Journal of neural engineering最新文献

{"title":"Enhancing EEG-based sleep staging efficiency with minimal channels through adversarial domain adaptation and active deep learning.","authors":"Roya Ghasemigarjan, Mohammad Mikaeili, Seyed Kamaledin Setarehdan, Arash Saboori","doi":"10.1088/1741-2552/adeec7","DOIUrl":"10.1088/1741-2552/adeec7","url":null,"abstract":"<p><p><i>Objective</i>. Accurate sleep-stage classification is crucial for advancing both sleep research and healthcare applications. Traditional deep learning (DL) and domain adaptation (DA) methods often struggle due to the limited availability of labeled data in the target domain and their inability to capture the subtle distinctions between sleep-stage classes, which hampers classification accuracy.<i>Approach</i>. To address these limitations, we introduce a novel framework, adversarial domain adaptation with active deep learning (ADAADL). This framework combines adversarial learning with active learning (AL) strategies to improve feature alignment and effectively leverage unlabeled data. ADAADL employs two separate sleep-stage classifiers as discriminators, allowing for a more refined consideration of class boundaries during the feature alignment process. Moreover, it incorporates entropy measures alongside cross-entropy loss during training to make better use of the information from unlabeled data. The AL component (ADL) further enhances performance by iteratively selecting and labeling the most informative data points, thereby reducing annotation efforts and improving generalization to unseen data.<i>Main results.</i>Experimental evaluations on three benchmark EEG datasets demonstrate that ADAADL produces robust, transferable features, significantly outperforming existing DA methods in classification accuracy. This research advances sleep-stage classification techniques, offering improved accuracy for real-world applications and contributing to a deeper understanding of sleep dynamics.<i>Significance</i>. The proposed ADAADL framework advances the state of the art in sleep-stage classification by effectively leveraging unlabeled data and reducing labeling costs. It offers a scalable and accurate solution for real-world sleep monitoring applications and contributes to a deeper understanding of sleep dynamics through improved modeling of sleep stages.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2025-08-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144621621","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":"xEEGNet: towards explainable AI in EEG dementia classification.","authors":"Andrea Zanola, Louis Fabrice Tshimanga, Federico Del Pup, Marco Baiesi, Manfredo Atzori","doi":"10.1088/1741-2552/adf6e6","DOIUrl":"10.1088/1741-2552/adf6e6","url":null,"abstract":"<p><p><i>Objective.</i>This work presents xEEGNet, a novel, compact, and explainable neural network for electroencephalography (EEG) data analysis. It is fully interpretable and reduces overfitting through a major parameter reduction.<i>Approach.</i>As an applicative use case to develop our model, we focused on the classification of common dementia conditions, Alzheimer's and frontotemporal dementia, versus controls. xEEGNet, however, is broadly applicable to other neurological conditions involving spectral alterations. We used ShallowNet, a simple and popular model in the EEGNet family, as a starting point. Its structure was analyzed and gradually modified to move from a 'black box' model to a more transparent one, without compromising performance. The learned kernels and weights were analyzed from a clinical standpoint to assess their medical significance. Model variants, including ShallowNet and the final xEEGNet, were evaluated using a robust nested-leave-n-subjects out cross-validation for unbiased performance estimates. Variability across data splits was explained using the embedded EEG representations, grouped by class and set, with pairwise separability to quantify group distinction. Overfitting was measured through training-validation loss correlation and training speed.<i>Main results.</i>xEEGNet uses only 168 parameters, 200 times fewer than ShallowNet, yet retains interpretability, resists overfitting, achieves comparable median performance (-1.5%), and reduces performance variability across splits. This variability is explained by the embedded EEG representations: higher accuracy correlates with greater separation between test-set controls and Alzheimer's cases, without significant influence from the training data.<i>Significance.</i>The capability of xEEGNet to filter specific EEG bands, learns band specific topographies and use the right EEG spectral bands for disease classification demonstrates its interpretability. While big deep learning models are typically prioritized for performance, this study shows that smaller architectures like xEEGNet can be equally effective in pathology classification, using EEG data.