Deep learning with convolutional neural networks for EEG decoding and visualization

GREEN: A lightweight architecture using learnable wavelets and Riemannian geometry for biomarker exploration with EEG signals

Spectral analysis using wavelets is widely used for identifying biomarkers in EEG signals. Recently, Riemannian geometry has provided an effective mathematical framework for predicting biomedical outcomes from multichannel electroencephalography (EEG) recordings while showing concord with neuroscientific domain knowledge. However, these methods rely on handcrafted rules and sequential optimization. In contrast, deep learning (DL) offers end-to-end trainable models achieving state-of-the-art performance on various prediction tasks but lacks interpretability and interoperability with established neuroscience concepts. We introduce Gabor Riemann EEGNet (GREEN), a lightweight neural network that integrates wavelet transforms and Riemannian geometry for processing raw EEG data. Benchmarking on six prediction tasks across four datasets with over 5,000 participants, GREEN outperformed non-deep state-of-the-art models and performed favorably against large DL models while using orders-of-magnitude fewer parameters. Computational experiments showed that GREEN facilitates learning sparse representations without compromising performance. By integrating domain knowledge, GREEN combines a desirable complexity-performance trade-off with interpretable representations.

The More, the Better? Evaluating the Role of EEG Preprocessing for Deep Learning Applications

The last decade has witnessed a notable surge in deep learning applications for electroencephalography (EEG) data analysis, showing promising improvements over conventional statistical techniques. However, deep learning models can underperform if trained with bad processed data. Preprocessing is crucial for EEG data analysis, yet there is no consensus on the optimal strategies in deep learning scenarios, leading to uncertainty about the extent of preprocessing required for optimal results. This study is the first to thoroughly investigate the effects of EEG preprocessing in deep learning applications, drafting guidelines for future research. It evaluates the effects of varying preprocessing levels, from raw and minimally filtered data to complex pipelines with automated artifact removal algorithms. Six classification tasks (eye blinking, motor imagery, Parkinson's, Alzheimer's disease, sleep deprivation, and first episode psychosis) and four established EEG architectures were considered for the evaluation. The analysis of 4800 trained models revealed statistical differences between preprocessing pipelines at the intra-task level for each model and at the inter-task level for the largest model. Models trained on raw data consistently performed poorly, always ranking last in average scores. In addition, models seem to benefit more from minimal pipelines without artifact handling methods. These findings suggest that EEG artifacts may affect the performance and generalizability of deep neural networks.

Cognitive load recognition in simulated flight missions: an EEG study

Cognitive load recognition (CLR) utilizing EEG signals has experienced significant advancement in recent years. However, current load-eliciting paradigms often rely on simplistic cognitive tasks such as arithmetic calculations, failing to adequately replicate real-world scenarios and lacking applicability. This study explores simulated flight missions over time to better reflect operational environments and investigate temporal dynamics of multiple load states. Thirty-six participants were recruited to perform simulated flight tasks with varying cognitive load levels of low, medium, and high. Throughout the experiments, we collected EEG load data from three sessions, pre- and post-task resting-state EEG data, subjective ratings, and objective performance metrics. Then, we employed several deep convolutional neural network (CNN) models, utilizing raw EEG data as model input, to assess cognitive load levels with six classification designs. Key findings from the study include (1) a notable distinction between resting-state and post-fatigue EEG data; (2) superior performance of shallow CNN models compared to more complex ones; and (3) temporal dynamics decline in CLR as the missions progressed. This paper establishes a potential foundation for assessing cognitive states during intricate simulated tasks across different individuals.

Different responses of MVL neurons when pigeons attend to local versus global information during object classification

Most prior studies have indicated that pigeons have a tendency to rely on local information for target categorization, yet there is a lack of electrophysiological evidence to support this claim. The mesopallium ventrolaterale (MVL) is believed to play a role in processing both local and global information during visual cognition. The difference between responses of MVL neurons when pigeons are focusing on local versus global information during visual object categorization remain unknown. In this study, pigeons were trained to categorize hierarchical stimuli that maintained consistency in local and global information. Subsequently, stimuli with different local and global components were presented to examine the pigeons' behavioral preferences. Not surprisingly, the behavioral findings revealed that pigeons predominantly attended to the local elements when performing categorization tasks. Moreover, MVL neurons exhibited significantly distinct responses when pigeons prioritized local versus global information. Specifically, most recording sites showed heightened gamma band power and increased nonlinear entropy values, indicating strong neural responses and rich information when pigeons concentrated on the local components of an object. Furthermore, neural population functional connectivity was weaker when the pigeons focused on local elements, suggesting that individual neurons operated more independently and effectively when focusing on local features. These findings offer electrophysiological evidence supporting the notion of pigeons displaying a behavioral preference for local information. The study provides valuable insight into the understanding of cognitive processes of pigeons when presented with complex objects, and further sheds light on the neural mechanisms underlying pigeons' behavioral preference for attending to local information.

Sparse knowledge sharing (SKS) for privacy-preserving domain incremental seizure detection

Objective. Epilepsy is a neurological disorder that affects millions of patients worldwide. Electroencephalogram-based seizure detection plays a crucial role in its timely diagnosis and effective monitoring. However, due to distribution shifts in patient data, existing seizure detection approaches are often patient-specific, which requires customized models for different patients. This paper considers privacy-preserving domain incremental learning (PP-DIL), where the model learns sequentially from each domain (patient) while only accessing the current domain data and previously trained models. This scenario has three main challenges: (1) catastrophic forgetting of previous domains, (2) privacy protection of previous domains, and (3) distribution shifts among domains.Approach. We propose a sparse knowledge sharing (SKS) approach. First, Euclidean alignment is employed to align data from different domains. Then, we propose an adaptive pruning approach for SKS to allocate subnet for each domain adaptively, allowing specific parameters to learn domain-specific knowledge while shared parameters to preserve knowledge from previous domains. Additionally, supervised contrastive learning is employed to enhance the model's ability to distinguish relevant features.Main Results. Experiments on two public seizure datasets demonstrated that SKS achieved superior performance in PP-DIL.Significance. SKS is a rehearsal-free privacy-preserving approach that effectively learns new domains while minimizing the impact on previously learned domains, achieving a better balance between plasticity and stability.

Distinguishing the activity of flexor digitorum brevis and soleus across standing postures with deep learning models

BACKGROUND

Electromyographic (EMG) recordings indicate that both the flexor digitorum brevis and soleus muscles contribute significantly to the control of standing balance, However, less is known about the adjustments in EMG activity of these two muscles across different postures.

RESEARCH QUESTION

The purpose of our study was to use deep-learning models to distinguish between the EMG activity of the flexor digitorum brevis and soleus muscles across four standing postures.

METHODS

Deep convolutional neural networks were employed to classify standing postures based on the temporal and spatial features embedded in high-density surface EMG signals. The EMG recordings were obtained with grid electrodes placed over the flexor digitorum brevis and soleus muscles of healthy young men during four standing tasks: bipedal, tandem, one-leg, and tip-toe.

RESULTS AND SIGNIFICANCE

Two-way repeated-measures analysis of variance demonstrated that the model achieved significantly greater classification accuracy, particularly during tandem stance, using EMG data from flexor digitorum brevis compared with soleus muscle. Average classification accuracy was 84.6 % for flexor digitorum brevis and 79.1 % for soleus. The classification accuracy of both muscles varied across the four postures. There were significant differences in classification accuracy for flexor digitorum brevis between bipedal and tandem stances compared with one-leg and tip-toe stances. In contrast, the EMG data for soleus were only significantly different between bipedal stance and one-leg stance. These findings indicate that flexor digitorum brevis exhibited more distinct adjustments than soleus in the temporo-spatial features of EMG activity across the four postures.

AdaptEEG: A Deep Subdomain Adaptation Network With Class Confusion Loss for Cross-Subject Mental Workload Classification

EEG signals exhibit non-stationary characteristics, particularly across different subjects, which presents significant challenges in the precise classification of mental workload levels when applying a trained model to new subjects. Domain adaptation techniques have shown effectiveness in enhancing the accuracy of cross-subject classification. However, current state-of-the-art methods for cross-subject mental workload classification primarily focus on global domain adaptation, which may lack fine-grained information and result in ambiguous classification boundaries. We proposed a novel approach called deep subdomain adaptation network with class confusion loss (DSAN-CCL) to enhance the performance of cross-subject mental workload classification. DSAN-CCL utilizes the local maximum mean discrepancy to align the feature distributions between the source domain and the target domain for each mental workload category. Moreover, the class confusion matrix was constructed by the product of the weighted class probabilities (class probabilities predicted by the label classifier) and the transpose of the class probabilities. The loss for maximizing diagonal elements and minimizing non-diagonal elements of the class confusion matrix was added to increase the credibility of pseudo-labels, thus improving the transfer performance. The proposed DSAN-CCL method was validated on two datasets, and the results indicate a significant improvement of 3∼10 percentage points compared to state-of-the-art domain adaptation methods. In addition, our proposed method is not dependent on a specific feature extractor. It can be replaced by any other feature extractor to fit new applications. This makes our approach universal to cross-domain classification problems.

