Journal of Neural Engineering - IOPscience

Journal of Neural Engineering - IOPscience
Journal of Neural Engineering
Purpose-led Publishing
is a coalition of three not-for-profit publishers in the field of physical sciences: AIP Publishing, the American Physical Society and IOP Publishing.
Together, as publishers that will always put purpose above profit, we have defined a set of industry standards that underpin high-quality, ethical scholarly communications.
We are proudly declaring that science is our only shareholder.
SUPPORTS OPEN ACCESS
Journal of Neural Engineering
was created to help scientists, clinicians and engineers to understand, replace, repair and enhance the nervous system.
Submit
an article
opens in new tab
Track my article
opens in new tab
RSS
Sign up for new issue notifications
Median submission to first decision before peer review
7 days
Median submission to first decision after peer review
66 days
Impact factor
3.8
Citescore
7.8
Full list of journal metrics
A review of classification algorithms for EEG-based brain–computer interfaces: a 10 year update
F Lotte
et al
2018
J. Neural Eng.
15
031005
View article
, A review of classification algorithms for EEG-based brain–computer interfaces: a 10 year update
PDF
, A review of classification algorithms for EEG-based brain–computer interfaces: a 10 year update
Objective
. Most current electroencephalography (EEG)-based brain–computer interfaces (BCIs) are based on machine learning algorithms. There is a large diversity of classifier types that are used in this field, as described in our 2007 review paper. Now, approximately ten years after this review publication, many new algorithms have been developed and tested to classify EEG signals in BCIs. The time is therefore ripe for an updated review of EEG classification algorithms for BCIs.
Approach
. We surveyed the BCI and machine learning literature from 2007 to 2017 to identify the new classification approaches that have been investigated to design BCIs. We synthesize these studies in order to present such algorithms, to report how they were used for BCIs, what were the outcomes, and to identify their pros and cons.
Main results
. We found that the recently designed classification algorithms for EEG-based BCIs can be divided into four main categories: adaptive classifiers, matrix and tensor classifiers, transfer learning and deep learning, plus a few other miscellaneous classifiers. Among these, adaptive classifiers were demonstrated to be generally superior to static ones, even with unsupervised adaptation. Transfer learning can also prove useful although the benefits of transfer learning remain unpredictable. Riemannian geometry-based methods have reached state-of-the-art performances on multiple BCI problems and deserve to be explored more thoroughly, along with tensor-based methods. Shrinkage linear discriminant analysis and random forests also appear particularly useful for small training samples settings. On the other hand, deep learning methods have not yet shown convincing improvement over state-of-the-art BCI methods.
Significance
. This paper provides a comprehensive overview of the modern classification algorithms used in EEG-based BCIs, presents the principles of these methods and guidelines on when and how to use them. It also identifies a number of challenges to further advance EEG classification in BCI.
The following article is
Open access
EEG foundation models: a critical review of current progress and future directions
Gayal Kuruppu
et al
2026
J. Neural Eng.
23
021001
View article
, EEG foundation models: a critical review of current progress and future directions
PDF
, EEG foundation models: a critical review of current progress and future directions
Premise.
Patterns of electrical brain activity recorded via electroencephalography (EEG) offer immense value for scientific and clinical investigations. The inability of supervised EEG encoders to learn robust EEG patterns and their over-reliance on expensive signal annotations have sparked a transition towards general-purpose self-supervised EEG encoders, i.e. EEG foundation models (EEG-FMs), for robust and scalable EEG feature extraction. However, the real-world readiness of early EEG-FMs and the rubrics for long-term research progress remain unclear.
Objective.
In this work, we conduct a review of ten early EEG-FMs to capture common trends and identify key directions for future development of EEG-FMs.
Methods.
We comparatively analyze each EEG-FM using three fundamental pillars of foundation modeling, namely the representation of input data, self-supervised modeling, and the evaluation strategy. Based on this analysis, we present a critical synthesis of EEG-FM methodology, empirical findings, and outstanding research gaps.
Results.
We find that most EEG-FMs adopt a sequence-based modeling scheme that relies on transformer-based backbones and the reconstruction of masked temporal EEG sequences for self-supervision. However, model evaluations remain heterogeneous and largely limited, making it challenging to assess their practical off-the-shelf utility. In addition to adopting standardized and realistic evaluations, future work should demonstrate more substantial scaling effects and make principled and trustworthy choices throughout the EEG representation learning pipeline.
Significance.
Our review indicates that the development of benchmarks, software tools, technical methodologies, and applications in collaboration with domain experts may advance the translational utility and real-world adoption of EEG-FMs.
The following article is
Open access
Deep learning for electroencephalogram (EEG) classification tasks: a review
Alexander Craik
et al
2019
J. Neural Eng.
16
031001
View article
, Deep learning for electroencephalogram (EEG) classification tasks: a review
PDF
, Deep learning for electroencephalogram (EEG) classification tasks: a review
Objective
. Electroencephalography (EEG) analysis has been an important tool in neuroscience with applications in neuroscience, neural engineering (e.g. Brain–computer interfaces, BCI’s), and even commercial applications. Many of the analytical tools used in EEG studies have used machine learning to uncover relevant information for neural classification and neuroimaging. Recently, the availability of large EEG data sets and advances in machine learning have both led to the deployment of deep learning architectures, especially in the analysis of EEG signals and in understanding the information it may contain for brain functionality. The robust automatic classification of these signals is an important step towards making the use of EEG more practical in many applications and less reliant on trained professionals. Towards this goal, a systematic review of the literature on deep learning applications to EEG classification was performed to address the following critical questions: (1) Which EEG classification tasks have been explored with deep learning? (2) What input formulations have been used for training the deep networks? (3) Are there specific deep learning network structures suitable for specific types of tasks?
Approach
. A systematic literature review of EEG classification using deep learning was performed on Web of Science and PubMed databases, resulting in 90 identified studies. Those studies were analyzed based on type of task, EEG preprocessing methods, input type, and deep learning architecture.
Main results
. For EEG classification tasks, convolutional neural networks, recurrent neural networks, deep belief networks outperform stacked auto-encoders and multi-layer perceptron neural networks in classification accuracy. The tasks that used deep learning fell into five general groups: emotion recognition, motor imagery, mental workload, seizure detection, event related potential detection, and sleep scoring. For each type of task, we describe the specific input formulation, major characteristics, and end classifier recommendations found through this review.
Significance
. This review summarizes the current practices and performance outcomes in the use of deep learning for EEG classification. Practical suggestions on the selection of many hyperparameters are provided in the hope that they will promote or guide the deployment of deep learning to EEG datasets in future research.
The following article is
Open access
Deep learning-based electroencephalography analysis: a systematic review
Yannick Roy
et al
2019
J. Neural Eng.
16
051001
View article
, Deep learning-based electroencephalography analysis: a systematic review
PDF
, Deep learning-based electroencephalography analysis: a systematic review
Context
. Electroencephalography (EEG) is a complex signal and can require several years of training, as well as advanced signal processing and feature extraction methodologies to be correctly interpreted. Recently, deep learning (DL) has shown great promise in helping make sense of EEG signals due to its capacity to learn good feature representations from raw data. Whether DL truly presents advantages as compared to more traditional EEG processing approaches, however, remains an open question.
Objective
. In this work, we review 154 papers that apply DL to EEG, published between January 2010 and July 2018, and spanning different application domains such as epilepsy, sleep, brain–computer interfacing, and cognitive and affective monitoring. We extract trends and highlight interesting approaches from this large body of literature in order to inform future research and formulate recommendations.
Methods
. Major databases spanning the fields of science and engineering were queried to identify relevant studies published in scientific journals, conferences, and electronic preprint repositories. Various data items were extracted for each study pertaining to (1) the data, (2) the preprocessing methodology, (3) the DL design choices, (4) the results, and (5) the reproducibility of the experiments. These items were then analyzed one by one to uncover trends.
Results
. Our analysis reveals that the amount of EEG data used across studies varies from less than ten minutes to thousands of hours, while the number of samples seen during training by a network varies from a few dozens to several millions, depending on how epochs are extracted. Interestingly, we saw that more than half the studies used publicly available data and that there has also been a clear shift from intra-subject to inter-subject approaches over the last few years. About
of the studies used convolutional neural networks (CNNs), while
used recurrent neural networks (RNNs), most often with a total of 3–10 layers. Moreover, almost one-half of the studies trained their models on raw or preprocessed EEG time series. Finally, the median gain in accuracy of DL approaches over traditional baselines was
across all relevant studies. More importantly, however, we noticed studies often suffer from poor reproducibility: a majority of papers would be hard or impossible to reproduce given the unavailability of their data and code.
Significance
. To help the community progress and share work more effectively, we provide a list of recommendations for future studies and emphasize the need for more reproducible research. We also make our summary table of DL and EEG papers available and invite authors of published work to contribute to it directly. A planned follow-up to this work will be an online public benchmarking portal listing reproducible results.
EEGNet: a compact convolutional neural network for EEG-based brain–computer interfaces
Vernon J Lawhern
et al
2018
J. Neural Eng.
15
056013
View article
, EEGNet: a compact convolutional neural network for EEG-based brain–computer interfaces
PDF
, EEGNet: a compact convolutional neural network for EEG-based brain–computer interfaces
Objective
. Brain–computer interfaces (BCI) enable direct communication with a computer, using neural activity as the control signal. This neural signal is generally chosen from a variety of well-studied electroencephalogram (EEG) signals. For a given BCI paradigm, feature extractors and classifiers are tailored to the distinct characteristics of its expected EEG control signal, limiting its application to that specific signal. Convolutional neural networks (CNNs), which have been used in computer vision and speech recognition to perform automatic feature extraction and classification, have successfully been applied to EEG-based BCIs; however, they have mainly been applied to single BCI paradigms and thus it remains unclear how these architectures generalize to other paradigms. Here, we ask if we can design a single CNN architecture to accurately classify EEG signals from different BCI paradigms, while simultaneously being as compact as possible.
Approach
. In this work we introduce EEGNet, a compact convolutional neural network for EEG-based BCIs. We introduce the use of depthwise and separable convolutions to construct an EEG-specific model which encapsulates well-known EEG feature extraction concepts for BCI. We compare EEGNet, both for within-subject and cross-subject classification, to current state-of-the-art approaches across four BCI paradigms: P300 visual-evoked potentials, error-related negativity responses (ERN), movement-related cortical potentials (MRCP), and sensory motor rhythms (SMR).