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2025-08-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144769473","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":"ML-STIM: Machine Learning for SubThalamic nucleus Intraoperative Mapping.","authors":"Fabrizio Sciscenti, Valentina Agostini, Laura Rizzi, Michele Lanotte, Marco Ghislieri","doi":"10.1088/1741-2552/adf579","DOIUrl":"10.1088/1741-2552/adf579","url":null,"abstract":"<p><p><i>Objective.</i>Deep Brain Stimulation (DBS) of the SubThalamic Nucleus (STN) is effective in alleviating motor symptoms in medication-refractory patients with Parkinson's Disease (PD). Intraoperative identification of the STN relies on MicroElectrode Recordings (MERs), typically analyzed by trained operators. However, this approach is time-consuming and subject to variability. For this reason, this study proposes ML-STIM (Machine Learning for SubThalamic nucleus Intraoperative Mapping), a ML pipeline designed to automate STN classification from MERs, ensuring high accuracy and real-time performance.<i>Approach.</i>ML-STIM consists of MERs pre-processing, feature extraction, and classification using a MultiLayer Perceptron. An adaptive artifact removal algorithm was optimized to balance artifacts identification and STN signal preservation, and the features were selected among those recommended in literature through correlation analysis and ReliefF ranking. The pipeline was trained and validated on a public dataset (Dataset A, 46 patients) and tested on an independent dataset (Dataset B, 36 patients), from a different surgical center, to assess generalizability. Dataset B is made publicly available as well.<i>Main Results.</i>ML-STIM achieved 87.8 ± 1.7% accuracy on Dataset A and 83.8 ± 1.6% accuracy on Dataset B, significantly outperforming a state-of-the-art deep learning model (ResNet-AT,<i>p</i>< 0.01). The artifact removal step significantly improved classification specificity (<i>p</i>< 0.001). ML-STIM processed raw 10-second recordings in 139.4 ± 2.1 ms, demonstrating real-time feasibility.<i>Significance.</i>These results confirm ML-STIM as an accurate, interpretable, and computationally efficient solution for intraoperative STN identification in DBS surgeries.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2025-08-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144746657","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":"Prioritized learning of cross-population neural dynamics.","authors":"Trisha Jha, Omid G Sani, Bijan Pesaran, Maryam M Shanechi","doi":"10.1088/1741-2552/ade569","DOIUrl":"10.1088/1741-2552/ade569","url":null,"abstract":"<p><p><i>Objective</i>. Improvements in recording technology for multi-region simultaneous recordings enable the study of interactions among distinct brain regions. However, a major computational challenge in studying cross-regional, or cross-population dynamics in general, is that the cross-population dynamics can be confounded or masked by within-population dynamics.<i>Approach</i>. Here, we propose cross-population prioritized linear dynamical modeling (CroP-LDM) to tackle this challenge. CroP-LDM learns the cross-population dynamics in terms of a set of latent states using a prioritized learning approach, such that they are not confounded by within-population dynamics. Further, CroP-LDM can infer the latent states both causally in time using only past neural activity and non-causally in time, unlike some prior dynamic methods whose inference is non-causal.<i>Main results</i>. First, through comparisons with various LDM methods, we show that the prioritized learning objective in CroP-LDM is key for accurate learning of cross-population dynamics. Second, using multi-regional bilateral motor and premotor cortical recordings during a naturalistic movement task, we demonstrate that CroP-LDM better learns cross-population dynamics compared to recent static and dynamic methods, even when using a low dimensionality. Finally, we demonstrate how CroP-LDM can quantify dominant interaction pathways across brain regions in an interpretable manner.<i>Significance</i>. Overall, these results show that our approach can be a useful framework for addressing challenges associated with modeling dynamics across brain regions.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2025-08-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12337745/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144319039","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":"Implications of shared motor and perceptual activations on the sensorimotor cortex for neuroprosthetic decoding.","authors":"Alexander B Silva, Jessie R Liu, Vanessa R Anderson, Cady M Kurtz-Miott, Irina P Hallinan, Kaylo T Littlejohn, Samantha C Brosler, Adelyn Tu-Chan, Karunesh Ganguly, David A Moses, Edward F Chang","doi":"10.1088/1741-2552/adf50e","DOIUrl":"10.1088/1741-2552/adf50e","url":null,"abstract":"<p><p><i>Objective.