Journal of neural engineering最新文献

{"title":"Harmonizing complexity and efficiency: Helix Fusion HarmonyNet's breakthrough in EEG-based postoperative delirium recognition.","authors":"Conghui Wei, Xiuqin Rao, Bo Hu, Juan Zeng, Pengcheng Yi, Huasu Shao, Qianqian Tan, Yu Xie, Liyuan Fang, Yan Gong, Fuzhou Hua","doi":"10.1088/1741-2552/aea18f","DOIUrl":"10.1088/1741-2552/aea18f","url":null,"abstract":"<p><p><i>Objective.</i>Postoperative delirium (POD) is a common perioperative complication involving central nervous system dysfunction, particularly among critically ill and elderly patients, yet its rapid and objective detection remains challenging in clinical settings. To develop and evaluate a lightweight, interpretable electroencephalography (EEG)-based deep learning framework for POD detection that remains accurate under sparse-electrode acquisition and practical for bedside or portable deployment.<i>Approach.</i>We propose Helix Fusion HarmonyNet (HFHN), a transform-domain architecture that explicitly separates EEG representations into amplitude and phase pathways. HFHN integrates multi-head self-attention, convolutional local feature extraction, learnable sinusoidal positional encoding, and controlled cross-pathway fusion to jointly model spectral intensity, temporal synchrony, local patterns, and long-range dependencies. The framework was evaluated against representative CNN-, Transformer-, and EEG-specific deep learning models across signal domains, fusion strategies, electrode configurations, ablation settings, and deployment-oriented efficiency tests.<i>Main Results.</i>In Fourier-domain POD detection, HFHN achieved 100% accuracy, precision, specificity,<i>F</i>1-score, and sensitivity. Ablation analysis confirmed the importance of amplitude-phase interaction, as removing cross-pathway fusion reduced accuracy from 100% to 95.38%. Under sparse-channel conditions, HFHN maintained 99.42% accuracy and a 99.33%<i>F</i>1-score with 8 electrodes, and 94.08% accuracy and a 93.26%<i>F</i>1-score with only 4 high-attention electrodes. Reducing the number of electrodes from 60 to 4 decreased FLOPs from 16.95 G to 0.99 G, while inference latency remained below 3.09 ms.<i>Significance.</i>HFHN provides an accurate, interpretable, and computationally efficient solution for EEG-based POD detection. Its robustness under sparse-channel acquisition and low-latency performance support potential integration into portable or bedside EEG systems for rapid postoperative monitoring in wards and intensive care units.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-10-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148883122","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Spectrally-resolved transfer entropy in resting-state SEEG reveals epileptogenic and non-epileptogenic communication in adults and children with refractory epilepsy.","authors":"Francois Okoroafor, Ghassan S Makhoul, Bruno Hidalgo Monroy Lerma, Anas Reda, Zekai Qiang, Addison C Cavender, Kate Wang, Emily Liao, Aswin Chari, Rory J Piper, Tim Denison, Rod C Scott, Victoria L Morgan, Torsten Baldeweg, Martin M Tisdall, Dario J Englot","doi":"10.1088/1741-2552/aeabf3","DOIUrl":"https://doi.org/10.1088/1741-2552/aeabf3","url":null,"abstract":"<p><p><i>Objective</i>. To characterise directed, frequency-resolved communication between seizure onset zone (SOZ) and non-involved zone (NIZ) regions in resting-state stereo-electroencephalography (SEEG) from adults and children with drug-resistant epilepsy (DRE).<i>Approach</i>. We analysed retrospective SEEG from 21 adults and 18 paediatric patients who underwent focal resective surgery. Channels were grouped into four link classes (SOZ → SOZ, NIZ → SOZ, SOZ → NIZ, NIZ → NIZ). Signals were decomposed into delta, theta, alpha, beta, gamma, and broadband bands, and transfer entropy (TE) was estimated with a Gaussian estimator. For each source-target channel pair, an optimal temporal delay was selected by scanning delays from 2-200 ms and choosing the delay that maximised broadband TE. Significance testing used 100 phase-randomised surrogates per source and per target; Benjamini-Hochberg false discovery rate correction (<i>α</i>= 0.05) was applied within each band-to-band TE grid (separately for source- and target-surrogate<i>p</i>-values).<i>Main results</i>. In NIZ → SOZ links median communication delays were longer in adults than children (median 87.9 ms vs 68.0 ms; Mann-Whitney<i>p</i>< 1 × 10^-10; Kolmogorov-Smirnov<i>p</i>< 1 × 10^-10). Across most link classes, dominant source frequencies were broadband, delta, and theta. A distinct SOZ → SOZ pattern emerged: adults showed broadband, alpha, and beta source dominance with prominent broadband → gamma transfer, whereas children showed broadband, delta, and theta dominance while retaining broadband → gamma contribution. Cross-link-class source and target frequency distributions differed significantly in both cohorts (<i>p</i>< 0.05). While intra-regional SOZ-SOZ communication revealed unique multiplex signatures, inter-regional NIZ → SOZ interactions did not show a distinct spectral profile from non-involved connectivity.