Main results
. We show that EEGNet generalizes across paradigms better than, and achieves comparably high performance to, the reference algorithms when only limited training data is available across all tested paradigms. In addition, we demonstrate three different approaches to visualize the contents of a trained EEGNet model to enable interpretation of the learned features.
Significance
. Our results suggest that EEGNet is robust enough to learn a wide variety of interpretable features over a range of BCI tasks. Our models can be found at:
The following article is
Open access
BCI for stroke rehabilitation: motor and beyond
Ravikiran Mane
et al
2020
J. Neural Eng.
17
041001
View article
, BCI for stroke rehabilitation: motor and beyond
PDF
, BCI for stroke rehabilitation: motor and beyond
Stroke is one of the leading causes of long-term disability among adults and contributes to major socio-economic burden globally. Stroke frequently results in multifaceted impairments including motor, cognitive and emotion deficits. In recent years, brain–computer interface (BCI)-based therapy has shown promising results for post-stroke motor rehabilitation. In spite of the success received by BCI-based interventions in the motor domain, non-motor impairments are yet to receive similar attention in research and clinical settings. Some preliminary encouraging results in post-stroke cognitive rehabilitation using BCI seem to suggest that it may also hold potential for treating non-motor deficits such as cognitive and emotion impairments. Moreover, past studies have shown an intricate relationship between motor, cognitive and emotion functions which might influence the overall post-stroke rehabilitation outcome. A number of studies highlight the inability of current treatment protocols to account for the implicit interplay between motor, cognitive and emotion functions. This indicates the necessity to explore an all-inclusive treatment plan targeting the synergistic influence of these standalone interventions. This approach may lead to better overall recovery than treating the individual deficits in isolation. In this paper, we review the recent advances in BCI-based post-stroke motor rehabilitation and highlight the potential for the use of BCI systems beyond the motor domain, in particular, in improving cognition and emotion of stroke patients. Building on the current results and findings of studies in individual domains, we next discuss the possibility of a holistic BCI system for motor, cognitive and affect rehabilitation which may synergistically promote restorative neuroplasticity. Such a system would provide an all-encompassing rehabilitation platform, leading to overarching clinical outcomes and transfer of these outcomes to a better quality of living. This is one of the first works to analyse the possibility of targeting cross-domain influence of post-stroke functional recovery enabled by BCI-based rehabilitation.
The following article is
Open access
Computational optimization of two-photon holographic stimulation sites
in vivo
Marcus A Triplett
et al
2026
J. Neural Eng.
23
026015
View article
, Computational optimization of two-photon holographic stimulation sites in vivo
PDF
, Computational optimization of two-photon holographic stimulation sites in vivo
Objective.
Determining the intricate structure and function of neural circuits requires the ability to precisely manipulate circuit activity. Two-photon holographic optogenetics has emerged as a powerful tool for achieving this via flexible excitation of user-defined neural ensembles. However, the precision of two-photon optogenetics has been constrained by off-target stimulation (OTS), an effect where proximal non-target neurons can be unintentionally activated due to imperfect spatial confinement of light onto target neurons. New approaches are therefore needed to resolve the OTS problem.
Approach.
Here, we introduce a real-time computational method for mitigating OTS that first empirically samples each neuron’s sensitivity to stimulation at proximal locations, and then optimizes stimulation sites using a fast, interpretable model based on adaptive non-negative basis function regression (NBFR).
Main results.
NBFR is highly scalable, completing model fitting for hundreds of neurons in just a few seconds and then optimizing stimulation sites in several hundred milliseconds per stimulus—fast enough for most closed-loop behavioral experiments. We characterize the performance of our approach in both simulations and
in vivo
experiments in mouse hippocampus, showing its efficacy under realistic experimental conditions.
Significance.
Our results thus establish NBFR-based photostimulus optimization as an important addition to an emerging computational toolkit for precise yet scalable holographic optogenetics.
The following article is
Open access
Prevalence of sympathetic fibers within the rat cervical vagus, and functional consequence on physiological effects mediated by vagus nerve stimulation (VNS)
Ashlesha Deshmukh
et al
2026
J. Neural Eng.
23
026021
View article
, Prevalence of sympathetic fibers within the rat cervical vagus, and functional consequence on physiological effects mediated by vagus nerve stimulation (VNS)
PDF
, Prevalence of sympathetic fibers within the rat cervical vagus, and functional consequence on physiological effects mediated by vagus nerve stimulation (VNS)
Objective.
Electrical stimulation of the vagus nerve (VNS) is an Food and Drug Administration approved therapy for epilepsy, depression and rehabilitation after stroke, with recent clinical trials to treat heart failure and inflammation. VNS is often assumed to activate either parasympathetic efferents projecting to visceral organs, and/or sensory afferents projecting from these organs, for its therapeutic effects. Recent studies in humans, swine and dogs have shown that sympathetic nerve fibers from the sympathetic trunk (ST) can frequently be found within the cervical vagus nerve (VN). However, the prevalence and functional consequence of sympathetic fibers on VNS have yet to be elucidated in the most common high throughput animal model to study disease, the rodent.
Approach.
We carefully traced ST from superior cervical ganglion (SCG) to find its location in the carotid sheath with reference to the VN in a cohort of Long Evans rats. We then assessed the prevalence of ST fibers with the cervical VN across the cohort using micro-computer tomography and immunohistochemistry. Finally, we stimulated the VN and the ST in isolation, and where they were conjoined, to evaluate the ST contribution to changes in heart rate (HR). VNS induced HR changes are a commonly used surrogate for changes in sympathetic/parasympathetic tone.
Main results.
The ST frequently runs in very close proximity to the VN in rats when traced caudally from the SCG. The ST is even conjoined with the VN for stretches within the carotid sheathe at the most common location to place an epineural cuff. Cross-connecting branches were found between the ST and the VN. VNS performed at locations where there was minimal ST crossover induced dose-dependent bradycardia (decrease in HR) across the cohort, with detectable bradycardia across the cohort beginning at 50
A (
= 8 right,
= 3 left). Conversely, stimulation of the isolated ST induced tachycardia (increase in HR) across the cohort beginning at ∼200
A (
= 7 right,
= 3 left).
Significance.
These data suggest that studies of VNS in the rodent model may also be stimulating sympathetic fibers from the ST in addition to canonical VN pathways. Concurrent sympathetic activation has profound implications for dissecting mechanisms of VNS for a host of diseases/disorders. As such, careful post-mortem assessment of the presence of ‘hitchhiking’ sympathetic fibers within the VN is critical for understanding sources of variability in VNS outcomes.
The following article is
Open access
Large-scale training data enhances silent speech decoding with around-ear EEG
Masakazu Inoue
et al
2026
J. Neural Eng.
23
026027
View article
, Large-scale training data enhances silent speech decoding with around-ear EEG
PDF
, Large-scale training data enhances silent speech decoding with around-ear EEG
Objective
. Silent speech decoding (SSD) offers a potential communication alternative for individuals with impaired vocalization. However, conventional multi-electrode electroencephalography (EEG) or facial electromyography (EMG) systems require cumbersome preparation and are unsuitable for daily use. This study evaluates the practicality of SSD using a wearable around-ear EEG device, focusing on data scaling, cross-subject transfer, vocabulary extensibility, and online decoding performance.
Approach
. We collected 72 h of around-ear EEG from 24 healthy participants and one individual with incomplete locked-in syndrome (LIS) during silent, vocalized, and attempted speech, and integrated these around-ear EEG recordings with prior EMG + high-density EEG datasets, yielding 282.4 total h of training data. Using a 64-word classification task as the evaluation metric, we assessed: (1) whether larger datasets improve around-ear EEG–based SSD, (2) whether healthy-participant data supplement limited LIS-participant data despite articulatory differences, (3) transferability to unseen vocabulary, and (4) online user-interface performance.
Main results
. Large-scale EEG/EMG data improved SSD accuracy in both healthy participants and the LIS participant. Training on the heterogeneous dataset achieved 56.6% accuracy for healthy users and 47.3% for the LIS participant. Fine-tuning this decoder for new vocabulary increased the accuracy by 22 percentage points relative to training from scratch. Regression analysis showed that, for decoding in the LIS participant, data from the LIS participant contributed approximately four times the weight of healthy-participant data, quantifying data strategies for SSD. Online experiments achieved top-1/top-5 accuracies of 47.2%/76.0% for healthy users and 26.5%/49.1% for the LIS participant.
Significance
. The results indicate that lightweight, commercially feasible around-ear EEG can enable practical SSD when combined with large-scale healthy-participant data, supporting online operation. Moreover, models trained on a 64-word vocabulary facilitate decoding of a new vocabulary, providing a path toward SSD systems requiring minimal LIS-participant data. This study advances non-invasive SSD systems suitable for everyday communication.
The following article is
Open access
MOABB: trustworthy algorithm benchmarking for BCIs
Vinay Jayaram and Alexandre Barachant 2018
J. Neural Eng.
15
066011
View article
, MOABB: trustworthy algorithm benchmarking for BCIs
PDF
, MOABB: trustworthy algorithm benchmarking for BCIs
Objective
. Brain–computer interface (BCI) algorithm development has long been hampered by two major issues: small sample sets and a lack of reproducibility. We offer a solution to both of these problems via a software suite that streamlines both the issues of finding and preprocessing data in a reliable manner, as well as that of using a consistent interface for machine learning methods.
Approach
. By building on recent advances in software for signal analysis implemented in the MNE toolkit, and the unified framework for machine learning offered by the scikit-learn project, we offer a system that can improve BCI algorithm development. This system is fully open-source under the BSD licence and available at
Main results
. We analyze a set of state-of-the-art decoding algorithms across 12 open access datasets, including over 250 subjects. Our results show that even for the best methods, there are datasets which do not show significant improvements, and further that many previously validated methods do not generalize well outside the datasets they were tested on.
Significance
. Our analysis confirms that BCI algorithms validated on single datasets are not representative, highlighting the need for more robust validation in the machine learning for BCIs community.