</i>Neuroprostheses can restore communicative ability to people with paralysis by decoding intended speech motor movements from the sensorimotor cortex (SMC). However, overlapping neural populations in the SMC are also engaged in visual and auditory perceptual processing. The nature of these shared motor and perceptual activations and their potential to interfere with decoding are particularly relevant questions for speech neuroprostheses, as reading and listening are essential daily functions.<i>Approach.</i>In two participants with vocal-tract paralysis and anarthria (ClinicalTrials.gov; NCT03698149), we developed an online electrocorticography (ECoG) based speech-decoding system that maintained accuracy and specificity to intended speech, even during common daily tasks like reading and listening. Offline, we studied the spectrotemporal characteristics and spatial distribution of reading, listening, and attempted-speech responses across our participants' ECoG arrays.<i>Main results.</i>Across participants, the speech-decoding system had zero false-positive activations during 63.2 min of attempted speech and perceptual tasks, maintaining accuracy and specificity to volitional speech attempts. Offline, though we observed shared neural populations that responded to attempted speech, listening, and reading, we found they leveraged different neural representations with differentiable spectrotemporal responses. Shared populations localized to the middle precentral gyrus and may have a distinct role in speech-motor planning.<i>Significance.</i>Potential neuroprosthesis users strongly desire reliable systems that will retain specificity to volitional speech attempts during daily use. These results demonstrate a decoding framework for speech neuroprostheses that maintains this specificity and further our understanding of shared perceptual and motor activity on the SMC.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2025-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12337102/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144736420","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":"Mixup-based data augmentation for enhancing few-shot SSVEP detection performance.","authors":"Jiayang Huang, Pengfei Yang, Bang Xiong, Yidan Lv, Quan Wang, Bo Wan, Zhi-Qiang Zhang","doi":"10.1088/1741-2552/adf467","DOIUrl":"10.1088/1741-2552/adf467","url":null,"abstract":"<p><p><i>Objective.</i>Few-shot steady-state visual evoked potential (SSVEP) detection remains a major challenge in brain-computer interface (BCI) systems, as limited calibration data often leads to degraded performance. This study aims to enhance few-shot SSVEP detection through an effective data augmentation (DA) strategy.<i>Approach.</i>We propose a mixup-based DA method that generates synthetic trials by linearly interpolating between real SSVEP signals extracted using a sliding window strategy. The interpolation weight is optimized by maximizing the similarity between the mixed signal and both the template and reference signals. The augmented data is then used to train spatial filters for improved SSVEP detection.<i>Main results.</i>The proposed method was evaluated on two benchmark SSVEP datasets using task-related component analysis and incorporating neighboring stimuli data as spatial filters. Results demonstrate that the mixup-based augmentation significantly improves detection accuracy under few-shot conditions, outperforming existing augmentation and baseline methods.<i>Significance.</i>The mixup-based method offers an effective and practical solution for enhancing SSVEP decoding with limited data, reducing calibration time, and improving BCI systems' usability in real-world scenarios.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2025-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144719293","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":"Revisiting convolutive blind source separation for identifying spiking motor neuron activity: From theory to practice.","authors":"Thomas Klotz, Robin Rohlén","doi":"10.1088/1741-2552/adf886","DOIUrl":"https://doi.org/10.1088/1741-2552/adf886","url":null,"abstract":"<p><strong>Objective: </strong>Identifying the spiking activity of alpha motor neurons (MNs) non-invasively is possible by decomposing signals from active muscles, e.g., obtained with surface electromyography (EMG) or ultrasound. The theoretical background of MN identification using these techniques is convolutive blind source separation (cBSS), in which different algorithms have been developed and validated. However, the existence and identifiability of inverse solutions and the corresponding estimation errors are not fully understood. In addition, the guidelines for selecting appropriate parameters are often built on empirical observations, limiting the translation to clinical applications and other modalities.