EEG-Deformer: A Dense Convolutional Transformer for Brain-Computer Interfaces

Effectively learning the temporal dynamics in electroencephalogram (EEG) signals is challenging yet essential for decoding brain activities using brain-computer interfaces (BCIs). Although Transformers are popular for their long-term sequential learning ability in the BCI field, most methods combining Transformers with convolutional neural networks (CNNs) fail to capture the coarse-to-fine temporal dynamics of EEG signals. To overcome this limitation, we introduce EEG-Deformer, which incorporates two main novel components into a CNN-Transformer: (1) a Hierarchical Coarse-to-Fine Transformer (HCT) block that integrates a Fine-grained Temporal Learning (FTL) branch into Transformers, effectively discerning coarse-to-fine temporal patterns; and (2) a Dense Information Purification (DIP) module, which utilizes multi-level, purified temporal information to enhance decoding accuracy. Comprehensive experiments on three representative cognitive tasksâcognitive attention, driving fatigue, and mental workload detectionâconsistently confirm the generalizability of our proposed EEG-Deformer, demonstrating that it either outperforms or performs comparably to existing state-of-the-art methods. Visualization results show that EEG-Deformer learns from neurophysiologically meaningful brain regions for the corresponding cognitive tasks.

Decoding thoughts, encoding ethics: A narrative review of the BCI-AI revolution

OBJECTIVES

This narrative review aims to analyze mechanisms underlying Brain-Computer Interface (BCI) and Artificial Intelligence (AI) integration, evaluate recent advances in signal acquisition and processing techniques, and assess AI-enhanced neural decoding strategies. The review identifies critical research gaps and examines emerging solutions across multiple domains of BCI-AI integration.

METHODS

A narrative review was conducted using major biomedical and scientific databases including PubMed, Web of Science, IEEE Xplore, and Scopus (2014-2024). Literature was analyzed to identify key developments in BCI-AI integration, with particular emphasis on recent advances (2019-2024). The review process involved thematic analysis of selected publications focusing on practical applications, technical innovations, and emerging challenges.

RESULTS

Recent advances demonstrate significant improvements in BCI-AI systems: 1) High-density electrode arrays achieve spatial resolution up to 5 mm, with stable recordings over 15 months; 2) Deep learning decoders show 40 % improvement in information transfer rates compared to traditional methods; 3) Adaptive algorithms maintain >90 % success rates in motor control tasks over 200-day periods without recalibration; 4) Novel closed-loop optimization frameworks reduce user training time by 55 % while improving accuracy. Latest developments in flexible neural interfaces and self-supervised learning approaches show promise in addressing long-term stability and cross-user generalization challenges.

CONCLUSIONS

BCI-AI integration shows remarkable progress in improving signal quality, decoding accuracy, and user adaptation. While challenges remain in long-term stability and user training, advances in adaptive algorithms and feedback mechanisms demonstrate the technology's growing viability for clinical applications. Recent innovations in electrode technology, AI architectures, and closed-loop systems, combined with emerging standardization frameworks, suggest accelerating progress toward widespread therapeutic use and human augmentation applications.

ESI-GAL: EEG source imaging-based trajectory estimation for grasp and lift task

BACKGROUND

Electroencephalogram (EEG) signals-based motor kinematics prediction (MKP) has been an active area of research to develop Brain-computer interface (BCI) systems such as exosuits, prostheses, and rehabilitation devices. However, EEG source imaging (ESI) based kinematics prediction is sparsely explored in the literature.

METHOD

In this study, pre-movement EEG features are utilized to predict three-dimensional (3D) hand kinematics for the grasp-and-lift motor task. A public dataset, WAY-EEG-GAL, is utilized for MKP analysis. In particular, sensor-domain (EEG data) and source-domain (ESI data) based features from the frontoparietal region are explored for MKP. Deep learning-based models are explored to achieve efficient kinematics decoding. Various time-lagged and window sizes are analyzed for hand kinematics prediction. Subsequently, intra-subject and inter-subject MKP analysis is performed to investigate the subject-specific and subject-independent motor-learning capabilities of the neural decoders. The Pearson correlation coefficient (PCC) is used as the performance metric for kinematics trajectory decoding.

RESULTS

The rEEGNet neural decoder achieved the best performance with sensor-domain and source-domain features with the time lag and window size of 100ms and 450ms, respectively. The highest mean PCC values of 0.790, 0.795, and 0.637 are achieved using sensor-domain features, while 0.769, 0.777, and 0.647 are achieved using source-domain features in x, y, and z-directions, respectively.

CONCLUSION

This study explores the feasibility of trajectory prediction using EEG sensor-domain and source-domain features for the grasp-and-lift task. Furthermore, inter-subject trajectory estimation is performed using the proposed deep learning decoder with EEG source domain features.

Deep Learning for Lumbar Disc Herniation Diagnosis and Treatment Decision-Making Using Magnetic Resonance Imagings: A Retrospective Study

BACKGROUND

Lumbar disc herniation (LDH) is a common cause of back and leg pain. Diagnosis relies on clinical history, physical exam, and imaging, with magnetic resonance imaging (MRI) being an important reference standard. While artificial intelligence (AI) has been explored for MRI image recognition in LDH, existing methods often focus solely on disc herniation presence.

METHODS

We retrospectively analyzed MRI images from patients assessed for surgery by specialists. We then trained deep learning convolutional neural networks to detect LDH on MRI images. This study compared pure AI, pure human, and AI-assisted approaches for diagnosis accuracy and decision time. Statistical analysis evaluated each method's effectiveness.

RESULTS

Our approach demonstrated the potential of deep learning to aid LDH diagnosis and treatment. The AI-assisted group achieved the highest accuracy (94.7%), outperforming both pure AI and pure human approaches. AI integration reduced decision time without compromising accuracy.

CONCLUSIONS

Convolutional neural networks effectively assist specialists in initial LDH diagnosis and treatment decisions based on MRI images. This synergy between AI and human expertise improves diagnostic accuracy and efficiency, highlighting the value of AI-assisted diagnosis in clinical practice.

Response coupling with an auxiliary neural signal for enhancing brain signal detection

Brain-computer interfaces (BCIs) offer an implicit, non-linguistic communication channel between users and machines. Despite their potential, BCIs are far from becoming a mainstream communication modality like text and speech. While non-invasive BCIs, such as Electroencephalography, are favored for their ease of use, their broader adoption is limited by challenges related to signal noise, artifacts, and variability across users. In this paper, we propose a novel method called response coupling, aimed at enhancing brain signal detection and reliability by pairing a brain signal with an artificially induced auxiliary signal and leveraging their interaction. Specifically, we use error-related potentials (ErrPs) as the primary signal and steady-state visual evoked potentials (SSVEPs) as the auxiliary signal. SSVEPs, known for their phase-locked responses to rhythmic stimuli, are selected because rhythmic neural activity plays a critical role in sensory and cognitive processes, with evidence suggesting that reinforcing these oscillations can improve neural performance. By exploring the interaction between these two signals, we demonstrate that response coupling significantly improves the detection accuracy of ErrPs, especially in the parietal and occipital regions. This method introduces a new paradigm for enhancing BCI performance, where the interaction between a primary and an auxiliary signal is harnessed to enhance the detection performance. Additionally, the phase-locking properties of SSVEPs allow for unsupervised rejection of suboptimal data, further increasing BCI reliability.

