Search Results (11,420)

Electroencephalography (EEG) is widely used to study cognitive states and clinical biomarkers, yet the contribution of stable between-subject differences to common EEG feature representations is rarely quantified directly or incorporated into evaluation design. Here, we examine subject-linked structure in spectral EEG features using a longitudinal four-session dataset spanning 199 days and replicate key findings across four public datasets covering 39 to 395 participants, multiple paradigms, and recording systems ranging from low-channel consumer devices to research-grade EEG. In spectral band-power representations, subject identity accounted for substantially more variance than session-related drift and supported extremely strong individual discriminability under subject-wise held-out protocols, with high verification performance and robust cross-session recognition over months. However, this identity structure was not fully invariant across paradigms, showing stronger transfer in richer, higher-channel recordings than in low-channel consumer EEG under larger task shifts. We further show that the same subject-linked structure can inflate downstream clinical classification when evaluation is performed with naive segment-level splits, demonstrating that apparent diagnostic performance can partly reflect identity leakage rather than biomarker learning. These findings highlight subject identity as a measurable and persistent source of structured variance in spectral EEG features and support routine use of subject-wise evaluation and explicit confound diagnostics in EEG machine-learning studies. Full article

Multimodal knowledge extraction has become increasingly important for applications that require compact and informative representations of text-image data, yet existing models often remain computationally expensive for practical use. To address this issue, we propose MKE, a lightweight framework for multimodal knowledge extraction based on knowledge distillation and efficient multimodal fusion. Specifically, MKE transfers linguistic knowledge from a large BART teacher model to a compact StuBART backbone and combines Flash-Attention-based cross-modal interaction with a lightweight nonlinear transformation module. Experiments on the MSMO English news dataset show that MKE achieves competitive multimodal summarization performance with 270M parameters and a computational cost of 155.63 GMACs (311.27 GFLOPs) per sample. We further evaluate inference efficiency under a controlled RTX 4090 setup, and we clarify that deployment on smartphones, IoT systems, and wearable devices remains a promising direction for future work rather than a scenario directly validated in this study. Full article

Background: Wearable sensors combined with machine learning (ML) offer new opportunities for movement analysis and performance monitoring in tennis and padel. However, the current evidence and potential applications remain unclear. This systematic review aimed to summarize the applications of ML to wearable sensor data in tennis and padel, focusing on movement recognition, performance monitoring, workload and fatigue assessment, injury-related applications, and rehabilitation. Methods: PubMed and Scopus were searched from inception to 1 August 2026. Studies were eligible if they investigated tennis or padel, used wearable sensors, and applied ML or deep learning methods. Study selection and data extraction were independently performed by two reviewers. Risk of bias was assessed across domains related to participants, predictors, outcomes, and analysis/validation. Due to substantial heterogeneity, a qualitative synthesis was performed. Results: Of 393 records identified, 16 studies were included. Fourteen investigated tennis, one padel, and one included tennis data within a multisport study. IMU-based systems were the most frequently used technology. Most studies focused on movement and stroke recognition, while fewer addressed performance, workload, fatigue, or injury-related applications. Reported model performance was generally high, but validation strategies were heterogeneous, and participant-independent or external validation was uncommon. Nine studies presented some concerns, and seven were classified as having high risk of bias. Evidence directly addressing rehabilitation and return-to-sport was particularly limited. Conclusions: ML applied to wearable sensor data shows promising potential for movement recognition and performance assessment in tennis and padel. However, methodological limitations, small and selected cohorts, and limited external validation restrict current generalizability and practical translation. Future research should prioritize larger and more diverse cohorts, subject-independent and external validation, and prospective studies addressing injury prevention, rehabilitation, and return-to-sport. Full article

