Welcome to the Future of Brain-Computer Interfaces
Explore cutting-edge BCI research and join our growing neurotech community
📄 Interpretable Dual-Filter Fuzzy Neural Networks for Affective Brain-Computer Interfaces
Title: Interpretable Dual-Filter Fuzzy Neural Networks for Affective Brain-Computer Interfaces
Authors: Xiaowei Jiang, Yanan Chen, Nikhil Ranjan Pal, Yu-Cheng Chang, Yunkai Yang, Thomas Do, Chin-Teng Lin
Published on: arXiv, 2025
Affiliations: GrapheneX-UTS Human-centric AI Centre, Australian AI Institute, School of Computer Science, Faculty of Engineering and Information Technology, University of Technology Sydney; Institute of Psychology and Behavior, Henan University; Indian Statistical Institute, Kolkata, India
Summary: This paper introduces iFuzzyAffectDuo, an interpretable fuzzy neural network architecture with dual filters (spatial and temporal) designed for affective brain-computer interfaces (BCIs). The model employs a novel fuzzy membership function based on the Laplace distribution, which enhances the accuracy and interpretability of emotion recognition from neuroimaging data compared to traditional deep learning approaches. By refining emotion-related neural signal features, the dual-filter design offers a more explainable decision-making process in detecting human emotional states. The authors validate iFuzzyAffectDuo on three neuroimaging datasets (using functional Near-Infrared Spectroscopy fNIRS and Electroencephalography EEG), where it achieves state-of-the-art accuracy and demonstrates its potential to advance affective computing and improve human–computer interaction.
📄 An efficient 3D column-only P300 speller paradigm utilizing few numbers of electrodes and flashings for practical BCI implementation
Title: An efficient 3D column-only P300 speller paradigm utilizing few numbers of electrodes and flashings for practical BCI implementation
Authors: Onur Erdem Korkmaz; Onder Aydemir; Emin Argun Oral; Ibrahim Yucel Ozbek
Published on: PLOS ONE (2022)
Affiliations: Atatürk University (Erzurum, Turkey); Karadeniz Technical University (Trabzon, Turkey)
Summary: This paper proposes a new P300 speller paradigm that presents stimuli in a 3D column-only format to improve spelling accuracy with fewer EEG electrodes and flashes. The method uses 3D animation and column-wise flashing (instead of traditional row–column flashing) to evoke stronger P300 responses. In experiments with an ANN classifier, the 3D column paradigm yielded higher average character recognition accuracy than the classic 2D approach – for example, improving one-flash spelling accuracy from ~89.97% to 93.90% (a 4.36% increase). Notably, the performance gains were especially pronounced when using a single or very few EEG electrodes, demonstrating the paradigm’s practicality for simpler, faster brain–computer interfaces.
📄 ECA-FusionNet: a hybrid EEG-fNIRS signals network for MI classification
Title: ECA-FusionNet: a hybrid EEG-fNIRS signals network for MI classification
Authors: Yuxin Qin, Baojiang Li, Wenlong Wang, Xingbin Shi, Cheng Peng, Xichao Wang, Haiyan Wang
Published on: Journal of Neural Engineering, 2025
Affiliations: School of Electrical Engineering, Shanghai DianJi University, Shanghai, China; Intelligent Decision and Control Technology Institute, Shanghai DianJi University, Shanghai, China
Summary: The authors propose ECA-FusionNet, a novel hybrid deep learning network that fuses electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) signals for motor imagery (MI) classification. The model extracts complementary spatio-temporal features from both EEG and fNIRS and combines them at both the feature level and decision level to leverage the strengths of each modality. Evaluated on a public EEG-fNIRS MI dataset, ECA-FusionNet achieved higher classification accuracy than single-modality (EEG-only or fNIRS-only) approaches and outperformed existing fusion methods, demonstrating improved adaptability and robustness for MI-based brain–computer interface tasks.
📄 Master classes of the tenth international brain-computer interface meeting: showcasing the research of BCI trainees
Title: Master classes of the tenth international brain–computer interface meeting: showcasing the research of BCI trainees
Authors: Stephanie Cernera, Tan Gemicioglu, Julia Berezutskaya, Richard Csaky, Maxime Verwoert, Daniel Polyakov, et al. (38 authors total)
Published on: Journal of Neural Engineering (2025)
Affiliations: Multi-institution collaboration (including University of California San Francisco, University of Oxford, Maastricht University, University of Lyon, among others)
Summary: This article is a comprehensive report from the 10th International BCI Meeting, highlighting the “master class” sessions where leading mentors guided trainees in advanced BCI research. It features contributions from dozens of early-career researchers (“BCI trainees”) across diverse topics presented at the meeting. The paper summarizes emerging trends and projects – from novel signal processing techniques to new BCI applications – that were showcased by the trainees. Key insights include the value of multidisciplinary mentorship and hands-on workshops in accelerating BCI innovations, and the piece emphasizes how these master classes helped shape future BCI research directions by sharing practical expertise and fostering a collaborative community of new researchers.
📄 fNIRS-based Monitoring of Functional Brain Network during Image Guided Hand Movements
Title: fNIRS-based Monitoring of Functional Brain Network during Image Guided Hand Movements
Authors: Wenyao Zheng; Yun-Hsuan Chen; Jiachen Wang; Mohamad Sawan
Published on: IEEE Biomedical Circuits and Systems Conference (BioCAS 2024) – Proceedings (2024)
Affiliations: Westlake University, Hangzhou, China
Summary: This conference paper explores using functional near-infrared spectroscopy (fNIRS) to monitor brain network activity during visually guided hand movements. The authors developed an image-guided motor task in which participants perform specific hand movements prompted by visual cues, while fNIRS sensors measure cortical activation in relevant brain regions. The approach allowed mapping of the functional brain network engaged by the hand movements, revealing characteristic hemodynamic patterns associated with the guided motor actions. The preliminary results show that fNIRS can successfully track connectivity and activation changes in the motor cortex during these tasks, demonstrating the feasibility of fNIRS for monitoring complex motor behaviors. This work lays groundwork for using portable fNIRS in neurorehabilitation or brain–machine interfaces to assess and train motor function in real-world settings.
📄 A Personalized Multimodal BCI–Soft Robotics System for Rehabilitating Upper Limb Function in Chronic Stroke Patients
Title: A Personalized Multimodal BCI–Soft Robotics System for Rehabilitating Upper Limb Function in Chronic Stroke Patients
Authors: Brian Premchand; Zhuo Zhang; Kai Keng Ang; Juanhong Yu; Isaac O. Tan; Josephine P. W. Lam; Anna X. Y. Choo; Ananda Sidarta; Patrick W. H. Kwong; Lau Ha Chloe Chung
Published on: Biomimetics (MDPI) – Special Issue Advances in BCI (2025)
Affiliations: Institute for Infocomm Research (A*STAR) & Nanyang Technological University (Singapore); Hong Kong Polytechnic University; Tan Tock Seng Hospital (Singapore)
Summary: This study presents a personalized stroke rehabilitation system that combines a multimodal EEG–fNIRS brain–computer interface (BCI) with a soft robotic exoskeleton for upper-limb therapy. The system aligns each patient’s specific abilities with tailored training: it uses EEG and fNIRS signals to drive a soft robotic sleeve assisting the patient’s wrist and arm movements. To improve reliability, the authors synchronized motor imagery cues with the patient’s breathing cycle and incorporated respiration sensors to filter fNIRS noise. In a pilot trial with four chronic stroke patients over 6 weeks, the BCI-robotic training led to notable gains in arm function – Fugl-Meyer scores improved by ~10 points and active wrist range-of-motion increased 20–50°. Patients also learned to modulate their brain signals (with 42–78% success during fNIRS neurofeedback) and achieved clinically significant improvements in daily arm tasks. This work demonstrates that combining multimodal BCI feedback with assistive robotics can enhance motor recovery in severely impaired stroke survivors.
