Objective : The field of rehabilitation engineering has evolved to incorporate Artificial Intelligence (AI) and Machine Learning (ML) into technologies and tools designed to assist and promote the rehabilitation of individuals suffering from stroke, paralysis, and other clinical conditions. Here, a systematic review was done to understand the opportunities, challenges, gaps, deficiencies, inconsistencies, and trends in the deployment of AI/ML in rehabilitation engineering. Approach : A search in IEEE Xplore, Web of Science and PubMed databases for the period 2011 to 2021 resulted in 110 identified studies that met the inclusion criteria. Findings are summarized with respect to clinical population, number of participants, rehabilitation task, rehabilitation device, model inputs, duration of intervention, AI/ML algorithm, model accuracy, clinical outcome and presence of feedback. Main Results : Most studies (63%) focused on upper limb or hand rehabilitation with 27% addressing lower-limb rehabilitation. Stroke had the largest representation (58%) among the clinical populations studied followed by spinal cord injury (20%). Classical ML algorithms accounted for 90% of the studies, followed by deep learning (DL). About half of the studies used electroencephalography (EEG) as input to the AI/ML model. The output of these models were used to control different types of virtual and physical devices. Although not all the studies reported the model accuracy, for those that reported results the accuracy ranged from 42% - 100%. Clinical outcome measures were not included in 66% of the studies. Less than one-fifth of the studies had open-loop protocols. Unfortunately, 18% of the studies were single session, with 26% reporting between 2-10 sessions. Significance : Future studies should 1) investigate applications of DL in the rehabilitation of motor function; 2) be closed-loop and 3) include multiple sessions to understand the longitudinal effects of therapy. Moreover, studies should address model explainability, and follow standard engineering, clinical, and end-user metrics.

Background: Democratized access to safe and effective robotic neurorehabilitation for stroke survivors requires innovative, affordable solutions that can be used not only in clinics but also at home. This requires the high usability of the devices involved to minimize costs associated with support from physical therapists or technicians. Methods: This paper describes the early findings of the NeuroExo brain–machine interface (BMI) with an upper-limb robotic exoskeleton for stroke neurorehabilitation. This early feasibility study consisted of a six-week protocol, with an initial training and BMI calibration phase at the clinic followed by 60 sessions of neuromotor therapy at the homes of the participants. Pre- and post-assessments were used to assess users’ compliance and system performance. Results: Participants achieved a compliance rate between 21% and 100%, with an average of 69%, while maintaining adequate signal quality and a positive perceived BMI performance during home usage with an average Likert scale score of four out of five. Moreover, adequate signal quality was maintained for four out of five participants throughout the protocol. These findings provide valuable insights into essential components for comprehensive rehabilitation therapy for stroke survivors. Furthermore, linear mixed-effects statistical models showed a significant reduction in trial duration (p-value < 0.02) and concomitant changes in brain patterns (p-value < 0.02). Conclusions: the analysis of these findings suggests that a low-cost, safe, simple-to-use BMI system for at-home stroke rehabilitation is feasible.

Cerebral Palsy (CP) is a lifelong neuromuscular condition requiring continuous physical assistance to manage symptoms, improve mobility, and enhance quality of life. While many passive and active exoskeleton technology advancements have shown promise, few solutions are tailored to meet the unique needs of children with CP. The Gross Motor Function Classification System (GMFCS) categorizes the functional ability of individuals with CP into five levels, ranging from level 1 (full mobility with some difficulties) to level 5 (limited independent mobility). Many assistive devices serve the needs of those with higher GMFCS levels (4-5), but children with GMFCS levels 1 to 3 have limited access to appropriate assistance. This study aims to identify the key factors that drive the design of a rehabilitative, assistive, and diagnostic wearable tailored to integrate into the daily lives of children with CP, GMFCS levels 1-3. A comprehensive literature review, interviews with physical and occupational therapists, and ethnographic research provided insights into the key factors influencing children with CP’s experiences in their daily lives. This collaborative and innovative study between industrial design and engineering programs have led to identifying the required technological and functional components that will bridge the gap between theory and practice in a holistic exoskeleton design approach. Finally, a series of prototypes initiated the first phase of concept generation, which explores the impact of aforementioned factors into an adaptive lifestyle-centered soft-suit that integrates seamlessly into children with CP’s lives. The final prototype is currently undergoing user and functional testing and development. Future implications of this work emphasize user-centered design, calling for the involvement of children and parents as stakeholders in the design process. As exoskeletons become more available in-home settings beyond clinical applications, it is essential to understand and meet children's physical, emotional, social, and mental needs with CP GMFCS levels 1-3 to create meaningful, lifestyle-aware solutions.

