Introduction
Electroencephalography (EEG) is a widely used technique for monitoring brain electrical activity. Advances in high-density EEG systems and advanced statistical methods have significantly enhanced its spatial resolution.1,2 Evoked potentials obtained from EEG are crucial for evaluating the brain’s response to sensory stimuli, offering key insights into neurological function. Consequently, EEG plays a pivotal role in diagnosing and monitoring a broad range of neurological disorders.3,4
The functional near-infrared spectroscopy (fNIRS) assesses changes in oxyhemoglobin and deoxyhemoglobin concentrations to infer neuronal activity.5–8 This technique provides valuable insights into functional brain connectivity and activation patterns during cognitive tasks. Its ability to capture cognitive, visual, and auditory signals expands the applicability of brain-computer interfaces (BCIs), particularly benefiting stroke patients through personalized therapeutic interventions.9 While fNIRS has gained attention for its capacity to measure hemodynamic responses, similar to functional magnetic resonance imaging, it is limited by relatively low spatial and depth resolution.10,11
EEG is characterized by high temporal resolution but limited spatial resolution, whereas fNIRS offers better spatial resolution with lower temporal precision. To address these individual limitations, the integration of EEG and fNIRS, often termed photoelectric synchronization, has been applied in research and clinical practices due to their complementary strengths. By combining electrical and hemodynamic data, this multimodal approach offers a more comprehensive understanding of brain function.12,13 Studies have demonstrated the unique value of EEG-fNIRS integration in exploring functional connectivity during resting and sleep states, investigating the neural mechanisms of emotional interaction, and analyzing complex neural activity patterns during motor execution, observation, and imagery.14,15 This synergistic approach enables high-resolution analysis of cortical activation patterns, facilitating the detection of subtle brain activity and connectivity features. It has significantly advanced our understanding of the neurophysiological mechanisms underlying motor and neurological disorders, providing robust support for diagnosis and therapeutic interventions. For instance, EEG-fNIRS integration has proven effective in mechanistic studies of mitochondrial epilepsy.16
The application of photoelectric synchronization technology extends further—to detecting brain activity in infants and children, assessing residual consciousness in patients with brain injuries, and advancing emotion recognition models in human-computer interaction. These applications highlight the potential of photoelectric synchronization as a frontier in neuroscience and clinical research. Despite its growing relevance, comprehensive evaluations of the progress and future directions in this field remain limited. Bibliometric analysis, a quantitative method for evaluating scientific literature, attempts to identify patterns, trends, and relationships within a body of research. This approach provides valuable insights into the evolution and structure of a particular scientific domain. This bibliometric analysis aimed to comprehensively examine the EEG and fNIRS research literature from 2005 to 2024. By evaluating integrated EEG-fNIRS studies, this work sought to identify emerging research trends, highlight pivotal findings, delineate clinical applications, and provide insights to inform and guide future research directions.
Results
General analysis
Based on the established search parameters, a total of 645 publications related to EEG-fNIRS were retrieved from the WoSCC for the period 2005–2024. These works originated from 55 countries or regions across Asia, Europe, and the Americas, involving 307 institutions, 481 authors, and 90 research categories, and were published in 213 journals. Regarding publication numbers, a clear upward trend in global research output is evident, with the annual number of publications increasing from a single paper (0.31%) in 2005 to 97 papers (15.03%) in 2023 (Fig. 2a). Concerning publication and citation trends in EEG and fNIRS literature, the overall pattern shows continuous growth from 2004 to 2023. Notably, the number of publications has increased significantly since 2015, rising from about 100 in 2015 to 228 in 2023, indicating that research activity in this field has intensified annually in recent years. The cumulative number of citations grew exponentially, reaching 23,866 in 2023 (Fig. 2b). This trend signifies growing scholarly interest in EEG-fNIRS research over recent years.