Geometric neural network based on phase space for BCI-EEG decoding

Objective.The integration of Deep Learning (DL) algorithms on brain signal analysis is still in its nascent stages compared to their success in fields like Computer Vision. This is particularly true for Brain-computer interface (BCI), where the brain activity is decoded to control external devices without requiring muscle control. Electroencephalography is a widely adopted choice for designing BCI systems due to its non-invasive and cost-effective nature and excellent temporal resolution. Still, it comes at the expense of limited training data, poor signal-to-noise, and a large variability across and within-subject recordings. Finally, setting up a BCI system with many electrodes takes a long time, hindering the widespread adoption of reliable DL architectures in BCIs outside research laboratories. To improve adoption, we need to improve user comfort using, for instance, reliable algorithms that operate with few electrodes.Approach.Our research aims to develop a DL algorithm that delivers effective results with a limited number of electrodes. Taking advantage of the Augmented Covariance Method and the framework of SPDNet, we propose the Phase-SPDNet architecture and analyze its performance and the interpretability of the results. The evaluation is conducted on 5-fold cross-validation, using only three electrodes positioned above the Motor Cortex. The methodology was tested on nearly 100 subjects from several open-source datasets using the Mother Of All BCI Benchmark framework.Main results.The results of our Phase-SPDNet demonstrate that the augmented approach combined with the SPDNet significantly outperforms all the current state-of-the-art DL architecture in MI decoding.Significance.This new architecture is explainable and with a low number of trainable parameters.

Testing machine learning algorithms to evaluate fluctuating and cognitive profiles in Parkinson's disease by motion sensors and EEG data

Objective

We aimed to test machine learning algorithms for classifying fluctuating and cognitive profiles in Parkinson's Disease (PD) by using multimodal instrumental data.

Methods

Data of motion transducers while performing instrumented Timed-Up-and-Go test (iTUG) (N = 30 subjects) and EEG (N = 49 subjects) from PD patients were collected. Study patients were classified based on cognitive profile ("mild cognitive impairment" by standardized criteria vs "normal cognition") and L-dopa acute motor response ("fluctuating" vs "stable") to be analyzed by machine learning algorithms and compared with historical control data from healthy subjects group-matched by age for both iTUG and EEG study (for iTUG: N = 31 subjects; for EEG: N = 27 subjects).

Results

Artificial Neural Network-based models revealed the best performances when applied to specific phases of the iTUG in differentiating PD vs controls (91 % accuracy) as well as in differentiating cognitive profile (95 % accuracy) and motor response status (96 % accuracy) among PD subjects. K-Nearest Neighbors revealed best performances when applied to EEG data in discriminating PD vs controls (85 % accuracy). Random Forest Classifier revealed best performances when applied to EEG data in differentiating cognitive profile (96 % accuracy) and motor response status (91 % accuracy) among PD subjects.

Conclusions

By processing multimodal instrumental data, specific machine learning algorithms have been identified which discriminated L-dopa responsiveness and cognitive profile in PD. Further studies are needed to validate them in independent samples using a user-friendly software interface created ad hoc.

Hybrid BCI for Meal-Assist Robot Using Dry-Type EEG and Pupillary Light Reflex

Brain-computer interface (BCI)-based assistive technologies enable intuitive and efficient user interaction, significantly enhancing the independence and quality of life of elderly and disabled individuals. Although existing wet EEG-based systems report high accuracy, they suffer from limited practicality. This study presents a hybrid BCI system combining dry-type EEG-based flash visual-evoked potentials (FVEP) and pupillary light reflex (PLR) designed to control an LED-based meal-assist robot. The hybrid system integrates dry-type EEG and eyewear-type infrared cameras, addressing the preparation challenges of wet electrodes, while maintaining practical usability and high classification performance. Offline experiments demonstrated an average accuracy of 88.59% and an information transfer rate (ITR) of 18.23 bit/min across the four target classifications. Real-time implementation uses PLR triggers to initiate the meal cycle and EMG triggers to detect chewing, indicating the completion of the cycle. These features allow intuitive and efficient operation of the meal-assist robot. This study advances the BCI-based assistive technologies by introducing a hybrid system optimized for real-world applications. The successful integration of the FVEP and PLR in a meal-assisted robot demonstrates the potential for robust and user-friendly solutions that empower the users with autonomy and dignity in their daily activities.

Brain state forecasting for precise brain stimulation: Current approaches and future perspectives

Transcranial magnetic stimulation (TMS) has the potential to yield insights into cortical functions and improve the treatment of neurological and psychiatric conditions. However, its reliability is hindered by a low reproducibility of results. Among other factors, such low reproducibility is due to structural and functional variability between individual brains. Informing stimulation protocols with individual neuroimaging data could mitigate this issue, ensuring accurate targeting of structural brain areas and functional brain states in a subject-by-subject fashion. However, this process poses a set of theoretical and technical challenges. We focus on the problem of online functional targeting, which requires collecting electroencephalography (EEG) data, extracting brain states, and using them to trigger TMS in real time. This stream of operations introduces hardware and software delays in the real time set-up, such that brain states of interest may vanish before TMS delivery. To compensate for delays, it is necessary to process the EEG signal in real time, forecast it, and instruct TMS devices to target forecasted - rather than measured - brain states. Recently, this approach has been adopted successfully in a number of studies, opening interesting opportunities for personalised brain stimulation treatments. However, little has been done to explore and overcome the limitations of current forecasting methods. After reviewing the state of the art in brain state-dependent stimulation, we will discuss two broad classes of forecasting methods and their suitability for application to EEG time series. Subsequently, we will review the evidence in favour of data-driven forecasting and discuss its potential contributions to TMS methodology and the scientific understanding of brain dynamics, highlighting the transformative potential of big open datasets.

Enhancing detection of SSVEPs using discriminant compacted network

Objective. Steady-state visual evoked potential-based brain-computer interfaces (SSVEP-BCIs) have gained significant attention due to their simplicity, high signal to noise ratio and high information transfer rates (ITRs). Currently, accurate detection is a critical issue for enhancing the performance of SSVEP-BCI systems.Approach.This study proposed a novel decoding method called Discriminant Compacted Network (Dis-ComNet), which exploited the advantages of both spatial filtering and deep learning (DL). Specifically, this study enhanced SSVEP features using global template alignment and discriminant spatial pattern, and then designed a compacted temporal-spatio module (CTSM) to extract finer features. The proposed method was evaluated on a self-collected high-frequency dataset, a public Benchmark dataset and a public wearable dataset.Main Results.The results showed that Dis-ComNet significantly outperformed state-of-the-art spatial filtering methods, DL methods, and other fusion methods. Remarkably, Dis-ComNet improved the classification accuracy by 3.9%, 3.5%, 3.2%, 13.3%, 17.4%, 37.5%, and 2.5% when comparing with eTRCA, eTRCA-R, TDCA, DNN, EEGnet, Ensemble-DNN, and TRCA-Net respectively in the high-frequency dataset. The achieved results were 4.7%, 4.6%, 23.6%, 52.5%, 31.7%, and 7.0% higher than those of eTRCA, eTRCA-R, DNN, EEGnet, Ensemble-DNN, and TRCA-Net, respectively, and were comparable to those of TDCA in Benchmark dataset. The accuracy of Dis-ComNet in the wearable dataset was 9.5%, 7.1%, 36.1%, 26.3%, 15.7% and 4.7% higher than eTRCA, eTRCA-R, DNN, EEGnet, Ensemble-DNN, and TRCA-Net respectively, and comparable to TDCA. Besides, our model achieved the ITRs up to 126.0 bits/min, 236.4 bits/min and 103.6 bits/min in the high-frequency, Benchmark and the wearable datasets respectively.Significance.This study develops an effective model for the detection of SSVEPs, facilitating the development of high accuracy SSVEP-BCI systems.

User-wise perturbations for user identity protection in EEG-based BCIs

Objective. An electroencephalogram (EEG)-based brain-computer interface (BCI) is a direct communication pathway between the human brain and a computer. Most research so far studied more accurate BCIs, but much less attention has been paid to the ethics of BCIs. Aside from task-specific information, EEG signals also contain rich private information, e.g. user identity, emotion, disorders, etc which should be protected.Approach. We show for the first time that adding user-wise perturbations can make identity information in EEG unlearnable. We propose four types of user-wise privacy-preserving perturbations, i.e. random noise, synthetic noise, error minimization noise, and error maximization noise. After adding the proposed perturbations to EEG training data, the user identity information in the data becomes unlearnable, while the BCI task information remains unaffected.Main results. Experiments on six EEG datasets using three neural network classifiers and various traditional machine learning models demonstrated the robustness and practicability of the proposed perturbations.Significance. Our research shows the feasibility of hiding user identity information in EEG data without impacting the primary BCI task information.

