A Transformer-Based Approach Combining Deep Learning Network and Spatial-Temporal Information for Raw EEG Classification

STaRNet: A spatio-temporal and Riemannian network for high-performance motor imagery decoding

Brain-computer interfaces (BCIs), representing a transformative form of human-computer interaction, empower users to interact directly with external environments through brain signals. In response to the demands for high accuracy, robustness, and end-to-end capabilities within BCIs based on motor imagery (MI), this paper introduces STaRNet, a novel model that integrates multi-scale spatio-temporal convolutional neural networks (CNNs) with Riemannian geometry. Initially, STaRNet integrates a multi-scale spatio-temporal feature extraction module that captures both global and local features, facilitating the construction of Riemannian manifolds from these comprehensive spatio-temporal features. Subsequently, a matrix logarithm operation transforms the manifold-based features into the tangent space, followed by a dense layer for classification. Without preprocessing, STaRNet surpasses state-of-the-art (SOTA) models by achieving an average decoding accuracy of 83.29% and a kappa value of 0.777 on the BCI Competition IV 2a dataset, and 95.45% accuracy with a kappa value of 0.939 on the High Gamma Dataset. Additionally, a comparative analysis between STaRNet and several SOTA models, focusing on the most challenging subjects from both datasets, highlights exceptional robustness of STaRNet. Finally, the visualizations of learned frequency bands demonstrate that temporal convolutions have learned MI-related frequency bands, and the t-SNE analyses of features across multiple layers of STaRNet exhibit strong feature extraction capabilities. We believe that the accurate, robust, and end-to-end capabilities of the STaRNet will facilitate the advancement of BCIs.

GCTNet: a graph convolutional transformer network for major depressive disorder detection based on EEG signals

Objective.Identifying major depressive disorder (MDD) using objective physiological signals has become a pressing challenge.Approach.Hence, this paper proposes a graph convolutional transformer network (GCTNet) for accurate and reliable MDD detection using electroencephalogram (EEG) signals. The developed framework integrates a residual graph convolutional network block to capture spatial information and a Transformer block to extract global temporal dynamics. Additionally, we introduce the contrastive cross-entropy (CCE) loss that combines contrastive learning to enhance the stability and discriminability of the extracted features, thereby improving classification performance.Main results. The effectiveness of the GCTNet model and CCE loss was assessed using EEG data from 41 MDD patients and 44 normal controls, in addition to a publicly available dataset. Utilizing a subject-independent data partitioning method and 10-fold cross-validation, the proposed method demonstrated significant performance, achieving an average Area Under the Curve of 0.7693 and 0.9755 across both datasets, respectively. Comparative analyses demonstrated the superiority of the GCTNet framework with CCE loss over state-of-the-art algorithms in MDD detection tasks.Significance. The proposed method offers an objective and effective approach to MDD detection, providing valuable support for clinical-assisted diagnosis.

A diagonal masking self-attention-based multi-scale network for motor imagery classification

Objective. Electroencephalography (EEG)-based motor imagery (MI) is a promising paradigm for brain-computer interface (BCI), but the non-stationarity and low signal-to-noise ratio of EEG signals make it a challenging task.Approach. To achieve high-precision MI classification, we propose a Diagonal Masking Self-Attention-based Multi-Scale Network (DMSA-MSNet) to fully develop, extract, and emphasize features from different scales. First, for local features, a multi-scale temporal-spatial block is proposed to extract features from different receptive fields. Second, an adaptive branch fusion block is specifically designed to bridge the semantic gap between these coded features from different scales. Finally, in order to analyze global information over long ranges, a diagonal masking self-attention block is introduced, which highlights the most valuable features in the data.Main results. The proposed DMSA-MSNet outperforms state-of-the-art models on the BCI Competition IV 2a and the BCI Competition IV 2b datasets.Significance. Our study achieves rich information extraction from EEG signals and provides an effective solution for MI classification.

Exploring the frontier: Transformer-based models in EEG signal analysis for brain-computer interfaces

This review systematically explores the application of transformer-based models in EEG signal processing and brain-computer interface (BCI) development, with a distinct focus on ensuring methodological rigour and adhering to empirical validations within the existing literature. By examining various transformer architectures, such as the Temporal Spatial Transformer Network (TSTN) and EEG Conformer, this review delineates their capabilities in mitigating challenges intrinsic to EEG data, such as noise and artifacts, and their subsequent implications on decoding and classification accuracies across disparate mental tasks. The analytical scope extends to a meticulous examination of attention mechanisms within transformer models, delineating their role in illuminating critical temporal and spatial EEG features and facilitating interpretability in model decision-making processes. The discourse additionally encapsulates emerging works that substantiate the efficacy of transformer models in noise reduction of EEG signals and diversifying applications beyond the conventional motor imagery paradigm. Furthermore, this review elucidates evident gaps and propounds exploratory avenues in the applications of pre-trained transformers in EEG analysis and the potential expansion into real-time and multi-task BCI applications. Collectively, this review distils extant knowledge, navigates through the empirical findings, and puts forward a structured synthesis, thereby serving as a conduit for informed future research endeavours in transformer-enhanced, EEG-based BCI systems.

NeuroDM: Decoding and visualizing human brain activity with EEG-guided diffusion model

BACKGROUND AND OBJECTIVE

Brain-Computer Interface (BCI) technology has recently been advancing rapidly, bringing significant hope for improving human health and quality of life. Decoding and visualizing visually evoked electroencephalography (EEG) signals into corresponding images plays a crucial role in the practical application of BCI technology. The recent emergence of diffusion models provides a good modeling basis for this work. However, the existing diffusion models still have great challenges in generating high-quality images from EEG, due to the low signal-to-noise ratio and strong randomness of EEG signals. The purpose of this study is to address the above-mentioned challenges by proposing a framework named NeuroDM that can decode human brain responses to visual stimuli from EEG-recorded brain activity.