Objective. Neurological disorders affecting speech production adversely impact quality of life for over 7 million individuals in the US. Traditional speech interfaces like eye-tracking devices and P300 spellers are slow and unnatural for these patients. An alternative solution, speech brain-computer interfaces (BCIs), directly decodes speech characteristics, offering a more natural communication mechanism. This research explores the feasibility of decoding speech features using non-invasive EEG.Approach. Nine neurologically intact participants were equipped with a 63-channel EEG system with additional sensors to eliminate eye artifacts. Participants read aloud sentences selected for phonetic similarity to the English language. Deep learning models, including Convolutional Neural Networks and Recurrent Neural Networks with and without attention modules, were optimized with a focus on minimizing trainable parameters and utilizing small input window sizes for real-time application. These models were employed for discrete and continuous speech decoding tasks.Main results. Statistically significant participant-independent decoding performance was achieved for discrete classes and continuous characteristics of the produced audio signal. A frequency sub-band analysis highlighted the significance of certain frequency bands (delta, theta, gamma) for decoding performance, and a perturbation analysis was used to identify crucial channels. Assessed channel selection methods did not significantly improve performance, suggesting a distributed representation of speech information encoded in the EEG signals. Leave-One-Out training demonstrated the feasibility of utilizing common speech neural correlates, reducing data collection requirements from individual participants.Significance. These findings contribute significantly to the development of EEG-enabled speech synthesis by demonstrating the feasibility of decoding both discrete and continuous speech features from EEG signals, even in the presence of EMG artifacts. By addressing the challenges of EMG interference and optimizing deep learning models for speech decoding, this study lays a strong foundation for EEG-based speech BCIs.

This longitudinal Mobile Brain-Body Imaging dataset was acquired during six rehearsal sessions and three public performances of a scene from a play with highly emotional components. Three student actor dyads (N=6), one theatre director (N=1) and three audience members (N=3) participated in this study. The MoBI data recorded includes mobile electroencephalography, electrooculography, blood volume pulse, heart rate, body temperature, electrodermal activity, triaxial arm and head acceleration. The visual data includes five streams of video. This article describes the experimental setup, the multi-modal data streams acquired using a hyperscanning methodology, and an assessment of the data quality.

Imagine unlocking the power of the mind to communicate, create, and even interact with the world around us. Recent breakthroughs in Artificial Intelligence (AI), especially in how machines "see" and "understand" language, are now fueling exciting progress in decoding brain signals from scalp electroencephalography (EEG). Prima facie, this opens the door to revolutionary brain-computer interfaces (BCIs) designed for real life, moving beyond traditional uses to envision Brain-to-Speech, Brain-to-Image, and even a Brain-to-Internet of Things (BCIoT). However, the journey is not as straightforward as it was for Computer Vision (CV) and Natural Language Processing (NLP). Applying AI to real-world EEG-based BCIs, particularly in building powerful foundational models, presents unique and intricate hurdles that could affect their reliability. Here, we unfold a guided exploration of this dynamic and rapidly evolving research area. Rather than barely outlining a map of current endeavors and results, the goal is to provide a principled navigation of this hot and cutting-edge research landscape. We consider the basic paradigms that emerge from a causal perspective and the attendant challenges presented to AI-based models. Looking ahead, we then discuss promising research avenues that could overcome today's technological, methodological, and ethical limitations. Our aim is to lay out a clear roadmap for creating truly practical and effective EEG-based BCI solutions that can thrive in everyday environments.

The MyoStep project, developed at the University of Houston (UH) and funded by the IEEE Electron Devices Society (EDS) Award, exemplifies the transformative potential of technology in addressing global challenges. This humanitarian initiative focuses on creating a lightweight, soft exoskeleton to assist children with cerebral palsy (CP), improving mobility and enhancing their quality of life. Beyond its humanitarian impact, MyoStep serves as an unparalleled educational platform, offering undergraduate students, particularly those from underrepresented groups in science, technology, engineering, and mathematics (STEM), hands-on experience in cutting-edge fields such as smart materials; wearable technology; and biomedical, electrical, and computer engineering. By merging technical excellence with a mission-driven purpose, MyoStep not only inspires the next generation of engineers but also embodies the IEEE EDS vision of advancing technology for the benefit of humanity.

