IEEE Transactions on Neural Systems and Rehabilitation Engineering最新文献

{"title":"Human Cortical Brain and Biomechanical Responses to Abrupt Changes in Exoskeleton Assistance While Walking","authors":"Seongmi Song;Courtney A. Haynes;J. Cortney Bradford","doi":"10.1109/TNSRE.2025.3610317","DOIUrl":"10.1109/TNSRE.2025.3610317","url":null,"abstract":"Robotic exoskeletons have advanced significantly, yet control systems still face challenges in delivering assistance that seamlessly aligns with the user’s musculoskeletal system. To enhance user-device interaction, it is essential to understand the neural mechanisms underlying human adaptation to exoskeleton assistance. Although cortical brain activity plays a critical role in locomotion, how it is modulated by exoskeleton assistance while walking remains poorly understood. This study investigates cortical dynamics during abrupt exoskeleton state changes to clarify the brain’s role in gait adaptation. EEG, kinematic, and muscle activity data were collected from 21 healthy adults walking on a treadmill with bilateral ankle exoskeletons. Results reveal that frontal theta-band activity could track exoskeleton state transitions. Adaptations occurred within two strides, reflected in changes in frontal theta, alpha, and beta activity, along with variations in knee and ankle range of motion and muscle activation patterns. These findings demonstrate EEG’s sensitivity to neural responses during exoskeleton transitions and highlight its potential application for enabling real-time feedback to optimize personalized assistance.","PeriodicalId":13419,"journal":{"name":"IEEE Transactions on Neural Systems and Rehabilitation Engineering","volume":"33 ","pages":"3770-3783"},"PeriodicalIF":5.2,"publicationDate":"2025-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11165471","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145075253","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Searching for Optimal EMG Latent Subspace With Discriminant DoF-Wise Distributions for Subject-Generic Model","authors":"Jiahao Fan;Yangyang Yuan;Tanying Su;Jionghui Liu;Chih-Hong Chou;Xinyu Jiang;Fumin Jia;Chenyun Dai","doi":"10.1109/TNSRE.2025.3608128","DOIUrl":"10.1109/TNSRE.2025.3608128","url":null,"abstract":"Recognizing hand gestures from surface electromyography (sEMG) signals is crucial for neural interfaces and human–machine interaction. However, developing subject-generic models remains challenging due to substantial inter-subject variability. Complicating matters further, the muscle groups driving gestures with varying degrees of freedom (DoFs) often overlap, producing highly convoluted feature distributions across subjects and DoFs. To address these challenges, we introduce a multi-branch autoencoder (AE) architecture that disentangles sEMG features into two latent subspaces: a DoF-specific (subject-invariant) space and a subject-specific (DoF-invariant) space. We systematically compare our approach against well-established feature projection methods: principal component analysis (PCA), kernel PCA (KPCA), linear discriminant analysis (LDA), kernel discriminant analysis (KDA), and a conventional AE, as well as two style-independent feature transformation methods: canonical correlation analysis (CCA) and spectral regression discriminant analysis (SRDA). Experimental results on 20 subjects across multiple days demonstrate that our multi-branch AE markedly improves DoF discrimination while maintaining subject invariance, leading to consistently higher inter-subject classification accuracy for all common classifiers. These findings underscore the potential of our approach for robust, user-independent sEMG-based gesture recognition.","PeriodicalId":13419,"journal":{"name":"IEEE Transactions on Neural Systems and Rehabilitation Engineering","volume":"33 ","pages":"3723-3733"},"PeriodicalIF":5.2,"publicationDate":"2025-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11156143","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145033184","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Lightweight CNN Approach for Hand Gesture Recognition via GAF Encoding of A-Mode Ultrasound Signals","authors":"Qican Shangguan;Yue Lian;Zhiwei Liao;Jinshui Chen;Yiru Song;Ligang Yao;Cai Jiang;Zongxing Lu;Zhonghua Lin","doi":"10.1109/TNSRE.2025.3608180","DOIUrl":"10.1109/TNSRE.2025.3608180","url":null,"abstract":"Hand gesture recognition(HGR) is a key technology in human-computer interaction and human communication. This paper presents a lightweight, parameter-free attention convolutional neural network (LPA-CNN) approach leveraging Gramian Angular Field(GAF)transformation of A-mode ultrasound signals for HGR. First, this paper maps 1-dimensional (1D) A-mode