METHODS

In NeuroDM, an EEG-Visual-Transformer (EV-Transformer) is used to extract the visual-related features with high classification accuracy from EEG signals, then an EEG-Guided Diffusion Model (EG-DM) is employed to synthesize high-quality images from the EEG visual-related features.

RESULTS

We conducted experiments on two EEG datasets (one is a forty-class dataset, and the other is a four-class dataset). In the task of EEG decoding, we achieved average accuracies of 99.80% and 92.07% on two datasets, respectively. In the task of EEG visualization, the Inception Score of the images generated by NeuroDM reached 15.04 and 8.67, respectively. All the above results outperform existing methods.

CONCLUSIONS

The experimental results on two EEG datasets demonstrate the effectiveness of the NeuroDM framework, achieving state-of-the-art performance in terms of classification accuracy and image quality. Furthermore, our NeuroDM exhibits strong generalization capabilities and the ability to generate diverse images.

Attention-based convolutional neural network with multi-modal temporal information fusion for motor imagery EEG decoding

Convolutional neural network (CNN) has been widely applied in motor imagery (MI)-based brain computer interface (BCI) to decode electroencephalography (EEG) signals. However, due to the limited perceptual field of convolutional kernel, CNN only extracts features from local region without considering long-term dependencies for EEG decoding. Apart from long-term dependencies, multi-modal temporal information is equally important for EEG decoding because it can offer a more comprehensive understanding of the temporal dynamics of neural processes. In this paper, we propose a novel deep learning network that combines CNN with self-attention mechanism to encapsulate multi-modal temporal information and global dependencies. The network first extracts multi-modal temporal information from two distinct perspectives: average and variance. A shared self-attention module is then designed to capture global dependencies along these two feature dimensions. We further design a convolutional encoder to explore the relationship between average-pooled and variance-pooled features and fuse them into more discriminative features. Moreover, a data augmentation method called signal segmentation and recombination is proposed to improve the generalization capability of the proposed network. The experimental results on the BCI Competition IV-2a (BCIC-IV-2a) and BCI Competition IV-2b (BCIC-IV-2b) datasets show that our proposed method outperforms the state-of-the-art methods and achieves 4-class average accuracy of 85.03% on the BCIC-IV-2a dataset. The proposed method implies the effectiveness of multi-modal temporal information fusion in attention-based deep learning networks and provides a new perspective for MI-EEG decoding. The code is available at https://github.com/Ma-Xinzhi/EEG-TransNet.

Contribution of low-level image statistics to EEG decoding of semantic content in multivariate and univariate models with feature optimization

Spatio-temporal patterns of evoked brain activity contain information that can be used to decode and categorize the semantic content of visual stimuli. However, this procedure can be biased by low-level image features independently of the semantic content present in the stimuli, prompting the need to understand the robustness of different models regarding these confounding factors. In this study, we trained machine learning models to distinguish between concepts included in the publicly available THINGS-EEG dataset using electroencephalography (EEG) data acquired during a rapid serial visual presentation paradigm. We investigated the contribution of low-level image features to decoding accuracy in a multivariate model, utilizing broadband data from all EEG channels. Additionally, we explored a univariate model obtained through data-driven feature selection applied to the spatial and frequency domains. While the univariate models exhibited better decoding accuracy, their predictions were less robust to the confounding effect of low-level image statistics. Notably, some of the models maintained their accuracy even after random replacement of the training dataset with semantically unrelated samples that presented similar low-level content. In conclusion, our findings suggest that model optimization impacts sensitivity to confounding factors, regardless of the resulting classification performance. Therefore, the choice of EEG features for semantic decoding should ideally be informed by criteria beyond classifier performance, such as the neurobiological mechanisms under study.

Multimodal brain-controlled system for rehabilitation training: Combining asynchronous online brain-computer interface and exoskeleton

BACKGROUND

Traditional therapist-based rehabilitation training for patients with movement impairment is laborious and expensive. In order to reduce the cost and improve the treatment effect of rehabilitation, many methods based on human-computer interaction (HCI) technology have been proposed, such as robot-assisted therapy and functional electrical stimulation (FES). However, due to the lack of active participation of brain, these methods have limited effects on the promotion of damaged nerve remodeling.

NEW METHOD

Based on the neurofeedback training provided by the combination of brain-computer interface (BCI) and exoskeleton, this paper proposes a multimodal brain-controlled active rehabilitation system to help improve limb function. The joint control mode of steady-state visual evoked potential (SSVEP) and motor imagery (MI) is adopted to achieve self-paced control and thus maximize the degree of brain involvement, and a requirement selection function based on SSVEP design is added to facilitate communication with aphasia patients.

COMPARISON WITH EXISTING METHODS

In addition, the Transformer is introduced as the MI decoder in the asynchronous online BCI to improve the global perception of electroencephalogram (EEG) signals and maintain the sensitivity and efficiency of the system.

RESULTS

In two multi-task online experiments for left hand, right hand, foot and idle states, subject achieves 91.25% and 92.50% best accuracy, respectively.

CONCLUSION

Compared with previous studies, this paper aims to establish a high-performance and low-latency brain-controlled rehabilitation system, and provide an independent and autonomous control mode of the brain, so as to improve the effect of neural remodeling. The performance of the proposed method is evaluated through offline and online experiments.