Electroencephalography (EEG) is a widely used non-invasive technique for monitoring brain activity, but low signal-to-noise ratios (SNR) due to various artifacts often compromise its utility. Conventional artifact removal methods require manual intervention or risk suppressing critical neural features during filtering/reconstruction. Recent advances in generative models, including Variational Autoencoders (VAEs) and Generative Adversarial Networks (GANs), have shown promise for EEG reconstruction; however, these approaches often lack integrated temporal-spectral-spatial sensitivity and are computationally intensive, limiting their suitability for real-time applications like brain-computer interfaces (BCIs). To overcome these challenges, we introduce EEGReXferNet, a lightweight Gen-AI framework for EEG subspace reconstruction via cross-subject transfer learning - developed using Keras TensorFlow (v2.15.1). EEGReXferNet employs a modular architecture that leverages volume conduction across neighboring channels, band-specific convolution encoding, and dynamic latent feature extraction through sliding windows. By integrating reference-based scaling, the framework ensures continuity across successive windows and generalizes effectively across subjects. This design improves spatial-temporal-spectral resolution (mean PSD correlation >= 0.95; mean spectrogram RV-Coefficient >= 0.85), reduces total weights by ~45% to mitigate overfitting, and maintains computational efficiency for robust, real-time EEG preprocessing in neurophysiological and BCI applications.

BACKGROUND:Dissecting the neurobiology of dance would shed light on a complex, yet ubiquitous, form of human communication. In this experiment, we sought to study, via mobile electroencephalography (EEG), the brain activity of five experienced dancers while dancing butoh, a postmodern dance that originated in Japan. RESULTS:We report the experimental design, methods, and practical execution of a highly interdisciplinary project that required the collaboration of dancers, engineers, neuroscientists, musicians, and multimedia artists, among others. We explain in detail how we technically validated all our EEG procedures (e.g., via impedance value monitoring) and minimized potential artifacts in our recordings (e.g., via electrooculography and inertial measurement units). We also describe the engineering details and hardware that enabled us to achieve synchronization between signals recorded at different sampling frequencies, along with a signal preprocessing and denoising pipeline that we used for data re-sampling and power line noise removal. As our experiment culminated in a live performance, where we generated a real-time visualization of the dancers' interbrain synchrony on a screen via an artistic brain-computer interface, we outline all the methodology (e.g., filtering, time-windows, equation) we used for online bispectrum estimations. Additionally, we provide access to all the raw EEG data and codes we used in our recordings. We, lastly, discuss how we envision that the data could be used to address several hypotheses, such as that of interbrain synchrony or the motor theory of vocal learning. CONCLUSIONS:Being, to our knowledge, the first study to report synchronous and simultaneous recording from five dancers, we expect that our findings will inform future art-science collaborations, as well as dance-movement therapies.

Although significant progress has been made in understanding the cortical correlates underlying balance control, these studies focused on a single task, limiting the ability to generalize the findings. Different balance tasks may elicit cortical activations in the same regions but show different levels of activation because of distinct underlying mechanisms. In this study, twenty young, neurotypical adults were instructed to maintain standing balance while the standing support surface was either translated or rotated. The differences in cortical activations in the frontocentral region between these two widely used tasks were examined using electroencephalography (EEG). Additionally, the study investigated whether transcranial magnetic stimulation could modulate these cortical activations during the platform translation task. Higher delta and lower alpha relative power were found over the frontocentral region during the platform translation task when compared to the platform rotation task, suggesting greater engagement of attentional and sensory integration resources for the former. Continuous theta burst stimulation over the supplementary motor area significantly reduced delta activity in the frontocentral region but did not alter alpha activity during the platform translation task. The results provide a direct comparison of neural activations between two commonly used balance tasks and are expected to lay a strong foundation for designing neurointerventions for balance improvements with effects generalizable across multiple balance scenarios.

Implanted brain–computer interfaces (iBCIs) translate brain activity recorded intracranially into commands for virtual or physical machines to restore or rehabilitate motor, sensory or speech functions. Currently, no iBCIs have been approved by regulatory agencies for the medical device market despite being in clinical trials since 1998, with little information available about their progress and outcomes. To address this gap, we conducted a review of all identified clinical trials of iBCIs for communication, motor control or restoration of tactile perception conducted between 1998 and 2023. We summarize findings from 21 research groups worldwide and their 67 participants who received implants to understand the challenges and opportunities in the iBCI field. This analysis highlights the importance of improving participant diversity, creating a participant registry to inform future research, regulatory and payer approvals, investor funding and new applications, adopting governed data sharing and standards, and boosting collaborative research. Information on clinical trial progress and outcomes of implanted brain–computer interfaces (iBCIs) remains limited. This Review provides a comprehensive survey of all identified clinical trials of iBCIs for communication, motor control or restoration of tactile perception conducted from the first chronic implantation in 1998 to December 2023.