Adaptive GCN and Bi-GRU-Based Dual Branch for Motor Imagery EEG Decoding

Decoding motor imagery electroencephalography (MI-EEG) signals presents significant challenges due to the difficulty in capturing the complex functional connectivity between channels and the temporal dependencies of EEG signals across different periods. These challenges are exacerbated by the low spatial resolution and high signal redundancy inherent in EEG signals, which traditional linear models struggle to address. To overcome these issues, we propose a novel dual-branch framework that integrates an adaptive graph convolutional network (Adaptive GCN) and bidirectional gated recurrent units (Bi-GRUs) to enhance the decoding performance of MI-EEG signals by effectively modeling both channel correlations and temporal dependencies. The Chebyshev Type II filter decomposes the signal into multiple sub-bands giving the model frequency domain insights. The Adaptive GCN, specifically designed for the MI-EEG context, captures functional connectivity between channels more effectively than conventional GCN models, enabling accurate spatial-spectral feature extraction. Furthermore, combining Bi-GRU and Multi-Head Attention (MHA) captures the temporal dependencies across different time segments to extract deep time-spectral features. Finally, feature fusion is performed to generate the final prediction results. Experimental results demonstrate that our method achieves an average classification accuracy of 80.38% on the BCI-IV Dataset 2a and 87.49% on the BCI-I Dataset 3a, outperforming other state-of-the-art decoding approaches. This approach lays the foundation for future exploration of personalized and adaptive brain-computer interface (BCI) systems.

The classification of absence seizures using power-to-power cross-frequency coupling analysis with a deep learning network

High frequency oscillations are important novel biomarkers of epileptic tissue. The interaction of oscillations across different time scales is revealed as cross-frequency coupling (CFC) representing a high-order structure in the functional organization of brain rhythms. Power-to-power coupling (PPC) is one form of coupling with significant research attesting to its neurobiological significance as well as its computational efficiency, yet has been hitherto unexplored within seizure classification literature. New artificial intelligence methods such as deep learning neural networks can provide powerful tools for automated analysis of EEG. Here we present a Stacked Sparse Autoencoder (SSAE) trained to classify absence seizure activity based on this important form of cross-frequency patterns within scalp EEG. The analysis is done on the EEG records from the Temple University Hospital database. Absence seizures (n = 94) from 12 patients were taken into analysis along with segments of background activity. Power-to-power coupling was calculated between all frequencies 2-120 Hz pairwise using the EEGLAB toolbox. The resulting CFC matrices were used as training or testing inputs to the autoencoder. The trained network was able to recognize background and seizure segments (not used in training) with a sensitivity of 93.1%, specificity of 99.5% and overall accuracy of 96.8%. The results provide evidence both for (1) the relevance of PPC for seizure classification, as well as (2) the efficacy of an approach combining PPC with SSAE neural networks for automated classification of absence seizures within scalp EEG.

Pattern Identification in Patients with Functional Dyspepsia Using Brain-Body Bio-Signals: Protocol of a Clinical Trial for AI Algorithm Development

Background: Functional dyspepsia (FD) is a common functional gastrointestinal disorder characterized by chronic digestive symptoms without identifiable structural abnormalities. FD affects approximately 8-46% of the population, leading to significant socioeconomic burdens due to reduced quality of life and productivity. Traditional medicine utilizes differential diagnosis through comprehensive examinations, which include observing and questioning, abdominal examination, and pulse diagnosis for functional gastrointestinal disorders. However, challenges persist in the standardization and objectivity of diagnostic protocols. Methods: This study aims to develop an artificial intelligence-based algorithm to predict identified patterns in patients with functional dyspepsia by integrating brain-body bio-signals, including brain activity measured by functional near-infrared spectroscopy, pulse wave, skin conductance response, and electrocardiography. We will conduct an observational cross-sectional study comprising 100 patients diagnosed according to the Rome IV criteria, collecting bio-signal data alongside differential diagnoses performed by licensed Korean medicine doctors. The study protocol was reviewed and approved by the Institutional Review Board of Kyung Hee University Hospital at Gangdong on 25 January 2024 (IRB no. KHNMCOH 2023-12-003-003) and was registered in the Korean Clinical Trial Registry (KCT0009275). Results: By creating AI algorithms based on bio-signals and integrating them into clinical practice, the objectivity and reliability of traditional diagnostics are expected to be enhanced. Conclusions: The integration of bio-signal analysis into the diagnostic process for patients with FD will improve clinical practices and support the broader acceptance of traditional-medicine diagnostic processes in healthcare.

MDD-SSTNet: detecting major depressive disorder by exploring spectral-spatial-temporal information on resting-state electroencephalography data based on deep neural network

Major depressive disorder (MDD) is a psychiatric disorder characterized by persistent lethargy that can lead to suicide in severe cases. Hence, timely and accurate diagnosis and treatment are crucial. Previous neuroscience studies have demonstrated that major depressive disorder subjects exhibit topological brain network changes and different temporal electroencephalography (EEG) characteristics compared to healthy controls. Based on these phenomena, we proposed a novel model, termed as MDD-SSTNet, for detecting major depressive disorder by exploring spectral-spatial-temporal information from resting-state EEG with deep convolutional neural network. Firstly, MDD-SSTNet used the Sinc filter to obtain specific frequency band features from pre-processed EEG data. Secondly, two parallel branches were used to extract temporal and spatial features through convolution and other operations. Finally, the model was trained with a combined loss function of center loss and Binary Cross-Entropy Loss. Using leave-one-subject-out cross-validation on the HUSM dataset and MODMA dataset, the MDD-SSTNet model outperformed six baseline models, achieving average classification accuracies of 93.85% and 65.08%, respectively. These results indicate that MDD-SSTNet could effectively mine spatial-temporal difference information between major depressive disorder subjects and healthy control subjects, and it holds promise to provide an efficient approach for MDD detection with EEG data.

TFTL: A Task-Free Transfer Learning Strategy for EEG-Based Cross-Subject and Cross-Dataset Motor Imagery BCI

OBJECTIVE

Motor imagery-based brain-computer interfaces (MI-BCIs) have been playing an increasingly vital role in neural rehabilitation. However, the long-term task-based calibration required for enhanced model performance leads to an unfriendly user experience, while the inadequacy of EEG data hinders the performance of deep learning models. To address these challenges, a task-free transfer learning strategy (TFTL) for EEG-based cross-subject & cross-dataset MI-BCI is proposed for calibration time reduction and multi-center data co-modeling.

METHODS

TFTL strategy consists of data alignment, shared feature extractor, and specific classifiers, in which the label predictor for MI tasks classification, as well as domain and dataset discriminator for inter-subject variability reduction are concurrently optimized for knowledge transfer from subjects across different datasets to the target subject. Moreover, only resting data of the target subject is used for subject-specific model construction to achieve task-free.

RESULTS

We employed three deep learning methods (ShallowConvNet, EEGNet, and TCNet-Fusion) as baseline approaches to evaluate the effectiveness of the proposed strategy on five datasets (BCIC IV Dataset 2a, Dataset 1, Physionet MI, Dreyer 2023, and OpenBMI). The results demonstrate a significant improvement with the inclusion of the TFTL strategy compared to the baseline methods, reaching a maximum enhancement of 15.67% with a statistical significance (p = 2.4e-5 < 0.05). Moreover, task-free resulted in MI trials needed for calibration being 0 for all datasets, which significantly alleviated the calibration burden for patients before usage.

CONCLUSION/SIGNIFICANCE

The proposed TFTL strategy effectively addresses challenges posed by prolonged calibration periods and insufficient EEG data, thus promoting MI-BCI from laboratory to clinical application.

A protocol for trustworthy EEG decoding with neural networks

Deep learning solutions have rapidly emerged for EEG decoding, achieving state-of-the-art performance on a variety of decoding tasks. Despite their high performance, existing solutions do not fully address the challenge posed by the introduction of many hyperparameters, defining data pre-processing, network architecture, network training, and data augmentation. Automatic hyperparameter search is rarely performed and limited to network-related hyperparameters. Moreover, pipelines are highly sensitive to performance fluctuations due to random initialization, hindering their reliability. Here, we design a comprehensive protocol for EEG decoding that explores the hyperparameters characterizing the entire pipeline and that includes multi-seed initialization for providing robust performance estimates. Our protocol is validated on 9 datasets about motor imagery, P300, SSVEP, including 204 participants and 26 recording sessions, and on different deep learning models. We accompany our protocol with extensive experiments on the main aspects influencing it, such as the number of participants used for hyperparameter search, the split into sequential simpler searches (multi-step search), the use of informed vs. non-informed search algorithms, and the number of random seeds for obtaining stable performance. The best protocol included 2-step hyperparameter search via an informed search algorithm, with the final training and evaluation performed using 10 random initializations. The optimal trade-off between performance and computational time was achieved by using a subset of 3-5 participants for hyperparameter search. Our protocol consistently outperformed baseline state-of-the-art pipelines, widely across datasets and models, and could represent a standard approach for neuroscientists for decoding EEG in a trustworthy and reliable way.

