Journal of neural engineering最新文献

{"title":"IMU-weighted multimodal fusion for robust gesture recognition across arm positions.","authors":"Shang Shi, Jianjun Meng, Zongtian Yin, Weichao Guo, Xiangyang Zhu","doi":"10.1088/1741-2552/aea36c","DOIUrl":"10.1088/1741-2552/aea36c","url":null,"abstract":"<p><p><i>Objective</i>. Human-machine interfaces (HMIs) based on surface electromyography (sEMG) and A-mode ultrasound have achieved high gesture recognition accuracy; however, most studies have been conducted under a single static arm position. In practice, arm position variation degrades performance and limits translation beyond fixed-position laboratory settings. Although A-mode ultrasound can outperform sEMG at fixed positions, it is more sensitive to positional changes, an issue that has rarely been examined. This study addresses the cross-position sensitivity of ultrasound-based HMIs by fusing ultrasound data with sEMG and inertial measurement unit (IMU) data.<i>Approach</i>. We propose an IMU-weighted multimodal fusion framework. A feature fusion model and an sEMG single-modality model are trained in parallel. At the prediction stage, the arm position is estimated from the IMU, and the angular offset relative to the training positions is used to compute a continuous weight for combining the two models' posterior probabilities. Near training positions, greater weight is placed on the high-accuracy fusion model; as the offset grows, the contribution progressively shifts toward the robust sEMG model to avoid further degradation caused by A-mode ultrasound.<i>Main results</i>. Experiments were conducted across nine arm positions ranging from forearm elevation to lowering. With only two training positions, the proposed IMU-weighted method achieved an average accuracy of 85.88% across all nine positions, significantly outperforming other models.<i>Significance</i>. The method effectively exploits the high static accuracy of A-mode ultrasound and the cross-position stability of sEMG. With a low training burden, the proposed method improves cross-arm-position decoding accuracy, providing a practical route toward robust gesture recognition based on A-mode ultrasound across varying arm positions.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-09-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148908510","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":"Continuous sliding pressure feedback enhances supernumerary robotic thumb proprioception over discrete vibrotactile feedback: behavioral and EEG evidence.","authors":"Haneen Alsuradi, Georgios Korres, Pi Ko, Joseph Hong, Mohamad Eid","doi":"10.1088/1741-2552/aea36f","DOIUrl":"10.1088/1741-2552/aea36f","url":null,"abstract":"<p><p><i>Objective</i>. Supernumerary robotic thumbs (SRTs) offer promising augmentation capabilities, yet their adoption remains limited by the lack of reliable proprioceptive awareness, which is critical for enabling intuitive control and cognitive integration. Vibrotactile (VF) and pressure feedback (PF) represent two widely used modalities for proprioception that differ fundamentally in their encoding of information: VF feedback is widely adopted due to its low cost and ease of wearability, but provides discrete, temporally patterned cues, whereas PF is expensive and bulky yet it delivers continuous, analog signals that more closely resemble natural proprioceptive input.<i>Approach</i>. This study compares discrete VF and continuous sliding PF for conveying SRT proprioceptive information in the absence of visual input, combining behavioral assessment with electroencephalography. A total of 42 participants were recruited and assigned to either a VF or PF group; they wore the SRT, received modality-specific sensory feedback during passive finger movements, and were asked to decode the start and end location of the finger's motion. Sensorimotor event-related desynchronization (ERD) in the alpha and beta bands and beta-band functional connectivity were analyzed to characterize cortical differences between modalities.<i>Main results</i>. Both modalities outperformed implicit feedback, with PF demonstrating superior accuracy compared to VF. VF elicited more sustained alpha-band ERD across both contralateral and ipsilateral sensorimotor cortices and sustained beta-band ERD in the ipsilateral sensorimotor cortex, suggesting prolonged cortical engagement and increased processing demand. PF elicited significantly higher beta-band small-worldness, indicating more efficient network-level integration of proprioceptive information.<i>Significance</i>. These findings demonstrate that continuous, analog feedback more effectively conveys proprioceptive information and sensorimotor integration than discrete, temporally encoded cues; informing the design of sensory augmentation systems that effectively integrate artificial limbs into the human body schema.