TASA: Temporal Attention With Spatial Autoencoder Network for Odor-Induced Emotion Classification Using EEG

The olfactory system enables humans to smell different odors, which are closely related to emotions. The high temporal resolution and non-invasiveness of Electroencephalogram (EEG) make it suitable to objectively study human preferences for odors. Effectively learning the temporal dynamics and spatial information from EEG is crucial for detecting odor-induced emotional valence. In this paper, we propose a deep learning architecture called Temporal Attention with Spatial Autoencoder Network (TASA) for predicting odor-induced emotions using EEG. TASA consists of a filter-bank layer, a spatial encoder, a time segmentation layer, a Long Short-Term Memory (LSTM) module, a multi-head self-attention (MSA) layer, and a fully connected layer. We improve upon the previous work by utilizing a two-phase learning framework, using the autoencoder module to learn the spatial information among electrodes by reconstructing the given input with a latent representation in the spatial dimension, which aims to minimize information loss compared to spatial filtering with CNN. The second improvement is inspired by the continuous nature of the olfactory process; we propose to use LSTM-MSA in TASA to capture its temporal dynamics by learning the intercorrelation among the time segments of the EEG. TASA is evaluated on an existing olfactory EEG dataset and compared with several existing deep learning architectures to demonstrate its effectiveness in predicting olfactory-triggered emotional responses. Interpretability analyses with DeepLIFT also suggest that TASA learns spatial-spectral features that are relevant to olfactory-induced emotion recognition.

Feasibility of decoding covert speech in ECoG with a Transformer trained on overt speech

Several attempts for speech brain-computer interfacing (BCI) have been made to decode phonemes, sub-words, words, or sentences using invasive measurements, such as the electrocorticogram (ECoG), during auditory speech perception, overt speech, or imagined (covert) speech. Decoding sentences from covert speech is a challenging task. Sixteen epilepsy patients with intracranially implanted electrodes participated in this study, and ECoGs were recorded during overt speech and covert speech of eight Japanese sentences, each consisting of three tokens. In particular, Transformer neural network model was applied to decode text sentences from covert speech, which was trained using ECoGs obtained during overt speech. We first examined the proposed Transformer model using the same task for training and testing, and then evaluated the model's performance when trained with overt task for decoding covert speech. The Transformer model trained on covert speech achieved an average token error rate (TER) of 46.6% for decoding covert speech, whereas the model trained on overt speech achieved a TER of 46.3% ( p > 0.05 ; d = 0.07 ) . Therefore, the challenge of collecting training data for covert speech can be addressed using overt speech. The performance of covert speech can improve by employing several overt speeches.

On the role of generative artificial intelligence in the development of brain-computer interfaces

Since their inception more than 50 years ago, Brain-Computer Interfaces (BCIs) have held promise to compensate for functions lost by people with disabilities through allowing direct communication between the brain and external devices. While research throughout the past decades has demonstrated the feasibility of BCI to act as a successful assistive technology, the widespread use of BCI outside the lab is still beyond reach. This can be attributed to a number of challenges that need to be addressed for BCI to be of practical use including limited data availability, limited temporal and spatial resolutions of brain signals recorded non-invasively and inter-subject variability. In addition, for a very long time, BCI development has been mainly confined to specific simple brain patterns, while developing other BCI applications relying on complex brain patterns has been proven infeasible. Generative Artificial Intelligence (GAI) has recently emerged as an artificial intelligence domain in which trained models can be used to generate new data with properties resembling that of available data. Given the enhancements observed in other domains that possess similar challenges to BCI development, GAI has been recently employed in a multitude of BCI development applications to generate synthetic brain activity; thereby, augmenting the recorded brain activity. Here, a brief review of the recent adoption of GAI techniques to overcome the aforementioned BCI challenges is provided demonstrating the enhancements achieved using GAI techniques in augmenting limited EEG data, enhancing the spatiotemporal resolution of recorded EEG data, enhancing cross-subject performance of BCI systems and implementing end-to-end BCI applications. GAI could represent the means by which BCI would be transformed into a prevalent assistive technology, thereby improving the quality of life of people with disabilities, and helping in adopting BCI as an emerging human-computer interaction technology for general use.

TBEEG: A Two-Branch Manifold Domain Enhanced Transformer Algorithm for Learning EEG Decoding

The electroencephalogram-based (EEG) brain-computer interface (BCI) has garnered significant attention in recent research. However, the practicality of EEG remains constrained by the lack of efficient EEG decoding technology. The challenge lies in effectively translating intricate EEG into meaningful, generalizable information. EEG signal decoding primarily relies on either time domain or frequency domain information. There lacks a method capable of simultaneously and effectively extracting both time and frequency domain features, as well as efficiently fuse these features. Addressing these limitations, a two-branch Manifold Domain enhanced transformer algorithm is designed to holistically capture EEG's spatio-temporal information. Our method projects the time-domain information of EEG signals into the Riemannian spaces to fully decode the time dependence of EEG signals. Using wavelet transform, the time domain information is converted into frequency domain information, and the spatial information contained in the frequency domain information of EEG signal is mined through the spectrogram. The effectiveness of the proposed TBEEG algorithm is validated on BCIC-IV-2a dataset and MAMEM-SSVEP-II datasets.

Graph neural network based on brain inspired forward-forward mechanism for motor imagery classification in brain-computer interfaces

Introduction

Within the development of brain-computer interface (BCI) systems, it is crucial to consider the impact of brain network dynamics and neural signal transmission mechanisms on electroencephalogram-based motor imagery (MI-EEG) tasks. However, conventional deep learning (DL) methods cannot reflect the topological relationship among electrodes, thereby hindering the effective decoding of brain activity.