📄 Effects of Selection of the Learning Set Formation Strategy and Filtration Method on the Effectiveness of a BCI Based on Near Infrared Spectrometry
Title: Effects of Selection of the Learning Set Formation Strategy and Filtration Method on the Effectiveness of a BCI Based on Near Infrared Spectrometry
Authors: M. R. Isaev; P. D. Bobrov
Published on: Neuroscience and Behavioral Physiology 53(3): 373–380 (2023)
Affiliations: Institute of Higher Nervous Activity & Neurophysiology, RAS (Moscow, Russia); Pirogov Russian National Research Medical University (Moscow, Russia)
Summary: This paper focuses on improving a motor-imagery BCI based on near-infrared spectroscopy (NIRS) by optimizing data selection and filtering strategies. The authors propose a custom algorithm for NIRS-based BCI control that includes a specialized signal filtration method to remove low-frequency drifts and sequential classification of rest vs. motor imagery states. They also introduce additional training using previously recorded sessions to adapt the classifier. Using data from three prior NIRS-MI experiments, they evaluate how these choices affect classification accuracy. The results show that removing low-frequency noise from NIRS signals significantly boosts accuracy, and incorporating “transfer learning” from prior sessions further improves performance. Moreover, they found the system could maintain high accuracy even after reducing the number of NIRS channels, indicating an efficient configuration.
Overall, the study demonstrates that carefully chosen training sets and tailored filtering can substantially enhance the reliability of NIRS-based BCIs for motor rehabilitation.
📄 Investigation of deep convolutional neural network for classification of motor imagery fNIRS signals
Title: Investigation of deep convolutional neural network for classification of motor imagery fNIRS signals
Authors: A. Janani; M. Sasikala; H. Chhabra; N. Shajil; G. Venkatasubramanian
Published on: Biomedical Signal Processing and Control 62: 102133 (2021)
Affiliations: Anna University (Chennai, India); National Institute of Mental Health and Neurosciences (Bangalore, India)
Summary: This work explores using deep learning (CNNs) to classify motor-imagery tasks from fNIRS signals. The authors designed a simple yet effective convolutional neural network to distinguish among multiple motor imagery states using only fNIRS data. They also experimented with different ways of representing the 1D fNIRS time-series as input “images” for the CNN to optimize performance. In tests, their CNN approach achieved an average classification accuracy of around 72.3% (±4.4%) for a four-class motor imagery task, outperforming traditional machine learning methods (like SVM and MLP) applied to the same data. The study demonstrates that even a lightweight CNN can learn meaningful spatiotemporal features from fNIRS signals, offering a promising path for improving fNIRS-based BCIs for motor intent detection in rehabilitation settings.
📄 Multi-classification of fNIRS Signals in Four Body Parts Motor Imagery Tasks Measured from Motor Cortex
Title: Multi-classification of fNIRS Signals in Four Body Parts Motor Imagery Tasks Measured from Motor Cortex
Authors: Yuan Li; Hui Shen; Yang Yu; Dewen Hu
Published on: Proceedings of the 4th Int. Conf. on Artificial Intelligence and Big Data (ICAIBD 2021) (Chengdu, China, 2021)
Affiliations: National University of Defense Technology (Changsha, China)
Summary: This paper evaluates a four-class motor imagery BCI using fNIRS to distinguish imagined movements of different body parts. Ten subjects were asked to perform motor imagery of the left hand, right hand, feet, and tongue (the standard “four limbs + tongue” classes) while fNIRS measured hemodynamic activity over the motor cortex. The authors applied machine-learning classifiers to the fNIRS features to identify which body-part was imagined. The system successfully classified the four imagery classes above chance level, demonstrating the ability of fNIRS to discern multiple distinct motor intentions. The results indicated that combining signals from appropriate motor cortex locations yields enough information to differentiate complex motor imagery tasks. This work underscores the potential of fNIRS for multi-class BCIs, such as hands-free control schemes with more than binary commands, although accuracy still trails EEG-based approaches and could be improved with enhanced feature extraction or multimodal data.
📄 Comparison of Classification Accuracies Between Different Brain Areas During a Two-Class Motor Imagery in an fNIRS-Based BCI
Title: Comparison of Classification Accuracies Between Different Brain Areas During a Two-Class Motor Imagery in a fNIRS-Based BCI
Authors: Amir H. Moslehi, T. Claire Davies
Published in: 2021 10th International IEEE/EMBS Conference on Neural Engineering (NER)
DOI: 10.1109/NER49283.2021.9441376
Summary: This study analyzed functional near-infrared spectroscopy (fNIRS) data from 29 participants to determine how brain region selection affects classification accuracy in motor imagery-based brain–computer interface (BCI) tasks. By examining concentration changes of oxygenated (HbO) and deoxygenated hemoglobin (HbR), and using both LDA and SVM classifiers, the research found that signals from the motor cortex led to significantly higher accuracy than those from frontal or occipital areas. However, using all channels yielded the best performance overall. Additionally, SVM outperformed LDA, while mHbO and mHbR features resulted in comparable classification outcomes. These findings emphasize the motor cortex's suitability for fNIRS-based MI BCI applications.
📄 Combined real-time fMRI and real-time fNIRS brain computer interface (BCI): Training of volitional wrist extension after stroke, a case series pilot study
Title: Combined real-time fMRI and real-time fNIRS brain computer interface (BCI): Training of volitional wrist extension after stroke, a case series pilot study
Authors: Avi K. Matarasso; Jake D. Rieke; Keith White; M. Minhal Yusufali; Janis J. Daly
Published on: PLOS ONE 16(5): e0250431 (2021)
Affiliations: University of Florida & Malcom Randall VA Medical Center (Gainesville, FL, USA)
Summary: This pilot study evaluated a combined fMRI+fNIRS neurofeedback BCI for rehabilitating wrist movement in stroke patients. Four individuals with chronic severe hemiparesis underwent multiple sessions in which they attempted wrist extensions while receiving real-time feedback from either fMRI or fNIRS signals of their brain activity. The training was interwoven with traditional therapy over several weeks. Results showed clinically meaningful improvements in arm function: on average, patients’ Arm Motor Ability Test and Fugl-Meyer scores improved significantly (e.g. Fugl-Meyer gains ~10 points) and active wrist range of motion increased by 20–50° after the intervention. Neuroimaging measures indicated patients learned to modulate their cortical activity – during fNIRS feedback, 42–78% of trials showed successful volitional brain signal changes in the targeted motor cortex. Despite fewer physical practice repetitions (some replaced by BCI feedback), patients’ motor outcomes were at least as good as standard high-intensity therapy. The study demonstrates the safety and potential benefit of combined fMRI–fNIRS BCI feedback in engaging stroke survivors to activate motor areas and improve volitional movement.
📄 Comparison of Brain Activation Between Different Modes of Motor Acquisition: A Functional Near‐Infrared Study
Title: Comparison of Brain Activation Between Different Modes of Motor Acquisition: A Functional Near-Infrared Study
Authors: Meng-Hsuan Tsou, Pei-Yun Chen, Yi-Ting Hung, Yong-Wei Lim, Shiuan-Ling Huang, Yan-Ci Liu
Published on: Brain and Behavior (2025)
Affiliations: 1) School and Graduate Institute of Physical Therapy, College of Medicine, National Taiwan University, Taipei, Taiwan; 2) Taipei First Girls High School, Taipei, Taiwan; 3) Physical Therapy Center, National Taiwan University Hospital, Taipei, Taiwan
Summary: Using functional near-infrared spectroscopy (fNIRS), researchers compared brain activation in healthy adults performing a reaching task under four conditions: actual movement, motor imagery, action observation, and mirror visual feedback. All conditions except action observation significantly activated key motor regions, with real movement and mirror feedback eliciting the strongest responses. Motor imagery also robustly engaged these motor areas, suggesting that imagery and mirror feedback can effectively substitute physical practice in rehabilitation when actual movement is not possible.