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-09-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148908465","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":"In vitro modeling of the peripheral nerve: cell culture embodiments and outcomes.","authors":"Varadraj N Vernekar, Karla Rodriguez Pequeno, Rahul Rishal Sharma, Nguyen Kha Ai Vo, Romtin Golabi","doi":"10.1088/1741-2552/aeaa0e","DOIUrl":"https://doi.org/10.1088/1741-2552/aeaa0e","url":null,"abstract":"<p><p>Peripheral nerve injuries (PNI) are common and significantly affect the quality of life of patients. In vitro models of the peripheral nerve have emerged as a critical tool for studying peripheral nerve injury and regeneration. These existing models are cost-effective, reproducible, and offer controlled experimental conditions. However, they fail to adequately model the peripheral nerve because they lack the complex structural and functional interactions present in its in vivo environment. Advancements in cell culturing systems, biomaterials, and fabrication techniques are expected to bring in vitro models closer to physiological and clinical relevance, helping guide repair strategies for treating nerve injuries. In contrast to earlier reviews that focus on cell types or fabrication methods, this review takes a function-oriented approach, organizing in vitro models by the specific nerve processes or structures that they aim to replicate. This review begins with a discussion of the structure of the peripheral nerve, followed by a brief literature search methodology section, and a discussion on the advantages and challenges in the in vitro modeling of this tissue next. This is followed by a discussion focused on the creation of various in vitro models of the peripheral nerve used for the investigation of various nerve processes including a) axonal outgrowth, b) myelination, c) innervation, and d) vascularization along with illustrative figures and comparative analysis tables. Discussion includes also key aspects to be considered when designing in vitro models of the peripheral nerve such as selection of cell types, biomaterials, scaffold fabrication, nano-topography/fabrication, microfluidics/lab-on-a-chip, 3-D printing, deployment of growth factors and stem cells, and structural and functional characterization of the models. Finally, the clinical relevance of these models is discussed and the paper concludes with some impactful advances and future directions in the field.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-09-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149013505","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":"Eccentricity confound in EEG-based visual attention decoding from gaze-fixated neural tracking of motion in natural videos.","authors":"Yuanyuan Yao, Celina Salamanca González, Simon Geirnaert, Céline R Gillebert, Tinne Tuytelaars, Alexander Bertrand","doi":"10.1088/1741-2552/aea449","DOIUrl":"10.1088/1741-2552/aea449","url":null,"abstract":"<p><p><i>Objective.</i>Decoding visual attention from brain signals during naturalistic video viewing has emerged as a new direction in brain-computer interface research. Current methods assume that stronger coupling between object motion and neural activity indicates higher attention, but this can be confounded by eye movement artifacts and stimulus properties. This study investigates how visual eccentricity-the distance between a visual object and the fixation point-affects neural responses when eye movement artifacts are controlled.<i>Approach.</i>Electroencephalography (EEG) signals were recorded across three tasks that manipulated object eccentricity and attention conditions while participants maintained gaze fixation. Correlation analysis and match-mismatch decoding were performed to quantify the neural tracking of object motion.<i>Main results.</i>The analysis supports three conclusions: (1) neural tracking of object motion in natural videos works under gaze fixation; (2) the strength of this tracking under gaze fixation is modulated by attention; and (3) there exists a significant eccentricity confound in the EEG responses, with poorer neural tracking of motion at larger eccentricities.<i>Significance.</i>These results indicate that findings from previous free-viewing studies also reflect genuine neural processing rather than mere oculomotor artifacts. However, the identified eccentricity effect highlights a major limitation for current decoding approaches that assume coupling strength reflects attention levels alone.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148914717","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":"Neuromodulation effects of HF-rTMS and rPMS on stroke patients: an fNIRS study.","authors":"Zhenhua Song, Kai Wei, Huizi Zeng, Xiaoxuan Zhan, Qianqian Fan, Xiaojun Wang, Yanfang Sui, Binbin Li, Zhijie Zhao, Fengyan Liang","doi":"10.1088/1741-2552/aea36a","DOIUrl":"10.1088/1741-2552/aea36a","url":null,"abstract":"<p><p><i>Objective.</i>This study aims to explore the immediate neuromodulatory effects of a single session of high-frequency repetitive transcranial magnetic stimulation (HF-rTMS) versus repetitive peripheral magnetic stimulation (rPMS) on cortical hemodynamics and functional connectivity (FC) in patients with stroke.<i>Approach.