Deep Multiview Module Adaption Transfer Network for Subject-Specific EEG Recognition

Transfer learning is one of the popular methods to solve the problem of insufficient data in subject-specific electroencephalogram (EEG) recognition tasks. However, most existing approaches ignore the difference between subjects and transfer the same feature representations from source domain to different target domains, resulting in poor transfer performance. To address this issue, we propose a novel subject-specific EEG recognition method named deep multiview module adaption transfer (DMV-MAT) network. First, we design a universal deep multiview (DMV) network to generate different types of discriminative features from multiple perspectives, which improves the generalization performance by extensive feature sets. Second, module adaption transfer (MAT) is designed to evaluate each module by the feature distributions of source and target samples, which can generate an optimal weight sharing strategy for each target subject and promote the model to learn domain-invariant and domain-specific features simultaneously. We conduct extensive experiments in two EEG recognition tasks, i.e., motor imagery (MI) and seizure prediction, on four datasets. Experimental results demonstrate that the proposed method achieves promising performance compared with the state-of-the-art methods, indicating a feasible solution for subject-specific EEG recognition tasks. Implementation codes are available at https://github.com/YangLibuaa/DMV-MAT.

Lightweight Source-Free Domain Adaptation Based on Adaptive Euclidean Alignment for Brain-Computer Interfaces

For privacy protection of subjects in electroencephalogram (EEG)-based brain-computer interfaces (BCIs), using source-free domain adaptation (SFDA) for cross-subject recognition has proven to be highly effective. However, updating and storing a model trained on source subjects for each new subject can be inconvenient. This paper extends Euclidean alignment (EA) to propose adaptive Euclidean alignment (AEA), which learns a projection matrix to align the distribution of the target subject with the source subjects, thus eliminating domain drift issues and improving model classification performance of subject-independent BCIs. Combining the proposed AEA with various existing SFDA methods, such as SHOT, GSFDA, and NRC, this paper presents three new methods: AEA-SHOT, AEA-GSFDA, and AEA-NRC. In our experimental studies, these AEA-based SFDA methods were applied to four well-known deep learning models (i.e., EEGNet, Shallow ConvNet, Deep ConvNet, and MSFBCNN) on two motor imagery (MI) datasets, one event-related potential (ERP) dataset and one steady-state visual evoked potentials (SSVEP) dataset. The advanced cross-subject EEG classification performance demonstrates the efficacy of our proposed methods. For example, AEA-SHOT achieved the best average accuracy of 81.4% on the PhysioNet dataset.

Single-channel electroencephalography decomposition by detector-atom network and its pre-trained model

Signal decomposition techniques utilizing multi-channel spatial features are critical for analyzing, denoising, and classifying electroencephalography (EEG) signals. To facilitate the decomposition of signals recorded with limited channels, this paper presents a novel single-channel decomposition approach that does not rely on multi-channel features. Our model posits that an EEG signal comprises short, shift-invariant waves, referred to as atoms. We design a decomposer as an artificial neural network aimed at estimating these atoms and detecting their time shifts and amplitude modulations within the input signal. The efficacy of our method was validated across various scenarios in brain-computer interfaces and neuroscience, demonstrating enhanced performance. Additionally, cross-dataset validation indicates the feasibility of a pre-trained model, enabling a plug-and-play signal decomposition module.

Multiclass Classification Framework of Motor Imagery EEG by Riemannian Geometry Networks

In motor imagery (MI) tasks for brain computer interfaces (BCIs), the spatial covariance matrix (SCM) of electroencephalogram (EEG) signals plays a critical role in accurate classification. Given that SCMs are symmetric positive definite (SPD), Riemannian geometry is widely utilized to extract classification features. However, calculating distances between SCMs is computationally intensive due to operations like eigenvalue decomposition, and classical optimization techniques, such as gradient descent, cannot be directly applied to Riemannian manifolds, making the computation of the Riemannian mean more complex and reliant on iterative methods or approximations. In this paper, we propose a novel multiclass classification framework that integrates Riemannian geometry and neural networks to mitigate these challenges. The framework comprises two modules: a Riemannian module with multiple branches and a classification module. During training, a fusion loss function is introduced to update the branch corresponding to the true label, while other branches are updated using different loss functions along with the classification module. Comprehensive experiments on four sets of MI EEG data demonstrate the efficiency and effectiveness of the proposed model.

Working-memory load decoding model inspired by brain cognition based on cross-frequency coupling

Working memory, a fundamental cognitive function of the brain, necessitates the evaluation of cognitive load intensity due to limited cognitive resources. Optimizing cognitive load can enhance task performance efficiency by preventing resource waste and overload. Therefore, identifying working memory load is an essential area of research. Deep learning models have demonstrated remarkable potential in identifying the intensity of working memory load. However, existing neural networks based on electroencephalogram (EEG) decoding primarily focus on temporal and spatial characteristics while neglecting frequency characteristics. Drawing inspiration from the role of cross-frequency coupling in the hippocampal region, which plays a crucial role in advanced cognitive processes such as working memory, this study proposes a Multi-Band Multi-Scale Hybrid Sinc Convolutional Neural Network (MBSincNex). This model integrates multi-frequency and multi-scale Sinc convolution to facilitate time-frequency conversion and extract time-frequency information from multiple rhythms and regions of the EEG data with the aim of effectively model the cross-frequency coupling across different cognitive domains. Due to its unique structural design, the proposed model proficiently extracts features in temporal, frequency, and spatial domains while its feature extraction capability is validated through post-hoc interpretability techniques. On our collected three-class working memory dataset, the proposed model achieved higher classification accuracy compared to other state-of-the-art methods. Furthermore, by analyzing the model's classification performance during different stages of working memory processes, this study emphasizes the significance of the encoding phase and confirms that behavioral response does not accurately reflect cognitive load.

MI-Mamba: A hybrid motor imagery electroencephalograph classification model with Mamba's global scanning

Deep learning has revolutionized electroencephalograph (EEG) decoding, with convolutional neural networks (CNNs) being a predominant tool. However, CNNs struggle with long-term dependencies in sequential EEG data. Models like long short-term memory and transformers improve performance but still face challenges of computational efficiency and long sequences. Mamba, a state space model-based method, excels in modeling long sequences. To overcome the limitations of existing EEG decoding models and exploit Mamba's potential in EEG analysis, we propose MI-Mamba, a model integrating CNN with Mamba for motor imagery (MI) data decoding. MI-Mamba processes multi-channel EEG signals through a single convolutional layer to capture spatial features in the local temporal domain, followed by a Mamba module that processes global temporal features. A fully connected, layer-based classifier is used to derive classification results. Evaluated on two public MI datasets, MI-Mamba achieves 80.59% accuracy in the four-class MI task of the BCI Competition IV 2a dataset and 84.42% in the two-class task of the BCI Competition IV 2b dataset, while reducing parameter count by nearly six times compared to the most advanced previous models. These results highlight MI-Mamba's effectiveness in MI decoding and its potential as a new backbone for general EEG decoding.

A Survey of Few-Shot Learning for Biomedical Time Series

Advancements in wearable sensor technologies and the digitization of medical records have contributed to the unprecedented ubiquity of biomedical time series data. Data-driven models have tremendous potential to assist clinical diagnosis and improve patient care by improving long-term monitoring capabilities, facilitating early disease detection and intervention, as well as promoting personalized healthcare delivery. However, accessing extensively labeled datasets to train data-hungry deep learning models encounters many barriers, such as long-tail distribution of rare diseases, cost of annotation, privacy and security concerns, data-sharing regulations, and ethical considerations. An emerging approach to overcome the scarcity of labeled data is to augment AI methods with human-like capabilities to leverage past experiences to learn new tasks with limited examples, called few-shot learning. This survey provides a comprehensive review and comparison of few-shot learning methods for biomedical time series applications. The clinical benefits and limitations of such methods are discussed in relation to traditional data-driven approaches. This paper aims to provide insights into the current landscape of few-shot learning for biomedical time series and its implications for future research and applications.