Methods

Inspired by the concept of brain neuronal forward-forward (F-F) mechanism, a novel DL framework based on Graph Neural Network combined forward-forward mechanism (F-FGCN) is presented. F-FGCN framework aims to enhance EEG signal decoding performance by applying functional topological relationships and signal propagation mechanism. The fusion process involves converting the multi-channel EEG into a sequence of signals and constructing a network grounded on the Pearson correlation coeffcient, effectively representing the associations between channels. Our model initially pre-trains the Graph Convolutional Network (GCN), and fine-tunes the output layer to obtain the feature vector. Moreover, the F-F model is used for advanced feature extraction and classification.

Results and discussion

Achievement of F-FGCN is assessed on the PhysioNet dataset for a four-class categorization, compared with various classical and state-of-the-art models. The learned features of the F-FGCN substantially amplify the performance of downstream classifiers, achieving the highest accuracy of 96.11% and 82.37% at the subject and group levels, respectively. Experimental results affirm the potency of FFGCN in enhancing EEG decoding performance, thus paving the way for BCI applications.

Cerebral asymmetry representation learning-based deep subdomain adaptation network for electroencephalogram-based emotion recognition

Objective.Extracting discriminative spatial information from multiple electrodes is a crucial and challenging problem for electroencephalogram (EEG)-based emotion recognition. Additionally, the domain shift caused by the individual differences degrades the performance of cross-subject EEG classification.Approach.To deal with the above problems, we propose the cerebral asymmetry representation learning-based deep subdomain adaptation network (CARL-DSAN) to enhance cross-subject EEG-based emotion recognition. Specifically, the CARL module is inspired by the neuroscience findings that asymmetrical activations of the left and right brain hemispheres occur during cognitive and affective processes. In the CARL module, we introduce a novel two-step strategy for extracting discriminative features through intra-hemisphere spatial learning and asymmetry representation learning. Moreover, the transformer encoders within the CARL module can emphasize the contributive electrodes and electrode pairs. Subsequently, the DSAN module, known for its superior performance over global domain adaptation, is adopted to mitigate domain shift and further improve the cross-subject performance by aligning relevant subdomains that share the same class samples.Main Results.To validate the effectiveness of the CARL-DSAN, we conduct subject-independent experiments on the DEAP database, achieving accuracies of 68.67% and 67.11% for arousal and valence classification, respectively, and corresponding accuracies of 67.70% and 67.18% on the MAHNOB-HCI database.Significance.The results demonstrate that CARL-DSAN can achieve an outstanding cross-subject performance in both arousal and valence classification.

Boosting lower-limb motor imagery performance through an ensemble method for gait rehabilitation

Lower-limb exoskeletons have been used extensively in many rehabilitation applications to assist disabled people with their therapies. Brain-machine interfaces (BMIs) further provide effective and natural control schemes. However, the limited performance of brain signal decoding from lower-limb kinematics restricts the broad growth of both BMI and rehabilitation industry. To address these challenges, we propose an ensemble method for lower-limb motor imagery (MI) classification. The proposed model employs multiple techniques to boost performance, including deep and shallow parts. Traditional wavelet transformation followed by filter-bank common spatial pattern (CSP) employs neurophysiologically reasonable patterns, while multi-head self-attention (MSA) followed by temporal convolutional network (TCN) extracts deeper encoded generalized patterns. Experimental results in a customized lower-limb exoskeleton on 8 subjects in 3 consecutive sessions showed that the proposed method achieved 60.27% and 64.20% for three (MI of left leg, MI of right leg, and rest) and two classes (lower-limb MI vs. rest), respectively. Besides, the proposed model achieves improvements of up to 4% and 2% accuracy for the subject-specific and subject-independent modes compared to the current state-of-the-art (SOTA) techniques, respectively. Finally, feature analysis was conducted to show discriminative brain patterns in each MI task and sessions with different feedback modalities. The proposed models integrated in the brain-actuated lower-limb exoskeleton established a potential BMI for gait training and neuroprosthesis.

MST-net: A multi-scale swin transformer network for EEG-based cognitive load assessment

Cognitive load assessment plays a crucial role in monitoring safe production, resource allocation, and subjective initiative in human-computer interaction. Due to its high time resolution and convenient acquisition, Electroencephalography (EEG) is widely applied in brain monitoring and cognitive state assessment. In this study, a multi-scale Swin Transformer network (MST-Net) was proposed for cognitive load assessment, which extracts local features with different sensory fields using a multi-scale parallel convolution model and introduces the attention mechanism of the Swin Transformer to obtain the feature correlations among multi-scale local features. The performance of the proposed network was validated using the EEG signals collected during cognitive tasks and N-back tasks with three different load levels. Results show that the MST-Net network achieved the best classification accuracy on both local and public datasets, and was higher than the mainstream Swin Transformer and CNN. Furthermore, results of ablation experiments and feature visualization revealed that the proposed MST-Net could well characterize different cognitive loads, which not only provided novel and powerful tools for cognitive load assessment but also showed potential for broad application in brain-computer interface (BCI) systems.

Deep temporal networks for EEG-based motor imagery recognition

The electroencephalogram (EEG) based motor imagery (MI) signal classification, also known as motion recognition, is a highly popular area of research due to its applications in robotics, gaming, and medical fields. However, the problem is ill-posed as these signals are non-stationary and noisy. Recently, a lot of efforts have been made to improve MI signal classification using a combination of signal decomposition and machine learning techniques but they fail to perform adequately on large multi-class datasets. Previously, researchers have implemented long short-term memory (LSTM), which is capable of learning the time-series information, on the MI-EEG dataset for motion recognition. However, it can not model very long-term dependencies present in the motion recognition data. With the advent of transformer networks in natural language processing (NLP), the long-term dependency issue has been widely addressed. Motivated by the success of transformer algorithms, in this article, we propose a transformer-based deep learning neural network architecture that performs motion recognition on the raw BCI competition III IVa and IV 2a datasets. The validation results show that the proposed method achieves superior performance than the existing state-of-the-art methods. The proposed method produces classification accuracy of 99.7% and 84% on the binary class and the multi-class datasets, respectively. Further, the performance of the proposed transformer-based model is also compared with LSTM.

