IEEE Transactions on Biomedical Engineering最新文献

{"title":"Multilevel Correlation-Aware and Modal-Aware Graph Convolutional Network for Diagnosing Neurodevelopmental Disorders.","authors":"Shijia Zuo, Yu Li, Yinbao Qi, Aiping Liu","doi":"10.1109/TBME.2025.3617348","DOIUrl":"10.1109/TBME.2025.3617348","url":null,"abstract":"<p><strong>Objective: </strong>Graph-based methods using resting-state functional magnetic resonance imaging demonstrate strong capabilities in modeling brain networks. However, existing graph-based methods often overlook inter-graph relationships, limiting their ability to capture the intrinsic features shared across individuals. Additionally, their simplistic integration strategies may fail to take full advantage of multimodal information. To address these challenges, this paper proposes a Multilevel Correlation-aware and Modal-aware Graph Convolutional Network (MCM-GCN) for the reliable diagnosis of neurodevelopmental disorders.</p><p><strong>Methods: </strong>At the individual level, we design a correlation-driven feature generation module that incorporates a pooling layer with external graph attention to perceive inter-graph correlations, generating discriminative brain embeddings and identifying disease-related regions. At the population level, to deeply integrate multimodal and multi-atlas information, a multimodal-decoupled feature enhancement module learns unique and shared embeddings from brain graphs and phenotypic data and then fuses them adaptively with graph channel attention for reliable disease classification.</p><p><strong>Results: </strong>Extensive experiments on two public datasets for Autism Spectrum Disorder (ASD) and Attention Deficit Hyperactivity Disorder (ADHD) demonstrate that MCM-GCN outperforms other competing methods, with an accuracy of 93.11% for ASD and 76.41% for ADHD.</p><p><strong>Conclusion: </strong>The MCM-GCN framework integrates individual-level and population-level analyses, offering a comprehensive perspective for neurodevelopmental disorder diagnosis, significantly improving diagnostic accuracy while identifying key indicators.</p><p><strong>Significance: </strong>These findings highlight the potential of the MCM-GCN for imaging-assisted diagnosis of neurodevelopmental diseases, advancing interpretable deep learning in medical imaging analysis.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":"1863-1876"},"PeriodicalIF":4.5,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145212283","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"External Validation Complexities: A Comparative Study of Late-Onset Sepsis Prediction Models Across Multiple Clinical Environments.","authors":"Zheng Peng, Janno S Schouten, Demi Silvertand, Xi Long, Douglas E Lake, H Rob Taal, Hendrik J Niemarkt, Peter Andriessen, Brynne Sullivan, Carola van Pul","doi":"10.1109/TBME.2025.3618080","DOIUrl":"10.1109/TBME.2025.3618080","url":null,"abstract":"<p><strong>Objective: </strong>Neonatal late-onset sepsis (LOS) is a life-threatening condition in preterm infants in neonatal intensive care units (NICUs), with early detection being crucial for improving outcomes. Despite advancements in data-driven prediction models, their generalizability remains uncertain due to a lack of independent validation, particularly on national and international scales. This study evaluates the performance of two LOS prediction models on multiple validation datasets to assess their reliability for clinical implementation.</p><p><strong>Methods: </strong>Two models were validated: (1) a multi-channel feature-based extreme gradient boosting model (MC-XGB) and (2) a deep neural network using raw RR intervals (RR-DNN). Validation was conducted on three NICU datasets: an internal dataset (68 LOS, 100 controls) from the model-development hospital in the Netherlands, a national external dataset (20 LOS, 20 controls) from another Dutch hospital, and an international external dataset (17 LOS, 17 controls) from a U.S. hospital. Model performance was assessed using the area under the receiver operating characteristic curve (AUC) across multiple prediction time windows, with an hourly risk analysis.</p><p><strong>Results: </strong>Both models achieved a peak AUC of 0.82 in the internal dataset, their predictive performance demonstrates variable declines in external datasets. The respective AUCs for RR-DNN and MC-XGB were 0.80 and 0.72 in the national dataset, and 0.69 and 0.60 in the international dataset. This may result from variations in clinical practices, patient demographics, and monitoring technologies.