Assessing occupational workload under dynamically unstable offshore conditions is challenging because physiological demands and biomechanical exposures may vary across tasks and may not be adequately characterized using a single measure. This exploratory field study characterized physiological, ergonomic, and kinematic workload during coastal gillnet fishing using an integrated assessment approach. Four fishermen (one captain and three crew members) were monitored during three fishing sessions under real offshore conditions. Heart rate, physical activity, ergonomic exposure using the Quick Exposure Check, and lumbar–pelvic kinematics using inertial measurement units were assessed across operational phases. Workload was temporally concentrated during hauling, with heart rate and physical activity increasing particularly during the mid-to-late hauling stages. Ergonomic exposure of the lower back, shoulder/arm, and neck also increased during hauling. Kinematic analysis revealed moderate trunk flexion accompanied by non-neutral flexion exposure of up to 28.62%, while the lumbar contribution to lumbar–pelvic motion frequently exceeded 50%. These findings suggest that integrating wearable sensing with ergonomic assessment may help characterize task-specific workload patterns that are not readily captured by average physiological intensity alone. This approach may support field-based occupational workload assessment in dynamically unstable working environments. Full article

Depression is increasingly investigated through wearable digital phenotyping because alterations in motor activity, daily routines, and circadian organisation may provide objective behavioural information relevant to mental-health screening. Wrist actigraphy enables continuous, non-invasive monitoring of such longitudinal activity patterns in natural settings. However, existing machine- and deep-learning approaches often rely on engineered summary features or isolated temporal windows, limiting representation of within-day and between-day behavioural variation. Small participant-level datasets also increase the risk of information leakage and unreliable probability estimates. This retrospective secondary-data proof-of-concept study analysed minute-level wrist actigraphy from 55 participants, comprising 23 participants experiencing unipolar or bipolar depressive episodes and 32 healthy controls. Model development and internal validation used nested leave-one-subject-out cross-validation, with preprocessing, self-supervised representation learning, hyperparameter selection, and calibration restricted to training participants within each fold. The novelty lies in integrating hierarchical temporal modelling, leakage-resistant representation learning, interpretable circadian information, and uncertainty-aware prediction within a single activity-only framework. Performance was evaluated using balanced accuracy, sensitivity, specificity, F1-score, MCC, AUROC, AUPRC, and calibration measures. The five-seed probability-averaged HC-MAT ensemble achieved balanced accuracy of 0.872, sensitivity of 0.870, specificity of 0.875, F1-score of 0.851, MCC of 0.741, and AUROC of 0.920; across individual seeds, balanced accuracy was 0.859 ± 0.010. Although HC-MAT achieved the highest numerical performance across several metrics, none of its comparisons with baseline or ablated models remained statistically significant after Holm correction. These results support HC-MAT as a proof-of-concept framework requiring validation in substantially larger, independent clinical cohorts before clinical deployment. Full article

Flatfoot (pes planus) is a common foot deformity associated with altered lower-extremity muscle activity during gait. Conventional foot orthoses and insoles can provide arch support but have limitations related to comfort, adherence, and everyday usability. This study evaluated the immediate effects of a thermoplastic polyurethane (TPU) arch pad sock on lower-extremity muscle activity in adults with flexible flatfoot. Twenty adults with flexible flatfoot participated. Surface electromyography signals were recorded bilaterally from the tibialis anterior (TA), peroneus longus (PL), medial gastrocnemius (MG), and lateral gastrocnemius (LG), which yielded eight recording channels. Muscle activity was measured before and immediately after the TPU arch pad sock application and normalized to percentage of maximum voluntary isometric contraction. Muscle activity was significantly reduced in five of the eight channels following the TPU arch pad sock application (all p < 0.05), including the right TA, right PL, right MG, left TA, and left PL. The largest effect was observed in the right MG (Cohen’s d = 1.157). Although these reductions were statistically significant, all five channels remained below the minimal detectable change (MDC95), indicating that the immediate effects did not reach the threshold for clinically meaningful change; longitudinal studies evaluating cumulative use are needed to establish clinical significance. Wearing the TPU arch pad sock was immediately associated with altered lower-extremity muscle activity in the subjects, although the single-arm design precludes establishing a definitive causal attribution. These findings suggest that the device may affect neuromuscular responses during walking by providing medial longitudinal arch support. Full article