<i>Significance</i>. Resting-state SEEG communication in DRE exhibits multiplex, cross-frequency directed interactions. Spectrally resolved TE reveals conserved and cohort-specific shifts in SOZ-NIZ communication motifs. While these differences are confounded by pathology and site variations, they provide a frequency-specific characterization of the epileptogenic network that may support grounded interpretation of resting-state dynamics.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":"23 5","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-10-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149129726","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Ordered onset of regional desynchronization and a transition in movement-class discriminability during cue-paced motor imagery.","authors":"Seho Lee","doi":"10.1088/1741-2552/aeb161","DOIUrl":"https://doi.org/10.1088/1741-2552/aeb161","url":null,"abstract":"<p><p>Motor imagery recruits a distributed sensorimotor network, yet the order in which its regions are engaged has not been tested statistically within the healthy brain, and the interval in which information about the imagined movement appears remains unidentified. Early post-cue classification peaks have been attributed to visually induced orienting rather than to imagery, an inference drawn from a condition contrast rather than direct measurement.Two independent public datasets were analysed under a matched pipeline of causal filtering, ocular regression and current source density: Weibo2014 as the primary dataset and BCI Competition IV dataset 2a for validation. Mu and beta desynchronization onset was ordered across cortex, from occipital through sensorimotor and midline pre-supplementary motor area to lateral prefrontal. The ordering held under all nine onset criteria examined, under channel-count matching and within-trial pairing, and after removal of the sensorimotor component. Class-discriminative information rose most steeply between 0.60 and 0.75 s and saturated within a further 0.3 s. This transition lay closer to sensorimotor onset than to cue presentation in every participant and fell at or after that onset in nearly all of them, although the two were not shown equivalent at the resolution of the decoding grid. Windows centred on each region's own onset classified equivalently to reassigned windows, so the ordering serves as a temporal landmark rather than a decodable feature, locating the interval in which information about the imagined movement becomes available.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-10-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149129734","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The effect of different observation modalities in motor imagery: immersive virtual reality produces stronger desynchronization than video and real-world presentations.","authors":"Daniel Carrillo-Martín, Mateo Fortich-Mora, J Ignacio Serrano, M Dolores Del Castillo","doi":"10.1088/1741-2552/aeabf4","DOIUrl":"10.1088/1741-2552/aeabf4","url":null,"abstract":"<p><p><i>Objective</i>. The combination of action observation and motor imagery (AO + MI) has emerged as a promising neurorehabilitation strategy to promote sensorimotor plasticity without real movement, making it particularly valuable for individuals whose motor capacity is limited or absent. Improving AO + MI protocols requires understanding whether different observation stimuli engage the sensorimotor system to varying degrees, as the quality of cortical activation during each training trial may directly influence cumulative neuroplastic outcomes. The growing accessibility of immersive technologies such as virtual reality head-mounted displays and robotic agents opens new possibilities for enriching the observation component of AO + MI, yet the neurophysiological impact of these modalities relative to conventional video approaches remains largely unexplored. Therefore, this study aimed to determine how the observation modality influences the activation of the sensorimotor cortex during synchronous AO + MI.<i>Approach</i>. This study investigated the effects of three observation modalities: immersive virtual reality (iVR), two-dimensional video and real-world observation of a robotic agent, on sensorimotor cortical engagement, using event-related desynchronization (ERD) as a neurophysiological marker. Electroencephalographic signals were recorded from fifteen healthy participants across three sessions, one per modality, while they observed a humanoid robot performing arm movements and simultaneously engaged in motor imagery of the same movements. ERD was computed for alpha and beta frequency bands over eight electrodes covering frontocentral, central, and centroparietal regions.