</p><p><strong>Conclusion: </strong>Model performance declined in external validations, highlighting the challenges of implementing predictive models across diverse clinical settings.</p><p><strong>Significance: </strong>This study emphasizes the need for standardized guidelines and improved data sharing to enhance model development and facilitate reliable integration into NICU workflow for improved LOS management.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":"1921-1932"},"PeriodicalIF":4.5,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145238465","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Extending Multiscale Characterization of Heart Rate Variability via Deep Learning for Mortality Risk Prediction.","authors":"Joao G S Kruse, Yudai Fujimoto, Sinyoung Lee, Eiichi Watanabe, Ken Kiyono","doi":"10.1109/TBME.2025.3614714","DOIUrl":"10.1109/TBME.2025.3614714","url":null,"abstract":"<p><strong>Objective: </strong>To improve mortality risk prediction from heart rate variability (HRV) signals by capturing nonlinear scaling patterns often overlooked by traditional linear analyses.</p><p><strong>Methods: </strong>This study combines detrended moving average (DMA) analysis with convolutional neural networks (CNNs). DMA curves were computed from 2-hour overlapping windows of 24-hour Holter ECG recordings in 916 survivors and 70 nonsurvivors. A CNN was trained to extract features from these curves and benchmarked against models using traditional HRV and clinical features.</p><p><strong>Results: </strong>The CNN achieved an ROC-AUC of 0.72 and an adjusted hazard ratio of 2.129 for daytime recordings, outperforming standard models. Two patient groups emerged based on DMA scaling patterns. Group 1, with dominant short-term scaling, exhibited reduced slopes in nonsurvivors, suggesting impaired autonomic adaptability. Group 2 showed earlier transitions between short- and long-term behavior, where reduced long-term slopes more strongly predicted mortality. Integrated gradients analysis identified key timescales in the DMA curve driving model predictions.</p><p><strong>Conclusion: </strong>DMA combined with CNNs enhances HRV-based mortality risk stratification and reveals distinct physiological scaling patterns associated with survival outcomes.</p><p><strong>Significance: </strong>This study highlights the potential of DMA and CNNs in improving mortality risk stratification and providing mechanistic insights into HRV dynamics, with implications for personalized health monitoring.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":"1771-1780"},"PeriodicalIF":4.5,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145174846","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Next-Generation Tactile Sensing and Machine Learning Integration for Robot-Assisted Minimally Invasive Surgery.","authors":"Dema N Govalla, Anish S Naidu, Dhrubo Ahmad, Jerzy W Rozenblit","doi":"10.1109/TBME.2025.3613757","DOIUrl":"10.1109/TBME.2025.3613757","url":null,"abstract":"<p><p>Tactile feedback in robot-assisted minimally invasive surgery (RAMIS) is crucial for surgeons when palpating subsurface tumors and other organ structures. The research presented here is a new approach for tactile sensation generation that aims to provide deformation and texture detection in RAMIS. The proposed solution comprises three phases: feature extraction, recognition and feedback. The feature extraction process is based on data acquisition from two micro-electromechanical systems (MEMS) sensors and a force-sensitive resistor (FSR) sensor attached to an EndoWrist thoracic grasper instrument compatible with the da Vinci Surgical System. The acquired data is processed using digital signal processing methods and utilized in the recognition phase. The recognition segment receives the features as inputs for training and testing two advanced machine learning algorithms. The first algorithm is a Reflex Fuzzy Min-Max Neural Network (RFMN); the other is a Time Series Classification - Learning Shapelets (TSC-LS) method. The machine learning algorithms aim to accurately recognize and classify physiological structures with different softness and roughness into a corresponding deformation or texture label. Lastly, a means of mechanically giving the labeled data as feedback to the surgeon via a visual-tactile display and a wearable device located on the surgeon's forearm is accomplished to mimic palpation feedback during RAMIS.