [39]	Hemodynamic response (changes in oxygenated and deoxygenated hemoglobin) [39]
Temporal Resolution	High (millisecond scale) [39]	Low (seconds scale) [39]
Spatial Resolution	Low (centimeter-level) [39]	Moderate (better than EEG, but limited to cortical surface) [39]
Key Placement Parameter	Electrode Density & Location [40]	Source-Detector Distance [3]
Primary Optimization Goal	Maximize source localization accuracy and signal-to-noise ratio.	Balance sensitivity to cerebral cortex vs. extracerebral tissues and signal strength.

Optimizing fNIRS Source-Detector Distances

Principles and Challenges

The source-detector distance is a primary determinant of sensitivity and spatial specificity in fNIRS. Near-infrared light from a source optode is scattered and absorbed through tissue before being measured by a detector optode. The resulting measurement sensitivity profile is often described as a "banana-shaped" path. Key principles and challenges include:

• Depth Sensitivity: The sensitivity to brain tissue increases with source-detector separation. Longer paths penetrate deeper but suffer from significant light attenuation [3].
• Extracerebral Contamination: fNIRS signals are highly sensitive to hemodynamic changes in the scalp (extracerebral tissue). A major challenge is distinguishing these confounding signals from cerebral brain activity [11].
• Spatial Specificity: Consistently and reliably targeting specific cortical regions of interest (ROIs) can be challenging due to variations in cap placement and limited anatomical information across repeated measurements [11].

Recommended Distance Parameters

Optimized distances are determined by the need to penetrate the skull and reach the cerebral cortex while maintaining an adequate signal-to-noise ratio.

Table 2: Guidelines for fNIRS source-detector distances.

Distance	Target Tissue Sensitivity	Primary Application & Rationale
~1.5 cm	Extracerebral layers (scalp, skull) only [3]	Used as a short-separation channel to regress out systemic physiological noise from scalp hemodynamics [3].
~3.0 cm	Cerebral cortex and extracerebral layers [3]	Standard distance for measuring brain activity in adults. Provides a balance between cortical sensitivity and signal amplitude [3].
Up to 4-6 cm	Deeper cortical areas	Used in high-density diffuse optical tomography (HD-DOT) arrays to improve depth localization and spatial resolution (~1 cm). Requires high-quality systems due to low signal levels [3].

Optimizing EEG Electrode Density

Principles and the Case for Density Optimization

While high-density EEG (HD-EEG) systems with 64 to 256+ electrodes are available and generally provide superior source localization accuracy, recent research demonstrates that optimized low-density montages can be sufficient for specific applications [40]. The motivations for optimizing density include:

• Practical Constraints: Reducing the number of electrodes decreases setup time, computational load, cost, and improves participant comfort, which is particularly important for vulnerable populations like neonates [41] or long-term studies.
• Task-Dependent Performance: The minimum number of electrodes required for accurate source localization or classification is highly dependent on the specific brain activity being studied [40].

Data-Driven Optimization Methodology

An automated methodology based on the Non-dominated Sorting Genetic Algorithm II (NSGA-II) has been proposed to identify the minimal electrode subsets that retain the localization accuracy of HD-EEG [40]. The workflow for this optimization is illustrated below.

EEG Electrode Optimization Workflow: The process identifies minimal electrode sets that preserve source localization accuracy [40].

Key Findings and Recommendations

• Single-Source Localization: For reconstructing a particular brain activity, optimal subsets with as few as 6 to 8 electrodes can attain an equal or better accuracy than HD-EEG with 231 electrodes in a significant majority of cases (e.g., >88% for synthetic signals) [40].
• Multiple-Source Localization: For a case with three distinct sources, optimized combinations of 8, 12, and 16 electrodes matched or surpassed the accuracy of the 231-electrode reference in at least 58%, 76%, and 82% of cases, respectively [40].
• Application-Specific Montages: The optimal electrode locations are not universal; they are specific to the brain activity being analyzed. Therefore, optimization should be performed for each unique experimental paradigm or clinical question [40].

Integrated Experimental Protocol for Simultaneous EEG-fNIRS

This protocol provides a step-by-step guide for setting up a compatible and optimized simultaneous EEG-fNIRS system.

Pre-Experimental Planning

• Define the Region of Interest (ROI): Based on your research question and existing literature, identify the cortical area(s) to be monitored (e.g., prefrontal cortex for cognitive workload, motor cortex for movement).
• Select the EEG Cap: Choose a cap that accommodates both EEG electrodes and fNIRS optodes. Integrated caps with pre-defined, compatible holder rings for both modalities are recommended [42]. Ensure the cap material is opaque to prevent ambient light from affecting the fNIRS signals [42].
• Design the Optode Layout:
  • Place fNIRS sources and detectors over the ROI to create channels with ~3.0 cm separation for primary brain measurements [3].
  • Incorporate short-separation channels (~1.5 cm) by placing additional detectors close to selected sources. These are crucial for noise correction [3].
  • Use the international 10-10 or 10-5 system for positioning to ensure standardization and compatibility with EEG [39] [40].
• Plan the EEG Montage:
  • If using a standard system (e.g., 32 or 64 channels), ensure coverage over your ROI.
  • For a minimal setup, refer to literature or perform a preliminary optimization (as in Section 4.2) to identify a task-relevant, low-density montage. Studies have shown success with configurations focused on specific hemispheres or regions [41].

Sensor Placement and Hardware Setup

• Fit the Cap: Position the integrated cap on the participant's head according to the nasion-inion and pre-auricular points. Align the Cz position (or other reference landmarks) correctly.
• Place fNIRS Optodes: Insert the fNIRS sources and detectors into their designated holders in the cap. Ensure good scalp contact. For fiber-based systems, secure the fiber cables to minimize strain and movement artifacts.
• Place EEG Electrodes: Insert the EEG electrodes into the holders. It is often practical to place the smaller EEG electrodes in between the larger fNIRS optodes to avoid physical interference [42]. Use conductive gel or saline solution as required by the electrode type (e.g., active wet electrodes can reduce preparation time) [42].
• Check Signal Quality:
  • For fNIRS, visually inspect the raw light intensity levels for all channels. Reject channels with insufficient signal or excessive noise.
  • For EEG, check impedance values. Aim for impedances below 20 kΩ for active electrode systems to ensure high-quality data.
• Synchronize Systems: Connect both the EEG amplifier and fNIRS system to a central computer. Use a shared hardware trigger (e.g., TTL pulse) at the start of the experiment or software synchronization to align the data streams temporally with millisecond precision [39] [42].

Data Acquisition and Quality Control

• Simultaneous Recording: Begin recording on both systems. Present experimental stimuli or tasks to the participant.
• Monitor in Real-Time: Observe the incoming data streams for stability. Note any periods of excessive movement or artifacts.
• Documentation: Record the exact positions of all EEG electrodes and fNIRS optodes (sources and detectors). This is critical for offline data analysis and replication [43].

The Scientist's Toolkit

Table 3: Essential research reagents and materials for simultaneous EEG-fNIRS research.

Item	Function/Application
Integrated EEG-fNIRS Cap	A head cap with dedicated holders for both EEG electrodes and fNIRS optodes, ensuring stable and co-registered sensor placement [42].
Active EEG Electrodes	Electrodes with built-in amplification that provide high-quality signals with higher tolerance for skin impedance, reducing preparation time [42].
fNIRS Short-Separation Detectors	Specialized detector optodes placed 1-1.5 cm from a source to measure systemic physiological noise from the scalp, enabling its regression from the cerebral signal [3].
Conductive Gel	Electrolyte gel used with wet EEG electrodes to establish a stable electrical connection between the scalp and the electrode, crucial for low impedance [42].
3D Digitizer	A portable electromagnetic or optical system to record the precise 3D locations of EEG electrodes and fNIRS optodes on the participant's head, improving spatial accuracy of the coregistration with anatomical models [43].
Synchronization Hardware (e.g., TTL)	A device that generates a precise, shared trigger pulse to synchronize the timing of data acquisition between the EEG and fNIRS systems [39].

This application note provides a detailed protocol for simultaneous electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) data acquisition, framed within the broader context of sensor placement compatibility for multimodal neuroimaging research. The complementary nature of EEG and fNIRS offers significant potential for brain-computer interfaces (BCIs), clinical neurology, and neurorehabilitation by combining EEG's millisecond temporal resolution with fNIRS's spatially localized hemodynamic responses [8] [44]. However, successful integration requires careful consideration of hardware compatibility, sensor placement strategies, and artifact mitigation to ensure data quality. This protocol synthesizes current methodologies to establish a standardized workflow for researchers investigating neural correlates in both healthy and clinical populations.

Research Reagent Solutions and Essential Materials

Table 1: Essential equipment and materials for simultaneous EEG-fNIRS research

Category	Specific Item/Model	Key Specifications	Function/Purpose
EEG System	g.HIamp amplifier [10]	32 channels, 256 Hz sampling rate [10]	Records electrical brain activity with high temporal resolution
fNIRS System	NirScan [10]	11 Hz sampling rate, 730 nm & 850 nm wavelengths [45] [10]	Measures hemodynamic changes via HbO and HbR concentration
Hybrid Cap	Custom Model M [10]	54-58 cm head circumference, 32 EEG electrodes, 32 fNIRS sources, 30 detectors [10]	Integrated platform for co-located EEG electrode and fNIRS optode placement
Stimulation Software	E-Prime 3.0 [10]	-	Presents experimental paradigms and sends synchronized event markers
Synchronization Interface	Custom trigger interface [10]	-	Ensures temporal alignment of EEG and fNIRS data streams
Auxiliary Sensors	Accelerometer [45]	-	Records head movement to assist with motion artifact correction

Participant Preparation Protocol

Screening and Consent

• Participant Recruitment: Recruit right-handed participants to standardize cortical activation patterns for lateralized tasks like motor imagery [10]. For semantic tasks involving silent naming, ensure participants are native speakers of the stimulus language to minimize variability in neural representation [5].
• Ethical Considerations: Obtain written informed consent after fully explaining procedures, objectives, and data sharing practices. Secure approval from an institutional ethics committee before study commencement [5] [10].

Pre-Experimental Preparation

• Cap Fitting: Use a hybrid EEG-fNIRS cap sized appropriately for head circumference (e.g., 54-58 cm) [10]. Position the cap to ensure the fNIRS sensor covers the hairless forehead region, with the middle pointing toward the nose and the bottom just above the eyebrows. Avoid placing side sensors over hairy scalp areas to maintain signal quality [45].
• Task Familiarization and Calibration: For motor imagery tasks, introduce a grip strength calibration procedure using a dynamometer or stress ball to enhance kinesthetic sensation and task vividness. This involves repeated maximal force exertions and grip training at a standardized rhythm (e.g., one contraction per second) [10].
• Environmental Setup: Seat participants approximately 25 cm from the visual display monitor in an ergonomic position. Conduct the experiment in a controlled environment to minimize external distractions [10].

Hardware Integration and Sensor Placement

Integrated Cap Configuration

The custom hybrid cap should systematically integrate both modalities:

• EEG Electrodes: Arrange 32 electrodes according to an expanded international 10-20 system for comprehensive cortical coverage [10].
• fNIRS Optodes: Configure 32 laser sources and 30 photodetectors in a geometric matrix, maintaining a standard source-detector separation distance of 3 cm to create approximately 90 measurement channels [10]. This topology enables hemodynamic monitoring across prefrontal, motor, and association cortices.

Synchronization Setup

Temporal synchronization between EEG and fNIRS is critical for multimodal fusion. Implement a hardware trigger system where stimulation software (e.g., E-Prime 3.0) sends simultaneous event markers to both EEG and fNIRS recording systems at the onset of each experimental trial [10]. This ensures precise temporal alignment of electrophysiological and hemodynamic data streams during subsequent analysis.

Experimental Paradigm and Data Acquisition

Standardized Experimental Protocol

A typical motor imagery paradigm suitable for both healthy individuals and clinical populations (e.g., intracerebral hemorrhage patients) follows this trial structure [10]:

• Baseline Recording (2 minutes): Begin with 1-minute eyes-closed followed by 1-minute eyes-open states, demarcated by an auditory cue (200 ms beep).
• Visual Cue (2 seconds): Present a directional arrow (left/right) indicating the required motor imagery task.
• Execution Phase (10 seconds): Display a central fixation cross following an auditory cue; participants perform kinesthetic motor imagery of grasping with the indicated hand at approximately one imagined grasp per second.
• Inter-trial Interval (15 seconds): Show a blank screen for participant rest.

Conduct at least two consecutive sessions per participant, with each session containing 15 trials per hand (60 trials total), incorporating sufficient inter-session rest intervals to mitigate fatigue [10].

Data Acquisition Parameters

• EEG Settings: Acquire data at 256 Hz sampling rate using a g.HIamp amplifier with 32 electrodes [10].
• fNIRS Settings: Acquire data at 11 Hz sampling rate using a continuous-wave system like NirScan, measuring optical density changes at 730 nm and 850 nm wavelengths to calculate relative changes in oxyhemoglobin (HbO) and deoxyhemoglobin (HbR) concentrations via the modified Beer-Lambert Law [45] [10].

Table 2: Key acquisition parameters for simultaneous EEG-fNIRS recording

Parameter	EEG Specification	fNIRS Specification	Notes
Sampling Rate	256 Hz [10]	11 Hz [10]	Higher EEG rate captures neural dynamics
Spatial Resolution	~2 cm [5]	1-3 cm [46]	fNIRS offers better spatial localization
Penetration Depth	Cortical surface	1-2 cm deep [45]	fNIRS accesses superficial cortical regions
Measured Signals	Electrical potentials	HbO and HbR concentration changes [10]	Complementary neurophysiological measures
Typical Hemodynamic Response	N/A	Within 2-6 seconds [45]	Consider this delay in paradigm design

Data Processing and Multimodal Fusion

Signal Processing Pipeline

• Artifact Removal: For EEG, address ocular (EOG) and muscle (EMG) artifacts using techniques like principal component analysis (PCA) [8] [45]. For fNIRS, correct for motion artifacts using accelerometer data and adaptive filtering, while cardiac and respiratory interference can be removed via filtering or PCA [8] [45].
• Feature Extraction: For EEG signals, extract spatiotemporal features using methods such as dual-scale temporal convolution and depthwise separable convolution [47]. For fNIRS signals, employ spatial convolution across channels to explore regional activation differences and parallel temporal convolution to capture hemodynamic dynamics [47].