📄 Multimodal motor imagery decoding method based on temporal spatial feature alignment and fusion
Title: Multimodal motor imagery decoding method based on temporal spatial feature alignment and fusion
Authors: Yukun Zhang, Shuang Qiu, Huiguang He
Published: March 13, 2023
Affiliations:
School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing, China
Laboratory of Brain Atlas and Brain-Inspired Intelligence, State Key Laboratory of Multimodal Artificial Intelligence Systems, Institute of Automation, Chinese Academy of Sciences, Beijing, China
Summary: This paper proposes a novel neural network for motor imagery-based brain–computer interfaces (MI-BCIs) that leverages multiple neuroimaging modalities (EEG and fNIRS) to improve decoding accuracy. The method introduces spatial feature alignment losses to better represent heterogeneous data from different modalities and an attention-based fusion module to align and combine features over time. The authors collected a new five-class MI dataset with simultaneous EEG and fNIRS recordings to evaluate the approach. Experiments show that the proposed multimodal decoder outperforms single-modality and other comparison methods on both a self-collected dataset and a public dataset, achieving higher classification accuracy. Ablation studies confirm that each component of the model contributes to performance gains, and the work demonstrates how effective feature alignment and fusion can enhance MI-BCI decoding.
📄 Classification of Brain Signals Collected During a Rule Learning Paradigm
Title: Classification of Brain Signals Collected During a Rule Learning Paradigm
Authors: Alicia Howell-Munson, Deniz Sonmez Unal, Theresa G. Mowad, Catherine M. Arrington, Erin Walker, Erin T. Solovey
Published: June 30, 2023
Affiliations:
Worcester Polytechnic Institute (WPI), Worcester, MA, USA
University of Pittsburgh, Pittsburgh, PA, USA
Lehigh University, Bethlehem, PA, USA
Summary: This work explores integrating brain signals with behavioral data to detect a learner’s cognitive state during an educational task. Using a controlled rule learning paradigm from cognitive psychology, the authors captured functional near-infrared spectroscopy (fNIRS) brain data alongside user interaction data as students learned rules across different domains. They trained machine learning models to classify phases of the rule-learning process (inductive reasoning stages) using the combined neural and contextual features. The study found that classifiers incorporating fNIRS features can recognize learning phases better than chance and aligned with expected behavioral patterns, indicating that brain data provide added predictive value. These results suggest that blending neural signals with traditional performance data can improve real-time detection of learning states, which could inform the design of intelligent tutoring systems and adaptive learning environments.
📄 Enhancing Classification Accuracy of fNIRS-BCI for Gait Rehabilitation
Title: Enhancing Classification Accuracy of fNIRS-BCI for Gait Rehabilitation
Authors: Hamza Shabbir Minhas; Hammad Nazeer; Noman Naseer; Umar Shahbaz Khan; Ali R. Ansari
Published on: IEEE Access 12: 117944–117958 (2024)
Affiliations: Air University (Islamabad, Pakistan); Gulf University (Bahrain)
Summary: This paper describes a lower-limb exoskeleton BCI that uses fNIRS to distinguish walking versus resting state, with an emphasis on maximizing classification accuracy for reliable gait control. Twenty healthy participants were recorded with fNIRS over the motor cortex during imagined gait versus rest, and the authors extracted six statistical features (mean, peak, variance, skewness, kurtosis, slope) from the oxygenation signals. By evaluating different feature combinations and classifiers, they identified an optimal feature set and a k-NN classifier (with k tuned via the elbow method) that yielded high accuracy. The system achieved an average offline accuracy of 88.19% for classifying walk vs. rest, and in a simulated real-time test (sliding window of 2.5s) it reached 97.5% accuracy. These results are a significant improvement over earlier approaches and indicate the method’s robustness. The study demonstrates that carefully selected time-domain features combined with a simple classifier can drive an fNIRS-based gait BCI with accuracy approaching EEG-based systems. This advancement supports more dependable neuro-controlled exoskeletons, potentially allowing smoother transitions between user-intended walking and stopping in rehabilitation devices.
📄 MECASA: Motor Execution Classification using Additive Self-Attention for Hybrid EEG-fNIRS Data
Title: MECASA: Motor Execution Classification using Additive Self-Attention for Hybrid EEG-fNIRS Data
Authors: Gourav Siddhad, Juhi Singh, Partha Pratim Roy
Published on: arXiv (2025)
Affiliations: Gourav Siddhad – Indian Institute of Technology Roorkee, India; Juhi Singh – Manipal University Jaipur, India; Partha Pratim Roy – Indian Institute of Technology Roorkee, India
Summary: This study proposes a novel deep learning architecture (MECASA) to improve brain–computer interface classification by fusing electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) signals. Using a motor execution task dataset, the authors compare single-modality classifiers (EEG or fNIRS alone) with a multimodal fusion approach. The MECASA model, which combines convolutional layers with an additive self-attention mechanism (inspired by a vision transformer architecture), consistently outperforms previous methods. Results show that fusing EEG and fNIRS data yields higher accuracy than either modality alone (with fNIRS generally more accurate than EEG by itself). An ablation analysis identifies optimal model settings (e.g. specific feature dimensions) and highlights that including both pain-free and pain-present data in training greatly improves classification performance. Overall, the work demonstrates the potential of deep learning, and specifically MECASA, to enhance hybrid EEG-fNIRS BCI systems for tasks like motor execution.
📄 Analysis of the Impact of the Presence of Physical Pain in fNIRS-based BCI Systems
Title: Analysis of the Impact of the Presence of Physical Pain in fNIRS-based BCI Systems
Authors: Ashwini Subramanian, Foroogh Shamsi, Laleh Najafzadeh
Published on: IEEE Signal Processing in Medicine and Biology Symposium (2022)
Affiliations: Ashwini Subramanian – Dept. of Electrical and Computer Engineering, Rutgers University, NJ, USA; Foroogh Shamsi – Dept. of Psychology, Northeastern University, MA, USA; Laleh Najafzadeh – Dept. of Electrical and Computer Engineering, Rutgers University, NJ, USA.
Summary: This conference paper investigates how acute physical pain affects the performance of a fNIRS-based brain–computer interface. Participants performed mental arithmetic tasks while their brain activity was recorded using fNIRS, both under normal conditions and while experiencing a mild pain stimulus (heat applied to the hand). A convolutional neural network (CNN) was used to classify the cognitive task from the fNIRS data. The results showed that when the model was trained on pain-free data but then applied while the user was in pain, the BCI’s classification accuracy dropped significantly. A multi-label classification approach (to simultaneously detect task and pain presence) further confirmed that distinguishing mental tasks becomes more challenging if the user is in pain. However, training the model on data from both pain-free and pain conditions mitigated this drop in accuracy. The study concludes that for assistive BCI systems, it is important to include pain-condition data during training to ensure robust performance if users might be in pain.
📄 The Use of NIRS in Monitoring Lower-Limb Motor Activation: A Study Comparing a Mobile and Research-Grade System
Title: The Use of NIRS in Monitoring Lower-Limb Motor Activation: A Study Comparing a Mobile and Research-Grade System
Authors: Christopher W. Holland
Published on: Dalhousie University (MSc Thesis, 2020)
Affiliations: Christopher W. Holland – Master of Science candidate, Dalhousie University, Halifax, Nova Scotia, Canada.