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":"1718-1733"},"PeriodicalIF":4.5,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145137335","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A 3D Numerical Model of Ultrasonic Transthoracic Propagation for Cardiac Focused Ultrasound Therapy.","authors":"Clara Magnier, Wojciech Kwiecinski, Daniel Suarez Escudero, Gauthier Amis, Guillaume Goudot, Elie Mousseaux, Emmanuel Messas, Mathieu Pernot","doi":"10.1109/TBME.2025.3618103","DOIUrl":"10.1109/TBME.2025.3618103","url":null,"abstract":"<p><strong>Objective: </strong>Non-invasive focused ultrasound therapies of abdominal organs, including the heart and the liver, have emerged in the last decades. Transthoracic focusing of ultrasound poses challenges such as pressure loss and aberrations. Numerical models of ultrasonic propagation have been developed to study the focalization in heterogeneous tissues, particularly for transcranial applications. However, ribcage models were less studied than skull models, and no experimental validation of ribcage models has been performed so far.</p><p><strong>Methods: </strong>Both linear and nonlinear k-space simulations were used to model the ultrasonic propagation from a clinical system dedicated to transthoracic cardiac therapy. Tissue acoustic properties were determined from computed tomography scans. Experimental model validation was performed with hydrophone measurements of pressure fields through in vitro human ribs and in vitro porcine flail chest.</p><p><strong>Results: </strong>An excellent agreement of pressure distribution between the acquired and simulated pressure fields was found for the linear propagation model with a mean correlation coefficient between the measured and simulated pressure fields of R² = 0.89±0.07. For the nonlinear propagation, the mean correlation coefficient was R² = 0.91±0.06. The feasibility of the simulations through the human thorax was demonstrated on 9 patients who underwent non-invasive therapy of the aortic valve. The global attenuation estimated numerically was correlated with the amplitude at the focus necessary to nucleate cavitation (R² = 0.64).</p><p><strong>Conclusion: </strong>The numerical model of transthoracic ultrasound propagation was validated and used on a human patient's thorax.</p><p><strong>Significance: </strong>With further development, this model could be used as a treatment planning tool for non-invasive ultrasonic cardiac therapy.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":"1933-1942"},"PeriodicalIF":4.5,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145238517","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Noise-Robust Model-Based Approach to T-Wave Amplitude Measurement and Alternans Detection.","authors":"Zuzana Koscova, Amit Shah, Ali Bahrami Rad, Qiao Li, Gari D Clifford, Reza Sameni","doi":"10.1109/TBME.2025.3616055","DOIUrl":"10.1109/TBME.2025.3616055","url":null,"abstract":"<p><strong>Objective: </strong>T-wave alternans (TWA) is a potential marker for sudden cardiac death, but its reliable analysis is often constrained to noise-free environments, limiting its utility in real-world settings. We explore model-based T-wave estimation and detection to mitigate noise effects on TWA.</p><p><strong>Methods: </strong>Detection was performed using a surrogate-based method as a benchmark and a new approach based on a Markov model state transition matrix (STM). Estimation employed a Modified Moving Average (MMA) and polynomial T-wave modeling to improve noise robustness. Methods were evaluated across signal-to-noise ratios (SNRs) from -5 to 30 dB and noise types: baseline wander, muscle artifacts (MA), electrode movement (EM), and respiratory modulation. Synthetic ECGs with known TWA levels were used: 0 $bm mu$V for TWA-free and 30-72 $bm mu$V for TWA-present.</p><p><strong>Results: </strong>T-wave modeling improved estimation accuracy under noisy conditions. With MA noise at SNRs of -5 and 5 dB, mean absolute error (MAE) dropped from 62 to 49 $bm mu$V and 27 to 25 $bm mu$V, respectively (Mann-Whitney U test, $bm {p < 0.05}$). Similar improvements occurred with EM noise: MAE decreased from 101 to 71 and 26 to 23 $bm mu$V. In detection, STM achieved sensitivity of 0.87, outperforming the surrogate-based method (0.70), though both struggled under EM noise at -5 dB. Detection performance also depended on the number of beats analyzed.</p><p><strong>Conclusion: </strong>These findings show that applying model-based estimation and STM detection could improve TWA analysis under noise, supporting application in ambulatory and wearable ECG monitoring.