Analysis of publications, citations, and collaborations by countries or regions
We conducted a detailed analysis of contributions from 55 countries or regions engaged in photoelectric synchronization research. Figure 3a depicts a collaboration map, where the size of each circle represents the number of publications from each country or region. The top ten countries by publication count are: the United States in first place with 186 papers (28.84% of the total), followed by China with 139 papers (21.56%), and Germany with 80 papers (12.40%). Other major contributors include Korea (10.39%), Italy (9.61%), the United Kingdom (9.15%), Canada (8.06%), Japan (5.74%), France (4.03%), and Australia (2.95%). Countries ranked 10th and below have published fewer than 20 papers (Table 1). Of particular note is the consistent growth in research output from the United States between 2015 and 2023 (excluding 2018), and the exponential growth from China between 2016 and 2023. In contrast, both Germany and South Korea have experienced stable but declining trends in annual publication numbers since 2017 (Fig. 3b). We analyzed co-authorship among 27 countries that have each published at least five papers in this field. In the collaboration network, node size indicates the number of publications per country, while line thickness reflects the level of cooperation between countries. The top five countries with the strongest overall collaborative strength are the United States (142), Germany (74), China (69), the United Kingdom (50), and Italy (36) (Fig. 3c). In terms of citation count, the United States leads with 6,519 citations, followed by Italy (3,509), Germany (2,930), South Korea (2,712), China (2,115), Japan (2,108), the United Kingdom (1,389), and Canada (1,003) (Table 1). Overall, the relatively low publication output from individual countries and the limited international collaboration suggests significant potential to strengthen partnerships in this field. Enhancing such collaborations could greatly contribute to the development and advancement of EEG-fNIRS research across different nations.
Table 1Top 10 countries or regions in terms of the number of articles
Rank | Countries/Regions | Count (n) | Percentage (%) | Centrality | Year | Total number of citations | Average number of citations | Total link strength |
---|
1 | USA | 186 | 28.84 | 0.32 | 2005 | 6,519 | 35.05 | 142 |
2 | China | 139 | 21.55 | 0.19 | 2010 | 2,115 | 15.22 | 69 |
3 | Germany | 80 | 12.40 | 0.06 | 2008 | 2,930 | 36.63 | 74 |
4 | South Korea | 67 | 10.39 | 0.08 | 2010 | 2,712 | 40.48 | 35 |
5 | Italy | 62 | 9.61 | 0.07 | 2009 | 3,509 | 56.60 | 36 |
6 | England | 59 | 9.15 | 0.46 | 2008 | 1,389 | 23.54 | 50 |
7 | Canada | 52 | 8.06 | 0.11 | 2009 | 1,003 | 19.29 | 35 |
8 | Japan | 37 | 5.74 | 0.14 | 2005 | 2,108 | 56.97 | 29 |
9 | France | 26 | 4.03 | 0.08 | 2015 | 893 | 34.35 | 35 |
10 | Australia | 19 | 2.95 | 0.02 | 2016 | 369 | 19.42 | 19 |
Analysis of institutional output and collaboration
The analysis of institutional collaborations reveals that a total of 307 institutions have contributed to EEG and fNIRS research. The top 10 institutions by publication count and related metrics were analyzed in detail (Table 2). Université de Montréal ranked first with 27 articles (4.19%), followed by the Catholic University of the Sacred Heart (3.26%), Pusan National University (3.10%), Beijing Normal University (2.95%), and the University of London (2.79%). The collaboration network diagram illustrates relationships among these 307 institutions. Each node represents an institution, with node size proportional to the number of publications. Larger nodes indicate higher research output, while node color reflects publication year. Université de Montréal occupies a central position in the network, with the highest number of connections, underscoring its role as a key hub for collaboration. Other high-output institutions, such as the Catholic University of the Sacred Heart, Pusan National University, and Beijing Normal University, also exhibit significant collaborative ties (Fig. 4a). For further analysis, 62 institutions with five or more publications were selected for co-authorship network analysis. The top five institutions by total link strength were Université de Montréal (22), University of Houston (15), McGill University (15), Drexel University (14), and University of Leipzig (14). The diagram demonstrates active international collaboration, particularly among leading universities in Europe, North America, and Asia. For example, Université de Montréal has established close collaborative links with McGill University, University of Houston, and University of London. Notably, these collaborations span multiple regions, highlighting the globalized nature of EEG and fNIRS research (Fig. 4b).