Multimodal Fusion Approaches

Implement fusion strategies that leverage the complementary nature of EEG and fNIRS signals:

• Data-Driven Fusion: Utilize unsupervised symmetric techniques for continuous brain imaging in naturalistic environments where precise stimulus timing is unavailable [8].
• Deep Learning Integration: Apply end-to-end signal fusion methods using convolutional networks for spatial feature extraction and gated recurrent units (GRU) for temporal dynamics, combined with evidence theory for decision-level fusion [47].
• Representation Learning: Employ multimodal representation models that learn both modality-specific and shared representations across EEG and fNIRS, enabling adaptation to single-modality datasets when necessary [48].

This application note provides a comprehensive workflow for simultaneous EEG-fNIRS research, emphasizing practical considerations for sensor placement compatibility and multimodal data integration. The protocol addresses the complete experimental pipeline from participant preparation to data acquisition and processing, with particular attention to the technical challenges of hardware synchronization, artifact mitigation, and multimodal fusion. By standardizing these methodologies, researchers can more effectively leverage the complementary strengths of EEG and fNIRS to investigate neural processes across diverse populations and settings, from laboratory environments to real-world applications. The integration of these modalities continues to show significant promise for advancing brain-computer interfaces, clinical diagnostics, and our fundamental understanding of brain function.

Electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) represent a powerful multimodal neuroimaging approach, leveraging their complementary strengths to investigate brain function. EEG provides direct measurement of neural electrical activity with millisecond temporal resolution, while fNIRS tracks hemodynamic responses with better spatial specificity [4] [49]. This combination is particularly valuable for studying complex cognitive paradigms where both rapid neural processing and metabolic demands are of interest. Within this context, appropriate sensor placement is crucial for capturing relevant neural signatures while ensuring compatibility between modalities.

This article presents detailed application notes and experimental protocols for two key research paradigms: motor imagery and semantic decoding. We focus on the practical aspects of simultaneous EEG-fNIRS data acquisition, including technical specifications, experimental design considerations, and sensor placement strategies to optimize signal quality and minimize interference between systems.

The Scientist's Toolkit: Essential Research Equipment

Table 1: Core Equipment for Simultaneous EEG-fNIRS Research

Component Category	Specific Equipment Examples	Key Functions & Specifications
EEG Acquisition System	Neuroscan SynAmps2 Amplifier [50]; ActiCHamp [51]; Brain Products EEG systems [49]	Measures electrical brain activity; typically uses 64+ electrodes [50]; sampling rates ≥ 1000 Hz; impedance kept below 10 kΩ [50].
fNIRS Acquisition System	NIRScout System [50] [51]; NIRx NIRScout XP [51]; Hitachi ETG-4000 [52]	Measures hemodynamic changes via near-infrared light; uses sources & detectors; sampling rates ~7.8-10 Hz [50] [52].
Integrated Head Cap	EasyCap with 128-160 slits [49] [51]; Custom 3D-printed helmets [4]; Elastic caps with 3D-printed sockets [53]	Holds both EEG electrodes and fNIRS optodes in stable positions; dark fabric reduces light contamination [49]; ensures proper source-detector distance (~3 cm) [53].
Synchronization Interface	Lab Streaming Layer (LSL) protocol [49]; Shared hardware triggers [49]	Ensures precise temporal alignment of EEG and fNIRS data streams, critical for multimodal analysis.
Stimulus Presentation Software	Custom scripts (e.g., Python, MATLAB)	Presents visual/auditory cues and records event markers synchronized with brain data acquisition.

Case Study 1: Motor Imagery Paradigm

Experimental Design and Protocol

Motor imagery (MI) involves the mental rehearsal of physical movements without actual execution. This paradigm is widely used in brain-computer interfaces (BCIs) and neurorehabilitation [50] [51].

Participant Preparation and Task Instructions:

• Participants: 18-30 right-handed, healthy adults with normal or corrected-to-normal vision [50] [51]. Participants should avoid caffeine and alcohol prior to testing.
• Pre-experiment Training: Conduct a familiarization session comprising motor execution blocks followed by motor imagery blocks to ensure proper task understanding [50].
• Task Instructions: Participants are instructed to perform kinesthetic motor imagery (feeling the movement) rather than visual imagery (visualizing the movement). During the 4-second imagery period, they must suppress blinking and swallowing to minimize artifacts [50].

Trial Structure and Timing: The trial structure follows a precise timeline to capture both rapid EEG responses and slower fNIRS hemodynamics:

• Fixation Cross (2 s): A white cross appears at screen center to focus attention.
• Cue Presentation (2 s): A text and video cue indicates the specific motor imagery task to perform.
• Motor Imagery Period (4 s): The video disappears, and "Start Imagining" text appears. Participants perform the cued MI task.
• Rest Period (10-12 s, randomized): "Rest" appears on screen. This extended rest accounts for the slow hemodynamic response, preventing overlap between trials [50].

Experimental Conditions: The protocol includes eight MI tasks focusing on different joints of the right upper limb: hand open/close, wrist flexion/extension, wrist abduction/adduction, elbow pronation/supination, elbow flexion/extension, shoulder pronation/supination, shoulder abduction/adduction, and shoulder flexion/extension [50]. Each session comprises 8 blocks, with 40 randomized trials per block (5 trials per task), totaling 320 trials per participant.

Sensor Placement and Data Acquisition

EEG Configuration:

• Utilize a 64-channel electrode cap arranged according to the international 10-20 system, covering the whole scalp [50]. Use the left mastoid (M1) as the reference and ensure electrode impedances are maintained below 10 kΩ [50].
• For MI, focus on sensors over the sensorimotor cortex (e.g., C3, Cz, C4) [51].
• Apply a 0.5–100 Hz band-pass filter and a 50 Hz notch filter during acquisition to remove line noise [50].

fNIRS Configuration:

• Employ a system with 8 sources and 8 detectors placed over the left hemisphere (contralateral to the imagined right-hand movements) according to the international 10-5 system [50].
• This configuration creates 24 measurement channels with a source-detector distance of approximately 3 cm, optimizing sensitivity to cortical hemodynamic changes [50].
• Co-register fNIRS probe locations with the EEG cap for integrated data analysis.

Synchronization:

• Implement the Lab Streaming Layer (LSL) protocol or shared hardware triggers to synchronize EEG and fNIRS data streams with millisecond precision [49].

The following diagram illustrates the signaling pathways and logical sequence of a single motor imagery trial.

Data Analysis and Expected Outcomes

EEG Analysis:

• Process EEG data to extract event-related desynchronization (ERD) in the sensorimotor rhythm (mu rhythm, 8-13 Hz) and beta band (13-30 Hz) over the contralateral sensorimotor cortex during imagination [51].
• Apply deep learning methods such as ShallowConvNet for classification, which can achieve accuracies around 65.49% for distinguishing between different joint movements like hand versus shoulder imagery [50].

fNIRS Analysis:

• Analyze oxygenated hemoglobin (HbO) and deoxygenated hemoglobin (HbR) concentration changes. Expect a task-related increase in HbO and a decrease in HbR in the primary motor cortex [50].
• Preprocess signals using bandpass filtering (e.g., 0.01-0.5 Hz) and principal component analysis (PCA) to remove physiological noise [52].

Multimodal Integration:

• Fuse EEG and fNIRS features at the decision level using frameworks like Dempster-Shafer Theory (DST) to model uncertainty, potentially improving classification accuracy beyond unimodal approaches (e.g., reaching 83.26% in some studies) [47].

Case Study 2: Semantic Decoding Paradigm

Experimental Design and Protocol

Semantic decoding aims to identify which semantic concepts an individual is processing based on their brain activity patterns, with applications in direct semantic communication BCIs [5].

Participant Preparation and Task Instructions:

• Participants: 12 native English speakers for semantic tasks involving silent naming [5]. Participants should have normal or corrected-to-normal vision.
• Stimuli: Use images from distinct semantic categories (e.g., 18 animals and 18 tools), converted to grayscale and presented on a neutral background [5].
• Task Instructions: Participants perform four distinct mental tasks in response to the image stimuli:
  • Silent Naming: Silently name the displayed object in their mind.
  • Visual Imagery: Visualize the object in their mind.
  • Auditory Imagery: Imagine the sounds associated with the object.
  • Tactile Imagery: Imagine the feeling of touching the object [5].
• Participants are instructed to remain engaged for the full task duration (3-5 seconds) and minimize physical movements.

Trial Structure and Timing:

• Stimulus Presentation (3 s): An image representing an animal or tool is displayed.
• Mental Task Period (3-5 s): Participants perform the cued mental task (e.g., silent naming, auditory imagery).
• Inter-stimulus Interval (6-9 s, jittered): A rest period with neutral stimuli (e.g., fireworks, music) allows the hemodynamic response to return to baseline [5] [52].

• The experiment typically consists of multiple blocks (e.g., 12 blocks) with randomized stimulus presentation.

Sensor Placement and Data Acquisition

EEG Configuration:

• Use a high-density 64-channel EEG system. For semantic processing, ensure coverage over temporal and frontal lobes, which are critical for semantic memory and language processing [5].

fNIRS Configuration:

• Deploy two fNIRS arrays: a posterior array covering the occipital lobe for visual processing and a left lateral array covering temporal, parietal, and prefrontal regions for language and semantic processing [52].
• A typical setup may include 24 channels in the posterior array and 18-22 channels in the lateral array [52].

Synchronization:

• As with the MI paradigm, use LSL or hardware triggers for precise temporal synchronization of stimulus onset, task periods, and brain data acquisition across both systems [49].

The workflow for the semantic decoding paradigm, from stimulus presentation to analysis, is outlined below.

Data Analysis and Expected Outcomes

EEG Analysis:

• Analyze event-related potentials (ERPs) and time-frequency responses to differentiate between semantic categories (animals vs. tools). Focus on components like the N400, which is sensitive to semantic processing [5].

fNIRS Analysis:

• Extract HbO and HbR concentration changes from channels covering the temporal lobe. Preprocess data using bandpass filtering (e.g., 0.01-1.0 Hz) and PCA to reduce global physiological noise [52].
• Apply channel stability analysis to identify and retain channels with reliable responses across blocks [52].

Multimodal Integration and Decoding:

• Use multivariate pattern analysis (MVPA) and representational similarity analysis to decode semantic categories from the combined EEG-fNIRS data patterns [5] [52].
• Between-subjects decoding and model-based decoding using semantic models can achieve classification accuracy significantly above chance levels [52].

Comparative Analysis and Sensor Placement Considerations

Table 2: Protocol Comparison: Motor Imagery vs. Semantic Decoding

Parameter	Motor Imagery Protocol	Semantic Decoding Protocol
Primary Brain Regions	Sensorimotor cortex (Contralateral to movement) [50] [51]	Temporal lobe, Frontal lobe, Occipital cortex [5] [52]
Key EEG Features	Event-Related Desynchronization (ERD) in Mu/Beta rhythms [51]	Event-Related Potentials (ERPs), Time-Frequency features [5]
Key fNIRS Features	HbO increase in primary motor cortex [50]	HbO/HbR changes in temporal and frontal regions [52]
Typical Trial Duration	18-20 seconds [50]	12-17 seconds [5] [52]
Task Duration	4 seconds of active imagination [50]	3-5 seconds of mental task [5]
Optimal Cap Coverage	Central scalp regions (C3, Cz, C4) [51]	Whole-head, with emphasis on temporal and occipital areas [5] [52]
Classification Performance	~65% (EEG only, 2-class joints) [50]; Up to ~83% (Multimodal fusion) [47]	Significantly above chance (Exact accuracy varies by study and model) [52]

Sensor Placement Compatibility Guidelines:

• Spatial Co-registration: Precisely document the 3D locations of all EEG electrodes and fNIRS optodes using digitization systems. This is essential for accurate source analysis and interpreting hemodynamic changes in the context of underlying electrical activity [4].
• Cap Design and Competition for Space: Use caps with a sufficient number of slits (e.g., 128-160) to accommodate both sensors flexibly [49] [51]. In regions where sensors compete for space (e.g., over the motor cortex for MI), prioritize placement based on the primary research question. 3D-printed custom helmets can offer a tailored solution for complex montages [4].
• Minimizing Interference: fNIRS optodes do not electrically interfere with EEG electrodes [49]. However, ensure fNIRS holder materials do not physically impede EEG electrode contact with the scalp. Using dark fabric caps helps reduce ambient light contamination for fNIRS [49].
• Subject Comfort and Signal Quality: The combined cap should be secure but comfortable for long recording sessions. Proper optode-scalp coupling is vital for fNIRS signal quality; spring-loaded optodes can help maintain consistent pressure [53]. For EEG, continue to use conductive gel and maintain impedances below 10 kΩ [50].

Solving Common Challenges: Artifacts, Interference, and Data Quality

Combining electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) provides a powerful multimodal approach to investigating brain function by capturing complementary electrophysiological and hemodynamic information [2] [4]. However, the physical integration of these technologies creates significant challenges for signal quality, primarily through three interference types: hair and skin characteristics, inconsistent optode pressure, and light bleed between emission and detection points [4] [54]. These physical interference factors can introduce artifacts, reduce signal-to-noise ratio, and compromise data quality, ultimately threatening the validity of research findings. This application note systematically quantifies these interference sources and provides evidence-based protocols for their mitigation, with particular emphasis on ensuring equitable signal quality across diverse participant populations [54] [55].

Hair and Skin Characteristics

Table 1: Impact of Hair and Skin Characteristics on fNIRS Signal Quality

Factor	Impact on Signal Quality	Quantitative Effect	Recommended Mitigation
Hair Density	Increased absorption and scattering of NIR light [54]	Dense hair can reduce detected light intensity by >50% [54]	Hair parting with cotton-tipped applicators; use of ultrasound gel [54]
Hair Color	Darker hair absorbs more light [54]	Significant reduction in signal amplitude compared to light hair [54]	Ensure consistent optode-scalp coupling regardless of color [54]
Hair Type	Curly/kinky hair affects optode-scalp coupling [54]	Increased signal variance due to inconsistent contact [54]	Customized cap designs; extended optimization time [54]
Skin Pigmentation	Higher melanin increases light absorption [54] [55]	Reduced signal intensity; not correlated with functional signal quality [54]	Signal quality optimization protocols; inclusive hardware development [54] [55]

Recent evidence indicates that these factors can disproportionately affect data quality from underrepresented populations, potentially introducing bias in neuroimaging research [55]. One study of stroke survivors found that fNIRS signals were generally worse for Black women compared to Black men and White individuals regardless of gender, highlighting the urgent need for hardware improvements to ensure equity in fNIRS research [55].