</i>57 patients with unilateral stroke at varying post-stroke durations received a single pulse-count-matched session of either HF-rTMS (5 Hz, 1500 pulses) or rPMS (25 Hz, 1500 pulses). Functional near-infrared spectroscopy (fNIRS) monitored cortical hemodynamics during a motor task immediately before and after the intervention. The acute changes in oxyhemoglobin (HbO) response magnitude (root-mean-square (RMS)), response dispersion (Root-mean-square error (RMSE)), laterality index, and region-to-region FC were measured.<i>Main results.</i>Within-group analyses showed that HF-rTMS reduced task-evoked HbO RMS in selected frontal region of interest (ROI)s and reduced RMSE in the dorsolateral prefrontal cortex (DLPFC) and frontopolar area (FPA), and showed an exploratory decrease in DLPFC-inferior frontal gyrus (IFG) FC and a descriptive negative laterality index tendency. Conversely, rPMS increased HbO RMS in the M1 ROI, which comprised M1 channels from both hemispheres, increased FPA RMSE, and showed an exploratory increase in S1-IFG FC with a descriptive positive laterality index tendency. No FC difference remained significant after false discovery rate correction. The disease-duration-adjusted FPA RMSE Time × Group interaction was significant (<i>F</i>(1,54) = 15.595,<i>p</i>< 0.001, partial<i>η</i>²= 0.224) and remained significant after exclusion of the participant with an extreme post-stroke duration.<i>Significance.</i>This study provides fNIRS evidence that HF-rTMS and rPMS are associated with different acute cortical response profiles in stroke. By directly comparing central and peripheral magnetic stimulation under a pulse-count-matched design, it characterizes their differing associations with post-stroke cortical hemodynamics and network-level responses.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148908489","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":"CORSICA: reproducible suppression of cochlear implant artifacts in EEG evoked by continuous speech.","authors":"Constantin Jehn, Clara Stiller, Niki Katerina Vavatzanidis, Tobias Reichenbach","doi":"10.1088/1741-2552/aea36e","DOIUrl":"10.1088/1741-2552/aea36e","url":null,"abstract":"<p><p><i>Objective.</i>Electroencephalography (EEG) is a key tool for studying auditory processing in cochlear implant (CI) users. In particular, EEG recordings obtained during continuous speech are becoming increasingly important for assessing speech and language processing in CI users, and may be utilized for neurofeedback. However, CIs also induce strong stimulation artifacts that are time-locked to the stimulus and mask the neural responses that have smaller magnitudes. Existing artifact reduction methods are typically based on event-related potentials or require manual component selection, making them unsuitable for naturalistic listening conditions or large datasets.<i>Approach.</i>We develop CORSICA (CORrelation-baSed ICA artifact rejection), a reproducible, parameter-efficient method for CI artifact reduction in EEG responses to continuous speech. CORSICA operates on independent components (ICs) obtained through Infomax ICA and requires no manual component labeling, with performance governed by a single tunable threshold. It exploits the observation that CI artifacts temporally follow the audio signal without delay, whereas neural responses have an inherent lag due to auditory pathway latencies. For each IC, CORSICA computes the cross-correlation with the speech stimulus. Artifacts are identified by a high signal-to-noise ratio (SNR) of the correlation peak near zero lag, and the component is rejected if this SNR exceeds a threshold. To benchmark CORSICA, we evaluate two alternatives: a TRF-based SNR method, in which temporal response functions are fitted to each IC and artifact-driven peaks near zero lag are used for rejection, and a variant replacing ICA with second-order blind identification (SOBI) as the source separation step.<i>Main results.</i>CORSICA effectively suppressed CI artifacts while preserving neural activity, enabling recovery of physiologically plausible TRFs with only 2% of ICs rejected. Both benchmark methods confirmed the validity of the SNR-based rejection framework, but CORSICA outperformed the TRF-based alternative in artifact suppression quality. Replacing ICA with SOBI as the source separation step required more ICs to be rejected, further supporting ICA as the preferred backbone for CORSICA.<i>Significance.</i>CORSICA provides a fully objective, label-free approach to identifying CI artifacts in speech-evoked EEG data, with no manual intervention required. By centering artifact rejection on a single interpretable threshold, it offers a reproducible preprocessing standard for future EEG studies on speech processing in CI users.<i>Conclusion.</i>Our findings demonstrate that objective CI artifact suppression in speech-evoked EEG data is feasible on the basis of the IC's temporal response patterns.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148908490","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":"Leakage-audited benchmarking reveals limited evidence for cross-subject auditory-evoked EEG vowel perception decoding.","authors":"Xiaoyang Li, Zeyan Tao","doi":"10.1088/1741-2552/aea4f7","DOIUrl":"10.1088/1741-2552/aea4f7","url":null,"abstract":"<p><p><i>Objective.