Summary: This Master’s thesis examines whether near-infrared spectroscopy (NIRS) can effectively monitor brain activity during lower-limb movement and compares a wearable mobile NIRS device to a laboratory-grade NIRS system. In the study, participants performed a marching-in-place task designed to activate the sensorimotor cortex, while brain activation was measured using both a high-end research NIRS system (NIRScout) and a portable NIRS headset (Axem). The author found that both the mobile and the research-grade NIRS devices could detect changes in cortical oxygenation associated with the leg movement task. The research-grade system, however, demonstrated a greater sensitivity in detecting those neural changes (identifying more significant activation), whereas the mobile device still captured the expected activation patterns but with lower overall signal magnitude. These findings suggest that mobile NIRS technology is capable of monitoring motor-related brain activation and could be useful for real-world or rehabilitation settings, although the more advanced system provides more robust measurements. Incorporating insights from this comparison can guide the development of mobile NIRS systems for reliable brain monitoring outside of lab environments.
📄 Evaluating Multimodal EEG–fNIRS Neurofeedback for Motor Imagery
Title: Evaluation of multimodal EEG-fNIRS neurofeedback for motor imagery
Authors: Camille O. Muller, Thomas Prampart, Elise Bannier, Isabelle Corouge, Pierre Maurel
Published on: Real-Time Functional Imaging and Neurofeedback (rtFIN) Meeting, 2024
Affiliations: Univ. Rennes, Inria, CNRS, IRISA, Rennes, France; CHU Rennes, Department of Radiology, Rennes, France.
Summary: This paper investigates whether combining electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) in a neurofeedback system can improve motor imagery training, with an eye towards post-stroke motor rehabilitation. The authors developed an experimental platform where 30 participants perform motor imagery tasks under three conditions (EEG-only feedback, fNIRS-only feedback, and combined EEG-fNIRS feedback). EEG and fNIRS sensors placed over the primary motor cortex provide signals to compute a real-time neurofeedback score, which is visualized to the user (e.g. as a moving gauge) to guide their motor imagery. By comparing participants’ performance and brain activity across the three conditions, the study evaluates if multimodal feedback leads to better self-regulation of brain activity than single-modality feedback. Findings/Hypothesis: The researchers hypothesize and preliminary observations suggest that dual-modal (EEG+fNIRS) feedback can yield higher control of the feedback signal (i.e. improved neurofeedback performance) and enhanced task-specific brain activation, compared to using EEG or fNIRS alone. Implications: If confirmed, these results imply that multimodal neurofeedback could boost neuroplasticity and effectiveness in neurorehabilitation, potentially improving outcomes for post-stroke motor recovery and other clinical applications.
📄A novel multimodal approach for hybrid brain-computer interface
Title: A novel multimodal approach for hybrid brain-computer interface
Authors: Zhe Sun, Zihao Huang, Feng Duan, Yu Liu
Published on: IEEE Access, 2020
Affiliations: College of Artificial Intelligence, Nankai University (Tianjin, China); Key Laboratory of Exercise and Health Sciences, Ministry of Education, Shanghai University of Sport (Shanghai, China)
Summary: This paper introduces three methods to fuse EEG and NIRS signals for a hybrid brain–computer interface, significantly improving classification accuracy over single-modality systems. In particular, a polynomial fusion strategy achieves state-of-the-art accuracy on motor imagery and mental arithmetic tasks (77.53% and 90.19% respectively, surpassing previous results of ~74% and ~88%), while remaining computationally efficient. The results demonstrate that the hybrid EEG-NIRS approach outperforms individual signals and offers a practical, high-accuracy BCI solution with low computational cost.
📄 Analysis of the Impact of the Presence of Physical Pain in fNIRS-based BCI Systems
Title: Analysis of the Impact of the Presence of Physical Pain in fNIRS-based BCI Systems
Authors: Ashwini Subramanian, Foroogh Shamsi, Laleh Najafzadeh
Published on: IEEE Signal Processing in Medicine and Biology Symposium (2022)
Affiliations: Ashwini Subramanian – Dept. of Electrical and Computer Engineering, Rutgers University, NJ, USA; Foroogh Shamsi – Dept. of Psychology, Northeastern University, MA, USA; Laleh Najafzadeh – Dept. of Electrical and Computer Engineering, Rutgers University, NJ, USA.
Summary: This conference paper investigates how acute physical pain can affect the performance of a fNIRS-based brain–computer interface. Participants performed mental arithmetic tasks while their brain activity was recorded via fNIRS, both under normal conditions and while experiencing a mild pain stimulus (heat applied to the hand). A convolutional neural network (CNN) was trained to classify the tasks from the fNIRS signals. The results showed that when the model was trained on pain-free data but tested on data recorded during pain, the BCI’s task classification accuracy dropped significantly. A multi-label classification (simultaneously detecting pain presence and mental task) further demonstrated that distinguishing between the cognitive tasks is much more difficult in the presence of pain. However, the authors found that including data from pain conditions in the training set largely mitigates this accuracy loss. The study concludes that for assistive BCI systems, it is critical to account for pain: incorporating pain-condition data during classifier training can improve robustness if users might experience pain while using the BCI.👉 Click here to access the full article
📄 Mental Fatigue Classification Aided by Machine Learning-Driven Model under the Influence of Foot and Auditory Binaural Beats Brain Massages via fNIRS
Title: Mental Fatigue Classification Aided by Machine Learning-Driven Model under the Influence of Foot and Auditory Binaural Beats Brain Massages via fNIRS
Authors: Nazo Haroon, Hamid Jabbar, Umar Shahbaz Khan, Taikyeong Ted Jeong, Noman Naseer
Published on: IEEE Access (2024)
Affiliations: Nazo Haroon, Hamid Jabbar, Umar S. Khan – Department of Mechatronics Engineering, National University of Sciences & Technology (NUST), Islamabad, Pakistan; Taikyeong Ted Jeong – Department of AI Convergence, Hallym University, Chuncheon, South Korea; Noman Naseer – Department of Mechatronics Engineering, Air University, Islamabad, Pakistan.
Summary: This paper focuses on detecting and evaluating mental fatigue using fNIRS brain signals, especially when potential countermeasures like foot massage and binaural audio “brain massages” are applied. Participants first underwent tasks to induce mental fatigue, then received an intervention combining a mechanical foot massage (using a massage chair) and auditory binaural beats, while their brain activity was monitored via fNIRS. The authors developed an ensemble machine learning model to classify the level of mental fatigue from the fNIRS data under these conditions. The results show that the model could successfully distinguish different fatigue levels, and importantly, the combined massage intervention had a measurable beneficial effect on participants: brain signal patterns indicated reduced mental fatigue and improved cognitive performance when both the foot massage and binaural beats were administered. In summary, the study demonstrates that fNIRS combined with machine learning can effectively monitor mental fatigue, and that using simultaneous physical and auditory stimuli can help alleviate fatigue. These findings suggest practical applications in fatigue management systems and cognitive wellness, where such non-invasive “brain massage” techniques could be used to improve alertness and mental state.
📄 Hemodynamic responses during standing and sitting activities: a study toward fNIRS-BCI
Title: Hemodynamic responses during standing and sitting activities: a study toward fNIRS-BCI
Authors: Latifah Almulla, Ibraheem Al-Naib, Murad Althobaiti
Published on: Biomedical Physics & Engineering Express, 2020
Affiliations: Biomedical Engineering Department, Imam Abdulrahman Bin Faisal University, Dammam, Saudi Arabia
Summary: Using fNIRS, this study examined hemodynamic responses in the motor cortex during standing vs. sitting tasks (including both actual movements and motor imagery). Nine subjects performed repeated trials of standing and sitting tasks. The results showed that sitting tasks induced a higher oxyhemoglobin activation than standing tasks, consistently across all measured channels and in both real and imagined conditions. The authors also extracted six features from the fNIRS signals and evaluated four classifiers, demonstrating the feasibility of using these tasks for future fNIRS-based BCI control.