Table 2The top 10 institutions contributing to publications in EEG and fNIRS
Rank | Institutions | Count (n) | Percentage (%) | Centrality | Year |
---|
1 | Universite de Montreal | 27 | 4.19 | 0.1 | 2011 |
2 | Catholic University of the Sacred Heart | 21 | 3.26 | 0 | 2015 |
3 | Pusan National University | 20 | 3.10 | 0.1 | 2010 |
4 | Beijing Normal University | 19 | 2.95 | 0.1 | 2010 |
5 | University of London | 18 | 2.79 | 0.1 | 2008 |
6 | University of Houston | 17 | 2.64 | 0.08 | 2013 |
7 | Eberhard Karls University of Tubingen | 29 | 4.50 | 0.03 | 2015 |
8 | Harvard University | 16 | 2.48 | 0.14 | 2011 |
9 | Drexel University | 14 | 2.17 | 0.11 | 2005 |
10 | Polytechnique Montreal | 12 | 1.86 | 0 | 2011 |
Analysis of author influence and collaboration
Table 3 highlights the top 16 authors and their respective H-indices, ranked by the number of articles they have contributed. Among them, Michela Balconi emerges as the most prolific researcher, with 21 articles, accounting for 3.26% of the total publications. Following closely are Hong Keum-Shik (2.33%), Pouliot Philippe (1.55%), Zhang Yingchun (1.40%), and Rossi Sonja (1.40%), as illustrated in Figure 5a. We analyzed the collaboration network of 50 authors who have co-authored more than five papers (Fig. 5b). The strength of collaboration among authors is indicated by the thickness of the connections between nodes. The authors with the highest total connection strength are Pouliot Philippe (54), Lesage Frederic (50), Vannasing Phetsamone (46), Tremblay Julie (45), and Dang Khoa Nguyen (40).
Table 3Top 16 authors with the most articles about EEG and fNIRS
Rank | Authors | Count | H-index | Institution |
---|
1 | Balconi, Michela | 21 | 38 | Catholic University of the Sacred Heart |
2 | Hong, Keum-Shik | 15 | 54 | Pusan National University |
3 | Pouliot, Philippe | 10 | 25 | Universite de Montreal |
4 | Zhang, Yingchun | 9 | 15 | Guangdong Prov Work Injury Rehabil Hosp |
5 | Rossi, Sonja | 9 | 13 | Medical University of Innsbruck |
6 | Vanutelli, Maria Elide | 9 | 17 | University of Milan |
7 | Lesage, Frederic | 8 | 45 | Polytechnique Montreal |
8 | Li, Rihui | 7 | 24 | University of Macau |
9 | Ayaz, Hasan | 7 | 37 | University of Pennsylvania |
10 | Lassonde, Maryse | 6 | 53 | Universite de Montreal |
11 | Wallois, Fabrice | 6 | 30 | Institut National de la Sante et de la Recherche Medicale |
12 | Dang Khoa Nguyen | 6 | 20 | Universite de Montreal |
13 | Vannasing, Phetsamone | 6 | 20 | Sainte Justine Univ Hosp Ctr |
14 | Tremblay, Julie | 6 | 17 | Sainte Justine Univ Hosp Ctr |
15 | Al-Shargie, Fares | 6 | 13 | Abu Dhabi University |
16 | Naseer, Noman | 6 | 21 | Air University Islamabad |
An author’s influence in a scientific field depends more on the number of citations. Coupled network analysis shows that the most cited author in this field is Hong Keum-Shik (1,615 citations), followed by Dan Ippeita (715 citations), Tsuzuki Daisuke (671 citations), Ayaz Hasan (627 citations), and Khan M. Jawad (579 citations) (Fig. 5c). A co-citation network was constructed to analyze intellectual relationships among highly cited authors (Fig. 5d). Six authors exceeded 100 co-citations: ANONYMOUS (149), SCHOLKMANN F (134), HUPPERT T (126), JCUI X (122), FERRARI M (113), and STRANGMAN G (102). Node size corresponds to citation count, while the thickness of connecting lines indicate co-citation frequency, mapping conceptual linkages within the field. Node color represents publication year, with warmer hues denoting more recent work.