</i>We evaluated cross-subject five-vowel perception decoding from auditory-evoked EEG with explicit trial definitions, participant-grouped validation, complete prediction coverage, and participant-level inference.<i>Approach.</i>We parsed all Study 2 event tables from OpenNeuro ds006104 version 1.0.1 and analysed the consonant-vowel (CV) pair task. Marker-stimulus pairing converted 7680 event rows into 3840 independent CV trials. Excluding active-condition trials left 1280 eligible control trials. A 400V maximum across-channel peak-to-peak criterion rejected 186 epochs, retaining 1094 epochs from 16 participants and 61 EEG channels. Thirteen unique implementations underwent leave-one-subject-out testing. Participant metrics came from 36 102 trial predictions across 33 complete replicas. Descriptive analyses assessed deep-model stability across five seeds, class-resolved errors, and retention sensitivity. An exploratory MDM analysis comprised 9616 genuine model refits across training cohorts of 3-15 participants.<i>Main results.</i>Two independent cohort-construction implementations preserved the same ordered 1094 trials and labels (maximum tensor difference,V). Thus, corrected accounting changed the reported exclusion decomposition but not the analysed cohort or model inputs. Random Forest achieved the highest mean balanced accuracy: 21.474% (95% participant-bootstrap interval, 19.526%-23.482%; five-class chance, 20%). The one-sided Wilcoxon, Bonferroni-adjusted, and sign-flipvalues were 0.090 897, 1.000 000, and 0.092 791, respectively. All 13 implementations remained non-significant after multiplicity correction. Deep-architecture means were near chance, with substantial within-participant seed ranges and low trial-label agreement for several architectures. Across 3-15 training participants, refitted MDM means ranged from 20.553% to 20.922%, without a monotonic increase.<i>Significance.</i>This dataset and protocol provide limited evidence for cross-subject five-vowel decoding. Explicit trial units, participant-grouped validation, complete seed reporting, prediction-level traceability, and multiplicity-adjusted inference sharpen the interpretation of small-cohort EEG benchmarks.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148922066","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":"Reliability of spatiotemporal neuromuscular activity patterns in magnetomyography across force levels.","authors":"Haodi Yang, Burak Senay, Benedict Kleiser, Justus Marquetand","doi":"10.1088/1741-2552/aea1be","DOIUrl":"10.1088/1741-2552/aea1be","url":null,"abstract":"<p><p><i>Objective.</i>Magnetomyography (MMG) using an optically pumped magnetometer (OPM) provides a contactless and non-invasive approach to assess neuromuscular activity. However, given the limited studies using OPM magnetomyography (OPM-MMG) and that those primarily used a single OPM, it remains to be characterized how stable individual neuromuscular activity patterns are over time and across different force levels when using multiple OPMs, specifically an array-based MMG.<i>Approach.</i>12 healthy subjects performed ramped isometric contractions at 10, 20, 40, 60, and 80% MVC, while a 2 × 3 OPM array recorded MMG signals from the tibialis anterior. Test-retest repeatability was selectively validated at 20% and 60% maximal voluntary contraction (MVC), with one repeated ramp performed at each force level. For each subject, Spearman's rank correlation was computed across all OPM sensors between force conditions to assess consistency of feature rankings, and within-subject spatial repeatability between repeated 20% and 60% MVC ramps was quantified using ICC (3, 1) (two-way mixed-effects, single-measure, absolute agreement).<i>Main results.</i>In 9 out of 12 subjects (75.0%; exact 95% CI: 42.8%-94.5%), Spearman's rank analyses showed generally high correlations across force levels (<i>ρ</i>= 0.37-0.91,<i>p</i>< 0.05), whereas in the remaining 3 subjects (25.0%; exact 95% CI: 5.5%-57.2%), spatial patterns were less stable (<i>ρ</i>= - 0.73-0.58,<i>p</i>< 0.05). ICC (3,1) between repeated ramps indicated high within-subject spatial repeatability of the array pattern in 9 out of 12 subjects, with ICC > 0.75 at 20% MVC (5 out of 9 subjects) and 60% MVC (6 out of 9 subjects), respectively. In contrast, the remaining 3 subjects showed lower ICCs (20% MVC: -0.28-0.51; 60% MVC: 0.22-0.80).<i>Significance.</i>Array-based OPM-MMG shows high within-subject stability of spatial patterns across force levels and strong test-retest repeatability (selectively validated at 20% and 60% MVC) in the majority of subjects (9 out of 12), supporting its use for characterizing force-dependent neuromuscular activity while acknowledging inter-subject variability.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148883157","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":"Measuring motor intent for BCI control-A comparative analysis of signal quality of simultaneously recorded vECoG and scalp EEG.","authors":"Nikole Chetty, James Bennett, Hunter R Schone, Adam Fry, Peter E Yoo, David Lacomis, Thomas J Oxley, Jennifer L Collinger, David F Putrino, Douglas J Weber","doi":"10.1088/1741-2552/ae9eef","DOIUrl":"10.1088/1741-2552/ae9eef","url":null,"abstract":"<p><p><i>Objective.