Non-Invasive Brain-Computer Interfaces: State of the Art and Trends

Brain-computer interface (BCI) is a rapidly evolving technology that has the potential to widely influence research, clinical and recreational use. Non-invasive BCI approaches are particularly common as they can impact a large number of participants safely and at a relatively low cost. Where traditional non-invasive BCIs were used for simple computer cursor tasks, it is now increasingly common for these systems to control robotic devices for complex tasks that may be useful in daily life. In this review, we provide an overview of the general BCI framework as well as the various methods that can be used to record neural activity, extract signals of interest, and decode brain states. In this context, we summarize the current state-of-the-art of non-invasive BCI research, focusing on trends in both the application of BCIs for controlling external devices and algorithm development to optimize their use. We also discuss various open-source BCI toolboxes and software, and describe their impact on the field at large.

CLTNet: A Hybrid Deep Learning Model for Motor Imagery Classification

BACKGROUND

Brain-computer interface (BCI) technology opens up new avenues for human-machine interaction and rehabilitation by connecting the brain to machines. Electroencephalography (EEG)-based motor imagery (MI) classification is a key component of BCI technology, which is capable of translating neural activity in the brain into commands for controlling external devices. Despite the great potential of BCI technology, the challenges of extracting and decoding brain signals limit its wide application.

METHODS

To address this challenge, this study proposes a novel hybrid deep learning model, CLTNet, which focuses on solving the feature extraction problem to improve the classification of MI-EEG signals. In the preliminary feature extraction stage, CLTNet uses a convolutional neural network (CNN) to extract time series, channel, and spatial features of EEG signals to obtain important local information. In the deep feature extraction stage, the model combines the long short-term memory (LSTM) network and the Transformer module to capture time-series data and global dependencies in the EEG. The LSTM explains the dynamics of the brain activity, while the Transformer's self-attention mechanism reveals the global features of the time series. Ultimately, the CLTNet model classifies motor imagery EEG signals through a fully connected layer.

RESULTS

The model achieved an average accuracy of 83.02% and a Kappa value of 0.77 on the BCI IV 2a dataset, and 87.11% and a Kappa value of 0.74 on the BCI IV 2b dataset, both of which outperformed the traditional methods.

CONCLUSIONS

The innovation of the CLTNet model is that it integrates multiple network architectures, which offers a more comprehensive understanding of the characteristics of the EEG signals during motor imagery, providing a more comprehensive perspective and establishing a new benchmark for future research in this area.

Stochasticity as a solution for overfitting-A new model and comparative study on non-invasive EEG prospects

The potential and utility of inner speech is pivotal for developing practical, everyday Brain-Computer Interface (BCI) applications, as it represents a type of brain signal that operates independently of external stimuli however it is largely underdeveloped due to the challenges faced in deciphering its signals. In this study, we evaluated the behaviors of various Machine Learning (ML) and Deep Learning (DL) models on a publicly available dataset, employing popular preprocessing methods as feature extractors to enhance model training. We face significant challenges like subject-dependent variability, high noise levels, and overfitting. To address overfitting in particular, we propose using "BruteExtraTree": a new classifier which relies on moderate stochasticity inherited from its base model, the ExtraTreeClassifier. This model not only matches the best DL model, ShallowFBCSPNet, in the subject-independent scenario in our experiments scoring 32% accuracy, but also surpasses the state-of-the-art by achieving 46.6% average per-subject accuracy in the subject-dependent case. Our results on the subject-dependent case show promise on the possibility of a new paradigm for using inner speech data inspired from LLM pretraining but we also highlight the crucial need for a drastic change in data recording or noise removal methods to open the way for more practical accuracies in the subject-independent case.

Determination of the Time-frequency Features for Impulse Components in EEG Signals

Accurately identifying the timing and frequency characteristics of impulse components in EEG signals is essential but limited by the Heisenberg uncertainty principle. Inspired by the visual system's ability to identify objects and their locations, we propose a new method that integrates a visual system model with wavelet analysis to calculate both time and frequency features of local impulses in EEG signals. We develop a mathematical model based on invariant pattern recognition by the visual system, combined with wavelet analysis using Krawtchouk functions as the mother wavelet. Our method precisely identifies the localization and frequency characteristics of the impulse components in EEG signals. Tested on task-related EEG data, it accurately detected blink components (0.5 to 1 Hz) and separated muscle artifacts (16 Hz). It also identified muscle response durations (298 ms) within the 1 to 31 Hz range in emotional reaction studies, offering insights into both individual and typical emotional responses. We further illustrated how the new method circumvents the uncertainty principle in low-frequency wavelet analysis. Unlike classical wavelet analysis, our method provides spectral characteristics of EEG impulses invariant to time shifts, improving the identification and classification of EEG components.

The 'Sandwich' meta-framework for architecture agnostic deep privacy-preserving transfer learning for non-invasive brainwave decoding

Objective. Machine learning has enhanced the performance of decoding signals indicating human behaviour. Electroencephalography (EEG) brainwave decoding, as an exemplar indicating neural activity and human thoughts non-invasively, has been helpful in neural activity analysis and aiding paralysed patients via brain-computer interfaces. However, training machine learning algorithms on EEG encounters two primary challenges: variability across data sets and privacy concerns using data from individuals and data centres. Our objective is to address these challenges by integrating transfer learning for data variability and federated learning for data privacy into a unified approach.Approach. We introduce the 'Sandwich' as a novel deep privacy-preserving meta-framework combining transfer learning and federated learning. The 'Sandwich' framework comprises three components: federated networks (first layers) that handle data set differences at the input level, a shared network (middle layer) learning common rules and applying transfer learning techniques, and individual classifiers (final layers) for specific brain tasks of each data set. This structure enables the central network (central server) to benefit from multiple data sets, while local branches (local servers) maintain data and label privacy.Main results. We evaluated the 'Sandwich' meta-architecture in various configurations using the BEETL motor imagery challenge, a benchmark for heterogeneous EEG data sets. Compared with baseline models likeShallow ConvNetandEEGInception, our 'Sandwich' implementations showed superior performance. The best-performing model, the Inception SanDwich with deep set alignment (Inception-SD-Deepset), exceeded baseline methods by 9%.Significance. The 'Sandwich' framework demonstrates advancements in federated deep transfer learning for diverse tasks and data sets. It outperforms conventional deep learning methods, showcasing the potential for effective use of larger, heterogeneous data sets with enhanced privacy. In addition, through its diverse implementations with various backbone architectures and transfer learning approaches, the 'Sandwich' framework shows the potential as a model-agnostic meta-framework for decoding time series data like EEG, suggesting a direction towards large-scale brainwave decoding by combining deep transfer learning with privacy-preserving federated learning.

AMEEGNet: attention-based multiscale EEGNet for effective motor imagery EEG decoding

Recently, electroencephalogram (EEG) based on motor imagery (MI) have gained significant traction in brain-computer interface (BCI) technology, particularly for the rehabilitation of paralyzed patients. But the low signal-to-noise ratio of MI EEG makes it difficult to decode effectively and hinders the development of BCI. In this paper, a method of attention-based multiscale EEGNet (AMEEGNet) was proposed to improve the decoding performance of MI-EEG. First, three parallel EEGNets with fusion transmission method were employed to extract the high-quality temporal-spatial feature of EEG data from multiple scales. Then, the efficient channel attention (ECA) module enhances the acquisition of more discriminative spatial features through a lightweight approach that weights critical channels. The experimental results demonstrated that the proposed model achieves decoding accuracies of 81.17, 89.83, and 95.49% on BCI-2a, 2b and HGD datasets. The results show that the proposed AMEEGNet effectively decodes temporal-spatial features, providing a novel perspective on MI-EEG decoding and advancing future BCI applications.