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":"1828-1838"},"PeriodicalIF":4.5,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145199180","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multi-Stage Respiratory Sound Analysis: Confidence-Driven Wheeze and Crackle Detection.","authors":"Annapurna Kala, Mounya Elhilali","doi":"10.1109/TBME.2025.3613489","DOIUrl":"10.1109/TBME.2025.3613489","url":null,"abstract":"<p><strong>Objective: </strong>Accurate detection of adventitious respiratory sounds, such as wheezes and crackles, is essential for diagnosing and managing respiratory conditions. This study introduces a multi-stage, confidence-driven framework for automated pediatric auscultation analysis, performing a three-way classification of normal, wheeze, and crackle sounds to improve diagnostic accuracy.</p><p><strong>Methods: </strong>We develop a comprehensive pipeline integrating anomaly-specific segment selection, segment-level classification, and confidence-based fusion. Our contrastive variational recurrent neural network (CVRNN) enhances feature extraction, while a confidence-weighted aggregation strategy refines final predictions. The system is validated using a diverse pediatric dataset from 742 subjects (aged 1-59 months) from seven countries.</p><p><strong>Results: </strong>The multi-level framework is evaluated across three stages. The anomaly-specific segment selection achieves 98.47 $%$ recall, identifying adventitious regions. Next, segment-level classifiers improve sensitivity, achieving balanced accuracies of 72.15 $%$ (wheeze) and 68.1 $%$ (crackle). This performance surpasses state of the art systems on the same dataset and demonstrates enhanced balanced performance in detecting both crackle and wheeze sounds, which present different challenges to automated systems given their markedly different acoustic profiles. Finally, the confidence-driven fusion outperforms traditional aggregation methods, yielding a final three-way classification of 62.12 $%$.</p><p><strong>Conclusion: </strong>Our confidence-based multi-stage approach enhances automated respiratory sound classification by prioritizing high-certainty segment predictions, aligning with expert physician annotations.</p><p><strong>Significance: </strong>This framework advances computer-aided respiratory diagnostics, improving early detection and monitoring of pediatric respiratory conditions. By integrating expert-inspired segmentation with machine learning-driven confidence estimation, it has potential to enhance clinical workflows and screening for pulmonary diseases, particularly in resource-limited settings.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":"1696-1704"},"PeriodicalIF":4.5,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145130723","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An Integrated Rigid-Flexible Body Dynamic Approach to Computationally Efficient Musculoskeletal Modeling and Muscle Recruitment Simulation of the Lumbosacral Spine and Torso.","authors":"Siril Teja Dukkipati, Mark Driscoll","doi":"10.1109/TBME.2025.3617301","DOIUrl":"10.1109/TBME.2025.3617301","url":null,"abstract":"<p><strong>Objective: </strong>In silico biomechanical models of the spine traditionally follow either rigid body dynamic (RBD) modeling (multibody modeling) or finite element (FE) modeling techniques. While RBD models lack robust representation for flexible tissues, FE models are computationally expensive. This study proposes an integrated rigid-flexible body dynamic (RFBD) architecture to address these limitations, and develops a full-torso human model, focusing spinal mechanical stability.</p><p><strong>Methods: </strong>The model consisted of L1-L5 lumbar vertebrae, pelvis, sacrum, a lumped thoracic spine with ribcage as rigid bodies, while the intervertebral discs (IVDs), abdominal cavity and thoracolumbar fascia (TLF) were modeled as deformable reduced-order flexible bodies. Spinal ligaments were represented as nonlinear tension-only springs, while the musculature was modeled as tension-only forces. Level-by-level spinal stiffness was validated under pure flexion moments up to 7.5 Nm against literature studies. The reduced-order implementation was also validated against an identical FE model. Spinal stability contribution of different tissues in flexion was systematically evaluated using six on-off cases.</p><p><strong>Results: </strong>Passive spine segmental stiffness profiles matched well with ex vivo and in silico comparators. The RFBD method demonstrated strong agreement with the FE solver, while significantly reducing computational demand. Stability analyses highlighted the role of intra-abdominal pressure in spinal unloading and generation of compressive loads along the spinal curvature through muscle recruitment.</p><p><strong>Conclusion: </strong>This parametric, fast-solving, high-fidelity spine simulation platform could be a useful biomechanical tool for spine researchers.