<i>Main Results</i>. Statistical analyses revealed significant modality-dependent differences in sensorimotor cortical engagement across multiple frequency bands, electrode locations, and movement types. iVR consistently elicited the strongest desynchronization, with robust effects in the beta band over frontocentral and right centroparietal regions, and in the alpha band over the right central motor cortex.<i>Significance</i>. These results demonstrate that the modality of action observation significantly shapes sensorimotor cortical engagement during AO + MI. Our findings also support the integration of iVR as a neurophysiological grounded tool for optimizing AO + MI-based neurorehabilitation protocols and brain-computer interface applications.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-10-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149032343","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Towards maximizing optical access through Smart Dura.","authors":"Sergio Montalvo Vargo, Nari Hong, Gaku Hatanaka, Vishal Jain, Hefu Pan, Ruikang K Wang, Wyeth Bair, Azadeh Yazdan-Shahmorad, Maysamreza Chamanzar","doi":"10.1088/1741-2552/ae8eb3","DOIUrl":"10.1088/1741-2552/ae8eb3","url":null,"abstract":"<p><p><i>Objective.</i>To overcome the limitations of optical access caused by opaque electrodes in multimodal neural interfaces used in large-animal models, and to enable scalable, high-density electrophysiology with unobstructed optical imaging and optogenetic stimulation.<i>Approach.</i>In this work, we redesigned our previously demonstrated multimodal cortical surface array, the Smart Dura, by incorporating transparent indium tin oxide (ITO) electrodes. ITO provides good electrical conductivity, while providing high levels of transparency. We discuss the design, microfabrication, benchtop characterization and<i>in vivo</i>demonstration of transparent Smart Dura for multimodal brain interfacing.<i>In vivo</i>experiments were designed to show the feasibility of simultaneous electrophysiological recording with localized functional optical imaging through transparent Smart Dura in response to visual stimulation.<i>Main results.</i>The implementation of ITO electrodes increased the effective optical access of Smart Dura to 100%, eliminating obstruction from traditional metal electrodes. Bench-top characterization confirmed the electrical and optical performance of the transparent electrodes.<i>In vivo</i>results demonstrate direct visual access underneath the ITO electrodes for optical recording and perturbation with co-located simultaneous high-fidelity electrophysiological recording, validating the multimodal capabilities of the transparent Smart Dura.<i>Significance.</i>The transparent Smart Dura design removes major barriers to optically integrated, high-density electrophysiology in large-animal models and the number and density of ITO electrodes on Smart Dura can be increased without compromising the optical access. This platform advances multimodal neural interfacing in non-human primates and provides unique opportunities to study brain function in health and disease and design therapeutics that can readily translate to humans.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":"23 5","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-10-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13639829/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149120967","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Influence of out-of-plane electric fields in an ex vivo model of transcranial electrical stimulation.","authors":"Yuhyun Lee, Vishal Jain, Yuxin Guo, Maysamreza Chamanzar, Pulkit Grover, Mats Forssell","doi":"10.1088/1741-2552/aeb04a","DOIUrl":"https://doi.org/10.1088/1741-2552/aeb04a","url":null,"abstract":"<p><strong>Objective: </strong>Electrical stimulation is the most commonly used modality for neuromodulation, with a number of clinical and scientific applications. Although the basic mechanisms of neural activation by electrical stimulation are well understood, important questions remain regarding neural responses to diffuse electric fields such as those generated by transcranial electrical stimulation. For such diffuse fields, the orientation of the electric field with respect to the neurons being stimulated matters, but previous ex vivo brain slice studies-commonly used to study how neurons respond to external electric fields-have only considered electric fields applied in the plane of the slice.</p><p><strong>Approach: </strong>We used two electrode arrays-one oriented parallel and the other perpendicular to coronal brain slices-to generate electric fields oriented orthogonally to the slice as well as along the cortical column. We measured responses in Layer 5 of the somatosensory cortex using several 2-electrode stimulation configurations, while also varying the distance between the electrodes and the slice.