NeuroFlex: Feasibility of EEG-Based Motor Imagery Control of a Soft Glove for Hand Rehabilitation

Motor impairments resulting from neurological disorders, such as strokes or spinal cord injuries, often impair hand and finger mobility, restricting a person's ability to grasp and perform fine motor tasks. Brain plasticity refers to the inherent capability of the central nervous system to functionally and structurally reorganize itself in response to stimulation, which underpins rehabilitation from brain injuries or strokes. Linking voluntary cortical activity with corresponding motor execution has been identified as effective in promoting adaptive plasticity. This study introduces NeuroFlex, a motion-intent-controlled soft robotic glove for hand rehabilitation. NeuroFlex utilizes a transformer-based deep learning (DL) architecture to decode motion intent from motor imagery (MI) EEG data and translate it into control inputs for the assistive glove. The glove's soft, lightweight, and flexible design enables users to perform rehabilitation exercises involving fist formation and grasping movements, aligning with natural hand functions for fine motor practices. The results show that the accuracy of decoding the intent of fingers making a fist from MI EEG can reach up to 85.3%, with an average AUC of 0.88. NeuroFlex demonstrates the feasibility of detecting and assisting the patient's attempted movements using pure thinking through a non-intrusive brain-computer interface (BCI). This EEG-based soft glove aims to enhance the effectiveness and user experience of rehabilitation protocols, providing the possibility of extending therapeutic opportunities outside clinical settings.

C. S. Parallel convolutional neural network and empirical mode decomposition for high accuracy in motor imagery EEG signal classification

In recent years, the utilization of motor imagery (MI) signals derived from electroencephalography (EEG) has shown promising applications in controlling various devices such as wheelchairs, assistive technologies, and driverless vehicles. However, decoding EEG signals poses significant challenges due to their complexity, dynamic nature, and low signal-to-noise ratio (SNR). Traditional EEG pattern recognition algorithms typically involve two key steps: feature extraction and feature classification, both crucial for accurate operation. In this work, we propose a novel method that addresses these challenges by employing empirical mode decomposition (EMD) for feature extraction and a parallel convolutional neural network (PCNN) for feature classification. This approach aims to mitigate non-stationary issues, improve performance speed, and enhance classification accuracy. We validate the effectiveness of our proposed method using datasets from the BCI competition IV, specifically datasets 2a and 2b, which contain motor imagery EEG signals. Our method focuses on identifying two- and four-class motor imagery EEG signal classifications. Additionally, we introduce a transfer learning technique to fine-tune the model for individual subjects, leveraging important features extracted from a group dataset. Our results demonstrate that the proposed EMD-PCNN method outperforms existing approaches in terms of classification accuracy. We conduct both qualitative and quantitative analyses to evaluate our method. Qualitatively, we employ confusion matrices and various performance metrics such as specificity, sensitivity, precision, accuracy, recall, and f1-score. Quantitatively, we compare the classification accuracies of our method with those of existing approaches. Our findings highlight the superiority of the proposed EMD-PCNN method in accurately classifying motor imagery EEG signals. The enhanced performance and robustness of our method underscore its potential for broader applicability in real-world scenarios.

Finding Neural Biomarkers for Motor Learning and Rehabilitation using an Explainable Graph Neural Network

Human motor learning is a neural process essential for acquiring new motor skills and adapting existing ones, which is fundamental to everyday activities. Neurological disorders such as Parkinson's Disease (PD) and stroke can significantly affect human motor functions. Identifying neural biomarkers for human motor learning is essential for advancing therapeutic strategies for such disorders. However, identifying specific neural biomarkers associated with motor learning has been challenging due to the complex nature of brain activity and the limitations of traditional data analysis techniques. In response to these challenges, we developed a novel Spatial Graph Neural Network (SGNN) model to predict motor learning outcomes from electroencephalogram (EEG) data using the spatial-temporal dynamics of brain activity. We used it to analyse EEG data collected during a visuomotor rotation (VMR) task designed to elicit distinct types of learning: error-based and reward-based. By doing so, we establish a controlled environment that allows for precisely investigating neural signatures associated with these learning processes. To understand the features learned by the SGNN, we used a set of spatial, spectral, and temporal explainability methods to identify the brain regions and temporal dynamics crucial for learning. These approaches offer comprehensive insights into the neural biomarkers, aligning with current literature and ablation studies, and pave the way for applying this methodology to find biomarkers from various brain signals and tasks.

A Novel CNN-Based Framework for Alzheimer's Disease Detection Using EEG Spectrogram Representations

Background: Alzheimer's disease (AD) is a progressive neurodegenerative disorder that poses critical challenges in global healthcare due to its increasing prevalence and severity. Diagnosing AD and other dementias, such as frontotemporal dementia (FTD), is slow and resource-intensive, underscoring the need for automated approaches. Methods: To address this gap, this study proposes a novel deep learning methodology for EEG classification of AD, FTD, and control (CN) signals. The approach incorporates advanced preprocessing techniques and CNN classification of FFT-based spectrograms and is evaluated using the leave-N-subjects-out validation, ensuring robust cross-subject generalizability. Results: The results indicate that the proposed methodology outperforms state-of-the-art machine learning and EEG-specific neural network models, achieving an accuracy of 79.45% for AD/CN classification and 80.69% for AD+FTD/CN classification. Conclusions: These results highlight the potential of EEG-based deep learning models for early dementia screening, enabling more efficient, scalable, and accessible diagnostic tools.

Human-Centric Spatial Cognition Detecting System Based on Drivers' Electroencephalogram Signals for Autonomous Driving

Existing autonomous driving systems face challenges in accurately capturing drivers' cognitive states, often resulting in decisions misaligned with drivers' intentions. To address this limitation, this study introduces a pioneering human-centric spatial cognition detecting system based on drivers' electroencephalogram (EEG) signals. Unlike conventional EEG-based systems that focus on intention recognition or hazard perception, the proposed system can further extract drivers' spatial cognition across two dimensions: relative distance and relative orientation. It consists of two components: EEG signal preprocessing and spatial cognition decoding, enabling the autonomous driving system to make more contextually aligned decisions regarding the targets drivers focus on. To enhance the detection accuracy of drivers' spatial cognition, we designed a novel EEG signal decoding method called a Dual-Time-Feature Network (DTFNet). This approach integrates coarse-grained and fine-grained temporal features of EEG signals across different scales and incorporates a Squeeze-and-Excitation module to evaluate the importance of electrodes. The DTFNet outperforms existing methods, achieving 65.67% and 50.65% accuracy in three-class tasks and 84.46% and 70.50% in binary tasks. Furthermore, we investigated the temporal dynamics of drivers' spatial cognition and observed that drivers' perception of relative distance occurs slightly later than their perception of relative orientation, providing valuable insights into the temporal aspects of cognitive processing.

A Distribution Adaptive Feedback Training Method to Improve Human Motor Imagery Ability

A brain-computer interface (BCI) based on motor imagery (MI) can translate users' subjective movement-related mental state without external stimulus, which has been successfully used for replacing and repairing motor function. In contrast with studies about decoding methods, less work was reported about training users to improve the performance of MI-BCIs. This study aimed to develop a novel MI feedback training method to enhance the ability of humans to use the MI-BCI system. In this study, an adaptive MI feedback training method was proposed to improve the effectiveness of the training process. The method updated the feedback model during training process and assigned different weights to the samples to better adapt the changes in the distribution of the Electroencephalograms (EEGs). An online feedback training system was established. Each of ten subjects participated in a three-day experiment involving three different feedback methods: no feedback algorithm update, feedback algorithm update, and feedback algorithm update using the proposed adaptive method. Comparison experiments were conducted on three different feedback methods. The experimental results showed that the feedback algorithm using the proposed method can most quickly improve the MI classification accuracy and has the largest increase in accuracy. This indicates that the proposed method can enhance the effectiveness of feedback training and improve the practicality of MI-BCI systems.

ID3RSNet: cross-subject driver drowsiness detection from raw single-channel EEG with an interpretable residual shrinkage network

Accurate monitoring of drowsy driving through electroencephalography (EEG) can effectively reduce traffic accidents. Developing a calibration-free drowsiness detection system with single-channel EEG alone is very challenging due to the non-stationarity of EEG signals, the heterogeneity among different individuals, and the relatively parsimonious compared to multi-channel EEG. Although deep learning-based approaches can effectively decode EEG signals, most deep learning models lack interpretability due to their black-box nature. To address these issues, we propose a novel interpretable residual shrinkage network, namely, ID3RSNet, for cross-subject driver drowsiness detection using single-channel EEG signals. First, a base feature extractor is employed to extract the essential features of EEG frequencies; to enhance the discriminative feature learning ability, the residual shrinkage building unit with attention mechanism is adopted to perform adaptive feature recalibration and soft threshold denoising inside the residual network is further applied to achieve automatic feature extraction. In addition, a fully connected layer with weight freezing is utilized to effectively suppress the negative influence of neurons on the model classification. With the global average pooling (GAP) layer incorporated in the residual shrinkage network structure, we introduce an EEG-based Class Activation Map (ECAM) interpretable method to enable visualization analysis of sample-wise learned patterns to effectively explain the model decision. Extensive experimental results demonstrate that the proposed method achieves the superior classification performance and has found neurophysiologically reliable evidence of classification.