</i>Stent-electrode arrays enable endovascular brain-computer interfaces (BCI) by recording cortical neural activity from within the superior sagittal sinus and have recently been evaluated in an early feasibility clinical trial in the United States (ClinicalTrials.gov: NCT05035823). For a BCI to be viable, the signals need to be high quality to enable accurate decoding of user intent. Compared to electrodes placed on the scalp for electroencephalography (EEG), stent-electrode arrays lie closer to the cortical surface and would presumably offer higher signal quality yet a direct comparison of intravascular and scalp-based neural recordings in humans has not yet been investigated.<i>Approach.</i>We directly compared the signal quality of vascular electrocorticography (vECoG) versus scalp EEG signals in one participant with severe upper limb paralysis due to ALS. During two experimental sessions, the participant underwent simultaneous recording with the stent-electrode array and a scalp EEG using a gel cap. The participant was visually cued to attempt motor tasks, such as repeated flexion and extension of the ankles. Signal quality was assessed by quantifying motor modulation strength, differentiation of movement effort, and spatial lateralization. Noise metrics evaluated the relative impact of artifacts including 60 Hz line noise, electrocardiogram contamination, eye blinks, jaw clenching, and vocalization.<i>Main results.</i>Both recording modalities exhibited significant modulation during attempted movement relative to rest, with vECoG generally demonstrating significantly stronger modulation per channel in some frequency bands and conditions. Motor modulation was significantly reduced during motor imagery compared to overt movement in both modalities. Spatial source localization between left and right ankle movement did not reach significance for either modality. Each modality was vulnerable to some artifacts while generally unaffected by others. Scalp EEG showed large susceptibility to ocular artifacts and cranial muscle activity due to its proximity to superficial physiological sources, whereas vECoG exhibited prominent cardiac activity.<i>Significance.</i>Within this participant, the large modulation during attempted movement recorded with vECoG, unaffected by the attenuating effects of the skull, coupled with fewer artifacts in the frequency bands of interest, provides preliminary evidence that the stent-electrode arrays can acquire high quality neural signals that could support BCI control.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":"23 5","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13584699/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148984448","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":"MTGNet: a task-oriented and spectrally guided framework for EEG denoising.","authors":"Jinchen Hu, Zhongke Gao, Yushi Hao, Shangbin Xie, Xiaoyang Li, Qing Cai, Jianpeng An","doi":"10.1088/1741-2552/ae9ef1","DOIUrl":"10.1088/1741-2552/ae9ef1","url":null,"abstract":"<p><p><i>Objective.</i>Electroencephalography (EEG) is widely used in brain-computer interfaces, but its microvolt-level signals are easily contaminated by electromyography (EMG), electrooculography (EOG), and mixed physiological artifacts. This study develops an EEG denoising framework that suppresses artifacts while preserving information used by downstream biomedical artificial intelligence tasks.<i>Approach.</i>We propose MTGNet, a task-oriented and spectrally guided EEG denoising framework. MTGNet combines low-rank adaptation (LoRA)-based task-aware consistency regularization, a spectrally aware Guidance Network, and parallel Mamba-Transformer backbone. A pretrained 11.97 M-parameter backbone learns to preserve intrinsic EEG characteristics from paired noisy-clean data, while 0.33 M LoRA parameters enable task-specific adaptation without requiring paired clean EEG references.<i>Main results.</i>On EEGDenoiseNet, MTGNet reduces spectral relative root-mean-square error by over 18.9%, 31.5%, and 14.0% for EMG, EOG, and hybrid artifacts, respectively (<i>p</i>0.001). On a real-world fatigue EEG dataset, it improves classification accuracy by 6.20-6.69 percentage points compared with unprocessed inputs (<i>p</i>0.05). Ablation, cross-classifier, and cross-dataset analyses validate the proposed components and support the transferability of MTGNet across the evaluated settings.<i>Significance.</i>The proposed framework provides a practical approach to task-aware EEG denoising, while future work should further validate its applicability across real-world EEG settings involving diverse tasks, artifact types, and acquisition conditions.</p>","PeriodicalId":94096,"journal":{"name":"Journal of neural engineering","volume":" ","pages":""},"PeriodicalIF":3.8,"publicationDate":"2026-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148835958","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}