📄 Calibration of Deep Learning Classification Models in fNIRS
Title: Calibration of Deep Learning Classification Models in fNIRS
Authors: Zhihao Cao, Zizhou Luo
Published on: 23 Feb 2024 (arXiv preprint)
Affiliations: Department of Mathematics, ETH Zurich, Switzerland; Department of Informatics, University of Zurich, Switzerland
Summary: This paper addresses the reliability (calibration) of deep learning models used for classifying fNIRS (functional near-infrared spectroscopy) brain data. The authors found that many existing fNIRS deep learning models have poor confidence calibration, meaning their predicted probabilities do not reflect true accuracy. They emphasize the importance of incorporating calibration techniques to improve the reliability of deep learning–based predictions in fNIRS brain–computer interface tasks.
📄 Hybrid Functional Near-Infrared Spectroscopy System and Electromyography for Prosthetic Knee Control
Title: Hybrid Functional Near-Infrared Spectroscopy System and Electromyography for Prosthetic Knee Control
Authors: Nouf Jubran AlQahtani, Ibraheem Al-Naib, Ijlal Shahrukh Ateeq, and Murad Althobaiti
Published on: Biosensors, 2024
Affiliations: Imam Abdulrahman Bin Faisal University; King Fahd University of Petroleum & Minerals (Saudi Arabia)
Summary: This paper introduces a hybrid system combining electromyography (EMG) and fNIRS to improve control of prosthetic knees. Nine participants (healthy males) performed above-knee movement tasks in two sessions: real execution and motor imagery. The system simultaneously recorded muscle signals (EMG) and brain signals (fNIRS) from the participants. The combined analysis of EMG and fNIRS data significantly improved classification accuracy of knee movement intentions compared to using either modality alone. Notably, the hybrid approach yielded higher accuracy, with SVM performing best in imagined movements (~49.6% accuracy) and LDA excelling in real movements (~89.7%). These results demonstrate the potential of integrating EMG and fNIRS for enhanced prosthetic knee control via motor imagery.
📄 Machine Learning Based Prediction of Motor Imagery and Motor Execution Tasks from Functional Near Infrared Spectroscopy Signals
Title: Machine Learning Based Prediction of Motor Imagery and Motor Execution Tasks from Functional Near Infrared Spectroscopy Signals
Authors: Oğuzhan Aslan, Kurt Kağan Kurtoğlu, Kutay Yeşilalan, and Sinem Burcu Erdoğan
Published on: Biophotonics Congress: Biomedical Optics (Optics and the Brain 2020), 2020
Affiliations: Department of Medical Engineering, Acıbadem Mehmet Ali Aydınlar University, İstanbul, Turkey
Summary: This conference paper explores the feasibility of using fNIRS signals for BCI applications. Hemodynamic features were extracted from fNIRS measurements during various motor execution and motor imagery tasks, and machine learning models were trained to predict the task type. The results showed that fNIRS data alone can be used to distinguish between real and imagined motor tasks with promising accuracy. Overall, the study demonstrates that fNIRS-based brain signals have potential for classifying motor imagery versus execution tasks, contributing to the development of fNIRS-only BCI systems.
📄 Shedding Light on Motor-Independent Communication: fNIRS-based Brain-Computer Interfacing for Everyday Life
Title: Shedding Light on Motor-Independent Communication: fNIRS-based Brain-Computer Interfacing for Everyday Life
Author: Laurien Nagels-Coune
Published on: Doctoral Thesis, Maastricht University (2024)
Affiliations: Maastricht University, The Netherlands
Summary: This doctoral research focuses on using fNIRS-based BCIs to enable communication without muscle movement (targeting users with severe paralysis, such as locked-in syndrome patients). The thesis encompasses three studies that develop and test novel communication paradigms: two binary yes/no fNIRS-BCI systems (using different mental imagery strategies) and one four-choice fNIRS-BCI system, all designed for practical daily use. The experiments showed that a substantial portion of participants could successfully communicate using these fNIRS-BCI paradigms; notably, in one study all participants used a multi-choice fNIRS-BCI across three days and even communicated in a real-world setting (a cafeteria). These findings highlight that fNIRS-based BCIs are a promising avenue for motor-independent communication in real-life clinical settings.
📄 Comparative Analysis of NIRS-EEG Motor Imagery Data Using Features from Spatial, Spectral and Temporal Domain
Title: Comparative Analysis of NIRS-EEG Motor Imagery Data Using Features from Spatial, Spectral and Temporal Domain
Authors: Hyun-Ji Kim, In-Nea Wang, Young-Tak Kim, Hakseung Kim, and Dong-Joo Kim
Published on: 8th International Winter Conference on Brain-Computer Interface (BCI 2020), 2020
Affiliations: Korea University, Seoul, South Korea
Summary: This study analyzed a hybrid EEG–fNIRS motor imagery dataset (left- vs. right-hand imagery from 29 healthy subjects) to compare different EEG feature domains for BCI performance. The authors extracted features from the spatial domain (e.g., common spatial patterns from EEG), spectral domain (frequency-based features), and temporal domain (time-series features), and evaluated their efficacy in classifying motor imagery tasks using the combined EEG-fNIRS data. The comparative analysis revealed that each feature domain contributes differently to classification accuracy in the hybrid BCI. These insights help identify which types of features are most useful for decoding motor imagery in a combined EEG–fNIRS brain-computer interface, guiding optimal feature selection for hybrid BCI systems.
📄 On the Effects of Pain on fNIRS Classification
Title: On the Effects of Pain on fNIRS Classification
Authors: Foroogh Shamsi and Laleh Najafizadeh
Published on: 20 April 2020 (Biophotonics Congress: Biomedical Optics 2020)
Affiliations: Department of Electrical and Computer Engineering, Rutgers University, NJ, USA
Summary: This conference paper presents the first study examining how pain affects the classification of fNIRS brain signals. The authors trained a model on pain-free fNIRS data and tested it on data recorded under painful stimuli, finding that the presence of pain can alter the brain signals and degrade classification performance. In particular, a model trained on pain-free data performs worse when classifying data with pain, indicating that pain influences fNIRS-based BCI accuracy. These results highlight the need to account for pain-induced changes in brain activity when developing reliable fNIRS classifiers for real-world applications.
📄 Toward a Hybrid Passive BCI for the Modulation of Sustained Attention Using EEG and fNIRS
Title: Toward a Hybrid Passive BCI for the Modulation of Sustained Attention Using EEG and fNIRS
Authors: Alexander J. Karran, Théophile Demazure, Pierre-Majorique Léger, Élise Labonté-LeMoyne, Sylvain Senecal, Marc Fredette, and Gilbert Babin
Published on: November 6, 2019
Affiliations: HEC Montréal, Université de Montréal, Montréal, QC, Canada
Summary: This study developed a hybrid passive BCI combining EEG and fNIRS to help users sustain attention during long-duration tasks in a realistic setting. Thirty participants were divided into three groups with different neurofeedback interventions (none, continuous, or event-triggered). Results showed that providing continuous EEG-based feedback to the user led to moderately better task performance and higher engagement than no feedback. In particular, the continuous feedback group made fewer errors and maintained attention longer than the control group. These findings suggest that a hybrid EEG–fNIRS BCI can effectively boost sustained attention and performance in prolonged tasks by adapting to the user’s mental state.