Analysis of journals and categories
A total of 645 articles were published across 213 journals. The top 10 journals with the most EEG and fNIRS articles and the top 10 most cited journals are listed in Table 4. Frontiers in Human Neuroscience accounted for the highest number of publications (48 articles, 7.44% of all articles), followed by Frontiers in Neuroscience (5.11%), Scientific Reports (4.03%), Neuroimage (3.88%), and IEEE Access (3.88%). In the co-citation analysis, 394 journals were identified as being co-cited in more than 50 publications. Neuroimage had the largest number of citations (595), followed by Front Hum Neurosci (435), PLOS One (394), Hum Brain Mapp (334), and Clin Neurophysiol (307).
Table 4Top 10 popular journals and cited journals
Rank | Popular journals | Records (n) | 2024 impact factor | 2024 JCR Partition | Cited journals | Citations | 2024 impact factor | 2024 JCR Partition |
---|
1 | Frontiers In Human Neuroscience | 48 | 2.9 | Q3 | Neuroimage | 595 | 5.7 | Q1 |
2 | Frontiers In Neuroscience | 33 | 4.3 | Q2 | Front Hum Neurosci | 435 | 2.9 | Q3 |
3 | Scientific Reports | 26 | 4.6 | Q2 | Plos One | 394 | 3.7 | Q2 |
4 | Neuroimage | 25 | 5.7 | Q1 | Hum Brain Mapp | 334 | 4.8 | Q1 |
5 | Ieee Access | 25 | 3.9 | Q2 | Clin Neurophysiol | 307 | 4.7 | Q1 |
6 | Sensors | 23 | 3.9 | Q2 | Scientific Reports | 285 | 4.6 | Q2 |
7 | Neurophotonics | 21 | 5.3 | Q1 | Front Neurosci-Switz | 274 | 4.3 | Q2 |
8 | Brain Sciences | 20 | 3.3 | Q3 | P Natl Acad Sci Usa | 273 | 11.1 | Q1 |
9 | Ieee Transactions On Neural Systems | 18 | 4.9 | Q2 | J Neurosci Meth | 254 | 3 | Q3 |
10 | Plos One | 18 | 3.7 | Q2 | J Neurosci | 248 | 5.3 | Q1 |
A total of 90 research areas are represented in the co-occurrence analysis of Web of Science categories. The most well-represented research area is Neurosciences, comprising 316 records (48.992% of all articles), followed by Biomedical Engineering (74 records, 11.473%), Electrical and Electronic Engineering (70, 10.853%), Psychology (62, 9.612%), and Radiology, Nuclear Medicine, and Medical Imaging (60, 9.302%) (Table 5). This indicates that the EEG-fNIRS dual-modality technique is a multifaceted and multidisciplinary field. Analyzing journals and research areas provides insight into the publications in the field of EEG and fNIRS research, highlighting important journals and major research domains. Researchers can use this information to explore existing literature and identify potential collaboration opportunities.