This editorial provides a background and overview of the interdisciplinary workshop on “The Social and Neural Bases of Creative Movement,” bringing together dancers, choreographers, musicians, artists, kinesiologists and neuroscientists to share perspectives and develop a common language to define and explore the relationship between dance and the brain.

The ability to hold objects relies on neural processes underlying grip force control during grasping. Brain activity lateralized to contralateral hemisphere averaged over trials is associated with grip force applied on an object. However, the involvement of neural variability within-trial during grip force control remains unclear. We examined dependence of neural variability over frontal, central, and parietal regions of interest (ROI) on grip force magnitude using noninvasive electroencephalography (EEG). We utilized our existing EEG dataset comprised of healthy young adults performing an isometric force control task, cued to exert 5, 10, or 15% of their maximum voluntary contraction (MVC) across trials and received visual feedback of their grip force. We quantified variability in EEG signal via sample entropy (sequence-dependent) and standard deviation (sequence-independent measure) over ROI. We found lateralized modulation in EEG sample entropy with force magnitude over central electrodes but not over frontal or parietal electrodes. However, modulation was not observed for standard deviation in the EEG activity. These findings highlight lateralized and spatially constrained modulation in sequence-dependent, but not sequence-independent component of EEG variability. We contextualize these findings in applications requiring finer precision (e.g., prosthesis), and propose directions for future studies investigating role of neural entropy in behavior.

This report contains a description of physiological and motion data, recorded simultaneously and in synchrony using the hyperscanning method from two professional dancers using wireless mobile brain-body imaging (MoBI) technology during rehearsals and public performances of “LiveWire” - a new composition comprised of five choreographed music and dance sections inspired by neuroscience principles. Brain and ocular activity were measured using 28-channel scalp electroencephalography (EEG), and 4-channel electrooculography (EOG), respectively; and head motion was recorded using an inertial measurement unit (IMU) placed on the forehead of each dancer. Video recordings were obtained for each session to allow for tagging of physiological and motion signals and for behavioral analysis. Data recordings were collected from 10 sessions over a 4-month period, in which the dancers rehearsed or performed (in front of an audience) choreographed expressive movements. A detailed explanation of the experimental set-up, the steps carried out for data collection, and an explanation on the usage are provided in this report.

Background This research focused on the development of a motor imagery (MI) based brain-machine interface (BMI) using deep learning algorithms to control a lower-limb robotic exoskeleton. The study aimed to overcome the limitations of traditional BMI approaches by leveraging the advantages of deep learning, such as automated feature extraction and transfer learning. The experimental protocol to evaluate the BMI was designed as asynchronous, allowing subjects to perform mental tasks at their own will.Methods A total of five healthy able-bodied subjects were enrolled in this study to participate in a series of experimental sessions. The brain signals from two of these sessions were used to develop a generic deep learning model through transfer learning. Subsequently, this model was fine-tuned during the remaining sessions and subjected to evaluation. Three distinct deep learning approaches were compared: one that did not undergo fine-tuning, another that fine-tuned all layers of the model, and a third one that fine-tuned only the last three layers. The evaluation phase involved the exclusive closed-loop control of the exoskeleton device by the participants' neural activity using the second deep learning approach for the decoding.Results The three deep learning approaches were assessed in comparison to an approach based on spatial features that was trained for each subject and experimental session, demonstrating their superior performance. Interestingly, the deep learning approach without fine-tuning achieved comparable performance to the features-based approach, indicating that a generic model trained on data from different individuals and previous sessions can yield similar efficacy. Among the three deep learning approaches compared, fine-tuning all layer weights demonstrated the highest performance.Conclusion This research represents an initial stride toward future calibration-free methods. Despite the efforts to diminish calibration time by leveraging data from other subjects, complete elimination proved unattainable. The study's discoveries hold notable significance for advancing calibration-free approaches, offering the promise of minimizing the need for training trials. Furthermore, the experimental evaluation protocol employed in this study aimed to replicate real-life scenarios, granting participants a higher degree of autonomy in decision-making regarding actions such as walking or stopping gait.

Balance control is an important indicator of mobility and independence in activities of daily living. How the functional coupling between the cortex and the muscle for balance control is affected following stroke remains to be known. We investigated the changes in coupling between the cortex and leg muscles during a challenging balance task over multiple frequency bands in chronic stroke survivors. Fourteen participants with stroke and ten healthy controls performed a challenging balance task. They stood on a computerized support surface that was either fixed (low difficulty condition) or sway-referenced with varying gain (medium and high difficulty conditions). We computed corticomuscular coherence between electrodes placed over the sensorimotor area (electroencephalography) and leg muscles (electromyography) and assessed balance performance using clinical and laboratory-based tests. We found significantly lower delta frequency band coherence in stroke participants when compared with healthy controls under medium difficulty condition, but not during low and high difficulty conditions. These differences were found for most of the distal but not for proximal leg muscle groups. No differences were found at other frequency bands. Participants with stroke showed poor balance clinical scores when compared with healthy controls, but no differences were found for laboratory-based tests. The observation of effects at distal but not at proximal muscle groups suggests differences in the (re)organization of the descending connections across two muscle groups for balance control. We argue that the observed group difference in delta band coherence indicates balance context-dependent alteration in mechanisms for the detection of somatosensory modulation resulting from sway-referencing of the support surface for balance maintenance following stroke.