📄 Investigation of the Performance of fNIRS-based BCIs for Assistive Systems in the Presence of Acute Pain
Title: Investigation of the Performance of fNIRS-based BCIs for Assistive Systems in the Presence of Acute Pain
Authors: Ashwini Subramanian, Foroogh Shamsi, and Laleh Najafizadeh
Published on: 10 February 2023
Affiliations: Department of Electrical and Computer Engineering, Rutgers University, NJ, USA; Department of Psychology, Northeastern University, MA, USA
Summary: This chapter examines how acute pain impacts the performance of an fNIRS-based BCI designed for assistive devices. The authors evaluated a four-condition paradigm where the BCI was trained and tested with or without pain in the user. They found that when the pain conditions during training and testing were mismatched (i.e. trained in pain-free conditions but used during pain, or vice versa), the BCI’s classification accuracy dropped to chance levels. In contrast, when training and testing both occurred under the same pain condition (either consistently pain-free or consistently in pain), accuracy remained high. These results indicate that the presence of pain can significantly degrade BCI performance, and underscore the importance of accounting for pain-induced changes in brain signals when deploying BCIs for patients
📄 fNIRS-EEG BCIs for Motor Rehabilitation: A Review
Title: fNIRS-EEG BCIs for Motor Rehabilitation: A Review
Authors: Jianan Chen, Yunjia Xia, Xinkai Zhou, Ernesto Vidal Rosas, Alexander Thomas, Rui Loureiro, Robert J. Cooper, Tom Carlson, and Hubin Zhao
Published on: December 6, 2023
Affiliations: University College London (multiple departments), London, UK; University of Southampton, Southampton, UK
Summary: This comprehensive review covers recent advances in hybrid fNIRS-EEG brain–computer interfaces for motor rehabilitation. The authors discuss how BCIs can help patients with severe motor impairments by translating brain activity into control signals for assistive devices. They highlight that combining EEG and fNIRS in a hybrid BCI leverages the strengths of both modalities – EEG’s high temporal resolution and fNIRS’s spatial specificity – to improve neurofeedback for motor recovery training. The review surveys current methodologies (especially those using motor imagery), including common system components, signal processing techniques, and machine learning algorithms used for classifying motor intentions. It also addresses key challenges (e.g. noise, feature selection) and future opportunities for developing more effective fNIRS-EEG BCI systems in clinical rehabilitation
📄 Current Implications of EEG and fNIRS as Functional Neuroimaging Techniques for Motor Recovery After Stroke
Title: Current implications of EEG and fNIRS as functional neuroimaging techniques for motor recovery after stroke
Authors: Xiaolong Sun, Chunqiu Dai, Xiangbo Wu, Tao Han, Qiaozhen Li, Yixing Lu, Xinyu Liu, and Hua Yuan
Published on: May 24, 2024
Affiliations: Xijing Hospital, Air Force Medical University (Fourth Military Medical University), Xi’an, China.
Summary: This article reviews the use of EEG and fNIRS neuroimaging for post-stroke motor recovery. It notes that persistent motor deficits after stroke are common, and understanding brain reorganization via functional neuroimaging is a promising but challenging approach. Quantitative EEG (qEEG) metrics (such as power ratio index, brain symmetry index, etc.) have shown correlations with motor recovery outcomes, but have limitations in pinpointing sources of activity. The authors discuss how integrating EEG with fNIRS can provide a more comprehensive view of brain activity by capturing both neural electrical signals and hemodynamic responses. Such combined EEG-fNIRS monitoring offers multi-modal insights into cortical plasticity during rehabilitation. The paper argues that leveraging both modalities could improve prognostic assessments and the development of personalized neurorehabilitation strategies for stroke patients.
📄 Exploring Aesthetic Perception in Impaired Aging: A Multimodal Brain–Computer Interface Study
Title: Exploring Aesthetic Perception in Impaired Aging: A Multimodal Brain–Computer Interface Study
Authors: Livio Clemente, Marianna La Rocca, Giulia Paparella, Marianna Delussi, Giusy Tancredi, Katia Ricci, Giuseppe Procida, Alessandro Introna, Antonio Brunetti, Paolo Taurisano, Vitoantonio Bevilacqua, and Marina de Tommaso
Published on: April 6, 2024
Affiliations: University of Bari Aldo Moro, Italy; University of Southern California, USA (and collaborators in Italy)
Summary: This study used a multimodal BCI (EEG + fNIRS) to investigate how aging with cognitive impairment affects the perception of art. Twenty older participants were divided into a healthy control group and an “impaired aging” group based on cognitive scores. While participants viewed visual art stimuli of varying pleasantness and dynamism in VR, their EEG P300 and LPP responses and fNIRS hemodynamics were recorded. The impaired older adults showed reduced EEG responses (smaller P300/LPP amplitudes) compared to controls, indicating blunted neural reactivity to aesthetic stimuli. However, the impaired group exhibited an increase in fNIRS oxyhemoglobin for pleasurable stimuli, suggesting a compensatory recruitment of brain resources. The findings demonstrate the effectiveness of hybrid EEG-fNIRS BCIs in identifying neural differences in aesthetic appreciation due to cognitive aging, and point toward potential applications in cognitive rehabilitation focused on aesthetic perception.
📄 A bimodal deep learning network based on CNN for fine motor imagery
Title: A bimodal deep learning network based on CNN for fine motor imagery
Authors: Chenyao Wu, Yu Wang, Shuang Qiu, and Huiguang He
Published on: 19 August 2024 (online; in Cognitive Neurodynamics 18(6):3791–3804, Dec 2024)
Affiliations: Institute of Automation, Chinese Academy of Sciences, Beijing, China; University of Chinese Academy of Sciences, Beijing, China
Summary: This paper proposes a hybrid EEG-fNIRS deep learning framework to classify fine motor imagery tasks. Unlike traditional MI paradigms that imagine whole-limb movements, the authors designed a “fine” MI paradigm with four classes (imagining hand, wrist, shoulder movements, or rest) to allow more intuitive control of prosthetics. They recorded synchronized EEG and fNIRS from 12 subjects performing these tasks and extracted features from each modality (including EEG event-related desynchronization patterns and fNIRS activation features). A bimodal convolutional neural network (CNN) was trained to fuse EEG and fNIRS features. The results showed that the EEG+fNIRS fusion model outperformed single-modality models, achieving about 58.96% accuracy in four-class classification – higher than using EEG or fNIRS alone. This demonstrates the feasibility of an EEG-fNIRS bimodal BCI for classifying fine motor imagery, and the proposed CNN fusion approach significantly improved decoding performance of nuanced motor intentions.