ultrasound signals, collected from the forearm muscles of 10 healthy participants, into 2-dimensional (2D) images. Second, GAF is selected owing to its higher sensitivity against Markov Transition Field (MTF) and Recurrence Plot (RP) in HGR. Third, a novel LPA-CNN consisting of four components, i.e., a convolution-pooling block, an attention mechanism, an inverted residual block, and a classification block, is proposed. Among them, the convolution-pooling block consists of convolutional and pooling layers, the attention mechanism is applied to generate 3-D weights, the inverted residual block consists of multiple channel shuffling units, and the classification block is performed through fully connected layers. Fourth, comparative experiments were conducted on GoogLeNet, MobileNet, and LPA-CNN to validate the effectiveness of the proposed method. Experimental results show that compared to GoogLeNet and MobileNet, LPA-CNN has a smaller model size and better recognition performance, achieving a classification accuracy of 0.98 ± 0.02. This paper achieves efficient and high-accuracy HGR by encoding A-mode ultrasound signals into 2D images and integrating the LPA-CNN model, providing a new technological approach for HGR based on ultrasonic signals.","PeriodicalId":13419,"journal":{"name":"IEEE Transactions on Neural Systems and Rehabilitation Engineering","volume":"33 ","pages":"3734-3743"},"PeriodicalIF":5.2,"publicationDate":"2025-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11156155","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145033032","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"UPFP-SG: A New Benchmark for Unilateral Peripheral Facial Paralysis Severity Grading","authors":"Wei Gan;Ruiqi Zhao;Ke Lu;Yuxuan Li;Guohong Hu;Zhenghui Lei;Dongmei Jiang;Tao Zhang;Jian Xue","doi":"10.1109/TNSRE.2025.3608463","DOIUrl":"10.1109/TNSRE.2025.3608463","url":null,"abstract":"Unilateral facial palsy, a common type of facial paralysis, profoundly impacts individuals’ daily functionality and quality of life. The current clinical diagnosis of facial paralysis primarily relies on the subjective judgment of doctors, and the development of automated detection methods is challenged by the lack of publicly available facial paralysis datasets and the inability to analyze different facial nerve branches. To address these problems, we propose a new benchmark named UPFG-SG for Unilateral Peripheral Facial Paralysis Severity Grading. First, we establish a dataset with an improved subjective evaluating system to assess the palsy severity of different peripheral facial nerve branches, which can be obtained via <uri>https://www.iiplab.net/upfp-sg/</uri>. Second, we propose a new method trained on this dataset which integrates different facial features to rate the facial palsy severity of each facial nerve region. Additionally, an enhanced regression module is designed to improve the accuracy of evaluation. With these improvements, our method effectively captures both subtle facial expression changes and fine local details. Experimental results based on our dataset demonstrate that the proposed method outperforms current deep learning methods in the field.","PeriodicalId":13419,"journal":{"name":"IEEE Transactions on Neural Systems and Rehabilitation Engineering","volume":"33 ","pages":"3744-3754"},"PeriodicalIF":5.2,"publicationDate":"2025-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11156141","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145033126","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"E-DANN: An Enhanced Domain Adaptation Network for Audio-EEG Feature Decoupling in Explainable Depression Recognition","authors":"Qinglin Zhao;Hua Jiang;Zhongqing Wu;Lixin Zhang;Kunbo Cui;Kai Zheng;Jingyu Liu;Ran Cai;Mingqi Zhao;Fuze Tian;Bin Hu","doi":"10.1109/TNSRE.2025.3608181","DOIUrl":"10.1109/TNSRE.2025.3608181","url":null,"abstract":"Given the significant global health burden caused by depression, numerous studies have utilized artificial intelligence techniques to objectively and automatically detect depression. However, existing research primarily focuses on improving the accuracy of depression recognition while overlooking the explainability of detection models and the evaluation of feature importance. In this paper, we propose a novel framework named Enhanced Domain Adversarial Neural Network (E-DANN) for depression detection. First, we extract joint features that combine audio-specific physical properties and electroencephalogram (EEG) responses to construct a multimodal feature space, which facilitates the understanding of how EEG signals dynamically respond to the physical properties of audio. Next, we employ the feature decoupling framework of E-DANN, which