Pressure and Mechanical Factors

Table 2: Pressure-Related Artifacts and Solutions

Factor	Impact	Solution	Experimental Evidence
Variable Pressure	Inconsistent scalp coupling; motion artifacts [4]	3D-printed or thermoplastic custom helmets [4]	Custom helmets improve signal stability by >30% compared to elastic caps [4]
Cap Movement	Signal drift; complete signal loss [54]	Chin strap stabilization; cable management arms [54]	Proper stabilization reduces motion artifacts by 25-40% [54]
Optode Angle	Inconsistent light penetration; channel reliability [4]	Consistent optode orientation perpendicular to scalp surface	Critical for reproducible measurements across sessions [4]

Light Bleed and Optical Interference

Light bleed occurs when near-infrared light travels directly from source to detector without penetrating brain tissue, typically when optodes are improperly spaced or insufficiently shielded [56] [57]. This superficial signal contaminates the cerebral hemodynamic data with systemic physiological noise from the scalp [57]. Research demonstrates that incorporating short-separation channels (approximately 8 mm source-detector distance) as regressors in the general linear model is the most effective method for removing this contaminating signal, significantly improving sensitivity and specificity in detecting true brain activity [57].

Experimental Protocols for Interference Mitigation

Protocol 1: Comprehensive Cap Placement and Optimization

Objective: Establish consistent optode-scalp coupling across diverse hair and skin types. Materials: fNIRS-EEG cap, alcohol pads, cotton-tipped applicators, ultrasound gel, chin strap, cable management system. Duration: 20-30 minutes

• Scalp Preparation: Clean the forehead with alcohol pads to reduce skin impedance and improve EEG electrode contact [54].
• Initial Cap Placement: Position the cap starting from the front and moving backward to prevent hair from accumulating under the optode areas [54].
• Cz Alignment: Ensure the cap's Cz marker is positioned midway between the nasion and inion, and equidistant from ear-to-ear for consistent placement [54].
• Stabilization: Secure the cap with a chin strap and use a cable management arm to relieve wire strain on the cap [54].
• Hair Management: Using cotton-tipped applicators, carefully part hair away from under optodes. For challenging cases, apply small amounts of ultrasound gel to displace hair and improve optode-scalp contact [54].
• Signal Optimization: Use the system's signal optimization software while making final adjustments to optode positioning [54].
• Light Bleed Mitigation: Place short-separation channels (<1 cm source-detector distance) strategically around key regions of interest to capture superficial signals for regression [57].

Protocol 2: Signal Quality Validation and Quality Control

Objective: Quantify and verify signal quality before experimental data collection. Materials: fNIRS-EEG system with signal visualization software, quality assessment toolbox. Duration: 5-10 minutes

• Signal Intensity Check: Verify that detected light intensity falls within the manufacturer's recommended range for all channels [54].
• Physiological Noise Inspection: Confirm the presence of cardiac (≈1 Hz) and respiratory (≈0.2-0.3 Hz) pulsations in the fNIRS signal, indicating adequate scalp coupling [57].
• SCI Calculation: Compute Scalp Coupling Index for all channels, retaining only those with SCI > 0.8 for optimal data quality [55].
• Visual Inspection: Check EEG signals for abnormal impedance values (>50 kΩ) or unusual noise patterns that may indicate poor contact [58].
• Ambient Light Check: Ensure all optodes are properly shielded from ambient light by placing an opaque shower cap over the fNIRS cap if necessary [54].

Visualization of Integrated Interference Mitigation

Systematic Interference Mitigation Workflow: This diagram illustrates the integrated protocol for addressing physical interference throughout the experimental pipeline, from initial planning through data processing.

The Scientist's Toolkit: Essential Research Reagents and Materials

Table 3: Key Materials for fNIRS-EEG Interference Mitigation

Item	Function	Application Notes
3D-Printed Custom Caps	Ensures consistent optode placement and pressure [4]	Superior to elastic caps for maintaining consistent source-detector separation [4]
Short-Separation Detectors	Measures superficial signals for regression [57]	8 mm source-detector distance optimal for scalp signal capture [57]
Ultrasound Gel	Improves optode-scalp coupling in dense hair [54]	More effective than conductive gels for optical coupling [54]
Cotton-Tipped Applicators	Precisely parts hair away from optode placement sites [54]	Essential for managing curly or dense hair types [54]
Chin Strap Stabilization	Reduces cap movement artifacts [54]	Particularly important for long-duration experiments [54]
Cable Management Arm	Prevents wire strain on cap and optodes [54]	Reduces motion-related signal artifacts [54]
Opaque Shower Cap	Blocks ambient light interference [54]	Simple solution for light bleed from external sources [54]
Quality Assessment Software	Quantifies signal quality metrics (e.g., SCI) [55]	Enables objective quality control standards [55]

Successful simultaneous EEG-fNIRS research requires systematic attention to physical interference factors throughout the experimental pipeline. Hair characteristics, variable pressure, and light bleed represent significant challenges that can be effectively mitigated through standardized protocols, customized hardware, and appropriate signal processing techniques [4] [54] [57]. The implementation of these evidence-based practices is essential not only for data quality but also for promoting inclusivity in neuroimaging research by ensuring reliable signal acquisition across diverse participant populations [54] [55]. Future hardware developments should focus on improving optode design to automatically compensate for variations in hair and skin characteristics, thereby reducing the need for extensive manual optimization and making advanced neuroimaging accessible to broader research communities.

The integration of electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) offers a powerful, multimodal approach to investigating brain function by capturing complementary electrophysiological and hemodynamic activities [2]. However, the signals from both modalities are highly susceptible to physiological and motion-related artifacts, which can severely compromise data quality and interpretation [59] [60]. These artifacts present a particularly complex challenge in simultaneous EEG-fNIRS recordings, where the optimal physical placement of sensors is crucial for maximizing signal quality while minimizing cross-modal interference [61] [4]. This application note provides detailed protocols and analytical frameworks for advanced artifact removal, specifically contextualized within the constraints of dual-modality sensor compatibility.

Physiological Noise Characteristics and Origins

Major Artifact Types in EEG and fNIRS

Table 1: Physiological Artifacts in EEG and fNIRS Signals

Artifact Type	Origin	Frequency Characteristics	Primary Impacted Modality
Ocular Artifacts	Eye movements and blinks [59]	Similar to EEG signals [59]	EEG
Muscle Artifacts	Head/neck muscle activity, talking, swallowing [59]	Broad distribution (0 Hz to >200 Hz) [59]	EEG
Cardiac Artifacts	Heartbeat (ECG) and pulse [59]	~1.2 Hz (pulse) [59]	EEG, fNIRS
Motion Artifacts	Subject movement, optode decoupling [60] [62]	Non-stationary [60]	EEG, fNIRS
Systemic Physiological Noise	Respiration, Mayer waves, blood pressure changes [62] [63]	0.1 Hz (Mayer), 0.25 Hz (respiration), 1 Hz (heartbeat) [62]	fNIRS

Impact on Dual-Modality Recordings

In simultaneous EEG-fNIRS recordings, artifacts can manifest with different temporal profiles and affect both modalities simultaneously or independently. The neurovascular coupling relationship—where neuronal electrical activity (measured by EEG) triggers hemodynamic responses (measured by fNIRS)—creates a complex interdependence that must be considered during artifact removal [2]. Furthermore, sensor placement constraints in integrated systems can create unique artifact profiles, as EEG electrodes and fNIRS optodes must coexist on the same scalp surface without compromising signal quality from either modality [61] [4].

Advanced Artifact Removal Methodologies

Quantitative Performance of Removal Algorithms

Table 2: Performance Comparison of Artifact Removal Techniques

Method	Modality	Performance Metrics	Best Performing Parameters
WPD (Single-stage)	EEG	ΔSNR: 29.44 dB, η: 53.48% [60]	db2 wavelet (ΔSNR), db1 wavelet (η) [60]
WPD-CCA (Two-stage)	EEG	ΔSNR: 30.76 dB, η: 59.51% [60]	db1 wavelet [60]
WPD (Single-stage)	fNIRS	ΔSNR: 16.11 dB, η: 26.40% [60]	fk4 wavelet [60]
WPD-CCA (Two-stage)	fNIRS	ΔSNR: 16.55 dB, η: 41.40% [60]	db1 (ΔSNR), fk8 (η) wavelets [60]
GLM with tCCA	fNIRS	Correlation: +45%, RMSE: -55% [63]	With temporal embedding & auxiliary signals [63]

Technical Approaches for Artifact Removal

Blind Source Separation (BSS) and Regression Methods

Regression methods represent traditional approaches for artifact removal, particularly for ocular artifacts in EEG, where they define amplitude relationships between reference channels and EEG channels using transmission factors [59]. The regression equation takes the form:

EEGcor = EEGraw − γF(HEOG) − δF(VEOG) [59]

where γ and δ represent transmission coefficients between EOG and EEG channels [59].

Blind Source Separation (BSS) methods, particularly Independent Component Analysis (ICA), have emerged as powerful alternatives that exploit the statistical independence between neural signals and artifacts [59]. These methods are especially effective for muscle artifacts in EEG, as EMG contamination and EEG demonstrate substantial statistical independence both temporally and spatially [59].

Wavelet-Based and Hybrid Approaches

Wavelet Packet Decomposition (WPD) represents a significant advancement for handling non-stationary artifacts in both EEG and fNIRS signals [60]. WPD decomposes signals into wavelet packet bases at multiple scales, allowing for precise time-frequency localization of artifacts [60]. The two-stage WPD-CCA (Canonical Correlation Analysis) method has demonstrated superior performance for motion artifact correction in both modalities, with percentage reduction in motion artifacts increasing by 11.28% for EEG and 56.82% for fNIRS compared to single-stage WPD [60].

For fNIRS specifically, the General Linear Model with temporally embedded Canonical Correlation Analysis (GLM with tCCA) has shown remarkable improvements in physiological noise regression [63]. This approach incorporates short-separation measurements and auxiliary signals within a temporally embedded framework to address challenging signal characteristics like non-instantaneous and non-constant coupling [63].

Experimental Protocols for Artifact Removal

Protocol 1: Two-Stage WPD-CCA for Motion Artifacts

Application: Removal of motion artifacts from single-channel EEG and fNIRS signals [60]

Materials and Setup:

• Wearable EEG/fNIRS acquisition system
• Computer with MATLAB/Python and signal processing toolboxes
• Benchmark dataset for validation [60]

Procedure:

• Signal Acquisition: Record single-channel EEG or fNIRS signals during both resting state and task conditions.
• Parameter Selection: Choose appropriate wavelet packet (db1 for EEG, fk8 for fNIRS recommended) [60].
• Wavelet Packet Decomposition: Decompose the signal using WPD to level 4-5 [60].
• Component Selection: Identify artifact-contaminated components using statistical metrics (kurtosis, entropy).
• Canonical Correlation Analysis: Apply CCA to remove correlated noise components [60].
• Signal Reconstruction: Reconstruct the signal using inverse WPD.
• Validation: Calculate performance metrics (ΔSNR and η) to quantify improvement [60].

Sensor Placement Considerations: Ensure stable optode and electrode placement using integrated caps that minimize movement-induced decoupling [61].

Protocol 2: GLM with tCCA for fNIRS Physiological Noise

Application: Advanced physiological noise regression in fNIRS signals [63]

Materials and Setup:

• fNIRS system with short-separation channels (≤8 mm)
• Auxiliary physiological monitors (ECG, respiration, accelerometer)
• Custom analysis scripts implementing tCCA

Procedure:

• Data Collection: Acquire fNIRS signals with simultaneous short-separation measurements and auxiliary physiological recordings [63].
• Temporal Embedding: Create temporally embedded versions of all nuisance signals [63].
• CCA Processing: Apply regularized temporally embedded CCA to extract optimal nuisance regressors [63].
• GLM Construction: Build GLM with designed experimental regressors and CCA-derived nuisance regressors.
• Parameter Estimation: Solve GLM to obtain cleaned hemodynamic response functions [63].
• Validation: Compare results with traditional SS regression using correlation and RMSE metrics [63].

Integration Considerations: When using with simultaneous EEG, coordinate short-separation fNIRS detector placement with EEG electrode locations to avoid signal interference [4].

Protocol 3: Integrated EEG-fNIRS Artifact Removal Framework

Application: Comprehensive artifact handling for dual-modality studies

Materials and Setup:

• Simultaneous EEG-fNIRS acquisition system
• 3D-printed or custom-fitted integrated helmet [4]
• Multi-modal data processing software (e.g., NIRS-KIT, Homer2, EEGLAB)

Procedure:

• Synchronized Acquisition: Record EEG and fNIRS signals with precise temporal synchronization [4].
• Modality-Specific Preprocessing: Apply tailored artifact removal to each modality (WPD-CCA for EEG motion artifacts, GLM with tCCA for fNIRS physiological noise).
• Cross-Modality Validation: Use neurovascular coupling principles to validate artifact removal efficacy [2].
• Joint Analysis: Perform integrated analysis using EEG-informed fNIRS or parallel modality analysis [2].

Sensor Compatibility Protocol: Use 3D-printed customized helmets or cryogenic thermoplastic sheets to ensure optimal sensor placement and stability for both modalities [4].

The Scientist's Toolkit

Table 3: Essential Research Reagents and Materials

Item	Function	Application Notes
Short-Separation fNIRS Detectors	Measures superficial layer hemodynamics for noise regression [62] [63]	Place 8-10mm from sources; critical for GLM with tCCA [63]
Active EEG Electrodes	Reduces motion artifacts and preparation time [61]	Essential for combined systems; enables faster application [61]
Auxiliary Physiological Sensors	Measures cardiac, respiratory, and motion signals [62]	Provides reference signals for advanced regression methods [63]
Customized Integration Helmets	Maintains stable sensor placement for both modalities [4]	3D-printed or thermoplastic designs improve sensor-scalp coupling [4]
Wavelet Packet Toolboxes	Implements WPD and WPD-CCA algorithms [60]	MATLAB Wavelet Toolbox or Python PyWavelets with custom CCA implementation [60]
Accelerometers	Detects head movements for motion artifact correction [63]	Integrated into sensor caps for direct motion measurement [63]

Effective artifact removal is fundamental to extracting meaningful information from simultaneous EEG-fNIRS recordings. The advanced methods outlined in this application note—particularly WPD-CCA for motion artifacts and GLM with tCCA for physiological noise—demonstrate significant improvements over traditional approaches. When implementing these protocols, careful consideration of sensor placement compatibility is essential, as the physical integration of EEG electrodes and fNIRS optodes directly influences both signal quality and artifact profiles. By adopting these sophisticated artifact removal strategies within an optimized sensor integration framework, researchers can significantly enhance the reliability and interpretability of multimodal brain imaging data.