Behind-the-ear electroencephalography (EEG) can now be recorded continuously outside hospital, but forecasting the preictal state is hard in a distributed deployment: raw traces cannot leave the clinical site, the wearable montage is chosen per patient, and part of the labelling comes from automated annotators rather than clinicians. Federated learning removes the need to move recordings, yet weighting each site by its sample count lets a poorly annotated site dominate the shared model. FedCareX answers this with a montage-agnostic tokenizer, a lightweight preictal Transformer encoder, an adaptive reliability aggregation rule based on annotation quality, gradient consistency and probe-set agreement, and a sparsified error-feedback codec that reduces uplink traffic. On a five-centre wearable corpus FedCareX attains $83.1\pm 0.6\%$ sensitivity with 0.38 false predictions per hour, a 2.8 point increase in sensitivity and a 13.6% relative reduction in false prediction rate compared with the best 2026 baseline, and on a scalp benchmark with 23 patient clients it reaches $92.8\pm 0.4\%$ sensitivity with 0.22 false predictions per hour. The scalp margin is significant under a paired Wilcoxon signed-rank test over 23 independent client pairs; the wearable margin rests on five centres and is reported as a hierarchical bootstrap interval of $[1.1,4.4]$ sensitivity points rather than as a pooled significance test. Uplink volume drops by 22.1× to 0.22 MB per client per round, and edge inference spends 28.9 mJ per window against the measured 55 mJ per window budget on the gateway platform. Full article

Artificial intelligence (AI) is increasingly used in biosensing, yet its role is often limited to post-processing or high-accuracy regression on simulation-generated datasets. This critical narrative review examines physics-informed and explainable AI for nanomaterial-based biosensors, emphasizing graphene and other two-dimensional materials, surface plasmon resonance (SPR), terahertz (THz) metasurfaces, electrochemical platforms, field-effect transistors, surface-enhanced Raman spectroscopy, and wearable systems. Recent peer-reviewed studies and relevant technical guidance were critically compared with respect to sensing physics, data provenance, AI task, interpretability, validation strategy, and evidence level. Across these modalities, AI supports forward surrogate modeling, inverse design, spectral interpretation, classification, calibration, and uncertainty-aware decision support, but predictive accuracy alone does not establish translational maturity. Particular attention is given to explainable AI, the distinction between physics-guided and genuinely physics-informed learning, small-data validation, fabrication tolerance, drift, and the simulation-to-experiment gap. Quantitative comparisons place reported SPR and THz sensitivities in context, while independent experimental studies provide benchmarks for real-device and biomedical validation. Clinically credible intelligent biosensors should combine mechanistic constraints, uncertainty quantification, device- and batch-aware validation, interpretable features, realistic biological matrices, and prospective evaluation. Full article

Flexible shielding polymers may undergo geometric thinning during tensile use or localized damage, but flat-sheet attenuation measurements cannot resolve the resulting spatial loss of protection. A radiation-transport surrogate comprising 60 wt% thermoplastic polyurethane and 40 wt% ${\mathrm{Bi}}_2$${\mathrm{O}}_3$ (modeled density, 1.835 g ${\mathrm{cm}}^{-3}$) was investigated using prescribed deformation, Geant4, PHITS, Phy-X/XCOM, and 60–120 kVp SpekPy spectra. This is a computational transport model, not a mechanically or experimentally validated formulation. Geant4 mass attenuation coefficients were verified against Phy-X at six energies (30–100 keV) and against XCOM at 90, 91, 92, and 95 keV across the Bi K-edge. Constant-volume uniform strain of 50% raised whole-field transmission by 14.8–32.8%; an illustrative transverse-response sensitivity quantified the dependence of thickness and attenuation on the prescribed kinematics. At 50% local thinning, whole-field transmission increased by only 5.66–15.0%, whereas 5-million-history event-based ROI scoring gave dose enhancement factors of $2.435\pm 0.106$, $1.823\pm 0.064$, $1.508\pm 0.048$, and $1.395\pm 0.042$ at 60, 80, 100, and 120 kVp, respectively. Additional 60 kVp calculations varied the defect dimensions, ROI size, and detector position, revealing the influence of spatial averaging. Beer–Lambert attenuation accounts for the primary response to prescribed uniform thinning; the Monte Carlo framework additionally resolves polychromatic scatter, spectral changes, and localized dose enhancement. These numerical results characterize idealized geometric perturbations and should not be interpreted as measurements of real garment durability or clinical dose. Full article