separates the extracted feature space into shared features and private features through adversarial training. The decoupled private features are then utilized for the binary classification of depression. Our experimental results validate the effectiveness of the proposed framework, which achieves accurate classification of normal controls and individuals with depression (accuracy: <inline-formula> <tex-math>$92.83~pm ~4.38$ </tex-math></inline-formula>%, specificity: <inline-formula> <tex-math>$93.56~pm ~7.25$ </tex-math></inline-formula>%, sensitivity: <inline-formula> <tex-math>$91.61~pm ~6.87$ </tex-math></inline-formula>%, and F1 score: <inline-formula> <tex-math>$91.81~pm ~4.52$ </tex-math></inline-formula>%). Furthermore, we employ an Explainable Artificial Intelligence (XAI) approach to hierarchically visualize feature importance and elucidate complex feature interaction patterns. In summary, this study provides a theoretical foundation for developing explainable diagnostic tools for depression and contributes to improving the clinical trustworthiness of AI-assisted diagnostic systems.","PeriodicalId":13419,"journal":{"name":"IEEE Transactions on Neural Systems and Rehabilitation Engineering","volume":"33 ","pages":"3647-3661"},"PeriodicalIF":5.2,"publicationDate":"2025-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11156140","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145033153","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Bio-Inspired Synergistic Model Predictive Control for Control Reallocation and Reduced Computational Cost in a Hybrid Exoskeleton","authors":"Krysten Lambeth;Noor Hakam;Nitin Sharma","doi":"10.1109/TNSRE.2025.3608567","DOIUrl":"10.1109/TNSRE.2025.3608567","url":null,"abstract":"Dynamic optimization is a versatile control tool to determine optimal control inputs in a redundantly actuated wearable robot. However, dynamic optimization requires high computational resources for real-time implementation. In this paper, we present a bio-inspired control approach, based on the principle of muscle synergies, to reduce the computational cost of optimization. The most important linear combinations of actuators, dubbed “artificial synergies,” were identified for the double support phase (DSP) and single support phase (SSP) of walking, allowing for hip, knee, and ankle actuation. In simulations, we compared the bio-inspired (input dimensionality reduced) model predictive control (MPC) with a conventional MPC using the full-dimensional actuation model. For both the DSP and SSP, incorporating synergies reduces the mean number of iterations per optimization step. A minimum number of synergies are indeed necessary to truly achieve redistribution of control effort across the other actuators when a primary muscle is fatigued. Additionally, we provide a practical approach to conduct real-time experiments with the bio-inspired MPC. A data-driven modeling approach is used to identify the nonlinear musculoskeletal dynamics and extract personalized artificial synergies from the experimental hybrid exoskeleton walking data. Synergistic MPC reduces computation time by an average of 28.16% (<inline-formula> <tex-math>${p}lt {0}.{03}$ </tex-math></inline-formula>) compared to full-dimensional MPC. Furthermore, we demonstrate control redistribution in response to varying cost function penalties on individual synergy activations. It is, to the authors’ knowledge, the first instance of artificial synergy-based MPC in real-time for a hybrid gait exoskeleton. This study provides insight into the use of bio-inspiration for hybrid exoskeleton control and other rehabilitation systems with redundant actuators.","PeriodicalId":13419,"journal":{"name":"IEEE Transactions on Neural Systems and Rehabilitation Engineering","volume":"33 ","pages":"3755-3769"},"PeriodicalIF":5.2,"publicationDate":"2025-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11156144","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145033155","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Static Force Prediction: Comparative Analysis and Fusion Strategies Between Surface Electromyography and Electrical Impedance Myography","authors":"Pan Xu;Junwei Zhou;Yuandong Zhuang;Xinyu Li;Zeljka Lucev Vasic;Mario Cifrek;Yuqing Liu;Yueming Gao","doi":"10.1109/TNSRE.2025.3607757","DOIUrl":"10.1109/TNSRE.2025.3607757","url":null,"abstract":"Force prediction is crucial for functional rehabilitation of the upper limb. Surface electromyography (sEMG) signals play a pivotal role in muscle force studies, but its non-stationarity challenges the reliability of sEMG-driven models. This problem may be alleviated by fusion with electrical impedance myography (EIM), an active sensing technique