A multi-branch, multi-scale, and multi-view CNN with lightweight temporal attention mechanism for EEG-based motor imagery decoding

Convolutional neural networks (CNNs) have been widely utilized for decoding motor imagery (MI) from electroencephalogram (EEG) signals. However, extracting discriminative spatial-temporal-spectral features from low signal-to-noise ratio EEG signals remains challenging. This paper proposes MBMSNet , a multi-branch, multi-scale, and multi-view CNN with a lightweight temporal attention mechanism for EEG-Based MI decoding. Specifically, MBMSNet first extracts multi-view representations from raw EEG signals, followed by independent branches to capture spatial, spectral, temporal-spatial, and temporal-spectral features. Each branch includes a domain-specific convolutional layer, a variance layer, and a temporal attention layer. Finally, the features derived from each branch are concatenated with weights and classified through a fully connected layer. Experiments demonstrate MBMSNet outperforms state-of-the-art models, achieving accuracies of 84.60% on BCI Competition IV 2a, 87.80% on 2b, and 74.58% on OpenBMI, showcasing its potential for robust BCI applications.

N2GNet tracks gait performance from subthalamic neural signals in Parkinson's disease

Adaptive deep brain stimulation (DBS) provides individualized therapy for people with Parkinson's disease (PWP) by adjusting the stimulation in real-time using neural signals that reflect their motor state. Current algorithms, however, utilize condensed and manually selected neural features which may result in a less robust and biased therapy. In this study, we propose Neural-to-Gait Neural network (N2GNet), a novel deep learning-based regression model capable of tracking real-time gait performance from subthalamic nucleus local field potentials (STN LFPs). The LFP data were acquired when eighteen PWP performed stepping in place, and the ground reaction forces were measured to track their weight shifts representing gait performance. By exhibiting a stronger correlation with weight shifts compared to the higher-correlation beta power from the two leads and outperforming other evaluated model designs, N2GNet effectively leverages a comprehensive frequency band, not limited to the beta range, to track gait performance solely from STN LFPs.

Leveraging deep learning for robust EEG analysis in mental health monitoring

Introduction

Mental health monitoring utilizing EEG analysis has garnered notable interest due to the non-invasive characteristics and rich temporal information encoded in EEG signals, which are indicative of cognitive and emotional conditions. Conventional methods for EEG-based mental health evaluation often depend on manually crafted features or basic machine learning approaches, like support vector classifiers or superficial neural networks. Despite the potential of these approaches, they often fall short in capturing the intricate spatiotemporal relationships within EEG data, leading to lower classification accuracy and poor adaptability across various populations and mental health scenarios.

Methods

To overcome these limitations, we introduce the EEG Mind-Transformer, an innovative deep learning architecture composed of a Dynamic Temporal Graph Attention Mechanism (DT-GAM), a Hierarchical Graph Representation and Analysis (HGRA) module, and a Spatial-Temporal Fusion Module (STFM). The DT-GAM is designed to dynamically extract temporal dependencies within EEG data, while the HGRA models the brain's hierarchical structure to capture both localized and global interactions among different brain regions. The STFM synthesizes spatial and temporal elements, generating a comprehensive representation of EEG signals.

Results and discussion

Our empirical results confirm that the EEG Mind-Transformer significantly surpasses conventional approaches, achieving an accuracy of 92.5%, a recall of 91.3%, an F1-score of 90.8%, and an AUC of 94.2% across several datasets. These findings underline the model's robustness and its generalizability to diverse mental health conditions. Moreover, the EEG Mind-Transformer not only pushes the boundaries of state-of-the-art EEG-based mental health monitoring but also offers meaningful insights into the underlying brain functions associated with mental disorders, solidifying its value for both research and clinical settings.

MFRC-Net: Multi-Scale Feature Residual Convolutional Neural Network for Motor Imagery Decoding

Motor imagery (MI) decoding is the basis of external device control via electroencephalogram (EEG). However, the majority of studies prioritize enhancing the accuracy of decoding methods, often overlooking the magnitude and computational resource demands of deep learning models. In this study, we propose a novel lightweight Multi-Scale Feature Residual Convolutional Neural Network (MFRC-Net). MFRC-Net primarily consists of two blocks: temporal multi-scale residual convolution blocks and cross-domain dual-stream spatial convolution blocks. The former captures dynamic changes in EEG signals across various time scales through multi-scale grouped convolution and backbone temporal convolution skip connections; the latter improves local spatial feature extraction and calibrates feature mapping through the introduction of cross-domain spatial filtering layers. Furthermore, by specifically optimizing the loss function, MFRC-Net effectively reduces sensitivity to outliers. Experiment results on the BCI Competition IV 2a dataset and the SHU dataset demonstrate that, with a parameter size of only 13 K, MFRC-Net achieves accuracy of 85.1% and 69.3%, respectively, surpassing current state-of-the-art models. The integration of temporal multi-scale residual convolution blocks and cross-domain dual-stream spatial convolution blocks in lightweight models significantly boosts performance, as evidenced by ablation studies and visualizations.

Multiscale Spatial-Temporal Feature Fusion Neural Network for Motor Imagery Brain-Computer Interfaces

Motor imagery, one of the main brain-computer interface (BCI) paradigms, has been extensively utilized in numerous BCI applications, such as the interaction between disabled people and external devices. Precise decoding, one of the most significant aspects of realizing efficient and stable interaction, has received a great deal of intensive research. However, the current decoding methods based on deep learning are still dominated by single-scale serial convolution, which leads to insufficient extraction of abundant information from motor imagery signals. To overcome such challenges, we propose a new end-to-end convolutional neural network based on multiscale spatial-temporal feature fusion (MSTFNet) for EEG classification of motor imagery. The architecture of MSTFNet consists of four distinct modules: feature enhancement module, multiscale temporal feature extraction module, spatial feature extraction module and feature fusion module, with the latter being further divided into the depthwise separable convolution block and efficient channel attention block. Moreover, we implement a straightforward yet potent data augmentation strategy to bolster the performance of MSTFNet significantly. To validate the performance of MSTFNet, we conduct cross-session experiments and leave-one-subject-out experiments. The cross-session experiment is conducted across two public datasets and one laboratory dataset. On the public datasets of BCI Competition IV 2a and BCI Competition IV 2b, MSTFNet achieves classification accuracies of 83.62% and 89.26%, respectively. On the laboratory dataset, MSTFNet achieves 86.68% classification accuracy. Besides, the leave-one-subject-out experiment is performed on the BCI Competition IV 2a dataset, and MSTFNet achieves 66.31% classification accuracy. These experimental results outperform several state-of-the-art methodologies, indicate the proposed MSTFNet's robust capability in decoding EEG signals associated with motor imagery.

A Novel Recognition and Classification Approach for Motor Imagery Based on Spatio-Temporal Features

Motor imagery, as a paradigm of brain-computer interface, holds vast potential in the field of medical rehabilitation. Addressing the challenges posed by the non-stationarity and low signal-to-noise ratio of EEG signals, the effective extraction of features from motor imagery signals for accurate recognition stands as a key focus in motor imagery brain-computer interface technology. This paper proposes a motor imagery EEG signal classification model that combines functional brain networks with graph convolutional networks. First, functional brain networks are constructed using different brain functional connectivity metrics, and graph theory features are calculated to deeply analyze the characteristics of brain networks under different motor tasks. Then, the constructed functional brain networks are combined with graph convolutional networks for the classification and recognition of motor imagery tasks. The analysis based on brain functional connectivity reveals that the functional connectivity strength during the both fists task is significantly higher than that of other motor imagery tasks, and the functional connectivity strength during actual movement is generally superior to that of motor imagery tasks. In experiments conducted on the Physionet public dataset, the proposed model achieved a classification accuracy of 88.39% under multi-subject conditions, significantly outperforming traditional methods. Under single-subject conditions, the model effectively addressed the issue of individual variability, achieving an average classification accuracy of 99.31%. These results indicate that the proposed model not only exhibits excellent performance in the classification of motor imagery tasks but also provides new insights into the functional connectivity characteristics of different motor tasks and their corresponding brain regions.