Hierarchical Transformer for Motor Imagery-Based Brain Computer Interface

In this paper, we propose a novel transformer-based classification algorithm for the brain computer interface (BCI) using a motor imagery (MI) electroencephalogram (EEG) signal. To design the MI classification algorithm, we apply an up-to-date deep learning model, the transformer, that has revolutionized the natural language processing (NLP) and successfully widened its application to many other domains such as the computer vision. Within a long MI trial spanning a few seconds, the classification algorithm should give more attention to the time periods during which the intended motor task is imagined by the subject without any artifact. To achieve this goal, we propose a hierarchical transformer architecture that consists of a high-level transformer (HLT) and a low-level transformer (LLT). We break down a long MI trial into a number of short-term intervals. The LLT extracts a feature from each short-term interval, and the HLT pays more attention to the features from more relevant short-term intervals by using the self-attention mechanism of the transformer. We have done extensive tests of the proposed scheme on four open MI datasets, and shown that the proposed hierarchical transformer excels in both the subject-dependent and subject-independent tests.

Classification of attention deficit/hyperactivity disorder based on EEG signals using a EEG-Transformer model∗

Objective. Attention-deficit/hyperactivity disorder (ADHD) is the most common neurodevelopmental disorder in adolescents that can seriously impair a person's attention function, cognitive processes, and learning ability. Currently, clinicians primarily diagnose patients based on the subjective assessments of the Diagnostic and Statistical Manual of Mental Disorders-5, which can lead to delayed diagnosis of ADHD and even misdiagnosis due to low diagnostic efficiency and lack of well-trained diagnostic experts. Deep learning of electroencephalogram (EEG) signals recorded from ADHD patients could provide an objective and accurate method to assist physicians in clinical diagnosis.Approach. This paper proposes the EEG-Transformer deep learning model, which is based on the attention mechanism in the traditional Transformer model, and can perform feature extraction and signal classification processing for the characteristics of EEG signals. A comprehensive comparison was made between the proposed transformer model and three existing convolutional neural network models.Main results. The results showed that the proposed EEG-Transformer model achieved an average accuracy of 95.85% and an average AUC value of 0.9926 with the fastest convergence speed, outperforming the other three models. The function and relationship of each module of the model are studied by ablation experiments. The model with optimal performance was identified by the optimization experiment.Significance. The EEG-Transformer model proposed in this paper can be used as an auxiliary tool for clinical diagnosis of ADHD, and at the same time provides a basic model for transferable learning in the field of EEG signal classification.

Transformers for autonomous recognition of psychiatric dysfunction via raw and imbalanced EEG signals

Early identification of mental disorders, based on subjective interviews, is extremely challenging in the clinical setting. There is a growing interest in developing automated screening tools for potential mental health problems based on biological markers. Here, we demonstrate the feasibility of an AI-powered diagnosis of different mental disorders using EEG data. Specifically, this work aims to classify different mental disorders in the following ecological context accurately: (1) using raw EEG data, (2) collected during rest, (3) during both eye open, and eye closed conditions, (4) at short 2-min duration, (5) on participants with different psychiatric conditions, (6) with some overlapping symptoms, and (7) with strongly imbalanced classes. To tackle this challenge, we designed and optimized a transformer-based architecture, where class imbalance is addressed through focal loss and class weight balancing. Using the recently released TDBRAIN dataset (n= 1274 participants), our method classifies each participant as either a neurotypical or suffering from major depressive disorder (MDD), attention deficit hyperactivity disorder (ADHD), subjective memory complaints (SMC), or obsessive-compulsive disorder (OCD). We evaluate the performance of the proposed architecture on both the window-level and the patient-level. The classification of the 2-min raw EEG data into five classes achieved a window-level accuracy of 63.2% and 65.8% for open and closed eye conditions, respectively. When the classification is limited to three main classes (MDD, ADHD, SMC), window level accuracy improved to 75.1% and 69.9% for eye open and eye closed conditions, respectively. Our work paves the way for developing novel AI-based methods for accurately diagnosing mental disorders using raw resting-state EEG data.

Multi-Band CNN With Band-Dependent Kernels and Amalgamated Cross Entropy Loss for Motor Imagery Classification

In this paper, we present a novel MI classification method based on multi-band convolutional neural network (CNN) with band-dependent kernel sizes, named MBK-CNN, to improve classification performance, by resolving the subject dependency issue of the widely used CNN-based approaches due to the kernel size optimization problem. The proposed structure exploits the frequency diversity of the EEG signals and simultaneously resolves the subject dependent kernel size issue. EEG signal is decomposed into overlapping multi-band and passed through multiple CNNs (termed 'branch-CNNs') with different kernel sizes to generate frequency dependent features, which are combined by a simple weighted sum. In contrast to the existing works where single-band multi-branch CNNs with different kernel sizes are used to resolve the subject dependency issue, a unique kernel size per frequency band is used. To prevent possible overfitting induced by a weighted sum, each branch-CNN is additionally trained by tentative cross entropy loss while overall network is optimized by the end-to-end cross entropy loss, which is named amalgamated cross entropy loss. In addition, we further propose multi-band CNN with enhanced spatial diversity, named MBK-LR-CNN, by replacing each branch-CNN with several sub branch-CNNs applied for channel subsets (termed 'local region') to improve the classification performance. We evaluated the performance of the proposed methods, MBK-CNN and MBK-LR-CNN, on publicly available datasets, BCI Competition IV dataset 2a and High Gamma Dataset. The experimental results confirm the performance improvement of the proposed methods compared to the currently existing MI classification methods.