The simultaneous acquisition of electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) represents a powerful, non-invasive approach to studying human brain function. These modalities offer complementary insights: EEG captures millisecond-scale electrical neural dynamics, while fNIRS tracks hemodynamic changes related to neural activity with better spatial resolution [64] [3]. This complementary nature makes them ideal candidates for multimodal fusion, which aims to provide a more comprehensive picture of brain activity than either modality could alone [4].

Data fusion techniques for these signals have evolved from simple concatenation to sophisticated source-decomposition methods. The integration of fNIRS and EEG is particularly valuable for investigating neurovascular coupling—the relationship between neural electrical activity and subsequent hemodynamic responses [3]. Furthermore, the portability and movement tolerance of these systems enable brain imaging in naturalistic scenarios and across diverse populations, from infants to clinical patients [65] [3].

Foundational Principles of fNIRS and EEG

Technical Basis and Complementarity

EEG and fNIRS measure fundamentally different physiological processes, which accounts for their complementary strengths when combined.

• EEG Fundamentals: EEG measures the brain's electrical activity via electrodes placed on the scalp, detecting voltage changes from synchronized firing of cortical neurons, primarily pyramidal cells. Its greatest strength is exceptional temporal resolution (millisecond scale), making it ideal for analyzing rapid cognitive processes. However, its spatial resolution is limited due to the dispersion of electrical signals through the skull and scalp [64].
• fNIRS Fundamentals: fNIRS monitors cerebral hemodynamic responses by measuring changes in oxygenated (HbO) and deoxygenated hemoglobin (HbR) using near-infrared light. It offers better spatial resolution than EEG for surface cortical areas and is more tolerant of movement artifacts. Its temporal resolution is slower (seconds) as it reflects the indirect hemodynamic response to neural activity [64].

Table 1: Comparison of EEG and fNIRS Technical Characteristics

Feature	EEG	fNIRS
What It Measures	Electrical activity of neurons	Hemodynamic response (blood oxygenation)
Signal Source	Postsynaptic potentials in cortical neurons	Changes in oxygenated and deoxygenated hemoglobin
Temporal Resolution	High (milliseconds)	Low (seconds)
Spatial Resolution	Low (centimeter-level)	Moderate (better than EEG)
Depth of Measurement	Cortical surface	Outer cortex (~1–2.5 cm deep)
Sensitivity to Motion	High - susceptible to movement artifacts	Low - more tolerant to subject movement
Best Use Cases	Fast cognitive tasks, ERP studies, sleep research	Naturalistic studies, child development, motor rehab

Neurophysiological Basis for Fusion

The simultaneous measurement of electrical and hemodynamic activity provides a window into the complex interplay between neural firing and metabolic support. The neurovascular coupling mechanism links the electrical activity measured by EEG to the hemodynamic responses measured by fNIRS [3]. However, this relationship is not perfectly spatiotemporal, and their interaction can be studied using combined technologies [3]. If a model of neurovascular coupling is assumed, increased accuracy of neural signal estimates can be obtained from multimodal measurements [3].

Data Fusion Techniques: From Simple to Advanced

Concatenation and Decision-Level Fusion

Data concatenation represents one of the simplest and most straightforward fusion approaches. This method combines features from both modalities into a single, high-dimensional feature vector that is then fed into a classifier or analysis pipeline. The fusion occurs at the feature level, before the classification stage [65] [66].

Decision-level fusion represents another relatively simple approach where each modality is processed through separate analysis pipelines, and their independent decisions or outputs are combined at the final stage. For example, in a classification task, EEG and fNIRS data might be processed separately, and their respective classification results combined through voting schemes or weighted averaging [65].

Model-Based Fusion Approaches

Model-based approaches incorporate physiological constraints or statistical models to integrate the two data types. These methods often leverage the hemodynamic response function to model the relationship between neural electrical activity (EEG) and the subsequent blood oxygenation changes (fNIRS) [65]. One example includes adaptive general linear models that incorporate both electrical and hemodynamic information to improve activation detection [65]. These approaches can be particularly powerful when incorporating neurovascular coupling models to constrain the solution space and improve the physiological plausibility of the results [3].

Source-Decomposition Techniques

Source-decomposition methods represent more advanced fusion approaches that aim to separate mixed signals into their underlying components or sources.

Joint Independent Component Analysis (jICA) is a blind source separation technique that extends ICA to multiple modalities, assuming that the same underlying spatial components mix differently across modalities [64]. This approach can identify linked patterns of activation across electrical and hemodynamic domains.

Structured Sparse Multiset Canonical Correlation Analysis (ssmCCA) is a sophisticated method that identifies multivariate relationships between multiple datasets. As applied in recent research, ssmCCA fuses neural electrical and hemodynamic responses to pinpoint brain regions consistently detected by both fNIRS and EEG [19]. This method has been successfully used to identify activation over the left inferior parietal lobe, superior marginal gyrus, and post-central gyrus during motor execution, observation, and imagery [19].

Canonical Correlation Analysis (CCA) is a classical method that finds linear combinations of variables from two datasets that have maximum correlation with each other. In multimodal fusion, CCA can identify common temporal patterns or spatial components that are shared between EEG and fNIRS signals [64].

Table 2: Data Fusion Techniques for fNIRS-EEG

Fusion Category	Specific Methods	Key Characteristics	Applications
Concatenation	Feature concatenation	Simple implementation, may ignore modality relationships	Basic brain-state classification [65]
Decision-Level	Voting schemes, weighted averaging	Preserves modality-specific processing, simpler integration	Hybrid brain-computer interfaces [65]
Model-Based	Adaptive general linear models, Kalman filters	Incorporates physiological constraints	Neurovascular coupling studies [65] [3]
Source-Decomposition	jICA, ssmCCA, CCA	Reveals latent components, handles high-dimensional data	Identifying shared neural regions [64] [19]

Advanced Machine Learning Approaches

Recent advances have introduced sophisticated deep learning architectures specifically designed for multimodal fusion:

Multi-Branch Convolutional Neural Networks with Attention (MBC-ATT) employ independent branch structures to process EEG and fNIRS signals separately, leveraging the advantages of each modality [66]. These networks incorporate cross-modal attention mechanisms to dynamically weight the importance of features from each modality, strengthening the model's ability to focus on relevant signals [66].

Time-Distributed CNN-LSTM frameworks combine convolutional neural networks for spatial feature extraction with long short-term memory networks for temporal modeling, effectively capturing spatiotemporal patterns across modalities [66].

Experimental Protocols and Application Notes

Sensor Placement and Montage Design

Successful multimodal fusion begins with proper experimental setup, particularly regarding sensor placement compatibility. Both systems often use the international 10–20 system for electrode/sensor placement [64]. To avoid interference:

• Use high-density EEG caps with pre-defined fNIRS-compatible openings [64]
• Some fNIRS systems are designed to be embedded within EEG caps or mounted using optode holders that avoid electrode contact points [64]
• Customized joint-acquisition helmets can be created using 3D printing technology or cryogenic thermoplastic sheets tailored to individual head sizes [4]

Optimal Montage Design: Computational approaches can optimize optode placement to target specific regions of interest. The NIRSTORM toolbox with Optimal Montage functionality uses realistic head models to simulate light propagation and determines optode positions that maximize sensitivity to predefined target regions while respecting constraints on the number of sources, detectors, and distances [67].

Diagram 1: Optimal Montage Design Workflow (47 characters)

Motor Imagery Paradigm Protocol

Motor imagery (MI) paradigms are widely used in both basic neuroscience and clinical applications, particularly for brain-computer interfaces and stroke rehabilitation.

Participant Preparation and Calibration:

• Participants sit approximately 25 cm from a display monitor in an ergonomic position [10]
• For studies involving patients or those unfamiliar with MI, introduce a grip strength calibration procedure using a dynamometer and stress ball to enhance kinesthetic sensation and MI vividness [10]
• Record baseline signals: 1-minute eyes-closed followed by 1-minute eyes-open states, demarcated by an auditory cue [10]

Trial Structure:

• Visual cue presentation (2 s): A directional arrow indicating the required MI (e.g., left/right hand) [10]
• Execution phase (10 s): Participants perform kinesthetic MI of the grasping movement at approximately one imagined grasp per second [10]
• Inter-trial interval (15 s): Blank screen for rest [10]
• Repeat for at least 30 trials per session, with multiple sessions and adequate rest intervals [10]

Data Acquisition Parameters:

• EEG sampling rate: 256 Hz [10]
• fNIRS sampling rate: 11 Hz [10]
• Synchronization: Use event markers from stimulus presentation software (e.g., E-Prime) to simultaneously trigger both recording systems [10]

Block Design for Cognitive Tasks

Block designs are commonly used in fNIRS and multimodal studies due to their high signal-to-noise ratio and statistical power [68].

Standard n-back Working Memory Protocol:

• Task Structure: Each block consists of a 2-s instruction display, followed by a 40-s task period, then a 20-s rest period [66]
• Stimulus Presentation: During the task period, a random one-digit number is displayed every 2 s for 0.5 s, followed by a 1.5-s fixation cross [66]
• Task Conditions: Include 0-back, 2-back, and 3-back conditions to vary working memory load [66]
• Recommendations: Include sufficient repetitions (e.g., 20 trials × 3 series × 3 sessions) to ensure data reliability [66]

Best Practices for Block Design:

• Repeat each task block multiple times to achieve stable and reliable responses [68]
• Choose appropriate rest and task durations considering the hemodynamic delay (2-6 seconds) [64] [68]
• Jitter rest periods slightly (e.g., 28-32 seconds instead of fixed 30 seconds) to avoid correlation with periodic physiological confounds like Mayer waves [68]
• Randomize task order when multiple conditions are included to mitigate habituation effects [68]

Diagram 2: Block Design Experimental Structure (44 characters)

Data Acquisition and Synchronization

Hardware Integration:

• Some vendors offer integrated EEG-fNIRS systems, while others require combining separate systems [64]
• For separate systems, synchronization can be achieved via TTL pulses, parallel ports, or shared clock systems [64]
• Custom-designed hybrid caps with predefined positions for both EEG electrodes and fNIRS optodes ensure proper coverage and co-registration [10]

Synchronization Methods:

• Simple synchronization: Separate systems synchronized during acquisition and analysis via a host computer [4]
• Unified processor: Simultaneous processing and acquisition of both signals with high synchronization precision [4]
• Event markers: Transmitted from stimulus presentation software to simultaneously trigger both recording systems [10]

The Scientist's Toolkit: Essential Research Reagents and Materials

Table 3: Essential Materials for Simultaneous fNIRS-EEG Research

Item	Function/Purpose	Example Specifications
Integrated EEG-fNIRS Cap	Provides stable platform for co-registered sensor placement	32 EEG electrodes, 32 fNIRS sources, 30 detectors forming 90 channels [10]
fNIRS System	Measures hemodynamic responses via near-infrared light	Continuous wave system, 2 wavelengths (695 & 830 nm), 10 Hz sampling [19]
EEG Amplifier	Records electrical brain activity	256 Hz sampling rate, high input impedance [10]
Stimulus Presentation Software	Prescribes experimental paradigm and records behavioral responses	E-Prime, PsychoPy, Presentation
3D Digitizer	Records precise sensor locations for accurate co-registration	Fastrak (Polhemus) or similar magnetic digitizer [19]
Computational Tools	For data fusion analysis and montage optimization	NIRSTORM with CPLEX optimization, MCXlab for light simulation [67]
Auxiliary Calibration Equipment	Enhances task performance and data quality	Dynamometer, stress ball for motor imagery calibration [10]

The evolution of fNIRS-EEG data fusion techniques from simple concatenation to advanced source-decomposition methods represents significant progress in multimodal brain imaging. Each fusion approach offers distinct advantages depending on the research question, with source-decomposition techniques like ssmCCA showing particular promise for identifying neural regions consistently activated across electrical and hemodynamic domains [19].

Successful implementation requires careful attention to experimental design, sensor compatibility, and appropriate analysis techniques. The continued development of sophisticated fusion methods, coupled with optimized experimental protocols, will further enhance our ability to study brain function in naturalistic settings and advance clinical applications in neurology and psychiatry.

Ensuring Participant Comfort and Signal Stability for Longitudinal Studies

Longitudinal neuroimaging studies, which track changes in brain function over time, are essential for understanding brain development, cognitive aging, and the progression of neurological disorders [69]. Simultaneous electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) recordings present a powerful multimodal approach for such investigations, combining excellent temporal resolution with improved spatial localization [5] [17]. However, maintaining both participant comfort and signal stability across multiple sessions presents unique methodological challenges that can significantly impact data quality and study validity. This application note provides detailed protocols and evidence-based recommendations for optimizing sensor placement compatibility to ensure the reliability and comfort of longitudinal EEG-fNIRS studies, framed within the context of a broader thesis on multimodal integration.

Table 1: Key study parameters and participant demographics from relevant EEG-fNIRS research.

Study Reference	Participant Count & Demographics	Primary Cognitive Tasks	Data Modalities	Key Stability & Comfort Considerations
Phukhachee et al. (2025) [18]	16 participants (14M, 2F) Age: 21-37 years	Intentional visual memory encoding	Simultaneous EEG-fNIRS	Experiment duration; minimization of movements during 3s stimulus presentation and 9s decision period
Semantic Decoding Dataset (2025) [5]	12 native English speakers (3M, 9F) Age: 20-57 years (mean 32.75)	Silent naming, Visual/Auditory/Tactile imagery of animals/tools	Simultaneous EEG-fNIRS	Instructions to minimize physical movements, eye movements, facial expressions; right-handed participants only
Shin et al. (2018) [70]	26 right-handed (9M, 17F) Age: 26.1±3.5 years	n-back, DSR, Word Generation	Simultaneous EEG-fNIRS	~3.5 hour session duration; comfortable armchair; fixed finger placement; instructions to minimize body movement
Li et al. (2020) [71]	29 subjects (>50 years) HC, MCI, MAD, MSAD	Random digit encoding-retrieval	Simultaneous EEG-fNIRS	Special considerations for clinical populations (AD patients); all subjects able to follow instructions independently

Table 2: Signal quality and performance metrics from hybrid EEG-fNIRS studies.