The following article is
Open access
Bridging innovation and adoption: a mixed-methods investigation into stroke survivors’ acceptability of brain-computer interface-based rehabilitation interventions
E Grevet
et al
2026
J. Neural Eng.
23
026037
View article
, Bridging innovation and adoption: a mixed-methods investigation into stroke survivors’ acceptability of brain-computer interface-based rehabilitation interventions
PDF
, Bridging innovation and adoption: a mixed-methods investigation into stroke survivors’ acceptability of brain-computer interface-based rehabilitation interventions
Objective
. Brain-computer interface (BCI)-based interventions show growing evidence of efficacy for post-stroke upper-limb rehabilitation by closing the sensorimotor loop and promoting neuroplasticity. Despite this promise, their clinical uptake remains limited. This study aimed to identify the determinants of BCI acceptability among stroke survivors in order to inform user-centred neurotechnology design and facilitate clinical adoption.
Approach
. We conducted a mixed-methods study combining a large-scale questionnaire (
= 140) and semi-structured interviews (
= 12) with stroke survivors. The questionnaire was grounded in a validated theoretical model of BCI acceptability and assessed the determinants of three core constructs:
intention to use
(IU),
perceived usefulness
(PU), and
perceived ease of use
(PEOU). Qualitative data were analysed using thematic framework analysis to enrich and contextualise the quantitative findings.
Main Results
. Overall, stroke survivors reported high acceptability of BCI-based rehabilitation, with strong IU (mean IU = 8.48/10). Quantitative analyses showed that IU was primarily driven by PU, itself strongly influenced by perceived scientific relevance and, to a lesser extent, by individual factors such as autonomy, self-efficacy, and technology-related anxiety. PEOU was mainly determined by ease of learning and computer playfulness, but did not directly predict IU. Qualitative findings complemented these results by highlighting the importance of perceived innovativeness, the perception of directly engaging brain activity, visibility of progress, and clear scientific evidence. Interviews also emphasised the need for intuitive interfaces, clear instructions, short sessions, and appropriate therapeutic support to sustain engagement.
Significance
. These findings underscore that successful adoption of BCI-based rehabilitation requires more than technological performance alone. Enhancing acceptability among stroke survivors calls for: (i) goal-oriented and evidence-based rehabilitation protocols; (ii) clear communication of scientific rationale and outcomes to reinforce trust and social acceptance; and (iii) user-friendly system design that supports learning, autonomy, and self-efficacy while minimising cognitive and physical burden.
The following article is
Open access
Solving the problem of inception: a cross-species perspective on strategies for a mechanistic refinement of intracortical microstimulation
Takashi D Y Kozai
et al
2026
J. Neural Eng.
23
023001
View article
, Solving the problem of inception: a cross-species perspective on strategies for a mechanistic refinement of intracortical microstimulation
PDF
, Solving the problem of inception: a cross-species perspective on strategies for a mechanistic refinement of intracortical microstimulation
Objective.
Microstimulation delivers electrical pulses directly into the brain, with one of its promises being to restore lost senses to millions of people. Yet a fundamental challenge remains: how do intracortical microstimulation (ICMS) patterns engage neural circuits to achieve the inception of specific experiences, such as vivid sensory percepts of touch and vision? Here, we define ‘inception’ as the initiation of percepts evoked by microstimulation through the mapping of stimulation to circuit-level activity that results in sensory experiences.
Approach.
This perspective proposes an integrated research framework that combines Reverse Translation, Forward Translation, and computational neuroscience to bridge insights between clinical observations and high-resolution animal studies.
Framework.
Our framework envisions the development and evaluation of ICMS strategies within a cross-species system that narrows the range of plausible underlying neural mechanisms and the set of evoked perceptual outcomes. Reverse Translation uses human perceptual reports about phosphenes, tones, and touch to guide investigations in rodents and non-human primates, mapping the cell types and circuits underlying each percept. Forward Translation leverages these biological insights to design refined ICMS protocols for selective circuit engagement. Bidirectional Translation weaves these approaches together through computational neuroscience, ensuring that experimental observations iteratively and continuously refine one another across species and experimental modalities.
Significance.
This integrated strategy aims to transform microstimulation research into a dynamic dialogue between fundamental science and human experience. Harnessing the Bidirectional Translation Framework can accelerate therapies that enhance quality of life for people with sensory or motor impairments, and contribute more broadly to systems neuroscience by uncovering the mechanisms by which causal manipulation changes activity in neurons and networks.
Physiologically inspired modeling of cortical dynamics through spiking neural networks
Dario Milea
et al
2026
J. Neural Eng.
23
026035
View article
, Physiologically inspired modeling of cortical dynamics through spiking neural networks
PDF
, Physiologically inspired modeling of cortical dynamics through spiking neural networks
Objective
. The characterization of neural activity underlying neurophysiological function presents a major challenge in computational neuroscience. Several methods have been proposed to investigate cortical network dynamics by reconstructing underlying neural activity from electroencephalography (EEG) signals. However, these methods generally pose significant mathematical challenges.
Approach
. This study introduces a novel framework to model the underlying brain activity network from a functional and physiologically-inspired perspective, combining spiking neural networks with EEG signal analysis. The dynamics of single neurons are described by the well-known Izhikevich model, and distinct populations of cortical inhibitory and excitatory neurons are employed to model experimental EEG recordings. Functional interactions among distinct populations are mathematically formalized through connective probabilities.
Main results
. The proposed framework is validated by testing it on synthetic data, as well as on two experimental datasets comprising data from 30 healthy subjects undergoing a cold-pressure test (CPT), and 36 subjects undergoing a mental arithmetic stressor. Experimental results suggest that the proposed framework provides novel and complementary insights into characterizing neuronal changes in comparison to standard EEG power analysis.
Significance
. The proposed framework constitutes a promising tool for functionally characterizing the underlying cortical dynamics under pathophysiological conditions.
An EEG-based framework for exploring adaptive rhythmic human–machine interaction
Wannes Van Ransbeeck
et al
2026
J. Neural Eng.
23
026034
View article
, An EEG-based framework for exploring adaptive rhythmic human–machine interaction
PDF
, An EEG-based framework for exploring adaptive rhythmic human–machine interaction
Objective.
Understanding rhythmic human–human interaction and its underlying mechanisms can enhance experiential value and enjoyment by providing a tailored experience and supporting applications in medical human–machine contexts. Existing experimental paradigms often lack a unified and holistic analysis, characterised by limited ecological validity in partner realism, active engagement, and visual interaction. These can produce hidebound insights due to variable partner behaviour, inflexible design, or insufficient user experience analysis. The study presents and validates a multimodal paradigm that addresses these limitations and enables controlled evaluation of human–human rhythm interaction and its extension to virtual AI agents.
Approach.
Participants completed a tapping paradigm with an audio–visual drum animation driven by either a human or AI-based partner under simple and complex (polyrhythmic) conditions. Portable electroencephalography (EEG) recordings and post-trial questionnaires assessed neural and subjective responses.
Main results.
The framework improves ecological validity relative to existing approaches and effectively masks partner identity (human vs AI) without reducing experienced flow, arousal, or enjoyment, which remained positive overall. Notably, the AI-based partner considered a first attempt to create a virtual AI-driven interacting drummer, suitable for future consideration of alternative algorithms. Additionally, the design supports unobtrusive, portable EEG measurement of neural modulation and temporal alignment with both performed and presented stimuli.
Significance.
This paradigm offers a flexible foundation for studying rhythmic interaction in human–machine systems, balancing ecological realism with experimental partner control while supporting future adaptive or biofeedback-driven systems that optimise rhythm interaction in real-time.
The following article is
Open access
Multi-dimensional characterization and tracking of motor unit action potentials
Lara McManus
et al
2026
J. Neural Eng.
23
026033
View article
, Multi-dimensional characterization and tracking of motor unit action potentials
PDF
, Multi-dimensional characterization and tracking of motor unit action potentials
Objective.
Decomposition of high-density surface electromyography (HDsEMG) signals allows identification of individual motor unit firing times and provides a spatiotemporal image of their action potential waveforms. The ability to reliably match and track motor unit action potentials (MUAPs) from the same motor unit across multiple recordings allows changes in their recruitment and firing properties to be identified, however, similarities in MUAP shape can present challenges for reliable tracking.
Approach.
A new method for matching MUAP waveforms using a multi-dimensional (MD) representation is presented. MUAPs are represented as trajectories in high-dimensional space, where each HDsEMG channel corresponds to a different dimension. Trajectories are compared using MD features to measure the similarity between pairs of MUAP waveforms. Feature reduction and clustering are then used to classify pairs of MUAPs as belonging to the same or different motor units. The ability of the MD method to correctly identify pairs of matching MUAPs was assessed using MUAPs from simulated and experimental datasets and compared with two-dimensional cross-correlation (CC) using a threshold of 0.7, 0.8 or 0.9.
Main results.
The proposed MD method resulted in significantly higher F1 scores and lower false positive and false negative rates in both simulated and experimental datasets (
< 0.001). Across all datasets examined, the MD method correctly identified a greater number of matching MUAP pairs (89.8 ± 18.4%) compared with the best performing CC threshold (73.3 ± 21.0%). This was accompanied by a 49.6% lower false positive rate for the MD method.
Significance.
This study demonstrates that MD representations of MUAP trajectories recorded from high density arrays can more accurately identify MUAPs from the same motor unit, improving motor unit tracking compared with traditional correlation based approaches.
Bayesian decoding and its application in reading out spatial memory from neural ensembles
Ning Wang
et al
2026
J. Neural Eng.
23
021003
View article
, Bayesian decoding and its application in reading out spatial memory from neural ensembles
PDF
, Bayesian decoding and its application in reading out spatial memory from neural ensembles
Spatial memory serves as a foundation to establish cognitive map, supporting navigation and decision-making processes across species. Essential brain regions such as the hippocampus and entorhinal cortex enable these functions through spatially tuned neurons, particularly place cells, which encode an animal’s precise location. The continuous spatial trajectories are then able to be represented by temporally sequential firing of these cells at neural ensemble level. Bayesian frameworks are powerful tools for reconstructing such ‘mind travel’. In this article, we focus on the principles and advances of Bayesian decoding methods for extracting spatial memory information from neural ensembles. First, we review non-recursive approaches and recursive point process filters, paying special attention to clusterless decoding strategies. We also discuss emerging approaches such as neural manifolds within Bayesian estimation. Next, we discuss the advanced application of Bayesian decoding in understanding the neuronal coding mechanisms of memory consolidation and planning, and in supporting computational model establishment and closed-loop manipulation. Finally, we discuss the limitations and challenges of recent approaches, highlighting the promising strategies that could raise the decoding efficiency and adapt the growing scale of neural data. We believe that the developing of Bayesian decoding approach would significantly benefit for techniques and applications of memory-related brain machine interface.