&#xD;Approach: We revisited the cBSS model for EMG-based MN identification, augmented it with new theoretical insights and derived a framework that can predict the existence of solutions for spike train estimates. This framework allows the quantification of source estimation errors due to the imperfect inversion of the motor unit action potentials (MUAP), physiological and non-physiological noise, and the ill-conditioning of the inverse problem. To bridge the gap between theory and practice, we used computer simulations.&#xD;Main results: (1) Increasing the similarity of MUAPs or the correlation between spike trains increases the bias for detecting MN spike trains linked with high amplitude MUAPs. (2) The optimal objective function depends on the expected spike amplitude, spike amplitude statistics and the amplitude of background spikes. (3) There is some wiggle room for MN detection given non-stationary MUAPs. (4) There is no connection between MUAP duration and extension factor, in contrast to previous guidelines. (5) Source quality metrics like the silhouette score (SIL) or the pulse-to-noise ratio (PNR) are highly correlated with a source's objective function output. (6) Considering established source quality measures, SIL is superior to PNR.&#xD;Significance: We expect these findings will guide cBSS algorithm developments tailored for MN identification and translation to clinical applications.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2025-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144796553","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":"Synergic practice with a body-machine interface: implications for individual and collective motor learning.","authors":"Amy Bellitto, Ferdinando A Mussa-Ivaldi, Camilla Pierella, Maura Casadio","doi":"10.1088/1741-2552/adeec9","DOIUrl":"10.1088/1741-2552/adeec9","url":null,"abstract":"<p><p><i>Objective</i>. Body-machine interfaces (BoMIs) translate human body movements into commands for controlling external devices, such as computer cursors. This process allows researchers to study the development and refinement of inverse models, which generate motor commands necessary for achieving desired movements. Traditionally, motor learning has focused on solo practice, but recent research has shifted towards exploring dyadic tasks, where two individuals practice together. Within dyadic tasks, synergic practice-where partners collaborate toward a shared goal-has shown promise in enhancing performance and reducing stress. However, the impact of contributions of each partner within synergic practice on individual and collective learning remains underexplored. This study aims to (i) investigate how different levels of contribution during synergic practice affect both individual and collective motor learning, and (ii) assess the impact of these contribution levels on individual performance when returning to solo practice.<i>Approach</i>. Forty naïve participants underwent individual practice, dyadic synergic practice, and a final round of individual practice using a BoMI to control a cursor. Participants were classified as high or low contributors based on their participation in the cursor trajectory during dyadic practice. We analyzed how these contribution levels influenced performance, motor strategies, and internal models during and after the dyadic phase.<i>Main results</i>. During dyadic practice, high contributors maintained motor strategies similar to their initial solo performance, while low contributors showed significant deviations. After returning to solo practice, high contributors retained better task performance, whereas low contributors initially regressed but quickly improved with additional practice, eventually matching high contributors' performance levels.<i>Significance</i>. This understanding holds practical implications for optimizing dyadic practice. Our study sheds light on the influence of synergic practice on subsequent individual motor performance, contributing to a clearer understanding of its advantages and limitations for optimal implementation.