incorporating tissue morphology information. This study designed a wearable multimodal physiological measurement system to acquire sEMG and EIM signals simultaneously. The feature quantification indexes were defined for quantitative analysis of the efficacy of EIM and sEMG in static force prediction. We finally proposed Self-Attention Convolutional Long Short-Term Memory (SACLSTM) network to capture the spatio-temporal information among EIM and sEMG features for cross-modal feature fusion. The results indicated that EIM exhibited greater sensitivity to variations in static force compared to sEMG, especially at low muscle activation levels. Furthermore, the proposed SACLSTM network is significantly superior to LSTM, ConvLSTM, and several other baseline methods. Compared to the LSTM and ConvLSTM networks, the SACLSTM model exhibits an <inline-formula> <tex-math>${R}^{{2}}$ </tex-math></inline-formula> improvement of 12.4% and 3%, respectively, and an root mean square error reduction of 63% and 29%. Especially for patients with upper limb dysfunction, the accuracy and stability of the multimodal model were significantly improved after feature fusion compared with using only EIM or sEMG unimodal features. This study emphasised the great potential of fusing EIM and sEMG features to improve performance in the muscle force prediction, opening up new practice paths in the field of functional motor rehabilitation.","PeriodicalId":13419,"journal":{"name":"IEEE Transactions on Neural Systems and Rehabilitation Engineering","volume":"33 ","pages":"3697-3708"},"PeriodicalIF":5.2,"publicationDate":"2025-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11154113","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145029781","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Resting-State EEG Functional Connectivity for Brain Function Analysis and Severity Classification in Obstructive Sleep Apnea","authors":"Minghui Liu;Ligang Zhou;Yalin Wang;Wentao Lin;Jingchun Luo;Cong Fu;Fengfei Ding;Wei Chen;Chen Chen","doi":"10.1109/TNSRE.2025.3607776","DOIUrl":"10.1109/TNSRE.2025.3607776","url":null,"abstract":"Obstructive sleep apnea (OSA), one of the most common sleep disorders globally, is closely linked to brain function. Resting-state electroencephalography (EEG), due to its convenience, cost-effectiveness, and high temporal resolution, serves as a valuable tool for exploring the human brain function. This study utilized a large cohort with 968 participants who joined in 15-minute daytime resting-state EEG acquisition and overnight polysomnography (PSG) monitoring. Participants were categorized into healthy controls and mild, moderate, and severe OSA groups based on apnea-hypopnea index (AHI) derived from PSG data. Resting-state EEG functional connectivity (FC) was estimated using correlation (Corr), coherence (Coh), phase-locking value (PLV), and phase lag index (PLI). Results showed that FC between most nodes increased with the OSA severity, which suggest the potential neural compensation. However, regional decreases emerged in the right central, right frontal, left central, and left parieto-occipital regions. Higher frequency bands exhibited fewer enhanced FC connections. Graph-theoretical analysis revealed reduced centrality, indicating weakened communication hubs and potential topological reorganization. Multivariate analysis with adjustment of age, sex, and BMI, was also used as a feature selection strategy, identified effective FC features of OSA severity (p value < adjusted significance threshold, 2.15e-5). These FC features were used in machine learning models for severity classification and enhanced interpretability. The Corr-based XGBoost model achieved the highest performance, with an accuracy of 0.79 and AUC of 0.90. These findings highlight OSA-related brain function alterations and demonstrate that resting-state EEG FC provides a non-invasive, task-free and interpretable tool for OSA severity classification without disrupting natural sleep.","PeriodicalId":13419,"journal":{"name":"IEEE Transactions on Neural Systems and Rehabilitation Engineering","volume":"33 ","pages":"3662-3673"},"PeriodicalIF":5.2,"publicationDate":"2025-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11153990","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145029798","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The Changes in Synergistic Patterns of Peri-Shoulder Muscles During Shoulder Abduction and Flexion in Patients With Rotator Cuff Tears","authors":"Miao-Qin Zhan;Jian-Ning Sun;Nan Zheng;Yu-Rong Li;Peng Chen","doi":"10.1109/TNSRE.2025.3606739","DOIUrl":"10.1109/TNSRE.2025.3606739","url":null,"abstract":"Understanding muscle synergy variability and its clinical relevance in rotator