Personalized transcranial electrical stimulation: a review of computational modeling and optimization
Mo Wang
et al
2026
J. Neural Eng.
23
021002
View article
, Personalized transcranial electrical stimulation: a review of computational modeling and optimization
PDF
, Personalized transcranial electrical stimulation: a review of computational modeling and optimization
Objective
. Personalized transcranial electrical stimulation (tES) has gained increasing attention due to the substantial inter-individual variability in brain anatomy and physiology. While previous reviews have discussed the physiological mechanisms and clinical applications of tES, there remains a critical gap in up-to-date syntheses focused on the computational modeling frameworks that enable individualized stimulation optimization.
Approach
. This review presents a comprehensive overview of recent advances in computational techniques supporting personalized tES. We systematically examine developments in forward modeling for simulating individualized electric fields, as well as inverse modeling approaches for optimizing stimulation parameters. We critically evaluate progress in head modeling pipelines, optimization algorithms, and the integration of multimodal brain data.
Main results
. Recent advances have substantially accelerated the construction of subject-specific head conductor models and expanded the landscape of optimization methods, including multi-objective optimization and brain network-informed optimization. These advances allow for dynamic and individualized stimulation planning, moving beyond empirical trial-and-error approaches.
Significance
. By integrating the latest developments in computational modeling for personalized tES, this review highlights current challenges, emerging opportunities, and future directions for achieving precision neuromodulation.
The following article is
Open access
EEG foundation models: a critical review of current progress and future directions
Gayal Kuruppu
et al
2026
J. Neural Eng.
23
021001
View article
, EEG foundation models: a critical review of current progress and future directions
PDF
, EEG foundation models: a critical review of current progress and future directions
Premise.
Patterns of electrical brain activity recorded via electroencephalography (EEG) offer immense value for scientific and clinical investigations. The inability of supervised EEG encoders to learn robust EEG patterns and their over-reliance on expensive signal annotations have sparked a transition towards general-purpose self-supervised EEG encoders, i.e. EEG foundation models (EEG-FMs), for robust and scalable EEG feature extraction. However, the real-world readiness of early EEG-FMs and the rubrics for long-term research progress remain unclear.
Objective.
In this work, we conduct a review of ten early EEG-FMs to capture common trends and identify key directions for future development of EEG-FMs.
Methods.
We comparatively analyze each EEG-FM using three fundamental pillars of foundation modeling, namely the representation of input data, self-supervised modeling, and the evaluation strategy. Based on this analysis, we present a critical synthesis of EEG-FM methodology, empirical findings, and outstanding research gaps.
Results.
We find that most EEG-FMs adopt a sequence-based modeling scheme that relies on transformer-based backbones and the reconstruction of masked temporal EEG sequences for self-supervision. However, model evaluations remain heterogeneous and largely limited, making it challenging to assess their practical off-the-shelf utility. In addition to adopting standardized and realistic evaluations, future work should demonstrate more substantial scaling effects and make principled and trustworthy choices throughout the EEG representation learning pipeline.
Significance.
Our review indicates that the development of benchmarks, software tools, technical methodologies, and applications in collaboration with domain experts may advance the translational utility and real-world adoption of EEG-FMs.
The following article is
Open access
Tissue response to deep brain stimulation electrodes: a review of animal and neurohistopathological studies
Dorothy X Zhao
et al
2026
J. Neural Eng.
23
011003
View article
, Tissue response to deep brain stimulation electrodes: a review of animal and neurohistopathological studies
PDF
, Tissue response to deep brain stimulation electrodes: a review of animal and neurohistopathological studies
Objective.
Deep brain stimulation (DBS) is a neuromodulation therapy widely used to treat various neurological and neuropsychiatric conditions, with thousands of patients undergoing the procedure every year. However, despite the immense improvement in quality of life that most patients experience after surgery, many questions still remain surrounding various elements of DBS, including how the brain tissue responds to DBS electrodes and how that interaction may affect the therapy.
Approach.
In this review, we build off a previous neurohistopathological review to encompass studies up to present date.
Main results.
We identified 33 cases with 63 electrodes from patients with various disease pathologies and DBS targets. We supplemented the findings with animal studies.
Significance.
These studies can provide evidence where neurohistopathological studies have not been performed. They can also offer predictions to guide future neurohistopathological studies. Better understanding of the tissue response to DBS electrodes can contribute to improved clinical outcomes.
The following article is
Open access
Inferring neural sources from electroencephalography: foundations and frontiers
A R Phillips
et al
2026
J. Neural Eng.
23
011002
View article
, Inferring neural sources from electroencephalography: foundations and frontiers
PDF
, Inferring neural sources from electroencephalography: foundations and frontiers
Electroencephalography (EEG) provides robust, cost-effective, and portable measurements of brain electrical activity. However, its spatial resolution is limited, constraining the localization and estimation of deep sources. Although methods exist to infer neural activity from scalp recordings, major challenges remain due to high dimensionality, temporal overlap among neural sources, and anatomical variability in head geometry. This topical review synthesizes inverse modeling approaches, with emphasis on nonlinear methods, multimodal integration, and high-density EEG systems that address these limitations. We also review the forward model and related background theory, summarize clinical applications, outline research directions, and identify available software tools and relevant publicly available datasets. Our goal is to help researchers understand traditional source estimation techniques and integrate advanced methods that may better capture the complexity of neurophysiological sources.
The following article is
Open access
Bayesian time-history modeling enhances Parkinsonian motor state classification for adaptive deep brain stimulation
Leung et al
View accepted manuscript
, Bayesian time-history modeling enhances Parkinsonian motor state classification for adaptive deep brain stimulation
PDF
, Bayesian time-history modeling enhances Parkinsonian motor state classification for adaptive deep brain stimulation
Objective: Adaptive deep brain stimulation (aDBS) for Parkinson’s disease is a recently-approved therapy that adjusts stimulation in response to neurophysiologic biomarkers of motor-symptom state. Most real-time implementations of aDBS rely on instantaneous, noise-susceptible classifiers that apply simple thresholds to neurophysiologic biomarkers. We examined whether incorporating temporal history through Bayesian state-space modeling improved motor-state classification compared to instantaneous discriminant classifiers. Approach: We analyzed naturalistic neural data from three patients with Parkinson’s disease chronically implanted with investigational sensing-enabled DBS systems, recording from both the subthalamic nucleus (STN) and sensorimotor cortex. Biomarkers were extracted across multiple window lengths and labeled using wearable-derived bradykinesia and dyskinesia scores. Classifier behavior was evaluated using two biomarkers (cortical stimulation-entrained gamma and STN beta oscillations) across a factorial combination of two conditions: (1) instantaneous discriminant analysis vs. Bayesian time-history modeling via hidden Markov models (HMMs), and (2) single Gaussian vs. Gaussian mixture modeling of each motor state’s biomarker distribution. Performance metrics included F1 scores, accuracy, prediction smoothness, latency, and computational load. Main Results: Using entrained-gamma biomarkers, incorporating time history via HMMs significantly improved hyperkinetic-state detection (F1: +12.9 ± 1.8%; accuracy: +30.0 ± 2.7%; both padj < 0.001) with modest decreases in hypokinetic-state performance, yielding a net increase in average F1 (+4.7 ± 0.9%, p < 0.001). HMMs also yielded smoother and more accurate predictions for a given latency compared to simply increasing the window length used to extract neurophysiologic biomarkers. Entrained-gamma biomarkers outperformed STN beta biomarkers across all classifiers (average F1: +12.9% ± 0.5%, p < 0.001). All methods operated within sub-millisecond prediction times and demonstrated sublinear empirical computational scaling. Significance: Bayesian time-history modeling enhanced motor-state classification while preserving the low latency and computational efficiency required for real-time aDBS. These findings, derived from chronic at-home recordings, support the translational potential of Bayesian state-space models for next-generation aDBS systems.
Multimodal analysis of hemodynamic and electrocortical causal interactions during auditory processing
Mclinden et al
View accepted manuscript
, Multimodal analysis of hemodynamic and electrocortical causal interactions during auditory processing
PDF
, Multimodal analysis of hemodynamic and electrocortical causal interactions during auditory processing
Objective: Electrocortical and hemodynamic signals measured by electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) can be used to provide complementary information about neural activity underpinning auditory processing. However, the causal interactions between these signals remain underexplored, partly due to several factors, including the complexity of disentangling cortical and systemic physiological components captured in fNIRS modality. Recent developments in the ability to estimate the influence of non-cortical sources on fNIRS signals provide new opportunities to address these confounds. In this study, we investigated causal interactions between vascular-hemodynamic and electrocortical dynamics through simultaneous recording of fNIRS and EEG during auditory processing. Approach: We employed multimodal multivariate Granger causal (MVGC) analysis to investigate causal interactions between EEG and fNIRS signals. To explore the role of systemic physiology on the obtained causal interactions, a temporally-embedded canonical correlation analysis-general linear model (tCCA-GLM)-based preprocessing approach was applied to fNIRS signals to correct for potential systemic physiology confounds. Main Results: Our results showed significantly stronger causal interactions originating from fNIRS in the right auditory cortex to frontocentral EEG in the 38-42 Hz frequency range during the auditory task relative to rest before correction. We additionally observed broad connectivity in the direction originating from fNIRS variables to EEG across frequency bands within participants throughout the dataset. These results suggest complex relationships between cortical vascular-hemodynamics, electrocortical oscillations, and systemic physiological components with roles in both task-related and spontaneous neural activity. Significance: This study highlights the complex causal interplay between electrical, cerebral hemodynamic, and systemic physiology components underpinning spontaneous electro-vascular dynamics associated with auditory processing.
Microwaves stimulate and inhibit neurons via modulation of TRPV and TREK channels
Marar et al
View accepted manuscript
, Microwaves stimulate and inhibit neurons via modulation of TRPV and TREK channels
PDF
, Microwaves stimulate and inhibit neurons via modulation of TRPV and TREK channels
Objective. Electrical neuromodulation, the current clinical standard, is invasive, expensive, and prone to malfunction. Electromagnetic waves can perform noninvasive neuromodulation, but existing methods are limited by the tradeoff between penetration depth and spatial precision. Microwaves in the 0.9 – 3 GHz range are widely used for telecommunications and can penetrate to the deep brain. Microwaves have been shown to nonthermally modulate neural activity, but the acute cellular bioeffects remain unclear and under-studied. Approach. Here, we employ a miniaturized microwave-powered neuromodulation implantable device (MINI) to investigate the roles of ion channels in microwave-mediated thermal stimulation and nonthermal inhibition. The MINI enabled electrophysiological recordings of neurons exposed to microwaves, which elucidated the differential effects of pulsed and continuous microwaves on neurons. We further investigated the cellular mechanisms of microwave neuromodulation using voltage imaging and channel blockers. Main results. We determined that inhibition occurs via nonthermal upregulation of TREK-1 channel activity while stimulation occurs via thermal activation of TRPV-1 channels. Significance. These findings build the foundation for developing microwave-based wireless neuromodulation devices for drug-free treatment of seizures and chronic pain. Additionally, they provide a basis for nonthermal microwave bioeffects which should be considered when setting exposure limits.