📄 Applications of Brain Computer Interface for Motor Imagery Using Deep Learning: Review on Recent Trends
Title: Applications of Brain Computer Interface for Motor Imagery Using Deep Learning: Review on Recent Trends
Authors: Ahmed Zakaria Talha, NourElDin S. Eissa, and Mohd Ibrahim Shapiai
Published on: February 2024
Affiliations: Arab Academy for Science, Technology and Maritime Transport, Egypt; Universiti Teknologi Malaysia (UTM), Malaysia
Summary: This article provides a review of recent advances in motor imagery BCI using deep learning. MI-BCI technology enables translating brain activity into computer commands, offering paralyzed patients a pathway to interact with their environment. The review notes that despite substantial research progress, MI-BCI is still largely confined to lab settings and rarely deployed in real-life applications. The authors survey the latest trends in the field, focusing on two key aspects: (1) feature selection algorithms that identify the most informative brain signal features, and (2) classification techniques (especially deep learning) for distinguishing different imagined movements. They also discuss major challenges such as noise removal, variability across users, and the need for robust, transferable models. Overall, the paper highlights state-of-the-art methods and suggests future directions to move MI-BCI from research to practical use, including improving accuracy, usability, and user training
📄 Investigating Feature Set Decisions for Mental State Decoding in Virtual Reality based Learning Environments
Title: Investigating Feature Set Decisions for Mental State Decoding in Virtual Reality based Learning Environments
Authors: Katharina Lingelbach, Daniel Diers, Michael Bui, and Mathias Vukelić
Published on: 2023 (AHFE Neuroergonomics and Cognitive Engineering Conference Proceedings)
Affiliations: Fraunhofer Institute for Industrial Engineering IAO, Stuttgart, Germany; Carl von Ossietzky University of Oldenburg, Germany; University of Stuttgart, Germany
Summary: This work explores how to best select features from fNIRS signals to decode mental states in VR-based training environments. Eleven participants performed a visuo-spatial working memory task (an n-back exercise) in Virtual Reality while fNIRS measured prefrontal cortex activity. The authors compared different feature sets (oxygenated vs. deoxygenated hemoglobin, and combinations thereof) and feature selection methods for classifying working memory load with conventional machine learning. They found that the highest decoding accuracy was achieved using a feature set that combined both HbO and HbR features, with a sequential forward feature selection method. In particular, peak-to-peak HbR features from premotor and dorsolateral prefrontal regions were among the most informative for distinguishing workload levels. The study underscores the importance of feature set choice in fNIRS-based mental state decoding and provides empirically supported recommendations to improve real-time mental workload classification in VR learning applications
📄 Improving Classification Performance of Four-Class fNIRS-BCI Using Mel Frequency Cepstral Coefficients (MFCC)
Title: Improving classification performance of four-class fNIRS-BCI using Mel Frequency Cepstral Coefficients (MFCC)
Authors: Muhammad Saad Bin Abdul Ghaffar, Umar S. Khan, Javaid Iqbal, Nasir Rashid, Amir Hamza, Waqar S. Qureshi, Mohsin I. Tiwana, and Uzair Izhar
Published on: January 2021 (Infrared Physics & Technology, Volume 112, Article 103589)
Affiliations: National University of Sciences and Technology (NUST), Pakistan; University of the Sunshine Coast, Australia
Summary: This paper introduces Mel-Frequency Cepstral Coefficients (MFCC) as novel features to improve multiclass fNIRS-BCI classification. Using an open dataset of 29 subjects performing four mental tasks (rest, mental arithmetic, left-hand motor imagery, right-hand motor imagery), the authors extracted six time-domain statistical features (mean, variance, etc.) and 13 MFCC features from the fNIRS signals. They evaluated three classifiers (LDA, SVM, KNN) on each feature set and in combination. The results showed that using frequency-domain MFCC features yielded high accuracy (up to 95.7% with SVM) and combining MFCC with time-domain features further improved performance. The best result was an average accuracy of 95.9% for four-class classification using an MFCC+statistics feature set with a neural network classifier in a follow-up analysis. This represents a substantial improvement over prior approaches and demonstrates that MFCCs are an effective feature for fNIRS signal classification, opening a new direction for enhancing BCI accuracy
📄 Applications of Brain Computer Interface for Motor Imagery Using Deep Learning: Review on Recent Trends
Title: Applications of Brain Computer Interface for Motor Imagery Using Deep Learning: Review on Recent Trends
Authors: Ahmed Zakaria Talha; Noureldin S. Eissa; Mohd Ibrahim Shapiai
Published on: Journal of Advanced Research in Applied Sciences and Engineering Technology (February 2024)
Affiliations: Arab Academy for Science, Technology and Maritime Transport (Cairo, Egypt); Malaysia-Japan International Institute of Technology, Universiti Teknologi Malaysia (Kuala Lumpur, Malaysia)
Summary: This article is a review of the latest trends in motor imagery-based brain–computer interfaces (MI-BCIs) using deep learning techniques. It outlines the major challenges in the field and surveys state-of-the-art methods for classifying brain signals generated during imagined movements. The authors discuss how advanced feature selection algorithms and modern classifiers can improve the detection of distinct motor imagery patterns, which could help translate MI-BCI technology from controlled lab settings to practical real-world applications.
📄 A Biofeedback-Driven Serious Game for Self-Regulation of Stress and Anxiety: A Controlled Study of Design, Efficacy and User Experience
Title: A Biofeedback-Driven Serious Game for Self-Regulation of Stress and Anxiety: A Controlled Study of Design, Efficacy and User Experience
Authors: Karen L. Blackmore, Shamus P. Smith, Jacqueline D. Bailey, and Benjamin Krynski
Published on: 2024
Journal: Simulation & Gaming (SAGE Journals)
Summary: This paper introduces a serious game that integrates real-time biofeedback to help users learn to manage their stress and anxiety. The authors designed the game so that players’ physiological signals (e.g. heart rate or breathing) directly influence gameplay, training players to recognize and control their arousal levels. In a controlled study, they evaluated the game’s design, its efficacy in reducing stress/anxiety, and the user experience. The results demonstrate that participants who played the biofeedback-driven game achieved better self-regulation of stress – showing reduced anxiety/stress levels and improved calmness – compared to controls. Participants also reported a positive and engaging user experience, suggesting that biofeedback-enhanced gaming can be an effective and enjoyable tool for stress and anxiety management.
📄 Drowsiness detection using combined neuroimaging: Overview and Challenges
Title: Drowsiness detection using combined neuroimaging: Overview and Challenges
Authors: A S M Sharifuzzaman Sagar; Tajken Salehen; Md Abdur Rob
Published on: 27 Feb 2022 (arXiv)
Affiliations:
Intelligent Mechatronics Engineering, Sejong University, Seoul, South Korea
American International University of Bangladesh, Dhaka, Bangladesh
School of Mechatronics Engineering, China University of Mining and Technology, Xuzhou, China
Summary: This paper presents a comprehensive overview of drowsiness detection methods that use combined neuroimaging modalities in brain-computer interfaces (BCIs). In particular, the authors review recent studies integrating signals from multiple brain-monitoring techniques – such as electroencephalography (EEG) combined with functional near-infrared spectroscopy (fNIRS) or functional MRI – to detect when a person is drowsy or experiencing sleep inertia. They highlight that these multi-modal approaches can capture complementary brain information and improve detection accuracy compared to single-modality systems. The paper also identifies key challenges for combined neuroimaging-based drowsiness detection (including data integration difficulties, individual variability in brain signals, and real-time implementation constraints) and discusses potential solutions and future research directions to advance more reliable drowsiness detection systems.
📄 Block-as-Domain Adaptation for Workload Prediction from fNIRS Data
Title: Block-as-Domain Adaptation for Workload Prediction from fNIRS Data
Authors: Jiyang Wang, Ayse Altay, Senem Velipasalar
Published on: 30 Apr 2024 (arXiv)
Affiliations: Department of Electrical Engineering & Computer Science, Syracuse University
Summary: This paper introduces CABA-DA (Class-Aware, Block-Aware Domain Adaptation), a novel approach for improving cognitive workload prediction from fNIRS data. By treating each data block as a separate domain, the method minimizes intra-class variance and boosts inter-class separation, improving generalization across subjects and sessions. The authors also propose an MLP-Mixer model that outperforms traditional CNNs in this context. Experiments on multiple public fNIRS datasets show strong improvements over baseline models.
📄 Evaluating the effects of multimodal EEG-fNIRS neurofeedback for motor imagery: An experimental platform and study protocol
Title: Evaluating the effects of multimodal EEG-fNIRS neurofeedback for motor imagery: An experimental platform and study protocol
Authors: Camille O. Muller, Thomas Prampart, Elise Bannier, Isabelle Corouge, Pierre Maurel
Published on: 2025
Affiliations:
Univ. Rennes, Inria, CNRS, IRISA, Rennes, France.
CHU Rennes, Department of Radiology, Rennes, France.
Summary: This paper presents a hybrid neurofeedback platform combining EEG (electroencephalography) and fNIRS (functional near-infrared spectroscopy) for motor imagery training, aimed at post-stroke upper-limb rehabilitation. It describes the experimental setup and study protocol to evaluate the system’s efficacy. Preliminary results indicate that simultaneous EEG-fNIRS feedback enhances the user’s sense of control and allows easier disengagement during rest periods, compared to using either modality alone.