Study Reference	Classification Accuracy	Data Quality Measures	Stability Measures
Li et al. (2020) [71]	Hybrid: 79.31% EEG-only: 65.52% fNIRS-only: 58.62%	PCCFS feature selection; right prefrontal and left parietal regions most informative for AD classification	Four-class classification (HC, MCI, MAD, MSAD) demonstrating stability across disease stages
MDNF Model (2024) [17]	Superior to state-of-the-art methods on two public datasets	STFT for EEG 2D image transformation; fNIRS spectral entropy; strategic EEG channel selection based on neuroanatomy	Enhanced classification across MI, n-back, DSR, and WG tasks
EEG-fNIRS Cognitive Motivation (2025) [18]	ERP amplitudes enhanced in motivated conditions (parietal/occipital channels, 300ms peak)	Time-frequency analysis (theta/low alpha power); Cohen's D and ANOVA for fNIRS statistical analysis	fNIRS showed variable HbO responses without significant differences between motivation conditions

Experimental Protocols for Simultaneous EEG-fNIRS Recordings

Protocol 1: Semantic Category Decoding During Mental Imagery

Background: This protocol investigates the feasibility of semantic neural decoding to develop brain-computer interfaces that communicate conceptual meaning directly, bypassing character-by-character spelling approaches [5].

Participant Preparation:

• Recruit right-handed native English speakers (for language-specific tasks)
• Conduct pre-experiment stimulus familiarization to ensure object recognition
• Obtain written informed consent for data sharing in anonymized form
• Compensate participants appropriately (£16 in original study)

Stimuli and Task Design:

• Utilize 18 animals and 18 tools as semantic categories
• Present images converted to grayscale (400×400 pixels) on white background
• Implement four mental tasks in randomized order across blocks:
  • Silent Naming: Participants silently name displayed object in their mind
  • Visual Imagery: Visualize the object in their mind
  • Auditory Imagery: Imagine sounds associated with the object
  • Tactile Imagery: Imagine the feeling of touching the object
• Use 3-second task periods with instructions to remain engaged while minimizing movements

Data Acquisition Parameters:

• Record simultaneous EEG and fNIRS from 12 participants
• Conduct follow-up experiment with 7 participants for auditory imagery only (5-second duration)
• Ensure proper synchronization between EEG and fNIRS systems

Protocol 2: Cognitive Motivation and Intentional Memory Encoding

Background: This protocol examines the neural correlates of cognitive motivation during intentional memory encoding, combining early EEG responses with slower fNIRS hemodynamic changes [18].

Participant Preparation:

• Screen for no history of visual perception or memory disorders
• Use Lemeshow method for sample size determination
• Obtain ethics approval for experimental procedures

Experimental Paradigm:

• Implement visual cognitive motivation task with scene images from SUN database
• Present stimuli for 3 seconds (attention span) followed by 9-second decision period
• Categorize trials based on motivation and subsequent recognition:
  • Want to Remember and Remembered (RR)
  • Want to Remember but Forgot (RF)
  • Did Not Want to Remember but Remembered (FR)
  • Did Not Want to Remember and Forgot (FF)

Data Acquisition and Analysis:

• Focus EEG analysis on ERPs during first second post-stimulus
• Conduct time-frequency analysis using wavelet transform (theta and low alpha power)
• Analyze fNIRS hemodynamic responses during subsequent 9-second decision period
• Use statistical analyses (Cohen's D, one-way ANOVA) for between-condition comparisons

Protocol 3: Multimodal Classification of Alzheimer's Disease Progression

Background: This protocol enables classification of Alzheimer's disease stages using complementary EEG and fNIRS information, demonstrating clinical application of hybrid monitoring [71].

Participant Recruitment:

• Recruit four participant groups: Healthy Controls (HC), Mild Cognitive Impairment (MCI), Mild AD (MAD), Moderate/Severe AD (MSAD)
• Include only right-handed participants above 50 years of age
• Obtain informed consent (from caregivers in severe cases)
• Ensure all patients can follow instructions independently

Task Design:

• Implement random digit encoding-retrieval task
• Focus on cortical regions implicated in AD progression (right prefrontal, left parietal)

Data Processing and Feature Selection:

• Extract EEG and fNIRS features separately
• Implement Pearson Correlation Coefficient-based Feature Selection (PCCFS)
• Use Linear Discriminant Analysis (LDA) classifier for performance evaluation
• Compare unimodal vs. hybrid classification accuracy

The Researcher's Toolkit: Essential Materials and Equipment

Table 3: Essential research reagents and materials for simultaneous EEG-fNIRS studies.

Item Name	Specification/Function	Application Notes
EEG Recording System	High-density caps (e.g., 32+ channels); electrode gel for impedance reduction	Strategic channel selection based on neuroanatomy; prioritize contralateral posterior sites for visual processing [72]
fNIRS Recording System	Paired source-detector optodes; wavelengths for HbO/HbR differentiation	Positioning for superficial cortical coverage; compatible with EEG cap design; minimal cross-talk effect >1cm [71]
EEG-fNIRS Compatible Caps	Integrated solutions with pre-configured electrode and optode holders	Critical for reproducible sensor placement across longitudinal sessions; ensures consistent spatial registration
Conductive Electrolyte Gel	Hypoallergenic formulation; stable impedance properties	Minimize skin irritation for extended recordings; ensure compatibility with fNIRS optodes
Abrasive Skin Prep Gel	Mild exfoliation to reduce electrode-skin impedance	Improve signal quality while maintaining participant comfort; pre-application protocol
Photo-Blocking Caps	Light-blocking material to prevent optode light leakage	Essential for signal fidelity; consider comfort and breathability for extended wear
Behavioral Task Software	Precisely timed stimulus presentation (e.g., Psychophysics Toolbox)	Ensure synchronization with neural data acquisition; millisecond timing precision
Data Synchronization System	Hardware/software solution for EEG-fNIRS temporal alignment	Critical for multimodal data fusion; TTL pulses or shared clock infrastructure

Signaling Pathways and Experimental Workflows

Experimental workflow for longitudinal EEG-fNIRS studies illustrating the integration of participant comfort measures throughout the study timeline.

Signal processing and analysis pathway for simultaneous EEG-fNIRS data, demonstrating the complementary nature of both modalities and their integration for longitudinal analysis.

Successful longitudinal EEG-fNIRS studies require meticulous attention to both participant comfort and signal stability across multiple sessions. The protocols and recommendations presented here provide a framework for optimizing sensor placement compatibility while maintaining data quality. Key considerations include strategic EEG channel selection based on neuroanatomy, implementation of comfortable yet stable sensor mounting systems, standardized protocols for minimizing movement artifacts, and careful temporal synchronization of multimodal data streams. By addressing these methodological challenges, researchers can leverage the complementary strengths of EEG and fNIRS to investigate dynamic brain processes across development, aging, and disease progression with enhanced reliability and participant tolerance.

Proving Efficacy: Validation Frameworks and Performance Metrics

Simultaneous electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) offers a powerful, non-invasive window into brain function by capturing complementary electrical and hemodynamic signals. A critical step in validating this multimodal approach involves benchmarking its signals against the established gold standard in hemodynamic imaging—functional magnetic resonance imaging (fMRI). This application note details the theoretical basis, quantitative benchmarks, and experimental protocols for correlating hybrid EEG-fNIRS data with fMRI, with a specific focus on implications for sensor placement compatibility. Such benchmarking is essential for refining integrated helmet design and ensuring that concurrent recordings accurately capture coupled brain activity, thereby building researcher confidence in the hybrid system's validity for basic research and clinical drug development.

The physiological link between these modalities is neurovascular coupling, the process where neural electrical activity triggers a metabolic demand, leading to a localized hemodynamic response [2]. fMRI measures the Blood Oxygen Level Dependent (BOLD) signal, which reflects the balance of oxygenated and deoxygenated hemoglobin. fNIRS directly quantifies this balance by measuring concentration changes in oxygenated hemoglobin (HbO) and deoxygenated hemoglobin (HbR) [2] [19]. This shared physiological origin creates a direct correlation between the fNIRS hemodynamic signal and the fMRI BOLD signal. Furthermore, the neuronal populations activated during a task generate both the electrical potentials measured by EEG and the metabolic demand that drives the hemodynamic response measured by fNIRS and fMRI, creating a temporal and spatial link between EEG features and the BOLD signal [73].

Quantitative Benchmarks and Correlations

Establishing quantitative benchmarks is crucial for evaluating the performance of EEG-fNIRS systems against fMRI. The following tables summarize key correlation metrics and quality control benchmarks derived from empirical studies.

Table 1: Correlation Metrics Between fNIRS and fMRI Hemodynamic Signals

Brain Region	Experimental Paradigm	Correlation Coefficient (fNIRS-fMRI)	Key Findings
Sensorimotor & Parietal Cortices [19]	Motor Execution, Observation, Imagery	HbO and BOLD show strong positive correlation (r > 0.70 in activated regions)	Multimodal fusion (ssmCCA) identified consistent activation in left inferior parietal lobe across conditions.
Prefrontal Cortex [2]	Cognitive Tasks (e.g., Mental Arithmetic)	HbO correlates with positive BOLD; HbR correlates with negative BOLD signal.	fNIRS provides a comparable hemodynamic measure to fMRI with superior motion tolerance.
General Cortical Areas [2]	Resting-State & Various Tasks	fNIRS signals are "similar to the BOLD response obtained by fMRI."	Confirms fNIRS as a valid surrogate for fMRI BOLD in superficial cortical layers.

Table 2: fMRI Quality Control Benchmarks for Reliable Functional Connectomics

Pipeline performance is critical for benchmarking. A systematic evaluation of 768 fMRI data-processing pipelines established that optimal pipelines must simultaneously minimize spurious test-retest discrepancies while remaining sensitive to biological effects of interest [74]. The following benchmarks are derived from this large-scale analysis:

Evaluation Criterion	Optimal Pipeline Performance Benchmark	Implication for EEG-fNIRS Correlation
Test-Retest Reliability	Minimizes Portrait Divergence (PDiv) between networks from repeated scans of the same individual.	High test-retest reliability in fMRI suggests a stable hemodynamic baseline against which to correlate fNIRS.
Sensitivity to Individual Differences	Network topology should reliably distinguish different subjects.	Ensures correlated signals reflect true inter-subject variability, not pipeline noise.
Sensitivity to Experimental Effects	Network topology should detect significant changes (e.g., pharmacological intervention like propofol).	Confirms that correlated fNIRS-EEG-fMRI changes are driven by the experimental manipulation.
Motion Confound Resistance	Topology is robust to varying levels of head motion.	Critical for comparing with fNIRS/EEG, which are also motion-tolerant, ensuring correlations are not motion-artifact driven.

Experimental Protocols for Multimodal Benchmarking

Protocol 1: Simultaneous fNIRS-EEG Recording with Post-Hoc fMRI Correlation

This protocol is designed for studies that cannot physically operate an fMRI scanner and an fNIRS-EEG system simultaneously. It relies on a standardized task performed by the same participants across different sessions.

1. Objective: To validate fNIRS hemodynamic responses and EEG-derived features against the fMRI BOLD signal acquired in a separate session using an identical task paradigm.

2. Materials and Reagents:

• Integrated fNIRS-EEG System: A compatible system with synchronized data acquisition, such as a NIRScout with BrainAMP or a unified processor for high-precision synchronization [4].
• fMRI Scanner: A 3T or higher MRI scanner equipped with a standard head coil.
• Stimulus Presentation Software: Software capable of delivering identical visual, auditory, or motor cues across both scanning environments (e.g., Presentation, PsychoPy).
• 3D Magnetic Space Digitizer (e.g., Polhemus Fastrak): For coregistering fNIRS optode and EEG electrode locations with structural MRI scans [19].

3. Procedure:

• Step 1: Participant Scheduling. Conduct the fNIRS-EEG recording session and the fMRI session on separate days, minimizing the interval between sessions to reduce biological variability.
• Step 2: Task Paradigm. Employ a block or event-related design that is easily replicable. Motor imagery tasks are highly suitable, as they involve the Action Observation Network and elicit robust responses in both electrical and hemodynamic domains [19]. Example: A "cup-moving" task with conditions for Motor Execution, Motor Observation, and Motor Imagery, each cued by audio commands [19].
• Step 3: fNIRS-EEG Acquisition.
  • Use a customized joint-acquisition helmet that integrates EEG electrodes and fNIRS optodes on a shared substrate to ensure stable relative positioning [4].
  • Place optodes over the targeted cortical regions (e.g., sensorimotor and parietal cortices for motor tasks).
  • Record EEG data continuously from 64+ channels according to the 10-10 system. Record fNIRS data at two wavelengths (e.g., 695 nm and 830 nm) to compute HbO and HbR concentration changes.
  • Digitize the positions of all optodes and electrodes in reference to anatomical landmarks (nasion, inion, preauricular points).
• Step 4: fMRI Acquisition.
  • Acquire a high-resolution T1-weighted anatomical scan.
  • Acquire T2*-weighted BOLD fMRI scans using parameters optimized for the target regions (e.g., TR/TE = 2000/30 ms, voxel size 3x3x3 mm³).
  • Present the identical task paradigm used in the fNIRS-EEG session.
• Step 5: Data Preprocessing.
  • fNIRS: Convert intensity to optical density, correct for motion artifacts (e.g., Savitzky-Golay filtering), bandpass filter (0.01–0.1 Hz), and convert to HbO/HbR concentration changes using the Modified Beer-Lambert Law [75].
  • EEG: Downsample to 250 Hz, high-pass filter (1 Hz), remove line noise, reject bad channels, and apply Artifact Subspace Reconstruction (ASR). Re-reference to the global average [75].
  • fMRI: Perform standard preprocessing including slice-time correction, realignment, coregistration to the anatomical scan, normalization to standard space (e.g., MNI), and spatial smoothing.
• Step 6: Data Correlation and Analysis.
  • Coregister fNIRS channels to the fMRI space using the digitized positions.
  • Extract the mean BOLD time series from the fMRI data within a sphere (~10 mm diameter) centered on each fNIRS channel location.
  • Extract the mean HbO time series from the corresponding fNIRS channels.
  • Calculate the Pearson correlation between the fNIRS HbO time series and the fMRI BOLD time series for each channel and subject.

Protocol 2: Data Fusion for Neural Activity Localization

This protocol utilizes advanced analytics to fuse EEG and fNIRS data, creating a unified metric that can be more powerfully correlated with fMRI-derived networks.

1. Objective: To leverage the high temporal resolution of EEG and the superior spatial specificity of fNIRS via data fusion, creating a composite biomarker for benchmarking against fMRI network topology.

2. Materials and Reagents:

• All materials from Protocol 1.
• Analysis Software: Custom scripts in MATLAB or Python for implementing structured sparse multiset Canonical Correlation Analysis (ssmCCA) [19].