cuff tear (RCT) patients is crucial for elucidating motor control mechanisms and informing rehabilitation. This study uses non-negative matrix factorization (NMF) to assess the influence of age and pathological factors on synergy patterns during abduction (ABD) and flexion (FL) tasks. Fifteen young controls (YC), fifteen elderly controls (EC), and twenty elderly RCT patients were recruited. Surface electromyography (sEMG) signals from eight shoulder muscles were recorded for NMF analysis, and the correlation with the Constant-Murley Score (CMS) was evaluated via regression. Results revealed that the number of synergies in the FL task was significantly lower in the RCT group compared to the EC (<inline-formula> <tex-math>$4.5~pm ~1.3$ </tex-math></inline-formula> vs. <inline-formula> <tex-math>$5.6~pm ~1.1$ </tex-math></inline-formula>, P=0.01) and YC groups (<inline-formula> <tex-math>$4.5~pm ~1.3$ </tex-math></inline-formula> vs. <inline-formula> <tex-math>$5.6~pm ~1.0$ </tex-math></inline-formula>,P=0.03), with no significant difference in the ABD task (P=0.10). In ABD, synergy pattern 1 was dominated by the upper trapezius and middle trapezius, while in the EC group, pattern 4 was driven by the supraspinatus, and in the YC group, both supraspinatus and posterior deltoid were involved. In FL, the middle deltoid weight was significantly increased in pattern 3 in the RCT group, suggesting compensatory activation or a pathological marker. The root mean square value of middle deltoid strongly correlated with CMS (ABD: R<inline-formula> <tex-math>${}^{2} =0.53$ </tex-math></inline-formula>,P< 0.001; FL: R<inline-formula> <tex-math>${}^{{2}} =0.46$ </tex-math></inline-formula>,P< 0.001). These findings reveal task-specific compensation mechanisms in RCT patients and provide insights for targeted rehabilitation strategies.","PeriodicalId":13419,"journal":{"name":"IEEE Transactions on Neural Systems and Rehabilitation Engineering","volume":"33 ","pages":"3626-3636"},"PeriodicalIF":5.2,"publicationDate":"2025-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11151991","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145006018","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"KAleep-Net: A Kolmogorov-Arnold Flash Attention Network for Sleep Stage Classification Using Single-Channel EEG With Explainability","authors":"Zubair Akbar;Farhad Hassan;Jingzhen Li;Ubaidullah Alias Kashif;Yuhang Liu;Jia Gu;Kaixin Zhou;Zedong Nie","doi":"10.1109/TNSRE.2025.3606128","DOIUrl":"10.1109/TNSRE.2025.3606128","url":null,"abstract":"Sleep monitoring is essential for assessing sleep quality and understanding its broader implications for overall health. Although electroencephalography (EEG) remains the gold standard for sleep analysis, multichannel techniques are often cumbersome and impractical for real-world application. As a more feasible alternative, single-channel EEG offers greater practicality but still faces several persistent challenges, including reduced spatial resolution, feature instability, and limited clinical interpretability. To address these limitations, we propose KAleep-Net (Kolmogorov-Arnold based Sleep Network) for sleep stage classification. It employs a Multispectral Feature Pipeline to extract both fine-grained and coarse-grained features from single-channel EEG signals. It integrates a Temporal Sequencing Network with Flash Attention to capture rich and stable features effectively. The proposed approach achieved an accuracy of 86.5%, an F1-score of 79.6%, and a Cohen’s <inline-formula> <tex-math>$kappa $ </tex-math></inline-formula> of 79.9% on the Sleep-EDF-20 dataset, along with a 41.7% improvement in training speed. For the Sleep-EDF-78 dataset, it attained 85.0% accuracy, 77.0% F1-score, 78.0% <inline-formula> <tex-math>$kappa $ </tex-math></inline-formula>, and a 67.5% gain in training efficiency. On the SHHS dataset, the model achieved 86.4% accuracy, an F1-score of 0.79, and a <inline-formula> <tex-math>$kappa $ </tex-math></inline-formula> of 0.81, with an 8.18% improvement in training speed. For interpretability, an integrated gradient technique was adopted to enhance decision transparency and promote clinical adoption. The framework offers an efficient solution for sleep staging in resource-constrained environments with clinically trusted insights for single-channel EEG-based sleep monitoring.","PeriodicalId":13419,"journal":{"name":"IEEE Transactions on Neural Systems and Rehabilitation Engineering","volume":"33 ","pages":"3685-3696"},"PeriodicalIF":5.2,"publicationDate":"2025-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11151617","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145000440","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}