</p><p><strong>Significance: </strong>A novel human torso model with integrated rigid and flexible bodies was presented in this study, providing insights into mechanical spine stability.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":"1877-1889"},"PeriodicalIF":4.5,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145212295","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Clinically Explainable Disease Diagnosis Based on Biomarker Activation Map.","authors":"Pengxiao Zang, Carol Wang, Tristan T Hormel, Steven T Bailey, Thomas S Hwang, Yali Jia","doi":"10.1109/TBME.2025.3614518","DOIUrl":"10.1109/TBME.2025.3614518","url":null,"abstract":"<p><strong>Objective: </strong>Artificial intelligence (AI)-based disease classifiers have achieved specialist-level performances in several diagnostic tasks. However, real-world adoption of these classifiers remains challenging due to the black box issue. Here, we report a novel biomarker activation map (BAM) generation framework that can provide clinically meaningful explainability to current AI-based disease classifiers.</p><p><strong>Methods: </strong>We designed the framework based on the concept of residual counterfactual explanation by generating counterfactual outputs that could reverse the decision-making of the disease classifier. The BAM was generated as the difference map between the counterfactual output and original input with postprocessing. We evaluated the BAM on four different disease classifiers, including an age-related macular degeneration classier based on fundus photography, a diabetic retinopathy classifier based on optical coherence tomography angiography, a brain tumor classifier based on magnetic resonance imaging (MRI), and a breast cancer classifier based on computerized tomography (CT) scans.</p><p><strong>Results: </strong>The highlighted regions in the BAM correlated highly with manually demarcated biomarkers of each disease.</p><p><strong>Conclusion/significance: </strong>The BAM can improve the clinical applicability of an AI-based disease classifier by providing intuitive output clinicians can use to understand and verify the diagnostic decision.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":"1746-1757"},"PeriodicalIF":4.5,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145148681","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Online Regulation of Task Difficulty Based on Neuro- and Motor-Feedback to Improve Engagement in Visual-Motor Task.","authors":"Yifan Li, Rong Song","doi":"10.1109/TBME.2025.3615733","DOIUrl":"10.1109/TBME.2025.3615733","url":null,"abstract":"<p><strong>Objective: </strong>Enhancing active engagement in post-stroke rehabilitation is critical for promoting neuroplasticity. Although adaptive feedback can optimize arousal to improve engagement, most approaches rely solely on motor or neural indicators, overlooking the integration of task-specific physical performance with neural adaptation. The purpose of this study is to validate the effectiveness of enhancing prefrontal cortex (PFC) neural activity through a closed-loop adaptive feedback system.</p><p><strong>Methods: </strong>In this study, a neuro- and motor-feedback (NMF) system is proposed. It utilizes functional near infrared spectroscopy (fNIRS) and tracking error to continuously monitor real-time neural activity and motor performance during a visual-motor task, and realizes online adaptive regulation of task difficulty through fuzzy logic controller. 10 healthy participants were recruited for a 5-day training program, during which each participant completed 15 task trials at both fixed and adaptive difficulty levels, serving as the control group and the NMF group.</p><p><strong>Results: </strong>Compared to the control group, the NMF group showed increased tracking errors as well as heightened neural activity in the PFC and the sensorimotor cortex (SMC), in both single-task trial and after 5 days of training. Moreover, the NMF group exhibited significantly increased strength of brain functional connections between the PFC and sensorimotor areas after training compared to the control group.</p><p><strong>Conclusion: </strong>Our findings suggest that the proposed NMF system can enable online neural activity regulation in visual-motor tasks and achieve enhanced integration between cognitive and sensorimotor areas, with the potential to improve the rehabilitation training outcomes.</p>","PeriodicalId":13245,"journal":{"name":"IEEE Transactions on Biomedical Engineering","volume":"PP ","pages":"1805-1816"},"PeriodicalIF":4.5,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145300081","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}