Temporal transformer-spatial graph convolutional network: an intelligent classification model for anti N-methyl-D-aspartate receptor encephalitis based on electroencephalogram signal

Encephalitis is a disease typically caused by viral infections or autoimmunity. The most common type of autoimmune encephalitis is anti-N-methyl-D-aspartate receptor (NMDAR) antibody-mediated, known as anti-NMDA receptor encephalitis, which is a rare disease. Specific EEG patterns, including "extreme delta brush" (EDB), have been reported in patients with anti-NMDA receptor encephalitis. The aim of this study was to develop an intelligent diagnostic model for encephalitis based on EEG signals. A total of 131 Participants were selected based on reasonable inclusion criteria and divided into three groups: health control (35 participants), viral encephalitis (58 participants), and anti NMDAR receptor encephalitis (55 participants). Due to the low prevalence of anti-NMDAR receptor encephalitis, it took several years to collect participants' EEG signals while they were in an awake state. EEG signals were collected and analyzed following the international 10-20 system layout. We proposed a model called Temporal Transformer-Spatial Graph Convolutional Network (TT-SGCN), which consists of a Preprocess Module, a Temporal Transformer Module (TTM), and a Spatial Graph Convolutional Module (SGCM). The raw EEG signal was preprocessed according to traditional procedures, including filtering, averaging, and Independent Component Analysis (ICA) method. The EEG signal was then segmented and transformed using short-time Fourier transform (STFT) to produce concatenated power density (CPD) maps, which served as inputs for the proposed model. TTM extracted the time-frequency features of each channel, and SGCM fused these features using graph convolutional methods based on the location of electrodes. The model was evaluated in two experiments: classification of the three groups and pairwise classification among the three groups. The model was trained using two stages and achieved the performance, with an accuracy of 82.23%, recall of 80.75%, precision of 82.51%, and F1 score of 81.23% in the classification of the three groups. The proposed model has the potential to become an intelligent auxiliary diagnostic tool for encephalitis.

Local and global convolutional transformer-based motor imagery EEG classification

Transformer, a deep learning model with the self-attention mechanism, combined with the convolution neural network (CNN) has been successfully applied for decoding electroencephalogram (EEG) signals in Motor Imagery (MI) Brain-Computer Interface (BCI). However, the extremely non-linear, nonstationary characteristics of the EEG signals limits the effectiveness and efficiency of the deep learning methods. In addition, the variety of subjects and the experimental sessions impact the model adaptability. In this study, we propose a local and global convolutional transformer-based approach for MI-EEG classification. The local transformer encoder is combined to dynamically extract temporal features and make up for the shortcomings of the CNN model. The spatial features from all channels and the difference in hemispheres are obtained to improve the robustness of the model. To acquire adequate temporal-spatial feature representations, we combine the global transformer encoder and Densely Connected Network to improve the information flow and reuse. To validate the performance of the proposed model, three scenarios including within-session, cross-session and two-session are designed. In the experiments, the proposed method achieves up to 1.46%, 7.49% and 7.46% accuracy improvement respectively in the three scenarios for the public Korean dataset compared with current state-of-the-art models. For the BCI competition IV 2a dataset, the proposed model also achieves a 2.12% and 2.21% improvement for the cross-session and two-session scenarios respectively. The results confirm that the proposed approach can effectively extract much richer set of MI features from the EEG signals and improve the performance in the BCI applications.

An improved model using convolutional sliding window-attention network for motor imagery EEG classification

Introduction

The classification model of motor imagery-based electroencephalogram (MI-EEG) is a new human-computer interface pattern and a new neural rehabilitation assessment method for diseases such as Parkinson's and stroke. However, existing MI-EEG models often suffer from insufficient richness of spatiotemporal feature extraction, learning ability, and dynamic selection ability.

Methods

To solve these problems, this work proposed a convolutional sliding window-attention network (CSANet) model composed of novel spatiotemporal convolution, sliding window, and two-stage attention blocks.

Results

The model outperformed existing state-of-the-art (SOTA) models in within- and between-individual classification tasks on commonly used MI-EEG datasets BCI-2a and Physionet MI-EEG, with classification accuracies improved by 4.22 and 2.02%, respectively.

Discussion

The experimental results also demonstrated that the proposed type token, sliding window, and local and global multi-head self-attention mechanisms can significantly improve the model's ability to construct, learn, and adaptively select multi-scale spatiotemporal features in MI-EEG signals, and accurately identify electroencephalogram signals in the unilateral motor area. This work provided a novel and accurate classification model for MI-EEG brain-computer interface tasks and proposed a feasible neural rehabilitation assessment scheme based on the model, which could promote the further development and application of MI-EEG methods in neural rehabilitation.

LMDA-Net:A lightweight multi-dimensional attention network for general EEG-based brain-computer interfaces and interpretability

Electroencephalography (EEG)-based brain-computer interfaces (BCIs) pose a challenge for decoding due to their low spatial resolution and signal-to-noise ratio. Typically, EEG-based recognition of activities and states involves the use of prior neuroscience knowledge to generate quantitative EEG features, which may limit BCI performance. Although neural network-based methods can effectively extract features, they often encounter issues such as poor generalization across datasets, high predicting volatility, and low model interpretability. To address these limitations, we propose a novel lightweight multi-dimensional attention network, called LMDA-Net. By incorporating two novel attention modules designed specifically for EEG signals, the channel attention module and the depth attention module, LMDA-Net is able to effectively integrate features from multiple dimensions, resulting in improved classification performance across various BCI tasks. LMDA-Net was evaluated on four high-impact public datasets, including motor imagery (MI) and P300-Speller, and was compared with other representative models. The experimental results demonstrate that LMDA-Net outperforms other representative methods in terms of classification accuracy and predicting volatility, achieving the highest accuracy in all datasets within 300 training epochs. Ablation experiments further confirm the effectiveness of the channel attention module and the depth attention module. To facilitate an in-depth understanding of the features extracted by LMDA-Net, we propose class-specific neural network feature interpretability algorithms that are suitable for evoked responses and endogenous activities. By mapping the output of the specific layer of LMDA-Net to the time or spatial domain through class activation maps, the resulting feature visualizations can provide interpretable analysis and establish connections with EEG time-spatial analysis in neuroscience. In summary, LMDA-Net shows great potential as a general decoding model for various EEG tasks.