3. Procedure:

• Steps 1-5: Follow the acquisition and preprocessing steps from Protocol 1.
• Step 6: Feature Extraction.
  • From EEG: Extract time-locked Event-Related Potentials (ERPs) (e.g., P300) or band power features (e.g., theta, alpha) from specific electrodes.
  • From fNIRS: Extract the mean HbO concentration from channels within the Region of Interest (ROI) during the task blocks.
• Step 7: Multimodal Data Fusion.
  • Apply ssmCCA to the extracted EEG and fNIRS features. This method finds a common latent representation that maximizes the correlation between the electrical and hemodynamic modalities, effectively fusing them into a single, spatially refined activation profile [19].
• Step 8: Correlation with fMRI Networks.
  • From the fMRI data, construct a functional brain network (connectome). Use an optimal pipeline as defined in Table 2. This involves:
    • Parcellation: Define nodes using a brain atlas (e.g., 200-node functionally-derived parcellation).
    • Edge Definition: Calculate functional connectivity between nodes using Pearson correlation of BOLD time series.
    • Network Filtering: Retain the top 10% of connections to create a sparse, weighted network [74].
  • Compare the fused EEG-fNIRS activation profile with the nodes and edges of the fMRI network. Statistically evaluate the overlap between the fused activation map and the hubs of the fMRI-derived functional connectome.

The Scientist's Toolkit: Essential Research Reagents and Materials

Table 3: Key Materials for Simultaneous EEG-fNIRS-fMRI Benchmarking Studies

Item Name	Function / Application	Example Specifications / Notes
Integrated fNIRS-EEG Helmet	Houses optodes and electrodes in a stable, co-registered configuration for simultaneous acquisition.	Custom designs using 3D-printing or cryogenic thermoplastic sheets improve fit and reduce cross-modality interference [4].
fNIRS System	Measures hemodynamic activity via changes in HbO and HbR concentration.	Continuous-wave systems (e.g., Hitachi ETG-4100) with 2+ wavelengths (695 nm, 830 nm); 24+ channels [19].
EEG System	Measures electrical brain activity with high temporal resolution.	64+ electrode systems (e.g., Electrical Geodesics); integrated amplifier with synchronization input [19] [73].
3D Magnetic Digitizer	Coregisters sensor locations with anatomical MRI scans for spatial accuracy.	Critical for mapping fNIRS/EEG data to fMRI space (e.g., Polhemus Fastrak) [19].
Structured Sparse Multiset CCA (ssmCCA)	Data fusion algorithm to integrate EEG and fNIRS signals into a unified activation profile.	Identifies brain regions consistently activated across both modalities, enhancing spatial localization [19].
HOMER3 Software Package	Standardized preprocessing and analysis of fNIRS data.	Converts raw intensity to hemoglobin concentrations, performs motion correction, and bandpass filtering [75].
EEGLab Toolbox	Open-source toolbox for preprocessing and analyzing EEG data.	Used for filtering, artifact removal (e.g., ASR), and ERP analysis [75].
Optimal fMRI Pipeline	Constructs reliable functional connectomes from BOLD data for benchmarking.	Involves specific combinations of parcellation, connectivity definition, and filtering as per systematic evaluation [74].

Implications for Sensor Placement Compatibility

The correlation strength between fNIRS-EEG and fMRI is highly dependent on accurate spatial co-registration of sensors. Benchmarking studies directly inform helmet design and sensor placement strategies:

• Informed Region of Interest (ROI) Selection: fMRI meta-analyses can identify core hubs of functional networks, such as the left frontal and prefrontal cortices that show age-related connectivity increases [76]. These hubs should be prioritized for optode and electrode placement in hybrid systems.
• Guiding Integrated Helmet Design: The finding that elastic fabric caps can lead to inconsistent optode-scalp contact and varying source-detector distances [4] underscores the need for rigid, customized helmets (e.g., 3D-printed) to maintain stable placement. This stability is a prerequisite for obtaining reliable correlations with fMRI.
• Validating Co-registration Accuracy: The process of digitizing sensor locations and coregistering them with individual anatomy [19] is not merely a procedural step; it is fundamental for ensuring that the fNIRS and EEG signals being correlated with fMRI originate from the same underlying cortical tissue. Inaccurate co-registration will systematically weaken observed correlations.

In conclusion, rigorous benchmarking of simultaneous EEG-fNIRS against the gold standard of fMRI is not an isolated validation exercise. It is an iterative process that directly refines the physical integration and sensor placement strategies of multimodal systems. By adhering to the detailed protocols and benchmarks outlined herein, researchers can enhance the validity, reliability, and spatial precision of their findings, thereby advancing the utility of hybrid EEG-fNIRS in neuroscience and drug development.

The integration of electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) represents a paradigm shift in brain-computer interface (BCI) development and clinical diagnostics. This multimodal approach leverages the complementary strengths of each modality: EEG provides millisecond-scale temporal resolution of electrical neural activity, while fNIRS offers centimeter-scale spatial resolution of hemodynamic responses [49] [19]. Within the context of sensor placement compatibility for simultaneous EEG-fNIRS research, quantifying the precise added value in classification accuracy and diagnostic precision becomes paramount for advancing robust BCI systems and neurological assessment tools. This protocol outlines standardized methodologies for quantifying these improvements through controlled experimentation and analytical validation.

Quantitative Comparison of Classification Accuracies

Table 1: Classification Accuracy Improvements with fNIRS-Specific Advanced Methods

Method	Task	Baseline Accuracy	Enhanced Accuracy	Improvement	Reference
Common Spatial Pattern (CSP) + SVM	Hand Motion/Motor Imagery	59.81%	81.63%	+21.82%	[77]
Common Spatial Pattern (CSP) + LDA	Hand Motion/Motor Imagery	69.00%	84.19%	+15.19%	[77]
z-score Channel Selection	Motor Imagery vs. Rest	~80% (est.)	87.2%	~7%	[78]
Kalman Filter/a-GMM	Mental Arithmetic vs. Mental Singing	Conventional Methods	Proposed Method	>10% (vs. BPF-statistical features)	[79]
Deep Learning Stack/FFT	Hand Gripping	85.16% (CNN)	90.11% (FFT)	+4.95%	[80] [81]

Table 2: Performance of Multimodal EEG-fNIRS Fusion Approaches

Fusion Level	Method	Task	Classification Accuracy	Reference
Feature-Level	Mutual Information Feature Selection	Visuo-Mental Task	Significantly improved vs. single modality	[82]
Feature-Level	Multi-Domain Features + Multi-Level Learning	Motor Imagery	96.74%	[83]
Feature-Level	Multi-Domain Features + Multi-Level Learning	Mental Arithmetic	98.42%	[83]
Decision-Level	SVM Classifier Fusion	Mental Stress Detection	+7.76% vs. EEG; +10.57% vs. fNIRS	[83]
Data-Level	Structured Sparse Multiset CCA	Motor Execution/Observation/Imagery	Identified shared AON regions	[19]

Experimental Protocols for Quantifying Added Value

Protocol 1: unimodal fNIRS-BCI Enhancement Validation

This protocol quantifies the added value of advanced signal processing for fNIRS-BCI, establishing a baseline for subsequent multimodal comparison.

• Objective: To evaluate the efficacy of the Common Spatial Pattern (CSP) algorithm and z-score-based channel selection in improving fNIRS classification accuracy.
• Participants: 15-29 healthy, right-handed adults [77] [78].
• Equipment:
  • Continuous-wave fNIRS system (e.g., 20-channel setup [77]).
  • Optodes placed over the motor cortex according to the 10-20 system [80].
• Task Paradigm:
  • Block Design: Participants perform tasks in randomized blocks (e.g., 10-20s task periods interspersed with 20-30s rest periods) [78] [80].
  • Tasks: Include left/right hand motor execution, motor imagery [77], and mental arithmetic [78].
• Data Processing & Analysis:
  • Preprocessing: Convert raw light intensity to oxy-hemoglobin (ΔHbO) and deoxy-hemoglobin (ΔHbR) concentration changes using the Modified Beer-Lambert Law [80]. Apply band-pass filtering to remove physiological noise.
  • Channel Selection (z-score method):
    • Generate a desired hemodynamic response function (dHRF).
    • Calculate cross-correlation between dHRF and each channel's averaged trial signal.
    • Form a vector of maximum correlation coefficients and compute their z-scores.
    • Select channels with a z-score > 0 for further analysis [78].
  • Feature Extraction: For selected channels, extract statistical features (mean, variance, slope) from the ΔHbO signal time series [77] [79].
  • Dimensionality Reduction & Classification:
    • Apply the CSP algorithm to spatial filter the feature set, reducing dimensionality [77].
    • Feed the reduced features into classifiers (SVM, LDA).
    • Use cross-validation to compute average classification accuracy.
• Quantification of Added Value: Compare the average classification accuracy achieved with CSP and/or z-score selection against the accuracy using all channels and standard features.

Protocol 2: Multimodal EEG-fNIRS Integration and Fusion

This protocol directly quantifies the value added by integrating EEG with fNIRS, with explicit consideration of sensor placement.

• Objective: To determine the classification accuracy gain from multimodal EEG-fNIRS fusion compared to unimodal systems.
• Participants: 20-60 healthy adults [19] [83].
• Equipment & Sensor Placement:
  • EEG System: High-density amplifier (e.g., 128-electrode cap from Brain Products [49]).
  • fNIRS System: Continuous-wave system (e.g., Hitachi ETG-4100 [19]).
  • Integrated Cap: Use a cap with sufficient slits (e.g., 128-160) and black fabric to host both EEG electrodes and fNIRS optodes, minimizing interference [49]. Pre-define the montage based on the region of interest (e.g., motor cortex for motor imagery).
  • Synchronization: Implement Lab Streaming Layer (LSL) protocol or shared hardware triggers to ensure temporal alignment of EEG and fNIRS data streams [49].
• Task Paradigm: Utilize established paradigms like left/right hand motor imagery and mental arithmetic versus rest [83].
• Data Processing & Analysis:
  • Unimodal Processing:
    • EEG: Preprocess (filter, remove artifacts), extract features (band power from sensorimotor rhythms, time-frequency features).
    • fNIRS: Preprocess (convert to ΔHbO/ΔHbR, filter), extract features (mean, slope, etc.).
  • Multimodal Fusion:
    • Feature-Level Fusion: Concatenate feature vectors from both modalities [83]. Optionally, employ advanced selection (e.g., Mutual Information [82] or Atomic Search Optimization [83]) to reduce redundancy and enhance complementarity.
    • Decision-Level Fusion: Train separate classifiers on EEG and fNIRS features. Fuse the classifier outputs (e.g., via weighted voting or a meta-classifier) [82] [83].
  • Classification: Train a classifier (e.g., SVM, LDA, or deep learning model) on the fused feature set or fused decisions.
• Quantification of Added Value: Added Value = Accuracy(EEG-fNIRS) - max( Accuracy(EEG), Accuracy(fNIRS) ) This formula directly quantifies the performance gain attributable to multimodal integration.

The Scientist's Toolkit: Essential Research Reagents & Materials

Table 3: Key Materials and Equipment for Simultaneous EEG-fNIRS Research

Item	Function/Description	Example/Specification
Integrated EEG-fNIRS Cap	Holds both EEG electrodes and fNIRS optodes in a stable, pre-defined montage. Black fabric reduces optical reflection.	actiCAP Xpress with 160 slits [49].
EEG Amplifier	Records electrical brain activity with high temporal resolution.	Brain Products actiCHamp Plus [49].
fNIRS System	Measures hemodynamic changes by emitting NIR light and detecting attenuation.	NIRSport2 (NIRx) [80]; Hitachi ETG-4100 [19].
Synchronization Interface	Ensures precise temporal alignment of multimodal data streams.	Lab Streaming Layer (LSL) software protocol [49].
3D Digitizer	Records the precise 3D locations of EEG electrodes and fNIRS optodes for accurate spatial coregistration.	Fastrak (Polhemus) [19].
Analysis Software	For data preprocessing, feature extraction, fusion, and classification.	nirsLAB [80], MATLAB with toolboxes (e.g., BBCI, Homer2).

Workflow and Signaling Pathways

Multimodal Integration and Analysis Workflow

This diagram visualizes the complete experimental and analytical pipeline for a simultaneous EEG-fNIRS study, from participant preparation to the quantification of added value.

Technical Implementation & Sensor Placement Logic

This diagram outlines the critical technical considerations and logical decisions involved in setting up a compatible sensor montage for simultaneous EEG-fNIRS recordings.

Stroke remains a leading cause of global disability and mortality, with intracerebral hemorrhage (ICH) representing a particularly severe stroke subtype accounting for 6.5%–19.6% of all stroke cases and contributing disproportionately to stroke-related mortality [10] [84]. Motor impairment, especially upper limb dysfunction, is a recognized sequelae in approximately 55%–75% of stroke survivors, critically impacting functional independence and socioeconomic participation [10]. Modern neurorehabilitation increasingly leverages technological systems to facilitate recovery, with unimodal and hybrid neuroimaging approaches emerging as promising tools for understanding and promoting neuroplasticity.

The fundamental distinction between unimodal and hybrid systems lies in their data acquisition architecture. Unimodal systems rely on a single neuroimaging modality—typically measuring either electrical neural activity (e.g., EEG, EMG) or hemodynamic responses (e.g., fNIRS)—while hybrid systems integrate complementary modalities to capture multifaceted brain activity [2] [4]. This analysis systematically compares the efficacy, technical requirements, and clinical applications of both approaches within stroke and ICH rehabilitation contexts, with particular emphasis on sensor placement compatibility for simultaneous EEG-fNIRS research.

Performance Comparison: Hybrid vs. Unimodal Systems

Quantitative Performance Metrics

Table 1: Comparative performance metrics of hybrid versus unimodal systems

System Type	Modalities	Classification Accuracy	Temporal Resolution	Spatial Resolution	Key Advantages
Hybrid	EEG-fNIRS	82.30%-87.24% [85]	Millisecond (EEG) [2]	~1-2 cm (fNIRS) [2]	Spatiotemporal complementarity, robust cross-subject generalization [85]
Hybrid	EMG-EEG	94.5% (vs. 88.5% EMG-only) [86]	Millisecond (both) [86]	N/A (EMG)	Fatigue-adaptive control, enhanced robustness [86]
Unimodal	EEG	Limited in ICH populations [10]	Millisecond [2]	~2 cm [5]	High temporal resolution, portable [4]
Unimodal	fNIRS	Limited in ICH populations [10]	~1s [2]	~1-2 cm [2]	Superior spatial resolution, motion artifact resistant [10]
Multimodal Treatment	Resistance + Endurance	N/A	N/A	N/A	Significantly improved knee-extensor strength (SMD=1.25) [87]

Figure 1: Complementary spatiotemporal resolution of EEG and fNIRS forms the foundation for hybrid system superiority in neurorehabilitation applications.