OBJECTIVE:Motor imagery-based brain-computer interface (MI-BCI) is a promising novel mode of stroke rehabilitation. The current study aims to investigate the feasibility of MI-BCI in upper limb rehabilitation of chronic stroke survivors and also to study the early event-related desynchronization after MI-BCI intervention. METHODS:Changes in the characteristics of sensorimotor rhythm modulations in response to a short brain-computer interface (BCI) intervention for upper limb rehabilitation of stroke-disabled hand and normal hand were examined. The participants were trained to modulate their brain rhythms through motor imagery or execution during calibration, and they played a virtual marble game during the feedback session, where the movement of the marble was controlled by their sensorimotor rhythm. RESULTS:Ipsilesional and contralesional activities were observed in the brain during the upper limb rehabilitation using BCI intervention. All the participants were able to successfully control the position of the virtual marble using their sensorimotor rhythm. CONCLUSIONS:The preliminary results support the feasibility of BCI in upper limb rehabilitation and unveil the capability of MI-BCI as a promising medical intervention. This study provides a strong platform for clinicians to build upon new strategies for stroke rehabilitation by integrating MI-BCI with various therapeutic options to induce neural plasticity and recovery.

Scalp electroencephalography (EEG) is a neural source signal that is extensively used in neuroengineering due to its non-invasive nature and ease of collection. However, a drawback to the use of EEG is the prevalence of physiological artifacts generated by eye movements and eye blinks that contaminate the brain signals. Previously, we have proposed and validated an H ∞ -based Adaptive Noise Cancellation (ANC) technique for the real-time identification, learning and removal of eye blinks, eye motions, amplitude drifts and recording biases from EEG simultaneously. However, the standard electroocu-lography (EOG) electrode configuration requires four electrodes for EOG measurement, which limits its applicability for reduced-channel mobile applications, such as brain-computer interfaces (BCI). Here, we assess multiple configurations with varying number of EOG electrodes and compare the ANC effectiveness of these configurations to the ideal four-electrode configuration. From an analysis of the root mean squared error (RMSE) and differences in signal to noise ratios (SNR) between the ideal four-electrode case and the alternative configurations, it is reported that several three-electrode alternative configurations were effective in essentially replicating the ability to remove EOG artifacts in an experimental cohort of ten healthy subjects. For nine subjects, it was shown that only two to three EOG electrodes were needed to achieve similar performance as compared to the four-electrode case. This study demonstrates that the typical four-electrode configuration for EOG recordings for adaptive noise cancellation of ocular artifacts may not be necessary; by using the proposed new EOG configurations it is possible to improve electrode allocation efficiency for EOG measurements in mobile EEG applications.

Brain-Computer Interface (BCI) and Internet of Things (IoT) systems have recently been amalgamated to create BCIoT. Most of the early applications have focused on the healthcare sector, and more recently, in education, virtual reality, smart homes, and smart vehicles, amongst others. While there are many transversal developing stages that can be satisfied by a single system, no common enabling technology or standards exist. These challenges are address in the proposed platform, Brain-eNet. This technology was developed considering the constraints-space defined by BCIoT real-time mobile applications. This is expected to enable the development of BCIoT systems by providing modular hardware and software resources. Two instances of this platform implementation are provided, a motor intent detection for rehabilitation and an emotion recognition system.

Classification of patterns of brain activity in neuroengineering research is an important tool for understanding the brain, developing neurodiagnostics, and designing closed-loop neural interfaces. Scalp electroencephalography (EEG), by virtue of its noninvasiveness and lower cost, has been used for neural signal classification, and researchers have utilized various machine learning methods. Recently, deep learning has gained popularity due to its ability to significantly increase the classification performance in numerous domains while elucidating the relevant features for classification. It is a natural step to deploy such promising techniques for EEG classification tasks. This book chapter aims to serve as a comprehensive reference source for both EEG and deep learning researchers interested in EEG-based deep learning studies. Potential pitfalls, challenges, and opportunities in the application of deep learning to EEG data are discussed.