Table 5Top 10 well-represented research areas
Rank | Research areas | Record (n) | % (of 654) |
---|
1 | Neurosciences | 316 | 48.992 |
2 | Engineering biomedical | 74 | 11.473 |
3 | Engineering electrical electronic | 70 | 10.853 |
4 | Psychology | 62 | 9.612 |
5 | Radiology nuclear medicine medical imaging | 60 | 9.302 |
6 | Multidisciplinary sciences | 51 | 7.907 |
7 | Optics | 49 | 7.597 |
8 | Neuroimaging | 43 | 6.667 |
9 | Computer science information systems | 38 | 5.891 |
10 | Instruments instrumentation | 35 | 5.426 |
Analysis of references and bursts in citations
Based on the citation analysis, 106 documents were cited more than 20 times. The collaboration network of these 106 documents is revealed (Fig. 6a). The top ten documents with the highest citation counts are listed in Table 6. Hong et al.24 conducted a review of a framework for brain therapy and BCI for individuals with locked-in syndrome, utilizing a hybrid multimodality approach combining EEG and fNIRS. The integration of multiple modalities in brain imaging and prosthesis control represents a novel advancement in the field.25 Furthermore, the authors introduced a hybrid EEG-fNIRS scheme aimed at decoding eight distinct brain commands from the frontal brain region for BCI applications.26
Table 6Top 10 citation analysis of documents on EEG and fNIRS research
Rank | Title | First author | Corresponding Author | Source | Publication year | Citations (n) |
---|
1 | Feature extraction and classification methods for hybrid fNIRS-EEG brain-computer interfaces | Hong, KS | Hong, MJ | Frontiers in Human Neuroscience | 2018 | 31 |
2 | Hybrid brain-computer interface techniques for improved classification accuracy and increased number of commands: a review | Hong, KS | Khan, MJ | Frontiers in Neurorobotics | 2017 | 29 |
3 | Hybrid EEG-fNIRS-Based Eight-Command Decoding for BCI: Application to Quadcopter Control | Khan, MJ | Hong, KS | Frontiers in Neurorobotics | 2017 | 21 |
4 | Early detection of hemodynamic responses using EEG: a hybrid EEG-fNIRS study | Khan, MJ | Hong, KS | Frontiers in Human Neuroscience | 2018 | 18 |
5 | A brain-computer interface based on a few-channel EEG-fNIRS bimodal system | Ge, S | Wang, H | IEEE Access | 2017 | 17 |
6 | Hybrid EEG-fNIRS bei fusion using multi-resolution singular value decomposition (MSVD) | Khan, MU | Hasan, MAH | Frontiers in Human Neuroscience | 2020 | 17 |
7 | Enhancing classification performance of fNIRS-BCI by identifying cortically active channels using the z-score method | Nazeer, H | Ayaz, Y | Sensors | 2020 | 16 |
8 | Improvement of information transfer rates using a hybrid EEG-NIRS brain-computer interface with a short trial length: offline and pseudo-online analyses | Shin, J | Hwang, HJ | Sensors | 2018 | 15 |
9 | Enhancing the performance of a hybrid EEG-fNIRS system using channel selection and early temporal features | Li, R | Zhang, Y | Frontiers in Human Neuroscience | 2017 | 15 |
10 | Shining a light on awareness: a review of functional near-infrared spectroscopy for prolonged disorders of consciousness | Rupawala, M | Cruse, D | Frontiers in Neurology | 2018 | 15 |
Using CiteSpace’s log-likelihood ratio algorithm, the 10 largest clusters were identified by consensus (Fig. 6b). Based on cluster size (number of documents), the 10 largest clusters are labeled: “information transfer rate”, “machine learning”, “using functional near-infrared spectroscopy”, “human epilepsy”, “arithmetic task”, “Stroop task”, “social interaction”, “I social neuroscience”, “binge-eating disorder”, and “walking intervention”. The top-ranked item by citation bursts is Pinti P in Cluster #5.27 The second is Naseer Noma in Cluster #0,28 followed by Khan MJ in Cluster #0,26 Hong KS in Cluster #4,24 and Buccino AP in Cluster #0 (Fig. 6c).29
We utilized VOSviewer to comprehensively analyze all cited references. The co-citation network shown in Figure 6d includes publications with co-citation frequencies exceeding 50. The thickness of connecting lines reflects the strength of co-citation links between references. Among the most frequently cited works, Fazli et al.,30 published in NeuroImage, ranks highest, with a link strength of 972 and a citation frequency of 92. Following closely is the paper by Scholkmann et al,31 also in NeuroImage, with a link strength of 861 and citation frequency of 93. Notably, Ferrari et al.,32 published in NeuroImage, made a significant impact, achieving the highest citation frequency of 103 and a link strength of 806.
Figure 6e illustrates the top 21 most cited references. The dark blue line indicates the citation duration from 2005 to 2024, while the red line indicates the range of mutations in citation duration. The reference with the highest number of citations and citation burst value is the article entitled fNIRS-based BCIs by Noman Naseer, with a citation burst from 2015 to 2020 of 13.69.28 The second most cited reference is by Fazli et al.,30 entitled Enhanced performance by a hybrid EEG-fNIRS-BCI, with a citation burst of 13.26.