Integration of iEEG local microstate dynamics and neuro rank graph score for accurate SOZ localization
Nazar et al
View accepted manuscript
, Integration of iEEG local microstate dynamics and neuro rank graph score for accurate SOZ localization
PDF
, Integration of iEEG local microstate dynamics and neuro rank graph score for accurate SOZ localization
Objective. Accurate localization of the seizure onset zone (SOZ) is critical for successful surgical intervention in drug-resistant epilepsy. High Frequency Oscillation (HFO)-based SOZ localization methods often exhibit limited accuracy, which constrains their ability to reliably track large-scale and dynamic brain network activity. Microstate analysis captures dynamic spatiotemporal patterns of epileptogenic networks, improving localization. Propagation zone electrodes can mimic SOZ activity and complicate mapping. This study aims to enhance SOZ precision by combining microstate analysis with a Neuro Rank scoring system to eliminate false positives including propagation zone electrodes while prioritizing true SOZ regions. Approach. We developed a framework integrating high-frequency oscillation detection, microstate analysis, and Neuro Rank-based propagation zone refinement. HFOs were detected and characterized using time-frequency features and clustered via fuzzy c-means. Microstate analysis identified dominant spatiotemporal patterns using global field power peaks. Electrodes identified by both methods were refined using Neuro Rank, which integrates PageRank, reverse PageRank, and Hyperlink-Induced Topic Search to down-weight propagation zone electrodes while highlighting true SOZ activity. Main results. Using a leave-one-patient-out (LOPO) evaluation, the proposed pipeline combining HFO, microstate features, and NeuroRank achieved the best performance on the Freiburg dataset with 95.71% accuracy, 74.83% sensitivity, 98.56% specificity, and an Area Under the ROC Curve (AUC) of 0.83, while demonstrating strong generalization on the external HUP dataset with 96.34% accuracy, 72.34% sensitivity, 97.88% specificity, and an AUC of 0.80 Significance. Combining local HFO biomarkers, microstate-derived network dynamics, and Neuro Rank suppression of propagation zone electrodes provides precise and reliable SOZ localization, supporting improved surgical planning and outcomes in drug-resistant epilepsy.
The following article is
Open access
Temporal State-space model for forecasting slow-wave EEG power in non-human primates
Jiang et al
View accepted manuscript
, Temporal State-space model for forecasting slow-wave EEG power in non-human primates
PDF
, Temporal State-space model for forecasting slow-wave EEG power in non-human primates
Being able to accurately forecast brain activity over a prolonged period of time can help us establish a baseline of neural dynamics before external intervention. In this study, we developed an individualized time series forecasting framework based on a Trigonometric, Box-Cox transformation with ARMA errors and Trend and Seasonal/Periodic Components (TBATS), to predict slow-wave activity (SWA) power in non-human primates. Compared to a naive baseline and traditional methods such as Holt-Winters (HW) and Seasonal ARIMA (SARIMA), TBATS demonstrated comparable and even better out-of-sample accuracy while offering the flexibility to capture subject-specific latent temporal structure. These results support the use of TBATS as a data-efficient, interpretable tool for individualized forecasting of longitudinal EEG dynamics.
More Accepted manuscripts
The following article is
Open access
Bridging innovation and adoption: a mixed-methods investigation into stroke survivors’ acceptability of brain-computer interface-based rehabilitation interventions
E Grevet
et al
2026
J. Neural Eng.
23
026037
View article
, Bridging innovation and adoption: a mixed-methods investigation into stroke survivors’ acceptability of brain-computer interface-based rehabilitation interventions
PDF
, Bridging innovation and adoption: a mixed-methods investigation into stroke survivors’ acceptability of brain-computer interface-based rehabilitation interventions
Objective
. Brain-computer interface (BCI)-based interventions show growing evidence of efficacy for post-stroke upper-limb rehabilitation by closing the sensorimotor loop and promoting neuroplasticity. Despite this promise, their clinical uptake remains limited. This study aimed to identify the determinants of BCI acceptability among stroke survivors in order to inform user-centred neurotechnology design and facilitate clinical adoption.
Approach
. We conducted a mixed-methods study combining a large-scale questionnaire (
= 140) and semi-structured interviews (
= 12) with stroke survivors. The questionnaire was grounded in a validated theoretical model of BCI acceptability and assessed the determinants of three core constructs:
intention to use
(IU),
perceived usefulness
(PU), and
perceived ease of use
(PEOU). Qualitative data were analysed using thematic framework analysis to enrich and contextualise the quantitative findings.
Main Results
. Overall, stroke survivors reported high acceptability of BCI-based rehabilitation, with strong IU (mean IU = 8.48/10). Quantitative analyses showed that IU was primarily driven by PU, itself strongly influenced by perceived scientific relevance and, to a lesser extent, by individual factors such as autonomy, self-efficacy, and technology-related anxiety. PEOU was mainly determined by ease of learning and computer playfulness, but did not directly predict IU. Qualitative findings complemented these results by highlighting the importance of perceived innovativeness, the perception of directly engaging brain activity, visibility of progress, and clear scientific evidence. Interviews also emphasised the need for intuitive interfaces, clear instructions, short sessions, and appropriate therapeutic support to sustain engagement.
Significance
. These findings underscore that successful adoption of BCI-based rehabilitation requires more than technological performance alone. Enhancing acceptability among stroke survivors calls for: (i) goal-oriented and evidence-based rehabilitation protocols; (ii) clear communication of scientific rationale and outcomes to reinforce trust and social acceptance; and (iii) user-friendly system design that supports learning, autonomy, and self-efficacy while minimising cognitive and physical burden.
The following article is
Open access
Solving the problem of inception: a cross-species perspective on strategies for a mechanistic refinement of intracortical microstimulation
Takashi D Y Kozai
et al
2026
J. Neural Eng.
23
023001
View article
, Solving the problem of inception: a cross-species perspective on strategies for a mechanistic refinement of intracortical microstimulation
PDF
, Solving the problem of inception: a cross-species perspective on strategies for a mechanistic refinement of intracortical microstimulation
Objective.
Microstimulation delivers electrical pulses directly into the brain, with one of its promises being to restore lost senses to millions of people. Yet a fundamental challenge remains: how do intracortical microstimulation (ICMS) patterns engage neural circuits to achieve the inception of specific experiences, such as vivid sensory percepts of touch and vision? Here, we define ‘inception’ as the initiation of percepts evoked by microstimulation through the mapping of stimulation to circuit-level activity that results in sensory experiences.
Approach.
This perspective proposes an integrated research framework that combines Reverse Translation, Forward Translation, and computational neuroscience to bridge insights between clinical observations and high-resolution animal studies.
Framework.
Our framework envisions the development and evaluation of ICMS strategies within a cross-species system that narrows the range of plausible underlying neural mechanisms and the set of evoked perceptual outcomes. Reverse Translation uses human perceptual reports about phosphenes, tones, and touch to guide investigations in rodents and non-human primates, mapping the cell types and circuits underlying each percept. Forward Translation leverages these biological insights to design refined ICMS protocols for selective circuit engagement. Bidirectional Translation weaves these approaches together through computational neuroscience, ensuring that experimental observations iteratively and continuously refine one another across species and experimental modalities.
Significance.
This integrated strategy aims to transform microstimulation research into a dynamic dialogue between fundamental science and human experience. Harnessing the Bidirectional Translation Framework can accelerate therapies that enhance quality of life for people with sensory or motor impairments, and contribute more broadly to systems neuroscience by uncovering the mechanisms by which causal manipulation changes activity in neurons and networks.
The following article is
Open access
Bayesian time-history modeling enhances Parkinsonian motor state classification for adaptive deep brain stimulation
Brianna Leung
et al
2026
J. Neural Eng.
View article
, Bayesian time-history modeling enhances Parkinsonian motor state classification for adaptive deep brain stimulation
PDF
, Bayesian time-history modeling enhances Parkinsonian motor state classification for adaptive deep brain stimulation
Objective: Adaptive deep brain stimulation (aDBS) for Parkinson’s disease is a recently-approved therapy that adjusts stimulation in response to neurophysiologic biomarkers of motor-symptom state. Most real-time implementations of aDBS rely on instantaneous, noise-susceptible classifiers that apply simple thresholds to neurophysiologic biomarkers. We examined whether incorporating temporal history through Bayesian state-space modeling improved motor-state classification compared to instantaneous discriminant classifiers. Approach: We analyzed naturalistic neural data from three patients with Parkinson’s disease chronically implanted with investigational sensing-enabled DBS systems, recording from both the subthalamic nucleus (STN) and sensorimotor cortex. Biomarkers were extracted across multiple window lengths and labeled using wearable-derived bradykinesia and dyskinesia scores. Classifier behavior was evaluated using two biomarkers (cortical stimulation-entrained gamma and STN beta oscillations) across a factorial combination of two conditions: (1) instantaneous discriminant analysis vs. Bayesian time-history modeling via hidden Markov models (HMMs), and (2) single Gaussian vs. Gaussian mixture modeling of each motor state’s biomarker distribution. Performance metrics included F1 scores, accuracy, prediction smoothness, latency, and computational load. Main Results: Using entrained-gamma biomarkers, incorporating time history via HMMs significantly improved hyperkinetic-state detection (F1: +12.9 ± 1.8%; accuracy: +30.0 ± 2.7%; both padj < 0.001) with modest decreases in hypokinetic-state performance, yielding a net increase in average F1 (+4.7 ± 0.9%, p < 0.001). HMMs also yielded smoother and more accurate predictions for a given latency compared to simply increasing the window length used to extract neurophysiologic biomarkers. Entrained-gamma biomarkers outperformed STN beta biomarkers across all classifiers (average F1: +12.9% ± 0.5%, p < 0.001). All methods operated within sub-millisecond prediction times and demonstrated sublinear empirical computational scaling. Significance: Bayesian time-history modeling enhanced motor-state classification while preserving the low latency and computational efficiency required for real-time aDBS. These findings, derived from chronic at-home recordings, support the translational potential of Bayesian state-space models for next-generation aDBS systems.