</p><p><strong>Main results: </strong>Electrode patterns generating a large electric field component orthogonal to the slice evoked the largest neural responses at every electrode-slice distance. Moreover, neural thresholds were > 5× lower for electric fields orthogonal to the slice compared to electric fields along the cortical column.</p><p><strong>Significance: </strong>These results provide a better understanding of the brain slice as a model for noninvasive stimulation and could lead to new electrode configurations for efficient transcranial neural stimulation.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-10-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149120992","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A statistical mixture-of-experts framework for EMG artifact removal in EEG: empirical insights and a proof-of-concept application.","authors":"Benjamin J Choi, Griffin Milsap, Clara A Scholl, Francesco Tenore, Mattson Ogg","doi":"10.1088/1741-2552/aea4f8","DOIUrl":"10.1088/1741-2552/aea4f8","url":null,"abstract":"<p><p><i>Objective.</i>Effective control of neural interfaces is limited by poor signal quality. While neural network-based electroencephalography (EEG) denoising methods for electromyogenic (EMG) artifacts have gained traction in recent years, existing models perform suboptimally in settings with high noise. Since neural interfaces rely almost universally on some form of signal processing, uncovering effective algorithmic insights into EEG denoising is crucial for future success.<i>Approach.</i>To address the shortcomings of current machine learning (ML)-based denoising algorithms, we present a single-channel signal filtration algorithm driven by a new mixture-of-experts (MoEs) framework. Our algorithm leverages three new statistical insights into the EEG-EMG denoising problem: (1) EMG artifacts can be partitioned into quantifiable types to aid downstream MoE classification, (2) local experts trained on narrower signal-to-noise ratio ranges can achieve performance increases through specialization, and (3) correlation-based objective functions, in conjunction with rescaling algorithms, can enable faster convergence in a neural network-based denoising context.<i>Main results.</i>We empirically demonstrate these three insights into EMG artifact removal and use our findings to create a new downstream MoE denoising algorithm consisting of convolutional and recurrent neural networks. We tested all results on a major benchmark dataset (EEGdenoiseNet) collected from 67 subjects. We found that our MoE denoising model achieved competitive overall performance with contemporary ML denoising algorithms and strong lower bound performance in high noise settings. We further validated the framework on the SEED EEG corpus, observing statistically significant downstream classification gains under high-noise contamination.<i>Significance.</i>These preliminary results highlight the promise of our MoE framework for enabling advances in EMG artifact removal in EEG, especially in high noise settings. Further experimentation will be necessary to assess our MoE framework on a wider range of test cases and explore its downstream potential to unlock more effective neural interfaces.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148922125","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"BRAINet: a brain-region-aware interaction network for EEG-based diagnosis of disorders of consciousness.","authors":"Haoxiang Chen, Sha Zhao, Jie Yu, Jiquan Wang, Yumeng Bai, Chuan Xu, Shijian Li, Benyan Luo, Gang Pan","doi":"10.1088/1741-2552/aea7db","DOIUrl":"10.1088/1741-2552/aea7db","url":null,"abstract":"<p><p><i>Objective.</i>Reliable assessment and stratification of disorders of consciousness (DOC) is essential for patient care and clinical treatment planning. Electroencephalography (EEG) provides a non-invasive approach to measure neural activity and has shown promise in DOC assessment. However, most existing EEG-based approaches focus on binary unresponsive wakefulness syndrome (UWS)/minimally conscious state (MCS) classification and often process EEG channels as a whole, without explicitly modeling anatomical brain-region organization. In this study, our goal is to distinguish among UWS, minimally conscious state minus (MCS), and minimally conscious state plus (MCS) using resting-state EEG signals.<i>Approach.</i>We propose BRAINet, a brain-region-aware EEG framework for three-class DOC classification. BRAINet partitions EEG channels into five anatomical brain regions, extracts region-specific spatiotemporal and spectral features, models cross-region interactions using a Transformer-based attention module, and fuses the learned representations with approximate entropy features for final classification. We evaluated BRAINet on a clinical resting-state EEG dataset comprising 22 UWS, 24 MCS, and 15 MCSpatients using patient-wise five-fold cross-validation and comparisons with representative machine-learning and deep-learning baselines. Statistical comparisons were based on paired patient-level out-of-fold (OOF) predictions.