Analysis of keywords and topics
Figure 7a demonstrates the top 11 keywords with the strongest citation bursts. In total, 538 keywords were collected (Fig. 7b). The five most frequently occurring keywords were: NIRS (239 occurrences), fNIRS (229), EEG (197), functional magnetic resonance imaging (94), and brain (85). A total of 118 author keywords appearing more than five times were analyzed (Fig. 8a). Figure 8b maps these keywords, with colors indicating average publication years; most keywords were published after 2019, indicated by greener colors. Following clustering of the cited networks, 18 clusters were obtained and labeled with title words or keywords from citing articles (Fig. 8c). Density visualization in Figure 8d shows the same keywords mapped by frequency.
Using CiteSpace’s log-likelihood ratio algorithm, the 10 largest clusters were identified by consensus (Fig. 8e). Along the timeline, nodes of different colors indicate references from different years within clusters, while nodes farther to the right indicate more recent references. The modularity Q value (0.9448) was greater than 0.3, and the average silhouette score (0.9605) was greater than 0.7, indicating that the clusters are convincing and well-structured. The five largest clusters were labeled as “motor imagery”, “brain mapping”, “optical topography”, “neurovascular coupling”, and “medical signal processing”. Examining the timing of keyword appearance allows us to understand the development of these fields over time and estimate recent trends and directions. Figure 8f shows the high-frequency subject word map from 2005 to 2024. The research focus has gradually shifted from disease diagnosis toward treatment research. Attention should be paid to recent hotspots such as “task analysis”, “feature extraction”, “deep learning”, and “coherence”.
Discussion
This study conducted a bibliometric analysis of literature extracted from public databases to identify research hotspots and future development directions in EEG and fNIRS. Notably, a significant rise in publication activity has been observed since 2020, likely driven by advances in technologies such as BCI and artificial intelligence, which have positioned EEG-fNIRS as a compelling research topic. Despite growing attention from the scientific community, substantial potential remains for further expansion in this field, particularly in the context of photoelectric synchronization. The observed trends suggest increasing interest, but the current scale of research indicates considerable room for development and broader exploration.
Additionally, we analyzed the authorship of articles published in the WoSCC and found that the USA dominates the application of EEG and fNIRS. Interdisciplinary and international collaborations play a crucial role in advancing synchronous research in photonics. Université de Montréal is the most active institution, focusing its research on applying EEG-fNIRS to stroke and epilepsy monitoring.33 In particular, it has played an important role in understanding the pathophysiology of temporal lobe epilepsy, revealing complex local and distal oxygenation changes during temporal lobe seizures and the non-linear hemodynamic response in refractory focal epilepsy in humans.34,35 The University of Montréal exhibits the highest intermediary centrality and connection strength, boasting the strongest connections to other institutions and the highest number of collaborative publications. Following closely are the University of Eberhard Karls in Tübingen, Germany, and Drexel University in the USA, ranked second and third in the collaboration analysis, respectively. Pusan National University, meanwhile, achieved the highest average citation count in this study.
Michela Balconi is the most active researcher in this field. As evidenced by 21 co-authored publications from 2015 to 2024, her research primarily focuses on detecting frontal cortex responses to emotions via EEG and fNIRS under different environmental conditions,36 such as resting state,37,38 COVID-19,39 painful stimuli,40 audiovisual stimuli,14 and interoceptive attentiveness.41,42 Additionally, Balconi et al.43 used EEG and fNIRS to reveal a facilitatory role for motor imagery and executive motion sensation, suggesting that readiness potential amplitude acts as a predictor of hemodynamic brain activity, modulated by task and gesture type factors. Recent research has focused on EEG-fNIRS multimodal hyperscanning techniques in social interactions, including cooperative and competitive relationships,44–47 as well as interpersonal relationships,48 highlighting brain network connectivity in neuroscience as a current hotspot and cutting-edge direction for the field.49 Keum-Shik Hong of Pusan National University, together with researchers from the University of Wisconsin-Madison and Sejong University, has made significant contributions to optoelectronic synchronization research from an EEG-fNIRS-BCI perspective.25,50 Their work involves detecting brain signals during various tasks, assessing activation of specific brain regions, and achieving precise classification of brain signals through multimodal techniques. Research in this area has advanced the development of wearable BCIs.28,51,52 Philippe Pouliot has also significantly contributed to advancing optoelectronic synchronization research from an EEG-fNIRS-BCI perspective. As a Professor of Neuroscience at the University of Montréal, his primary focus has been on enhancing algorithms for NIRS technology to improve visualization of functional brain connectivity.53,54
Clinical applications of EEG-fNIRS
According to keyword and theme analyses, Alzheimer’s disease, epilepsy,55 attention-deficit hyperactivity disorder, stroke,56 consciousness disorders, and Parkinson’s disease are among the most prevalent clinical applications of EEG-fNIRS.57 The technique was initially used to study cognitive processes such as attention, working memory, target categorization, problem-solving, and human waste processing.58 In recent years, it has been extensively applied in cognitive neurosciences,59 neurological disorders,60 psychiatric disorders,61 mental health, pediatrics,62 sociology,63 and psychology. This section focuses on the use of EEG-fNIRS techniques in clinical disorders.