The following article is
Open access
Temporal State-space model for forecasting slow-wave EEG power in non-human primates
Ruitong Jiang
et al
2026
J. Neural Eng.
View article
, Temporal State-space model for forecasting slow-wave EEG power in non-human primates
PDF
, Temporal State-space model for forecasting slow-wave EEG power in non-human primates
Being able to accurately forecast brain activity over a prolonged period of time can help us establish a baseline of neural dynamics before external intervention. In this study, we developed an individualized time series forecasting framework based on a Trigonometric, Box-Cox transformation with ARMA errors and Trend and Seasonal/Periodic Components (TBATS), to predict slow-wave activity (SWA) power in non-human primates. Compared to a naive baseline and traditional methods such as Holt-Winters (HW) and Seasonal ARIMA (SARIMA), TBATS demonstrated comparable and even better out-of-sample accuracy while offering the flexibility to capture subject-specific latent temporal structure. These results support the use of TBATS as a data-efficient, interpretable tool for individualized forecasting of longitudinal EEG dynamics.
The following article is
Open access
Adapting frozen foundation models for montage-agnostic high-resolution EEG event segmentation
Jun Ma and Tuukka Ruotsalo 2026
J. Neural Eng.
View article
, Adapting frozen foundation models for montage-agnostic high-resolution EEG event segmentation
PDF
, Adapting frozen foundation models for montage-agnostic high-resolution EEG event segmentation
Objective. Deploying brain-computer interfaces (BCIs) outside controlled laboratories requires detecting neural events in continuous electroencephalography (EEG) without relying on time-locked synchronization, while simultaneously generalizing across the diverse electrode montages encountered with different acquisition hardware. We investigate whether frozen EEG foundation models can be adapted to perform high temporal resolution event segmentation across unseen montages and datasets without subject-specific calibration. Approach. We introduce a lightweight, parameter-efficient preprocessing layer that interpolates learned channel embeddings based on electrode coordinates, enabling any frozen foundation model backbone to accept arbitrary montages. A shallow segmentation head is attached to produce a label every 4\,ms of continuous EEG, and overlapping predictions are consolidated via sliding-window majority voting. Main results. Evaluated on eight public corpora spanning P300, steady-state visually evoked potential (SSVEP) and motor imagery (MI) paradigms, our method consistently outperforms the original foundation models (BIOT, EEGPT) and classical baselines (EEGNet), achieving a mean macro F1 of 0.492 and Intersection over Union (IoU) of 0.361 in cross-subject evaluation, and F1\,=\,0.462, IoU\,=\,0.319 in calibration-free cross-dataset generalization. Significance. By decoupling the electrode montage from the pre-trained feature extractor through a plug-in adapter rather than massive retraining, our framework enables practical, resource-efficient BCI applications that operate without time-locked synchronization or montage-specific calibration, laying the groundwork for bridging the lab-to-field gap. The code and pre-processed datasets are available at: https://anonymous.4open.science/r/VewOdXnk669E17342jch-F1BD
The following article is
Open access
Multi-dimensional characterization and tracking of motor unit action potentials
Lara McManus
et al
2026
J. Neural Eng.
23
026033
View article
, Multi-dimensional characterization and tracking of motor unit action potentials
PDF
, Multi-dimensional characterization and tracking of motor unit action potentials
Objective.
Decomposition of high-density surface electromyography (HDsEMG) signals allows identification of individual motor unit firing times and provides a spatiotemporal image of their action potential waveforms. The ability to reliably match and track motor unit action potentials (MUAPs) from the same motor unit across multiple recordings allows changes in their recruitment and firing properties to be identified, however, similarities in MUAP shape can present challenges for reliable tracking.
Approach.
A new method for matching MUAP waveforms using a multi-dimensional (MD) representation is presented. MUAPs are represented as trajectories in high-dimensional space, where each HDsEMG channel corresponds to a different dimension. Trajectories are compared using MD features to measure the similarity between pairs of MUAP waveforms. Feature reduction and clustering are then used to classify pairs of MUAPs as belonging to the same or different motor units. The ability of the MD method to correctly identify pairs of matching MUAPs was assessed using MUAPs from simulated and experimental datasets and compared with two-dimensional cross-correlation (CC) using a threshold of 0.7, 0.8 or 0.9.
Main results.
The proposed MD method resulted in significantly higher F1 scores and lower false positive and false negative rates in both simulated and experimental datasets (
< 0.001). Across all datasets examined, the MD method correctly identified a greater number of matching MUAP pairs (89.8 ± 18.4%) compared with the best performing CC threshold (73.3 ± 21.0%). This was accompanied by a 49.6% lower false positive rate for the MD method.
Significance.
This study demonstrates that MD representations of MUAP trajectories recorded from high density arrays can more accurately identify MUAPs from the same motor unit, improving motor unit tracking compared with traditional correlation based approaches.
The following article is
Open access
Characterization of a flexible cap for simultaneous OPM-MEG and EEG measurements
Paul Anders
et al
2026
J. Neural Eng.
View article
, Characterization of a flexible cap for simultaneous OPM-MEG and EEG measurements
PDF
, Characterization of a flexible cap for simultaneous OPM-MEG and EEG measurements
Objective. Combining Magnetoencephalography (MEG) and Electroencephalography (EEG) can lead to more accurate source reconstruction results. Previous efforts of simultaneous EEG and MEG using optically pumped magnetometers (OPM-MEG) used a two-layer setup, where a helmet or cap with mounted OPMs was worn on top of an EEG cap, resulting in decreased OPM signal amplitudes. This is inevitable for rigid OPM-MEG helmets used by most research groups, which are often heavy and need fixation, require time-consuming customization or do not adapt to different head shapes, diminishing the advantages of OPM-MEG. In this study, we therefore aimed to create a flexible cap that enables fully integrated simultaneous OPM-MEG and EEG measurements, which are suitable for source reconstruction and yield the same signal quality as each individual modality. Approach. We integrated OPM mounts into a commercial EEG cap and implemented a sensor localization pipeline based on co-registering the known OPM mount geometry to 3D scans of participants wearing the cap. In a multi-paradigm experiment with five volunteers, we investigated how much OPMs shift during an experiment due to head movements of the participant. We also compared signal-to-noise ratios and artifact strengths between single-modality and multimodal measurements. Main results. The distribution of changes in OPM positions and orientations are within acceptable limits for source reconstruction as published in a simulation study. Occipital OPMs experienced larger shifts. No systematic difference in signal-to-noise ratios or artifact strengths between single-modality and multimodal acquisition could be observed, indicating no signal degradation due to the presence of the other modality. Significance. Measuring simultaneous OPM-MEG and EEG using a flexible cap could improve source reconstruction results and help to bring OPM-MEG to clinical settings due to easier comparison and verification with the established EEG, which is relevant for presurgical epilepsy evaluation.
The following article is
Open access
Detrended fluctuation analysis of amygdala-hippocampal beta synchrony reveals network rigidity in depression associated with temporal lobe epilepsy
Sebastian Hanna
et al
2026
J. Neural Eng.
View article
, Detrended fluctuation analysis of amygdala-hippocampal beta synchrony reveals network rigidity in depression associated with temporal lobe epilepsy
PDF
, Detrended fluctuation analysis of amygdala-hippocampal beta synchrony reveals network rigidity in depression associated with temporal lobe epilepsy
Objective Depressive symptoms are common in individuals with temporal lobe epilepsy (TLE), yet the network dynamics linking limbic circuitry to mood disturbance in TLE remain poorly understood. Here, we investigated whether the temporal organization of amygdala-hippocampal beta synchrony reflects depressive symptom burden in individuals with TLE undergoing intracranial EEG (iEEG) monitoring. Approach We analyzed iEEG recordings from 14 adults with TLE who underwent intracranial monitoring as part of routine pre-surgical evaluation. Dynamic functional connectivity (dFC) in the beta band was calculated between the amygdala and the hippocampus, and the temporal structure of the dFC was quantified using detrended fluctuation analysis. Random permutation of epochs was performed to assess the influence of interictal epileptiform discharges (IEDs) on the network dynamics. We additionally extracted the ultra-slow fluctuations in dFC and assessed whether the timing of the IEDs was phase-locked to these ultra-slow rhythms. Main results TLE patients with depression exhibited higher α than non-depressed TLE subjects (P = 0.007), indicating greater temporal persistence of limbic beta synchrony. Across individuals, α was positively associated with depressive symptom severity (r=0.73, P=0.003). Disrupting the temporal alignment of IED-containing epochs reduced α, whereas permuting IED-containing epochs among themselves preserved α, suggesting that the timing of IEDs contributes to the observed long-range temporal structure. In depressed TLE subjects, IEDs preferentially occurred during the rising phase of ultra-slow dFC fluctuations (P = 0.004). Significance Depressive symptoms in TLE are associated with more temporally persistent, less flexible beta band limbic network dynamics. These findings highlight the temporal organization of amygdala-hippocampal synchrony as a potential network-level signature of affective dysfunction within epilepsy and may inform future approaches to monitoring and modulating mood-related circuit activity in this population.