<i>Main results.</i>BRAINet achieved the highest numerical performance among the compared methods. Across the five folds, its mean balanced accuracy (BA) was 54.07% at the epoch level and 59.89% at the subject level. Based on pooled patient-level OOF predictions, BRAINet achieved significantly higher BA than the best-performing baseline, Conformer (two-sided paired permutation test, Holm-adjusted).<i>Significance.</i>These results suggest that brain-region-aware EEG modeling may provide useful information for fine-grained UWS/MCS/MCSclassification. BRAINet provides an interpretable framework for exploring region-specific EEG representations in DOC and may support future studies on patient stratification and prognostic assessment.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148978847","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Influences of design aspects on recording functionality of cuff-type peripheral nerve interfaces.","authors":"Vera A Oppelt, Emma Mueller, Fabian Pfeiffer, Seline Mahler, Ronny Pfeifer, Martin Schuettler, Thomas Stieglitz","doi":"10.1088/1741-2552/aea36d","DOIUrl":"10.1088/1741-2552/aea36d","url":null,"abstract":"<p><p><i>Objective.</i>High-quality recording of peripheral nerve signals is a critical requirement for implantable neural interfaces, yet the influence of contact configuration and electrical insulation on cuff electrode recording quality remains poorly understood.<i>Approach.</i>Using a physical nerve model<i>in vitro</i>, the transfer function (TF) of commercially available split-cylinder and spiral cuff electrodes was measured for systematically varied contact configurations, including contact size, spacing, and size ratio of cathode to anode, as well as degree of electrical insulation. To derive single-fiber and compound action potentials, the TF was convolved with modeled action currents, enabling direct comparison of recording quality across designs.<i>Main results.</i>Contact edge spacing was found to be more decisive than contact surface area for recording amplitude, regardless of contact segmentation. In quasi-tripolar configurations, common mode suppression additionally reduced sensitivity to geometric variations. Both the length of the longitudinal insulation and the integrity of the electrode seal were found to strongly influence recording amplitude, with complete removal of the seal reducing amplitude by approximately 50%, outweighing all geometric design parameters combined.<i>Significance.</i>These findings establish electrical insulation integrity as the primary design criterion for cuff electrode recording performance, shifting focus away from contact geometry alone, and provide a practical framework for the development of next-generation peripheral nerve interfaces.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148908480","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An open-set learning framework for EEG-based sleep stage classification.","authors":"Qiunan Feng, Chang Li, Hu Peng, Heyuan Qiao, Rencheng Song, Xun Chen","doi":"10.1088/1741-2552/aea59e","DOIUrl":"10.1088/1741-2552/aea59e","url":null,"abstract":"<p><p><i>Objective.</i>Existing sleep stage classification methods generally rely on a closed set assumption. They assume all sleep categories are predefined, which overlooks unknown sleep patterns encountered in real world electroencephalogram (EEG) analysis. This study overcomes this constraint by identifying both known and unknown sleep stages.<i>Approach.</i>We developed a two stage open set learning framework for EEG sleep staging. First, an adaptive Gaussian method models each known category to compute Gaussian scores, utilizing class specific quantile adaptive thresholds to separate known from unknown samples. Second, we integrate a trainable decision boundary OpenMax (TDBOM) module. TDBOM optimizes decision boundaries during training to suppress samples with large boundary deviations, while calibrating normal samples to yield reliable probability estimates.<i>Main results.</i>Evaluated on three public EEG datasets, our framework accurately classifies known sleep stages and reliably detects unknown categories, significantly boosting open-set recognition performance.<i>Significance.</i>This work provides a practical and robust solution for EEG-based sleep stage classification under open-set conditions, facilitating more reliable sleep analysis and supporting potential applications in real-world clinical monitoring and sleep disorder assessment.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-09-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148932250","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}