EEG dynamically monitors brain activity through electrical signals with high temporal resolution, while fNIRS detects cerebral blood oxygenation changes using near-infrared light, offering high spatial resolution. These complementary, non-interfering techniques provide a more comprehensive view of brain function. Jindal et al.64 used fNIRS-EEG integration to study changes in brain activity during stroke recovery. They found that changes in rSO2 were negatively correlated with EEG power changes following electrical stimulation, indicating that decreased EEG power corresponded to increased corticospinal excitability. This study expands the application of fNIRS-EEG in assessing brain activity patterns in stroke patients and provides a quantitative method for evaluating therapeutic interventions.64 Similarly, Li et al.13 developed a multimodal neuroimaging technique combining EEG and fNIRS to assess motor deficits post-stroke. In their study involving 18 stroke patients and nine healthy controls, they demonstrated reduced task-induced activity in the somatosensory cortex of stroke patients and highlighted how improved functional connectivity in motor-related areas correlated with motor recovery during rehabilitation. Baseline connectivity patterns also predicted recovery outcomes, emphasizing the potential of fNIRS-EEG integration in monitoring and predicting rehabilitation success.13 Overall, fNIRS-EEG integration leverages the advantages of high temporal and spatial resolution, offering new insights into neurovascular coupling, brain network plasticity, and motor recovery in stroke rehabilitation. This multimodal approach not only enhances understanding of brain activity mechanisms but also promotes precise clinical applications and rehabilitation strategies.
Li and colleagues integrated EEG and fNIRS data with machine learning techniques to enhance the precision of depression diagnosis.65 They collected resting-state EEG signals and hemodynamic data from the forehead of 25 patients diagnosed with depression and 30 healthy individuals. By leveraging a support vector machine model, they achieved a classification accuracy of 92.7% using hybrid EEG-fNIRS features, compared to 81.8% with EEG data alone. Key biomarkers distinguishing depressive states included enhanced local efficiency in the delta band, hemispheric asymmetry in the theta band, and brain oxygen entropy.65 These findings underscore the potential of integrating EEG and fNIRS with machine learning as a reliable, non-invasive tool for individual-level depression diagnosis, offering both spatial and temporal insights into underlying neural mechanisms.
The integration of EEG and fNIRS has also shown increasing relevance in epilepsy research. Nourhashemi et al.66 employed this multimodal approach to investigate 25 episodes of childhood absence seizures in eight pediatric patients. Approximately 20 seconds before the onset of spike-wave discharges, transient shifts in direct current potentials were detected, correlating with preictal hemodynamic changes in cerebral blood flow and hemoglobin levels. These findings revealed intricate neurovascular interactions preceding seizures, providing a deeper understanding of mechanisms underlying absence epilepsy.66 Peng et al.67 further explored the role of EEG-fNIRS integration in epilepsy, focusing on both focal seizures and interictal epileptiform discharges. Their review highlighted the potential of these technologies to evaluate hemodynamic and neuronal changes associated with epileptic events, improving the localization and lateralization of seizure foci, particularly in focal epilepsy. Specific patterns of cerebral blood volume and oxygenation changes were associated with various seizure types, such as temporal and frontal lobe epilepsy. Notably, fNIRS could detect pre-seizure hemodynamic alterations, offering new insights into seizure initiation.67 Collectively, these studies demonstrate the potential of EEG-fNIRS integration as a robust tool for understanding the neurovascular mechanisms of epilepsy and advancing diagnostic and therapeutic strategies. Future research should prioritize expanding the clinical applicability of this multimodal approach.