The following article is
Open access
Measures of fatigue and performance are related to user interface and task in a communication BCI
Daniel Klee
et al
2026
J. Neural Eng.
View article
, Measures of fatigue and performance are related to user interface and task in a communication BCI
PDF
, Measures of fatigue and performance are related to user interface and task in a communication BCI
Objective. This exploratory study compared two non-implantable Communication Brain-Computer Interfaces (cBCIs) to determine whether physiologic and self-report measures of mental fatigue, effort, and boredom were greater during calibration than during copy-spelling and whether there were differences between two common cBCI interfaces, Rapid Serial Visual Presentation (RSVP) and Single-Character Presentation Matrix (SCP-Matrix). Approach. Twenty-three healthy adults successfully utilized both RSVP and SCP-Matrix speller cBCIs in a single experimental session. Participants completed a calibration task and three online (closed-loop) copy-spelling tasks for each interface and provided self-report data on state mental fatigue, effort, and boredom. Physiological measures included EEG recordings alongside autonomic markers, including blood pressure, heart rate, respiration rate, and pulse rate variability (PRV). Main Results. Participants reported significant increases in perceived mental fatigue, effort, boredom, and sleepiness during the session, with significant increases during calibration compared to copy-spelling. On average, users typed 1.5 more correct characters per copy-spelling phase using the SCP-Matrix interface than when using RSVP. Results for autonomic and self-report metrics were consistent with fatigue being increased during calibration tasks relative to copy-spelling. EEG measures showed increased absolute and relative alpha activity and decreased relative theta activity during calibrations compared to copy-spelling, and increased absolute and relative alpha activity and decreased relative theta activity during RSVP, compared to Matrix. P300 amplitude on average was greater during copy spelling tasks than during calibrations. Significance. Participants demonstrated increased fatigue while using non-implantable cBCIs. Evidence suggested that calibration tasks for both interfaces were more fatiguing, required more mental effort, and were less engaging than copy-spelling tasks. Increased user fatigue and perceived mental effort remain significant barriers to sustained use of non-implantable cBCI systems. Though limited, the current study enhances our understanding of user experience with cBCIs and emphasizes the need to design more engaging and concise calibration procedures.
The following article is
Open access
A novel neurophysiological approach to evaluate the impact of virtual training on skills acquisition
Vincenzo Ronca
et al
2026
J. Neural Eng.
23
026032
View article
, A novel neurophysiological approach to evaluate the impact of virtual training on skills acquisition
PDF
, A novel neurophysiological approach to evaluate the impact of virtual training on skills acquisition
Objective.
Objective evaluation of operational training remains challenging, particularly when comparing physical training with virtual reality (VR) training. Behavioral and subjective measures capture performance and perceived workload but provide limited insight into the underlying neurocognitive adaptation. This study proposes and validates a synthetic EEG-derived cognitive training index (CTI) that integrates multiple training-sensitive neurometrics into a single composite metric, with two aims: to validate CTI as an objective marker of training progression in a physical training group, and to test whether VR-based training leads to neurocognitive adaptation comparable to physical training when participants subsequently execute the task in the real environment.
Approach.
Participants completed three repetitions of a realistic confined-space industrial maintenance procedure structured into Entrance, Rolling shutter, and Pipe replacement phases. Behavioral outcomes (execution time, procedural errors) and subjective workload NASA Task Load Index (NASA-TLX) were collected after each trial. Wearable EEG was recorded continuously; six neurometrics were computed from cleaned EEG using band-specific Global Field Power over predefined electrode sets. CTI was obtained via principal component analysis applied at the trial level to the multivariate neurometric time series, followed by segment-wise analysis. Temporal specificity of CTI was assessed through phase-randomized surrogate-data analyses.
Main results.
Execution time decreased across trials (
= 14.82,
< .001,
= 0.35), procedural errors decreased similarly (
= 10.31,
< .001,
= 0.27), and NASA-TLX declined (
= 11.94,
< .001,
= 0.30). CTI increased significantly with training (
= 18.21,
< .001,
= 0.39) and was selectively higher in learning-relevant phases (Entrance and Pipe replacement), while remaining near baseline in the low-demand Rolling shutter phase. Surrogate analyses showed that real CTI exceeded null expectations, supporting temporal specificity. Repeated-measures correlations indicated significant negative associations between CTI and procedural errors (
= − 0.42,
< .01) and between CTI and NASA-TLX (
= − 0.38,
< .01), while CTI was not significantly related to execution time (
= 0.18,
= .12).
Significance.
CTI provides a psychologically grounded, multivariate EEG marker that tracks training progression and aligns with operationally meaningful outcomes, namely procedural accuracy and perceived workload, rather than task speed alone. This approach supports objective comparison of cognitive adaptation across training modalities and provides a practical basis for neurophysiologically informed training evaluation and future adaptive training systems.
More Open Access articles
EEGNet: a compact convolutional neural network for EEG-based brain–computer interfaces
Vernon J Lawhern
et al
2018
J. Neural Eng.
15
056013
View article
, EEGNet: a compact convolutional neural network for EEG-based brain–computer interfaces
PDF
, EEGNet: a compact convolutional neural network for EEG-based brain–computer interfaces
Objective
. Brain–computer interfaces (BCI) enable direct communication with a computer, using neural activity as the control signal. This neural signal is generally chosen from a variety of well-studied electroencephalogram (EEG) signals. For a given BCI paradigm, feature extractors and classifiers are tailored to the distinct characteristics of its expected EEG control signal, limiting its application to that specific signal. Convolutional neural networks (CNNs), which have been used in computer vision and speech recognition to perform automatic feature extraction and classification, have successfully been applied to EEG-based BCIs; however, they have mainly been applied to single BCI paradigms and thus it remains unclear how these architectures generalize to other paradigms. Here, we ask if we can design a single CNN architecture to accurately classify EEG signals from different BCI paradigms, while simultaneously being as compact as possible.
Approach
. In this work we introduce EEGNet, a compact convolutional neural network for EEG-based BCIs. We introduce the use of depthwise and separable convolutions to construct an EEG-specific model which encapsulates well-known EEG feature extraction concepts for BCI. We compare EEGNet, both for within-subject and cross-subject classification, to current state-of-the-art approaches across four BCI paradigms: P300 visual-evoked potentials, error-related negativity responses (ERN), movement-related cortical potentials (MRCP), and sensory motor rhythms (SMR).
Main results
. We show that EEGNet generalizes across paradigms better than, and achieves comparably high performance to, the reference algorithms when only limited training data is available across all tested paradigms. In addition, we demonstrate three different approaches to visualize the contents of a trained EEGNet model to enable interpretation of the learned features.
Significance
. Our results suggest that EEGNet is robust enough to learn a wide variety of interpretable features over a range of BCI tasks. Our models can be found at:
The following article is
Open access
Deep learning for electroencephalogram (EEG) classification tasks: a review
Alexander Craik
et al
2019
J. Neural Eng.
16
031001
View article
, Deep learning for electroencephalogram (EEG) classification tasks: a review
PDF
, Deep learning for electroencephalogram (EEG) classification tasks: a review
Objective
. Electroencephalography (EEG) analysis has been an important tool in neuroscience with applications in neuroscience, neural engineering (e.g. Brain–computer interfaces, BCI’s), and even commercial applications. Many of the analytical tools used in EEG studies have used machine learning to uncover relevant information for neural classification and neuroimaging. Recently, the availability of large EEG data sets and advances in machine learning have both led to the deployment of deep learning architectures, especially in the analysis of EEG signals and in understanding the information it may contain for brain functionality. The robust automatic classification of these signals is an important step towards making the use of EEG more practical in many applications and less reliant on trained professionals. Towards this goal, a systematic review of the literature on deep learning applications to EEG classification was performed to address the following critical questions: (1) Which EEG classification tasks have been explored with deep learning? (2) What input formulations have been used for training the deep networks? (3) Are there specific deep learning network structures suitable for specific types of tasks?
Approach
. A systematic literature review of EEG classification using deep learning was performed on Web of Science and PubMed databases, resulting in 90 identified studies. Those studies were analyzed based on type of task, EEG preprocessing methods, input type, and deep learning architecture.
Main results
. For EEG classification tasks, convolutional neural networks, recurrent neural networks, deep belief networks outperform stacked auto-encoders and multi-layer perceptron neural networks in classification accuracy. The tasks that used deep learning fell into five general groups: emotion recognition, motor imagery, mental workload, seizure detection, event related potential detection, and sleep scoring. For each type of task, we describe the specific input formulation, major characteristics, and end classifier recommendations found through this review.
Significance
. This review summarizes the current practices and performance outcomes in the use of deep learning for EEG classification. Practical suggestions on the selection of many hyperparameters are provided in the hope that they will promote or guide the deployment of deep learning to EEG datasets in future research.
The following article is
Open access
Deep learning-based electroencephalography analysis: a systematic review
Yannick Roy
et al
2019
J. Neural Eng.
16
051001
View article
, Deep learning-based electroencephalography analysis: a systematic review
PDF
, Deep learning-based electroencephalography analysis: a systematic review
Context
. Electroencephalography (EEG) is a complex signal and can require several years of training, as well as advanced signal processing and feature extraction methodologies to be correctly interpreted. Recently, deep learning (DL) has shown great promise in helping make sense of EEG signals due to its capacity to learn good feature representations from raw data. Whether DL truly presents advantages as compared to more traditional EEG processing approaches, however, remains an open question.
Objective
. In this work, we review 154 papers that apply DL to EEG, published between January 2010 and July 2018, and spanning different application domains such as epilepsy, sleep, brain–computer interfacing, and cognitive and affective monitoring. We extract trends and highlight interesting approaches from this large body of literature in order to inform future research and formulate recommendations.
Methods
. Major databases spanning the fields of science and engineering were queried to identify relevant studies published in scientific journals, conferences, and electronic preprint repositories. Various data items were extracted for each study pertaining to (1) the data, (2) the preprocessing methodology, (3) the DL design choices, (4) the results, and (5) the reproducibility of the experiments. These items were then analyzed one by one to uncover trends.
Results
. Our analysis reveals that the amount of EEG data used across studies varies from less than ten minutes to thousands of hours, while the number of samples seen during training by a network varies from a few dozens to several millions, depending on how epochs are extracted. Interestingly, we saw that more than half the studies used publicly available data and that there has also been a clear shift from intra-subject to inter-subject approaches over the last few years. About
of the studies used convolutional neural networks (CNNs), while
used recurrent neural networks (RNNs), most often with a total of 3–10 layers. Moreover, almost one-half of the studies trained their models on raw or preprocessed EEG time series. Finally, the median gain in accuracy of DL approaches over traditional baselines was
across all relevant studies. More importantly, however, we noticed studies often suffer from poor reproducibility: a majority of papers would be hard or impossible to reproduce given the unavailability of their data and code.
Significance
. To help the community progress and share work more effectively, we provide a list of recommendations for future studies and emphasize the need for more reproducible research. We also make our summary table of DL and EEG papers available and invite authors of published work to contribute to it directly. A planned follow-up to this work will be an online public benchmarking portal listing reproducible results.
A review of classification algorithms for EEG-based brain–computer interfaces
F Lotte
et al
2007
J. Neural Eng.
R1
View article
, A review of classification algorithms for EEG-based brain–computer interfaces
PDF
, A review of classification algorithms for EEG-based brain–computer interfaces
In this paper we review classification algorithms used to design brain–computer interface (BCI) systems based on electroencephalography (EEG). We briefly present the commonly employed algorithms and describe their critical properties. Based on the literature, we compare them in terms of performance and provide guidelines to choose the suitable classification algorithm(s) for a specific BCI.