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2025-08-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144621622","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 two-stage EEG zero-shot classification algorithm guided by class reconstruction.","authors":"Li Li, Baofa Wei","doi":"10.1088/1741-2552/adeaea","DOIUrl":"10.1088/1741-2552/adeaea","url":null,"abstract":"<p><p><i>Objective</i>. Researchers have long been dedicated to decoding human visual representations from neural signals. These studies are crucial in uncovering the mechanisms of visual processing in the human brain. Electroencephalogram (EEG) signals have garnered widespread attention recently due to their non-invasive nature and low cost. EEG classification is one of the most popular topics in brain-computer interface research. However, most traditional EEG classification algorithms are difficult to generalize to unseen classes that were not involved in the training phase. The main objective of this work is to improve the performance of these EEG classification algorithms for unseen classes.<i>Approach</i>. In this work, we propose a two-stage zero-shot EEG classification algorithm guided by class reconstruction. The method is specifically designed with a two-stage training strategy based on class reconstruction. This structure and training strategy enable the model to thoroughly learn the relations and distinctions among EEG embeddings of different classes. The contrastive language-image pre-training (CLIP) model has a well-aligned latent space and powerful cross-modality generalization ability.<i>Main results</i>. We conducted experiments on the ImageStimulus-EEG dataset to evaluate the performance of the proposed method. Meanwhile, it was compared with the state-of-the-art model and the baseline model. The experimental results demonstrate that our model achieves superior performance in among Top-1, Top-3, and Top-5 classification accuracy for a 50-way zero-shot classification task, reaching 17.77%, 38.76% and 54.75%, respectively.<i>Significance</i>. The proposed method bridges the modality gap between EEG, images, and text using CLIP features. It significantly improves the model's performance in unseen classes. The experimental results validate the effectiveness of it in EEG zero-shot classification.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2025-08-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144556291","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":"Preictal high-connectivity states in epilepsy: evidence of intracranial EEG, interplay with the seizure onset zone and network modeling.","authors":"Nicolas Medina, Manel Vila-Vidal, Ana Tost, Mariam Khawaja, Mar Carreño, Pedro Roldán, Jordi Rumià, María Centeno, Estefanía Conde, Antonio Donaire, Adrià Tauste Campo","doi":"10.1088/1741-2552/adf097","DOIUrl":"10.1088/1741-2552/adf097","url":null,"abstract":"<p><p><i>Objective.</i>Epilepsy affects around 50 million people worldwide, and reliable pre-seizure biomarkers could significantly improve neuromodulation therapies for drug-resistant patients. Recent research using stereo-electroencephalography (sEEG) has revealed transient changes in network dynamics preceding seizures. In particular, our previous work showed that these alterations are driven by recurrent, short-lasting (0.6 s) high-connectivity network configurations-termed high-connectivity states (HCSs). Here, we aim to replicate and further characterize HCS as a biomarker in a multicentric patient cohort, assess its robustness across recording modalities and montages, explore its relationship with interpretable physiological variables, and examine its network-level association with seizure-onset zone (SOZ) dynamics.<i>Approach.</i>We analyzed long-term intracranial EEG recordings from 12 patients with sEEG and electrocorticography. In two patients with extensive clinical information, we examined the interplay between HCS and SOZ dynamics. We also developed a low-dimensional stochastic network model to investigate mechanistic rationales of HCS emergence. Additionally, we compared HCS dynamics with gamma-band activity and heart rate, and tested robustness across different montage configurations.<i>Main Results.</i>In most patients, HCS probability reliably increased hours before seizure onset. In the two deeply characterized patients, this increase was specifically linked to an increased network centrality within the SOZ. The network model revealed that changes in HCS probability stem primarily from topological reconfigurations rather than changes in mean connectivity, underscoring the importance of dynamic interactions between epileptogenic and non-epileptogenic regions.<i>Significance.</i>These results support HCS probability as a promising biomarker for early seizure prediction and offer mechanistic insights into pre-seizure brain network dynamics.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2025-08-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144651617","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}