Clinical Efficacy in Rehabilitation Outcomes

Table 2: Clinical efficacy measures across rehabilitation system types

System Type	Patient Population	Primary Outcome Measures	Clinical Efficacy	Limitations
Hybrid EEG-fNIRS	ICH patients [85]	Motor imagery classification accuracy	74.87% mean accuracy with transfer learning [85]	Neurophysiological heterogeneity, complex setup
Hybrid EMG-EEG	Post-stroke (evaluated with healthy participants) [86]	Intention classification accuracy during fatigue	91.4% vs. 83.1% (EMG-only) under high fatigue [86]	Limited validation in clinical populations
Unimodal EEG	Stroke patients [10]	Motor imagery decoding	Variable, signal quality issues in ICH [10]	Susceptibility to motion artifacts, poor spatial specificity [10]
Unimodal fNIRS	Stroke patients [10]	Hemodynamic response detection	Variable, limited temporal resolution [2]	Poor temporal resolution, limited penetration depth [2]
Multimodal Exercise	Post-stroke patients [87]	Knee-extensor strength, gait speed, endurance	Significant improvement in lower limb function [87]	Requires physical capability, equipment access

Figure 2: Hybrid systems demonstrate particular advantage in ICH rehabilitation where neurovascular uncoupling challenges unimodal approaches, creating a compelling case for multimodal integration.

Technical Protocols for Hybrid System Implementation

Simultaneous EEG-fNIRS Acquisition Protocol

Objective: To achieve synchronized acquisition of electrophysiological (EEG) and hemodynamic (fNIRS) activity during motor imagery tasks for ICH rehabilitation.

Equipment Requirements:

• EEG system with active electrodes (e.g., g.tec g.HIamp, g.Nautilus) [88]
• fNIRS system (e.g., NirScan, NIRSport2) [10] [88]
• Custom hybrid EEG-fNIRS cap with integrated optode and electrode holders [10] [88]
• Synchronization interface (e.g., E-Prime 3.0 for event marking) [10]

Sensor Placement Configuration:

• Follow international 10-20 system for EEG electrode placement [10]
• Position fNIRS optodes to cover motor cortex regions (C3, Cz, C4) [10]
• Maintain 3cm source-detector distance for fNIRS channels [10]
• Integrate EEG electrodes within fNIRS optode matrix using flexible cap [88]
• Ensure scalp-contact pressure is sufficient but comfortable to minimize artifacts [4]

Experimental Paradigm (Adapted from HEFMI-ICH Dataset [10]):

• Preparatory Phase: Grip strength calibration using dynamometer and stress ball to enhance motor imagery vividness
• Baseline Recording: 1-minute eyes-closed followed by 1-minute eyes-open states
• Task Block Structure:
  • Visual cue presentation (2s): Directional arrow indicating left/right hand MI
  • Execution phase (10s): Kinesthetic MI of grasping movement at 1Hz rate
  • Inter-trial interval (15s): Rest period with blank screen
• Session Configuration: Minimum 2 sessions, 15 trials per hand per session

Data Synchronization and Processing:

• Utilize common trigger signals from stimulus presentation software to both systems [10]
• Employ hardware synchronization with EEG amplifier as master device [88]
• Preprocess EEG (0.5-40Hz bandpass filter, artifact removal) and fNIRS (hemoglobin concentration conversion via Modified Beer-Lambert Law) [2]

Fatigue-Adaptive Hybrid EMG-EEG Protocol for Robotic Rehabilitation

Objective: To implement real-time fatigue-adaptive intention detection for upper-limb robotic rehabilitation using hybrid EMG-EEG signals.

Equipment Requirements:

• Surface EMG system (wearable armband for biceps/triceps brachii) [86]
• EEG system (e.g., OpenBCI with motor cortex electrodes) [86]
• Lightweight robotic elbow rehabilitation device [86]
• Real-time signal processing platform (e.g., MATLAB/Simulink with custom algorithms) [86]

Sensor Placement Configuration:

• EMG electrodes positioned on biceps brachii and triceps brachii muscle bellies
• EEG electrodes focused on motor cortex (C3, Cz, C4 per 10-20 system)
• Ensure minimal interference between EMG and EEG sensor locations

Signal Processing and Fusion Pipeline:

• EMG Processing: Feature extraction (mean absolute value, waveform length), SVM classification [86]
• EEG Processing: Common Spatial Pattern filtering, SVM classification [86]
• Fatigue Estimation: k-NN classifier on EMG spectral features (median frequency, mean power frequency) [86]
• Adaptive Bayesian Fusion: Dynamic weighting based on fatigue level: α(f) = 0.2 + 0.6 × f(x) [86]

Experimental Implementation:

• Participants perform elbow flexion/extension tasks with robotic assistance
• Real-time classification output controls robotic movement intent
• System dynamically increases EEG weighting as fatigue escalates [86]

The Scientist's Toolkit: Essential Research Reagents and Materials

Table 3: Essential research materials for hybrid neuroimaging in stroke rehabilitation

Item	Function	Example Products/Specifications
Hybrid EEG-fNIRS Cap	Integrated sensor placement for simultaneous acquisition	Custom designs with 32 EEG electrodes, 32 fNIRS sources, 30 detectors [10]
Biosignal Amplifiers	Signal acquisition and preprocessing	g.HIamp, g.Nautilus (g.tec), NIRSport2 (NIRx) [88]
Synchronization Interface	Temporal alignment of multimodal data	E-Prime 3.0, LabStreamingLayer [10]
EMG-EEG Fusion Algorithm	Real-time intention classification	Adaptive Bayesian fusion with fatigue estimation [86]
Transfer Learning Framework	Cross-subject generalization for ICH patients	Wasserstein metric-driven source domain selection [85]
Motor Imagery Paradigm	Standardized cognitive task protocol	HEFMI-ICH dataset paradigm [10]
Robotic Rehabilitation Interface	Physical implementation of intention detection	Portable elbow rehabilitation robot with bilateral arm support [86]

Hybrid neuroimaging systems demonstrate significant advantages over unimodal approaches for stroke and ICH rehabilitation, primarily through their ability to capture complementary neural signatures and maintain robust performance under challenging clinical conditions. The integration of EEG with fNIRS or EMG creates synergistic systems that overcome individual modality limitations, particularly relevant for ICH populations where neurovascular uncoupling complicates unimodal assessment. Successful implementation requires careful attention to sensor placement compatibility, synchronization protocols, and adaptive fusion algorithms tailored to the unique neurophysiological presentations of stroke patients. Future developments should focus on standardized acquisition hardware, automated sensor placement systems, and clinically viable protocols to translate hybrid neuroimaging benefits from research laboratories to rehabilitation clinics.

Leveraging Public Datasets for Methodological Validation and Benchmarking

The integration of electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) offers a powerful multimodal approach to non-invasive brain imaging. EEG provides millisecond-level temporal resolution for tracking rapid neural electrical activity, while fNIRS delivers superior spatial localization of slower hemodynamic responses through measurements of oxygenated (HbO) and deoxygenated hemoglobin (HbR) concentrations [10] [89]. This complementary relationship enables more comprehensive investigation of brain function, particularly for applications in brain-computer interfaces (BCIs), motor rehabilitation, and cognitive neuroscience [90] [91].

A significant challenge in simultaneous EEG-fNIRS research involves optimizing sensor placement compatibility to ensure both modalities capture neural activity from the same cortical regions without technical interference. Publicly available datasets provide essential resources for methodological validation, algorithm benchmarking, and protocol standardization before researchers invest in costly original data collection [10] [91]. This application note outlines how to strategically leverage these resources, with a specific focus on addressing sensor placement considerations within multimodal neuroimaging research.

Available Public EEG-fNIRS Datasets

Several research groups have recently contributed high-quality, publicly available datasets that combine EEG and fNIRS recordings across various experimental paradigms. These resources provide valuable benchmarks for methodological development.

Table 1: Publicly Available Multimodal EEG-fNIRS Datasets for Methodological Validation

Dataset Name	Primary Focus	Participants	EEG Channels	fNIRS Channels	Key Features	Sensor Placement Documentation
HEFMI-ICH [10]	Motor Imagery for Intracerebral Hemorrhage Rehabilitation	17 healthy, 20 ICH patients	32	90	Clinical population, synchronized acquisition	3D coordinates under MNI template for both modalities
Multi-Joint MI Dataset [91]	Upper Limb Motor Imagery	18 healthy	64	20 fNIRS channels (8 sources, 8 detectors)	8 MI tasks across 4 joint types	Diagram of integrated cap layout with co-located sensors
Semantic Decoding Dataset [5]	Imagery of Animals vs. Tools	12 healthy	128	86	Semantic category decoding	Co-registered using standard 10-20 system
Implicit Learning Dataset [6]	Serial Reaction Time Task	30 healthy	32	48 (24 per hemisphere)	Learning classification	Reference file mapping fNIRS to EEG 10-20 system
EEG-fNIRS NFB Protocol [89] [51]	Motor Imagery Neurofeedback	30 healthy	32 (19 over sensorimotor)	Custom layout targeting sensorimotor cortex	Real-time NFB platform	Custom cap integrating both systems; code publicly available

Table 2: Technical Specifications Across Featured EEG-fNIRS Datasets

Dataset	EEG Sampling Rate	fNIRS Sampling Rate	fNIRS Measurement Types	Trigger Synchronization	Data Formats
HEFMI-ICH [10]	256 Hz	11 Hz	HbO, HbR	E-Prime 3.0 event markers	Raw, preprocessed, feature-engineered
Multi-Joint MI Dataset [91]	1000 Hz	7.8125 Hz	HbO, HbR	Simultaneous recording system	European Data Format (.edf), CSV
Semantic Decoding Dataset [5]	500 Hz	10 Hz	HbO, HbR	Not specified	BIDS format
Implicit Learning Dataset [6]	500 Hz or 1000 Hz	10 Hz	HbO, HbR, Total Hb	Event markers in separate files	.edf, CSV with event markers
EEG-fNIRS NFB Protocol [51]	Not specified	Not specified	HbO, HbR	Custom real-time platform	Real-time streaming compatible

Experimental Protocols for Sensor Compatibility Validation

Protocol 1: Cross-Modal Signal Verification Using Motor Imagery Paradigms

Purpose: To validate that EEG and fNIRS sensors are capturing complementary neural activity from the same cortical regions during motor imagery tasks.

Materials and Equipment:

• Simultaneous EEG-fNIRS recording system with integrated cap
• Experimental paradigm presentation software (e.g., E-Prime, PsychoPy)
• Grip strength dynamometer or stress ball (for motor imagery calibration) [10]

Procedure:

• Participant Preparation: Recruit right-handed participants (healthy adults or clinical populations based on research goals). Prepare scalp according to standard EEG protocols, ensuring optode-scalp contact quality for fNIRS.
• Motor Imagery Training: Implement a grip strength calibration procedure to enhance kinesthetic motor imagery vividness [10]. Have participants perform maximal force exertions using a dynamometer followed by equivalent force applications with a stress ball.
• Experimental Paradigm:
  • Position participant 60-100 cm from visual display monitor
  • Conduct baseline recordings: 1-minute eyes-closed followed by 1-minute eyes-open [10]
  • Implement block design: Each trial consists of:
    • Visual cue presentation (2 s): Directional arrow indicating left or right hand MI
    • Execution phase (10 s): Participants perform kinesthetic MI of grasping movement
    • Inter-trial interval (15 s): Blank screen for rest [10]
  • Include minimum of 30 trials per session (15 left/right hand MI each)
• Data Quality Assessment:
  • Verify EEG signal quality: Impedances <10 kΩ, check for artifacts
  • Confirm fNIRS signal quality: Assess signal-to-noise ratio, check for motion artifacts
  • Validate temporal synchronization between systems using event markers

Analysis:

• For EEG: Compute event-related desynchronization (ERD) in mu (8-13 Hz) and beta (13-30 Hz) rhythms over sensorimotor cortex
• For fNIRS: Calculate HbO and HbR concentration changes during MI periods versus baseline
• Assess spatial correspondence: Compare activation foci across modalities using coregistration with standard brain space

Protocol 2: Sensor Placement Reproducibility Assessment

Purpose: To evaluate the consistency of multimodal measurements across repeated sessions with sensor replacement.

Background: fNIRS signals show better reproducibility for HbO than HbR, and source localization improves reliability [92]. Optode placement shifts reduce spatial overlap across sessions.

Procedure:

• Initial Session Setup:
  • Apply integrated EEG-fNIRS cap according to manufacturer specifications
  • Digitize electrode and optode positions using 3D digitizer
  • Record motor imagery or visual tasks as described in Protocol 1
• Repeated Sessions (minimum 3 sessions on separate days):
  • Remove and reapply cap between sessions
  • Redigitize sensor positions for each session
  • Maintain identical experimental paradigm across sessions
• Data Collection:
  • Collect minimum of 10 sessions for robust reproducibility assessment [92]
  • Ensure consistent task performance across sessions

Analysis:

• Calculate intraclass correlation coefficients (ICC) for both EEG and fNIRS features across sessions
• Quantify spatial overlap of activated regions using Dice coefficient or similar metrics
• Correlate optode placement shifts with signal reproducibility measures

Research Workflow for Sensor Placement Validation

Signaling Pathways in Simultaneous EEG-fNIRS

The physiological relationship between EEG and fNIRS signals is governed by neurovascular coupling, the process by which neural activity triggers hemodynamic responses. Understanding these pathways is essential for interpreting multimodal data.

Neurovascular Coupling Between EEG and fNIRS

The Scientist's Toolkit: Essential Research Reagents and Materials

Table 3: Essential Materials for Simultaneous EEG-fNIRS Research

Item	Specification	Research Function	Compatibility Considerations
Integrated EEG-fNIRS Caps	Custom designs with 32-64 EEG electrodes, 16-32 fNIRS optodes	Ensures consistent spatial relationship between modalities	Look for MNI coordinate documentation; consider head size variations
fNIRS Systems	Continuous wave systems with 760/850 nm wavelengths	Measures HbO and HbR concentration changes	Ensure synchronization capability with EEG systems
EEG Amplifiers	High-impedance (>1 GΩ) with 256+ Hz sampling rate	Records electrical brain activity	Select systems with optical isolation to reduce fNIRS interference
Abrasive Preparatory Gels	EEG-specific abrasive gels	Reduces electrode impedances	Ensure compatibility with fNIRS optodes (non-interfering)
3D Digitizers	Magnetic or optical digitization systems	Documents precise sensor locations	Essential for coregistration with anatomical scans
Synchronization Hardware	TTL pulse generators, parallel port interfaces	Maintains temporal alignment between systems	Verify compatibility with both EEG and fNIRS equipment
Paradigm Presentation Software	E-Prime, PsychoPy, Presentation	Prescribes experimental tasks	Must output synchronous triggers to both recording systems

Public multimodal EEG-fNIRS datasets provide critical resources for validating sensor placement methodologies and analytical approaches before undertaking original research. The protocols and resources outlined in this application note enable researchers to systematically address the technical challenges of simultaneous data acquisition, particularly the spatial compatibility between EEG electrodes and fNIRS optodes. As multimodal neuroimaging continues to advance, standardized validation approaches using these public resources will enhance methodological rigor and reproducibility across the field.