A comprehensive review of EEG-based brain–computer interface paradigms
Reza Abiri
et al
2019
J. Neural Eng.
16
011001
View article
, A comprehensive review of EEG-based brain–computer interface paradigms
PDF
, A comprehensive review of EEG-based brain–computer interface paradigms
Advances in brain science and computer technology in the past decade have led to exciting developments in brain–computer interface (BCI), thereby making BCI a top research area in applied science. The renaissance of BCI opens new methods of neurorehabilitation for physically disabled people (e.g. paralyzed patients and amputees) and patients with brain injuries (e.g. stroke patients). Recent technological advances such as wireless recording, machine learning analysis, and real-time temporal resolution have increased interest in electroencephalographic (EEG) based BCI approaches. Many BCI studies have focused on decoding EEG signals associated with whole-body kinematics/kinetics, motor imagery, and various senses. Thus, there is a need to understand the various experimental paradigms used in EEG-based BCI systems. Moreover, given that there are many available options, it is essential to choose the most appropriate BCI application to properly manipulate a neuroprosthetic or neurorehabilitation device. The current review evaluates EEG-based BCI paradigms regarding their advantages and disadvantages from a variety of perspectives. For each paradigm, various EEG decoding algorithms and classification methods are evaluated. The applications of these paradigms with targeted patients are summarized. Finally, potential problems with EEG-based BCI systems are discussed, and possible solutions are proposed.
Filter bank canonical correlation analysis for implementing a high-speed SSVEP-based brain–computer interface
Xiaogang Chen
et al
2015
J. Neural Eng.
12
046008
View article
, Filter bank canonical correlation analysis for implementing a high-speed SSVEP-based brain–computer interface
PDF
, Filter bank canonical correlation analysis for implementing a high-speed SSVEP-based brain–computer interface
Objective
. Recently, canonical correlation analysis (CCA) has been widely used in steady-state visual evoked potential (SSVEP)-based brain–computer interfaces (BCIs) due to its high efficiency, robustness, and simple implementation. However, a method with which to make use of harmonic SSVEP components to enhance the CCA-based frequency detection has not been well established.
Approach.
This study proposed a filter bank canonical correlation analysis (FBCCA) method to incorporate fundamental and harmonic frequency components to improve the detection of SSVEPs. A 40-target BCI speller based on frequency coding (frequency range: 8–15.8 Hz, frequency interval: 0.2 Hz) was used for performance evaluation. To optimize the filter bank design, three methods (M
: sub-bands with equally spaced bandwidths; M
: sub-bands corresponding to individual harmonic frequency bands; M
: sub-bands covering multiple harmonic frequency bands) were proposed for comparison. Classification accuracy and information transfer rate (ITR) of the three FBCCA methods and the standard CCA method were estimated using an offline dataset from 12 subjects. Furthermore, an online BCI speller adopting the optimal FBCCA method was tested with a group of 10 subjects.
Main results
. The FBCCA methods significantly outperformed the standard CCA method. The method M
achieved the highest classification performance. At a spelling rate of ∼33.3 characters/min, the online BCI speller obtained an average ITR of 151.18 ± 20.34 bits min
−1
Significance.
By incorporating the fundamental and harmonic SSVEP components in target identification, the proposed FBCCA method significantly improves the performance of the SSVEP-based BCI, and thereby facilitates its practical applications such as high-speed spelling.
EEG artifact removal—state-of-the-art and guidelines
Jose Antonio Urigüen and Begoña Garcia-Zapirain 2015
J. Neural Eng.
12
031001
View article
, EEG artifact removal—state-of-the-art and guidelines
PDF
, EEG artifact removal—state-of-the-art and guidelines
This paper presents an extensive review on the artifact removal algorithms used to remove the main sources of interference encountered in the electroencephalogram (EEG), specifically ocular, muscular and cardiac artifacts. We first introduce background knowledge on the characteristics of EEG activity, of the artifacts and of the EEG measurement model. Then, we present algorithms commonly employed in the literature and describe their key features. Lastly, principally on the basis of the results provided by various researchers, but also supported by our own experience, we compare the state-of-the-art methods in terms of reported performance, and provide guidelines on how to choose a suitable artifact removal algorithm for a given scenario. With this review we have concluded that, without prior knowledge of the recorded EEG signal or the contaminants, the safest approach is to correct the measured EEG using independent component analysis—to be precise, an algorithm based on second-order statistics such as second-order blind identification (SOBI). Other effective alternatives include extended information maximization (InfoMax) and an adaptive mixture of independent component analyzers (AMICA), based on higher order statistics. All of these algorithms have proved particularly effective with simulations and, more importantly, with data collected in controlled recording conditions. Moreover, whenever prior knowledge is available, then a constrained form of the chosen method should be used in order to incorporate such additional information. Finally, since which algorithm is the best performing is highly dependent on the type of the EEG signal, the artifacts and the signal to contaminant ratio, we believe that the optimal method for removing artifacts from the EEG consists in combining more than one algorithm to correct the signal using multiple processing stages, even though this is an option largely unexplored by researchers in the area.
The following article is
Open access
BCI for stroke rehabilitation: motor and beyond
Ravikiran Mane
et al
2020
J. Neural Eng.
17
041001
View article
, BCI for stroke rehabilitation: motor and beyond
PDF
, BCI for stroke rehabilitation: motor and beyond
Stroke is one of the leading causes of long-term disability among adults and contributes to major socio-economic burden globally. Stroke frequently results in multifaceted impairments including motor, cognitive and emotion deficits. In recent years, brain–computer interface (BCI)-based therapy has shown promising results for post-stroke motor rehabilitation. In spite of the success received by BCI-based interventions in the motor domain, non-motor impairments are yet to receive similar attention in research and clinical settings. Some preliminary encouraging results in post-stroke cognitive rehabilitation using BCI seem to suggest that it may also hold potential for treating non-motor deficits such as cognitive and emotion impairments. Moreover, past studies have shown an intricate relationship between motor, cognitive and emotion functions which might influence the overall post-stroke rehabilitation outcome. A number of studies highlight the inability of current treatment protocols to account for the implicit interplay between motor, cognitive and emotion functions. This indicates the necessity to explore an all-inclusive treatment plan targeting the synergistic influence of these standalone interventions. This approach may lead to better overall recovery than treating the individual deficits in isolation. In this paper, we review the recent advances in BCI-based post-stroke motor rehabilitation and highlight the potential for the use of BCI systems beyond the motor domain, in particular, in improving cognition and emotion of stroke patients. Building on the current results and findings of studies in individual domains, we next discuss the possibility of a holistic BCI system for motor, cognitive and affect rehabilitation which may synergistically promote restorative neuroplasticity. Such a system would provide an all-encompassing rehabilitation platform, leading to overarching clinical outcomes and transfer of these outcomes to a better quality of living. This is one of the first works to analyse the possibility of targeting cross-domain influence of post-stroke functional recovery enabled by BCI-based rehabilitation.
Multi-channel intramuscular and surface EMG decomposition by convolutive blind source separation
Francesco Negro
et al
2016
J. Neural Eng.
13
026027
View article
, Multi-channel intramuscular and surface EMG decomposition by convolutive blind source separation
PDF
, Multi-channel intramuscular and surface EMG decomposition by convolutive blind source separation
Objective.
The study of motor unit behavior has been classically performed by selective recording systems of muscle electrical activity (EMG signals) and decomposition algorithms able to discriminate between individual motor unit action potentials from multi-unit signals. In this study, we provide a general framework for the decomposition of multi-channel intramuscular and surface EMG signals and we extensively validate this approach with experimental recordings.
Approach.
First, we describe the conditions under which the assumptions of the convolutive blind separation model are satisfied. Second, we propose an approach of convolutive sphering of the observations followed by an iterative extraction of the sources. This approach is then validated using intramuscular signals recorded by novel multi-channel thin-film electrodes on the Abductor Digiti Minimi of the hand and Tibilias Anterior muscles, as well as on high-density surface EMG signals recorded by electrode grids on the First Dorsal Interosseous muscle. The validation was based on the comparison with the gold standard of manual decomposition (for intramuscular recordings) and on the two-source method (for comparison of intramuscular and surface EMG recordings) for the three human muscles and contraction forces of up to 90% MVC.
Main results.
The average number of common sources identified for the validation was 14 ± 7 (averaged across all trials and subjects and all comparisons), with a rate of agreement in their discharge timings of 92.8 ± 3.2%. The average Decomposability Index, calculated on the automatic decomposed signals, was 16.0 ± 2.2 (7.3–44.1). For comparison, the same index calculated on the manual decomposed signals was 15.0 ± 3.0 (6.3–76.6).
Significance.
These results show that the method provides a solid framework for the decomposition of multi-channel invasive and non-invasive EMG signals that allows the study of the behavior of a large number of concurrently active motor units.
The following article is
Open access
Large-scale neuromorphic computing systems
Steve Furber 2016
J. Neural Eng.
13
051001
View article
, Large-scale neuromorphic computing systems
PDF
, Large-scale neuromorphic computing systems
Neuromorphic computing covers a diverse range of approaches to information processing all of which demonstrate some degree of neurobiological inspiration that differentiates them from mainstream conventional computing systems. The philosophy behind neuromorphic computing has its origins in the seminal work carried out by Carver Mead at Caltech in the late 1980s. This early work influenced others to carry developments forward, and advances in VLSI technology supported steady growth in the scale and capability of neuromorphic devices. Recently, a number of large-scale neuromorphic projects have emerged, taking the approach to unprecedented scales and capabilities. These large-scale projects are associated with major new funding initiatives for brain-related research, creating a sense that the time and circumstances are right for progress in our understanding of information processing in the brain. In this review we present a brief history of neuromorphic engineering then focus on some of the principal current large-scale projects, their main features, how their approaches are complementary and distinct, their advantages and drawbacks, and highlight the sorts of capabilities that each can deliver to neural modellers.
Journal links
Submit an article
About the journal
Editorial Board
Author guidelines
Review for this journal
Publication charges
Awards
Journal collections
Pricing and ordering
Journal information
2004-present
Journal of Neural Engineering
doi: 10.1088/issn.1741-2552
Online ISSN: 1741-2552
Print ISSN: 1741-2560