Strengths and weaknesses of the study
This research conducted a comprehensive search and analysis of literature encompassing various terms related to EEG and fNIRS. The analysis utilized two distinct econometric software tools, potentially introducing bias due to variations in calculation methods between the software. Given that scalp EEG predominantly captures cortical signals perpendicular to the recording electrodes, while fNIRS detects hemodynamic changes through oxy/deoxyhemoglobin concentration variations to indirectly reflect neural activity, the clinical implementation of EEG-fNIRS integration remains largely confined to research settings. This approach currently offers limited practical utility for clinicians engaged in the non-invasive diagnosis and treatment of brain disorders.
This study represents the first bibliometric analysis to systematically document and evaluate clinically treated EEG and fNIRS. An extensive literature review was conducted, though some limitations exist. Primarily, the search and analysis relied predominantly on the WoSCC core database, neglecting other databases such as PubMed and Scopus, potentially overlooking pertinent literature. Nevertheless, it is important to acknowledge that WoSCC is a widely utilized database in scientometrics, with most bibliometric software compatible with its format. Furthermore, incomplete keyword extraction may have impacted the outcomes of keyword analysis.
Future research directions
Challenges persist in optoelectronic synchronization, including limited spatial resolution and the inability to capture subcortical data. To address these limitations, potential strategies include incorporating supplementary sensors, integrating complementary modalities such as EEG, and augmenting fNIRS optoelectronic probes. Additionally, further optimization of signal processing and algorithms for EEG-fNIRS applications is needed.5,6,68 A significant additional limitation pertains to the temporal characteristics of the signals. EEG signals exhibit high temporal resolution, enabling analysis of both spectral and temporal dynamics. In contrast, fNIRS signals reflect slower hemodynamic changes, introducing an inherent physiological latency relative to the underlying neural activity.69,70 Therefore, achieving precise temporal synchronization (neurovascular coupling) between the fast electrical events measured by EEG and the slower hemodynamic responses measured by fNIRS presents a critical methodological challenge.
Wearable, integrated EEG-fNIRS technology represents a critical advancement in this field. Supporting large channel counts, extensive dynamic ranges, and rapid data acquisition and transmission requires enhanced data interfaces and control schemes. Furthermore, dry EEG electrodes with high gain and input impedance may be the optimal choice for extended periodic monitoring.12 The integration of EEG and fNIRS technologies faces inherent technical limitations. Although commercial devices capable of simultaneous EEG-fNIRS recording are available, achieving spatially co-localized measurements of neuronal activity remains challenging. This difficulty arises primarily from fundamental differences in signal acquisition mechanisms and the physical constraints of the hardware. EEG detects electrical potentials generated by neuronal populations, with relatively small electrodes achieving good contact with the scalp surface.71 In contrast, fNIRS measures hemodynamic responses by detecting changes in near-infrared light absorption between optical sources (emitters) and detectors (optodes). These optodes possess a finite size and require specific spacing to measure hemodynamic responses within underlying tissue. Consequently, to attempt spatial correspondence (co-channel configuration), EEG electrodes must be placed within the gaps between fNIRS optodes. This placement inevitably compromises spatial coverage and the fidelity of EEG signal acquisition due to physical obstruction and suboptimal electrode positioning. These spatial constraints also hinder comprehensive coverage of the entire brain region with truly co-localized EEG-fNIRS channels.72
BCI research frequently utilizes integrated EEG-fNIRS systems to enhance the amount of accessible brain information and improve classification accuracy.73,74 Non-invasive BCIs have demonstrated potential in stroke rehabilitation and are commonly applied in motor, sensory, and cognitive functions; however, limited spatial resolution remains a recognized constraint.75 Moreover, the utilization of non-invasive BCIs has been found to augment the quantity of brain-derived information.