MI-CAT: A transformer-based domain adaptation network for motor imagery classification

Due to its convenience and safety, electroencephalography (EEG) data is one of the most widely used signals in motor imagery (MI) brain-computer interfaces (BCIs). In recent years, methods based on deep learning have been widely applied to the field of BCIs, and some studies have gradually tried to apply Transformer to EEG signal decoding due to its superior global information focusing ability. However, EEG signals vary from subject to subject. Based on Transformer, how to effectively use data from other subjects (source domain) to improve the classification performance of a single subject (target domain) remains a challenge. To fill this gap, we propose a novel architecture called MI-CAT. The architecture innovatively utilizes Transformer's self-attention and cross-attention mechanisms to interact features to resolve differential distribution between different domains. Specifically, we adopt a patch embedding layer for the extracted source and target features to divide the features into multiple patches. Then, we comprehensively focus on the intra-domain and inter-domain features by stacked multiple Cross-Transformer Blocks (CTBs), which can adaptively conduct bidirectional knowledge transfer and information exchange between domains. Furthermore, we also utilize two non-shared domain-based attention blocks to efficiently capture domain-dependent information, optimizing the features extracted from the source and target domains to assist in feature alignment. To evaluate our method, we conduct extensive experiments on two real public EEG datasets, Dataset IIb and Dataset IIa, achieving competitive performance with an average classification accuracy of 85.26% and 76.81%, respectively. Experimental results demonstrate that our method is a powerful model for decoding EEG signals and facilitates the development of the Transformer for brain-computer interfaces (BCIs).

Transformer Based Cross-Subject Mental Workload Classification Using FNIRS for Real-World Application

Mental state monitoring is a hot topic especially in neurorehabilitation, skill training, etc, for which the functional near-infrared spectroscopy (fNIRS) has been suggested to be used, and fewer detection channels and cross-subject performance are usually required for real-world application. To this goal, we propose a transformer-based method for cross-subject mental workload classification using fewer channels of fNIRS. Firstly, the input fNIRS signals in a window are divided into patches in the temporal order and transformed into embeddings, to which a classification token and learnable position embeddings are added. Then, a transformer encoder is used to learn the long-range dependencies among the embeddings, of which the output classification token is sent to a multilayer perceptron (MLP) head. Mental workload classification results can be represented by the outputs of the MLP head. Finally, comparison experiments were conducted on the open-access fNIRS2MW dataset. The results show that, the proposed method can outperform previous methods in cross-subject classification accuracy, and relatively efficient computation can be obtained.

Recognize enhanced temporal-spatial-spectral features with a parallel multi-branch CNN and GRU

Deep learning has been applied to the recognition of motor imagery electroencephalograms (MI-EEG) in brain-computer interface, and the performance results depend on data representation as well as neural network structure. Especially, MI-EEG is so complex with the characteristics of non-stationarity, specific rhythms, and uneven distribution; however, its multidimensional feature information is difficult to be fused and enhanced simultaneously in the existing recognition methods. In this paper, a novel channel importance (NCI) based on time-frequency analysis is proposed to develop an image sequence generation method (NCI-ISG) for enhancing the integrity of data representation and highlighting the contribution inequalities of different channels as well. Each electrode of MI-EEG is converted to a time-frequency spectrum by utilizing short-time Fourier transform; the corresponding part to 8-30 Hz is combined with random forest algorithm for computing NCI; and it is further divided into three sub-images covered by α (8-13 Hz), β₁ (13-21 Hz), and β₂ (21-30 Hz) bands; their spectral powers are further weighted by NCI and interpolated to 2-dimensional electrode coordinates, producing three main sub-band image sequences. Then, a parallel multi-branch convolutional neural network and gate recurrent unit (PMBCG) is designed to successively extract and identify the spatial-spectral and temporal features from the image sequences. Two public four-class MI-EEG datasets are adopted; the proposed classification method respectively achieves the average accuracies of 98.26% and 80.62% by 10-fold cross-validation experiment; and its statistical performance is also evaluated by multi-indexes, such as Kappa value, confusion matrix, and ROC curve. Extensive experiment results show that NCI-ISG + PMBCG can yield great performance on MI-EEG classification compared to state-of-the-art methods. The proposed NCI-ISG can enhance the feature representation of time-frequency-space domains and match well with PMBCG, which improves the recognition accuracies of MI tasks and demonstrates the preferable reliability and distinguishable ability. This paper proposes a novel channel importance (NCI) based on time-frequency analysis to develop an image sequences generation method (NCI-ISG) for enhancing the integrity of data representation and highlighting the contribution inequalities of different channels as well. Then, a parallel multi-branch convolutional neural network and gate recurrent unit (PMBCG) is designed to successively extract and identify the spatial-spectral and temporal features from the image sequences.