Conclusion

The strategic integration of EEG and fNIRS through compatible sensor placement creates a powerful tool that transcends the limitations of either modality alone. By understanding the foundational principles, implementing robust methodological designs, proactively troubleshooting artifacts, and adhering to rigorous validation standards, researchers can unlock a more complete picture of brain function. The future of this multimodal approach is exceptionally promising, driven by advancements in machine learning for data fusion, the development of more sophisticated and comfortable wearable systems, and the translation of these technologies into personalized clinical diagnostics and neurorehabilitation protocols for conditions like intracerebral hemorrhage and stroke. This progression will fundamentally enhance how we monitor therapeutic efficacy and understand neurological disorders in drug development and clinical practice.

References

• 1. fNIRS vs EEG: Comparing Non-Invasive Brain Imaging ... [https://imotions.com/blog/learning/research-fundamentals/fnirs-vs-eeg/?srsltid=AfmBOor-LEkPiUr3njh-FTCysPsqPVHgu8Sa__VZCwJ2sspLP83-3cAc]
• 2. Concurrent fNIRS and EEG for Brain Function Investigation [https://pmc.ncbi.nlm.nih.gov/articles/PMC9371171/]
• 3. Simultaneous functional near-infrared spectroscopy and ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC5566595/]
• 4. A Cross-Disciplinary Review of the fNIRS-EEG Dual-Modality ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11506513/]
• 5. Simultaneous EEG and fNIRS recordings for semantic ... [https://www.nature.com/articles/s41597-025-04967-0]
• 6. Simultaneous EEG-fNIRS Data on Learning Capability via ... [https://www.mdpi.com/2306-5729/10/8/131]
• 7. EEG Biomarkers: Transforming CNS Drug Discovery [https://www.neuroelectrics.com/blog/eeg-biomarkers-transforming-cns-drug-discovery]
• 8. Multimodal fNIRS–EEG sensor fusion: Review of data-driven ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12592382/]
• 9. Comparing structure–function relationships in brain ... [https://www.nature.com/articles/s41598-024-79817-x]
• 10. HEFMI-ICH: a hybrid EEG-fNIRS motor imagery dataset for ... [https://www.nature.com/articles/s41597-025-06100-7]
• 11. Optimizing spatial specificity and signal quality in fNIRS [https://www.frontiersin.org/journals/neuroergonomics/articles/10.3389/fnrgo.2024.1286586/full]
• 12. A Practical Guide for Combined fNIRS & EEG Recordings [https://www.gtec.at/2021/04/12/combined-fnirs-eeg-recordings/?srsltid=AfmBOoo4kKsoNhwK09qd9JC-rXis4a2jl32wIEgpZ9jwYEhY5XBjr3Z4]
• 13. Multimodal fNIRS-EEG measurements — Integration on the ... [https://www.artinis.com/blogpost-all/2023/multimodal-fnirs-eeg-measurements-integration-on-the-head]
• 14. A quantitative model for human neurovascular coupling with ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC9821752/]
• 15. A quantitative model for human neurovascular coupling with ... [https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1010818]
• 16. fNIRS vs EEG: Comparing Non-Invasive Brain Imaging ... [https://imotions.com/blog/learning/research-fundamentals/fnirs-vs-eeg/?srsltid=AfmBOopAIN0objnsRHPAUbHgIVjcJHDrzLRgkyyfb_PhGuo4SsnybLqX]
• 17. Simultaneous EEG-fNIRS Data Classification Through ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11379445/]
• 18. Simultaneous EEG-fNIRS study of visual cognitive processing [https://pmc.ncbi.nlm.nih.gov/articles/PMC12157042/]
• 19. Simultaneous multimodal fNIRS-EEG recordings reveal ... [https://www.nature.com/articles/s41598-023-31609-5]
• 20. An EEG-fNIRS neurovascular coupling analysis method to ... [https://pubmed.ncbi.nlm.nih.gov/37196454/]
• 21. Stage-specific functional networks associated with ... [https://www.frontiersin.org/journals/aging-neuroscience/articles/10.3389/fnagi.2025.1562203/full]
• 22. A Practical Guide for Combined fNIRS & EEG Recordings [https://www.gtec.at/2021/04/12/combined-fnirs-eeg-recordings/?srsltid=AfmBOopnxwtC8PUwlnj39kncpwPsXWN0Sh-jlxqHvAzvko5HGsCS14L2]
• 23. Brain Activity Within Prefrontal Cortex: A Resting‐State fNIRS ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12620040/]
• 24. Wearable, Integrated EEG–fNIRS Technologies: A Review [https://pmc.ncbi.nlm.nih.gov/articles/PMC8469799/]
• 25. fNIRS vs EEG: Comparing Non-Invasive Brain ... [https://imotions.com/blog/learning/research-fundamentals/fnirs-vs-eeg/?srsltid=AfmBOooE8xqp6AtZPTAisZNpUjh2N32oGmTqTwsCcNvGqs_lEG3eaHRH]
• 26. Hybrid Integrated Wearable Patch for Brain EEG-fNIRS ... [https://www.mdpi.com/1424-8220/24/15/4847?]
• 27. Developing customized NIRS-EEG for infant sleep ... [https://www.spiedigitallibrary.org/journals/neurophotonics/volume-10/issue-3/035010/Developing-customized-NIRS-EEG-for-infant-sleep-research--methodological/10.1117/1.NPh.10.3.035010.full]
• 28. Easycap EEG Recording Caps: An Overview (2025) [https://pressrelease.brainproducts.com/easycap-cap-overview-2025/]
• 29. A Guide to Choosing the Right Headcap for fNIRS ... [https://www.artinis.com/blogpost-all/choosing-the-right-headcap-for-fnirs-studies]
• 30. Development of an Integrated EEG/fNIRS Brain Function ... [https://www.mdpi.com/1424-8220/21/22/7703]
• 31. Innovative wireless EEG and fNIRS device for 64 channels [https://www.gtec.at/product/g-nautilus-wireless-eeg-and-fnirs-headset/?srsltid=AfmBOoo48DcYVIYEgHBiFYUcUE8PGvH-r-ySQO1K6Mw6hXiOkfr09A0G]
• 32. NIRS Caps & Probes | NIRScaps fNIRS data ... [https://nirx.net/nirscap]
• 33. MedelOpt+ fNIRS Mobility Systems [https://www.biopac.com/product/medelopt-m/]
• 34. Semi-automatic 10/20 Identification Method for MRI-Free ... [https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2017.00004/full]
• 35. A Practical Guide for Combined fNIRS & EEG Recordings [https://www.gtec.at/2021/04/12/combined-fnirs-eeg-recordings/?srsltid=AfmBOoonvHYWGK1jLnm4uFi2AC_L-EATVXJZ7Ef5WLIv2QsewofSN1BV]
• 36. Spatial co-registration of functional near-infrared spectroscopy ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC5581202/]
• 37. fNIRS Optodes' Location Decider (fOLD): a toolbox for ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC5820343/]
• 38. Guiding functional near-infrared spectroscopy optode- ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC8211086/]
• 39. fNIRS vs EEG: Comparing Non-Invasive Brain Imaging ... [https://imotions.com/blog/learning/research-fundamentals/fnirs-vs-eeg/?srsltid=AfmBOops-nyRw8jYSDHiOj5_fYcOLCtliQbCKSzLkQIebwBVs1Q408Wu]
• 40. Automated methodology for optimal selection of minimum ... [https://www.nature.com/articles/s41598-022-15252-0]
• 41. EEG electrode setup optimization using feature extraction ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11825521/]
• 42. A Practical Guide for Combined fNIRS & EEG Recordings [https://www.gtec.at/2021/04/12/combined-fnirs-eeg-recordings/?srsltid=AfmBOopea04Z8o5onXdDJtef0FhtOmSAS960hbfx7ZfMdIRnWdlhMGw7]
• 43. Best practices for fNIRS publications - PMC - PubMed Central [https://pmc.ncbi.nlm.nih.gov/articles/PMC7793571/]
• 44. Advancing brain-computer interfaces through fNIRS and ... [https://www.frontiersin.org/research-topics/71646/advancing-brain-computer-interfaces-through-fnirs-and-eeg-integration]
• 45. What should I know about fNIRS imaging? [https://www.biopac.com/knowledge-base/fnir-faq/]
• 46. Applications and advances of combined fMRI-fNIRs ... [https://www.frontiersin.org/journals/neurology/articles/10.3389/fneur.2025.1542075/full]
• 47. EEG–fNIRS signal integration for motor imagery ... [https://www.sciencedirect.com/science/article/pii/S2772528625000299]
• 48. A Multimodal EEG-fNIRS Representation-learning Model ... [https://pubmed.ncbi.nlm.nih.gov/41223650/]
• 49. EEG-fNIRS Cookbook [https://www.brainproducts.com/support-resources/eeg-fnirs-cookbook/]
• 50. A multi-modal dataset of electroencephalography and ... [https://www.nature.com/articles/s41597-025-05286-0]
• 51. Evaluating the effects of multimodal EEG-fNIRS ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12425317/]
• 52. Decoding semantic representations from functional near- ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC5568915/]
• 53. Multichannel wearable fNIRS‐EEG system for long‐term ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC6866376/]
• 54. Quantifying the Impact of Hair and Skin Characteristics on ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11565806/]
• 55. The effects of protocol factors and participant ... [https://www.sciencedirect.com/science/article/pii/S2666956025000443]
• 56. Functional Near-Infrared Spectroscopy and Its Clinical ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC7364176/]
• 57. Quantitative comparison of correction techniques for ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC7511246/]
• 58. A Practical Guide for Combined fNIRS & EEG Recordings [https://www.gtec.at/2021/04/12/combined-fnirs-eeg-recordings/?srsltid=AfmBOorMa0EvqnT6ev2fhhY5ILUSJ7pm_3Jl2YEPFh7X4mDpB8JVVxSE]
• 59. Removal of Artifacts from EEG Signals: A Review - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC6427454/]
• 60. Motion Artifacts Correction from Single-Channel EEG and ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC9102309/]
• 61. A Practical Guide for Combined fNIRS & EEG Recordings [https://www.gtec.at/2021/04/12/combined-fnirs-eeg-recordings/?srsltid=AfmBOopuMZS_VL8xM4XgP_tXddoUAZqBGKp0FduSk-R6maGAVTUFOJGZ]
• 62. Adaptive filtering of physiological noises in fNIRS data [https://biomedical-engineering-online.biomedcentral.com/articles/10.1186/s12938-018-0613-2]
• 63. Improved physiological noise regression in fNIRS [https://pmc.ncbi.nlm.nih.gov/articles/PMC7703677/]
• 64. fNIRS vs EEG: Comparing Non-Invasive Brain Imaging ... [https://imotions.com/blog/learning/research-fundamentals/fnirs-vs-eeg/?srsltid=AfmBOop436vAv5zl1uuiqntSSnb6Xxk_Yp1IKNYfEI5AbfqHCR027RIN]
• 65. Multimodal fNIRS-EEG sensor fusion: Review of data- ... [https://pubmed.ncbi.nlm.nih.gov/41211102/]
• 66. Multimodal MBC-ATT: cross-modality attentional fusion of ... [https://www.frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2025.1660532/full]
• 67. Tutorials/NIRS_Optimal_montage - Brainstorm [https://neuroimage.usc.edu/brainstorm/Tutorials/NIRS_Optimal_montage]
• 68. Which experimental designs to use in fNIRS — Block design [https://www.artinis.com/blogpost-all/fnirs-block-design]
• 69. Current methods and limitations for longitudinal fMRI analysis ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC5963980/]
• 70. Simultaneous acquisition of EEG and NIRS during ... [https://www.nature.com/articles/sdata20183]
• 71. An EEG-fNIRS Hybridization Technique in the Four-Class ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC7376762/]
• 72. Decoding visual colour from scalp electroencephalography ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC8285579/]
• 73. Simultaneous EEG-fNIRS study of visual cognitive processing [https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0325017]
• 74. Systematic evaluation of fMRI data-processing pipelines for ... [https://www.nature.com/articles/s41467-024-48781-5]
• 75. fNIRS–EEG data preprocessing [https://bio-protocol.org/exchange/minidetail?id=18156217&type=30]
• 76. a meta-analysis of resting-state fMRI [https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2025.1576932/full]
• 77. Improvement of classification accuracy of functional near- ... [https://www.sciencedirect.com/science/article/pii/S2667102625000245]
• 78. Enhancing Classification Performance of fNIRS-BCI by ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC7730208/]
• 79. Combining robust level extraction and unsupervised adaptive... [https://pmc.ncbi.nlm.nih.gov/articles/PMC9589108/]
• 80. Improved performance of fNIRS-BCI by stacking of deep ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12005509/]
• 81. Improved performance of fNIRS-BCI by stacking of deep ... [https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0314447]
• 82. Multimodal fusion of EEG-fNIRS: a mutual information-based ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC7984774/]
• 83. Improved classification performance of EEG-fNIRS ... [https://www.frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2022.973959/full]
• 84. Global regional and national burden of intracerebral ... [https://pubmed.ncbi.nlm.nih.gov/39880918/]
• 85. A Novel Transfer Learning-Based Hybrid EEG-fNIRS Brain ... [https://pubmed.ncbi.nlm.nih.gov/40831229/]
• 86. A hybrid EMG–EEG interface for robust intention detection ... [https://www.nature.com/articles/s41598-025-24831-w]
• 87. Multi-Modal versus Uni-Modal Treatment for the Recovery of ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10815740/]
• 88. A Practical Guide for Combined fNIRS & EEG Recordings [https://www.gtec.at/2021/04/12/combined-fnirs-eeg-recordings/?srsltid=AfmBOorthAiPSBrp0r9RcGfK2Pyc6cFAfvqfEl5ZDW5ZNNcpBm4IXy96]
• 89. Evaluating the effects of multimodal EEG-fNIRS ... [https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0331177]
• 90. The role of fNIRS in assessing motor function: A bibliometric ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12338178/]
• 91. A multi-modal dataset of electroencephalography and ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12144169/]
• 92. Reproducibility of fNIRS within Subject for Visual and Motor ... [https://www.sciencedirect.com/science/article/pii/S1053811925004951]