📄 Intellectual Property in Brain-Computer Interfacing Technology Systems
Title: Navigating Intellectual Property in Brain-Computer Interfacing Technology Systems
Authors: Dr. Anil S. M., Dr. N. Vani Shree
Published on: 2024, JSS Journal for Legal Studies and Research (Vol. 10, No. 2, pp. 117–137)
Affiliations:
Dr. Anil S. M. – Student of LLM, JSS Law College, Mysuru, India
Dr. N. Vani Shree – Principal and Faculty of Law, JSS Law College, Mysuru, India
Summary: This paper explores the legal challenges surrounding intellectual property (IP) in Brain–Computer Interface (BCI) technologies. It addresses key issues like patentability of BCI hardware and software, copyright ownership of BCI-generated content, neural data privacy, and the role of Standard-Essential Patents (SEPs) in BCI innovation. The authors propose adapting IP laws, enforcing Fair, Reasonable, and Non-Discriminatory (FRAND) licensing, and encouraging open innovation to balance protection, competition, and user rights in the evolving BCI landscape.
📄 A multiple session dataset of NIRS recordings from stroke patients controlling brain–computer interface
Title: A multiple session dataset of NIRS recordings from stroke patients controlling brain–computer interface
Authors: Mikhail R. Isaev, Olesya A. Mokienko, Roman Kh. Lyukmanov, Ekaterina S. Ikonnikova, Anastasiia N. Cherkasova, Natalia A. Suponeva, Michael A. Piradov, Pavel D. Bobrov
Published on: 25 October 2024
Affiliations:
Institute of Higher Nervous Activity and Neurophysiology, Russian Academy of Sciences, Moscow, Russia.
Research Center of Neurology, Moscow, Russia.
Summary: This paper introduces an open-access dataset of functional near-infrared spectroscopy (fNIRS) recordings collected from 15 stroke patients during 237 brain–computer interface (BCI) training sessions. In these sessions, patients performed motor imagery (imagining hand movements) to control a BCI, with visual feedback provided in real time. The dataset comprises over 50 hours of NIRS data (spanning thousands of trials) along with patient demographics, clinical scores (e.g., ARAT and Fugl-Meyer), and recorded BCI performance metrics. As the first large NIRS dataset in a stroke population, it offers a valuable resource for developing and evaluating signal processing and machine learning techniques for post-stroke BCI rehabilitation.
👉 Click here to read the full article
📄 A Platform Technology for VR Biofeedback Training under Operant Conditioning for Functional Lower Limb Weakness
Title: A Platform Technology for VR Biofeedback Training under Operant Conditioning for Functional Lower Limb Weakness
Authors: Anirban Dutta, Abhijit Das
Published on: 31 July 2024
Affiliations:
Institute of Metabolism and Systems Research, University of Birmingham, Birmingham, UK.
Lancashire Teaching Hospitals NHS Foundation Trust, and School of Medicine, University of Central Lancashire, Preston, UK.
Summary: This preprint proposes a virtual reality (VR) biofeedback platform designed to rehabilitate patients with functional lower limb weakness (a form of functional neurological disorder). The system applies operant conditioning principles in a VR environment, meaning patients receive real-time feedback and positive reinforcement in response to attempted leg movements. By engaging users in immersive VR tasks that reward desired motor imagery or movement attempts, the platform aims to help patients “relearn” voluntary control of their affected limbs. The article outlines the platform’s design and suggests that such VR-based operant conditioning can be a promising therapeutic approach to improve lower limb function in patients without overt neurological damage.
👉 Click here to read the full article
📄 Topology-Aware Multimodal Fusion for Neural Dynamics Representation Learning and Classification
Title: Topology-Aware Multimodal Fusion for Neural Dynamics Representation Learning and Classification
Authors: Neela Rahimi, Chetan Kumar, John McLinden, Sarah M. Ismail Hosni, Seyyed Bahram Borgheai, Yalda Shahriari, Ming Shao
Published on: July 2024
Affiliations:
Department of Electrical, Computer, and Biomedical Engineering, University of Rhode Island, Kingston, RI, USA.
Interdisciplinary Neuroscience Program, University of Rhode Island, Kingston, RI, USA.
Neurology Department, Emory University, Atlanta, GA, USA.
Neurology Department, Massachusetts General Hospital, Boston, MA, USA.
Summary: This study presents a graph-based multimodal fusion framework (TaGMF) for classifying neural conditions, demonstrated on differentiating amyotrophic lateral sclerosis (ALS) patients from healthy controls. The approach combines electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) data, leveraging graph neural networks to incorporate the topological relationships of neural signals. By learning joint EEG-fNIRS representations of neural dynamics, the model achieved a substantial improvement (about 22.6% higher accuracy) in classification performance compared to conventional methods. The results indicate that topology-aware fusion of multimodal brain data can enhance the accuracy and robustness of BCI systems and neurological diagnostics.
👉 Click here to read the full article
📄 Multimodal Pre-screening Can Predict BCI Performance Variability: A Novel Subject-Specific Experimental Scheme
Title: Multimodal Pre-screening Can Predict BCI Performance Variability: A Novel Subject-Specific Experimental Scheme
Authors: Seyyed Bahram Borgheai, Alyssa Hillary Zisk, John McLinden, James McIntyre, Reza Sadjadi, Yalda Shahriari
Published on: January 2024
Affiliations:
Department of Electrical, Computer, and Biomedical Engineering, University of Rhode Island, Kingston, RI, USA.
Neurology Department, Emory University, Atlanta, GA, USA.
Interdisciplinary Neuroscience Program, University of Rhode Island, Kingston, RI, USA.
Neurology Department, Massachusetts General Hospital, Boston, MA, USA.
Summary: This article addresses the challenge of inconsistent performance in brain–computer interface (BCI) users by introducing a personalized, multimodal pre-screening approach. The authors used a combination of EEG and fNIRS measurements in a short pre-training session to identify individual factors that influence BCI control performance. Based on features from this pre-screening, they built predictive models that could forecast how well a user would perform under various BCI task conditions. In subsequent test sessions, these models guided adaptive strategies—such as selecting an optimal task paradigm for each user—to improve BCI accuracy and reliability. The results demonstrate that individualized pre-assessment can successfully predict BCI performance and that adjusting the BCI setup to a user’s profile can mitigate performance variability, paving the way for more robust long-term BCI use.
👉 Click here to read the full article
📄 It takes two (seconds): decreasing encoding time for two-choice functional near-infrared spectroscopy brain–computer interface communication
Title: It takes two (seconds): decreasing encoding time for two-choice functional near-infrared spectroscopy brain–computer interface communication
Authors: Anna Vorreuther, Lisa Bastian, Amaia Benitez Andonegui, Danielle Evenblij, Lars Riecke, Michael Lührs, Bettina Sorger
Published on: 2 November 2023
Affiliations:
Institute of Human Factors and Technology Management IAT, University of Stuttgart, Stuttgart, Germany.
Department of Cognitive Neuroscience, Faculty of Psychology and Neuroscience, Maastricht University, Maastricht, The Netherlands.
Summary: This study demonstrates an efficient two-choice communication BCI using functional near-infrared spectroscopy (fNIRS) with an encoding time of only 2 seconds per selection. Ten healthy participants answered yes/no biographical questions by performing combined motor and speech imagery tasks in short 2-second bursts, guided by auditory cues. The fNIRS signals were analyzed in real time, and the system achieved an average online accuracy of about 86% in decoding the participants’ answers (with single-trial accuracies around 68% offline). Notably, the optimal information transfer rate was reached with just four repeats of the 2-second signal. The findings show that even very brief fNIRS measurement periods can enable fast and reliable binary communication, highlighting a promising direction for improving the speed and convenience of fNIRS-based BCI systems for users (e.g., patients with severe motor impairments).
👉 Click here to read the full article