A transformer-based deep neural network model for SSVEP classification

Steady-state visual evoked potential (SSVEP) is one of the most commonly used control signals in the brain-computer interface (BCI) systems. However, the conventional spatial filtering methods for SSVEP classification highly depend on the subject-specific calibration data. The need for the methods that can alleviate the demand for the calibration data becomes urgent. In recent years, developing the methods that can work in inter-subject scenario has become a promising new direction. As a popular deep learning model nowadays, Transformer has been used in EEG signal classification tasks owing to its excellent performance. Therefore, in this study, we proposed a deep learning model for SSVEP classification based on Transformer architecture in inter-subject scenario, termed as SSVEPformer, which was the first application of Transformer on the SSVEP classification. Inspired by previous studies, we adopted the complex spectrum features of SSVEP data as the model input, which could enable the model to simultaneously explore the spectral and spatial information for classification. Furthermore, to fully utilize the harmonic information, an extended SSVEPformer based on the filter bank technology (FB-SSVEPformer) was proposed to improve the classification performance. Experiments were conducted using two open datasets (Dataset 1: 10 subjects, 12 targets; Dataset 2: 35 subjects, 40 targets). The experimental results show that the proposed models could achieve better results in terms of classification accuracy and information transfer rate than other baseline methods. The proposed models validate the feasibility of deep learning models based on Transformer architecture for SSVEP data classification, and could serve as potential models to alleviate the calibration procedure in the practical application of SSVEP-based BCI systems.

Motor imagery EEG signal classification with a multivariate time series approach

BACKGROUND

Electroencephalogram (EEG) signals record electrical activity on the scalp. Measured signals, especially EEG motor imagery signals, are often inconsistent or distorted, which compromises their classification accuracy. Achieving a reliable classification of motor imagery EEG signals opens the door to possibilities such as the assessment of consciousness, brain computer interfaces or diagnostic tools. We seek a method that works with a reduced number of variables, in order to avoid overfitting and to improve interpretability. This work aims to enhance EEG signal classification accuracy by using methods based on time series analysis. Previous work on this line, usually took a univariate approach, thus losing the possibility to take advantage of the correlation information existing within the time series provided by the different electrodes. To overcome this problem, we propose a multivariate approach that can fully capture the relationships among the different time series included in the EEG data. To perform the multivariate time series analysis, we use a multi-resolution analysis approach based on the discrete wavelet transform, together with a stepwise discriminant that selects the most discriminant variables provided by the discrete wavelet transform analysis RESULTS: Applying this methodology to EEG data to differentiate between the motor imagery tasks of moving either hands or feet has yielded very good classification results, achieving in some cases up to 100% of accuracy for this 2-class pre-processed dataset. Besides, the fact that these results were achieved using a reduced number of variables (55 out of 22,176) can shed light on the relevance and impact of those variables.

CONCLUSIONS

This work has a potentially large impact, as it enables classification of EEG data based on multivariate time series analysis in an interpretable way with high accuracy. The method allows a model with a reduced number of features, facilitating its interpretability and improving overfitting. Future work will extend the application of this classification method to help in diagnosis procedures for detecting brain pathologies and for its use in brain computer interfaces. In addition, the results presented here suggest that this method could be applied to other fields for the successful analysis of multivariate temporal data.

An attention-based deep learning approach for the classification of subjective cognitive decline and mild cognitive impairment using resting-state EEG

Objective. This study aims to design and implement the first deep learning (DL) model to classify subjects in the prodromic states of Alzheimer's disease (AD) based on resting-state electroencephalographic (EEG) signals.Approach. EEG recordings of 17 healthy controls (HCs), 56 subjective cognitive decline (SCD) and 45 mild cognitive impairment (MCI) subjects were acquired at resting state. After preprocessing, we selected sections corresponding to eyes-closed condition. Five different datasets were created by extracting delta, theta, alpha, beta and delta-to-theta frequency bands using bandpass filters. To classify SCDvsMCI and HCvsSCDvsMCI, we propose a framework based on the transformer architecture, which uses multi-head attention to focus on the most relevant parts of the input signals. We trained and validated the model on each dataset with a leave-one-subject-out cross-validation approach, splitting the signals into 10 s epochs. Subjects were assigned to the same class as the majority of their epochs. Classification performances of the transformer were assessed for both epochs and subjects and compared with other DL models.Main results. Results showed that the delta dataset allowed our model to achieve the best performances for the discrimination of SCD and MCI, reaching an Area Under the ROC Curve (AUC) of 0.807, while the highest results for the HCvsSCDvsMCI classification were obtained on alpha and theta with a micro-AUC higher than 0.74.Significance. We demonstrated that DL approaches can support the adoption of non-invasive and economic techniques as EEG to stratify patients in the clinical population at risk for AD. This result was achieved since the attention mechanism was able to learn temporal dependencies of the signal, focusing on the most discriminative patterns, achieving state-of-the-art results by using a deep model of reduced complexity. Our results were consistent with clinical evidence that changes in brain activity are progressive when considering early stages of AD.

Brain Age Prediction/Classification through Recurrent Deep Learning with Electroencephalogram Recordings of Seizure Subjects

With modern population growth and an increase in the average lifespan, more patients are becoming afflicted with neurodegenerative diseases such as dementia and Alzheimer's. Patients with a history of epilepsy, drug abuse, and mental health disorders such as depression have a larger risk of developing Alzheimer's and other neurodegenerative diseases later in life. Utilizing recordings of patients' brain waves obtained from the Temple University abnormal electroencephalogram (EEG) corpus, deep leaning long short-term memory neural networks are utilized to classify and predict patients' brain ages. The proposed deep learning neural network model structure and brain wave-processing methodology leads to an accuracy of 90% in patients' brain age classification across six age groups, with a mean absolute error value of 7 years for the brain age regression analysis. The achieved results demonstrate that the use of raw patient-sourced brain wave information leads to higher performance metrics than methods utilizing other brain wave-preprocessing methods and outperforms other deep learning models such as convolutional neural networks.