The persistent threat of chemical warfare agents (CWAs) drives the need for advanced detection and protective technologies. Polymers, with their tunable structures, ease of functionalization, and lightweight nature, have emerged as a versatile platform in this field. This review systematically summarizes recent progress in polymer-based materials for CWA sensing and protection, with emphasis on interaction mechanisms and structure–property relationships. In the realm of sensing, the working principles of polymer-based systems are rooted in electron transfer, hydrogen bonding, fluorescence quenching, and colorimetric response. Conductive polymers enable chemiresistive detection through charge transfer; hydrogen-bond acidic polymers provide selective recognition of organophosphorus agents; conjugated polymers exploit fluorescence quenching via the “molecular wire” effect; and polydiacetylenes offer visible color changes for naked-eye detection. Representative materials and their performance metrics are critically compared. For protection and decontamination, current polymer systems are designed around four synergistic mechanisms: barrier action, physical adsorption, filtration, and catalytic degradation. Barrier layers based on crosslinked networks or graphene/MOF composites suppress agent permeation while maintaining breathability. Porous polymers such as polymers of intrinsic microporosity (PIMs) and coordination polymers provide high-capacity adsorption through tailored surface functionality. Electrospun nanofiber membranes effectively filter aerosolized agents with low air resistance. Catalytic composites incorporating Zr-MOFs or single-atom catalysts enable hydrolysis of nerve agents and oxidation of blister agents under ambient conditions, with recent advances achieving self-buffering and solid-state operation. Despite significant advances, challenges remain in selectivity, environmental stability, and balancing protection with wearer comfort. Future directions point toward multifunctional systems that integrate detection, protection, and self-detoxification within wearable polymer platforms for next-generation chemical defense. Full article

Population aging has contributed to an increased prevalence of conditions that impair lower-limb mobility in older adults, resulting in reduced functional independence and a greater risk of falls. In this context, assistive and rehabilitation exoskeletons have emerged as a technological approach to maintaining and restoring mobility. This study reviews 135 publications from 2016 to 2026 on lower-limb exoskeletons for older adults. The studies were classified according to design, actuation method, assisted joint, application, and evaluation evidence. The results of this analysis show that, within the reviewed literature, multi-joint and ankle exoskeletons are the most frequently represented systems. Among the studies reporting the actuation method, active electric systems and passive unpowered mechanisms are the most common, whereas pneumatic and hybrid solutions are reported less frequently. The reviewed developments also emphasize lightweight design and portability, together with control strategies intended to adapt assistance to the user’s needs. Gait assistance and motor rehabilitation account for the largest body of evidence, whereas sit-to-stand transitions, stair locomotion, and fall prevention remain underexplored. The reported evaluations employ functional, metabolic, spatiotemporal, and electromyographic metrics, with several studies reporting improvements in mobility-related outcomes. However, further clinical validation and more standardized assessments of safety, stability, and long-term performance are needed. Additional progress in regulatory guidance may also facilitate the clinical and everyday use of these technologies. Full article

Functional hydrogels are increasingly used in neural interfaces. As hydrophilic polymer networks that combine tissue-like softness with tunable conductivity and adhesion, they relieve two fundamental limitations of rigid conventional electrodes within a single material: the mechanical mismatch that drives the chronic foreign body response, and the electrical mismatch that raises interfacial impedance. This review discusses functional hydrogels through their core interfacial properties of biocompatibility, adhesion, and interfacial impedance/conductivity, and highlights how these requirements differ between skin-mounted wearable and implantable devices. Their applications in neural recording, electrical and ultrasound neuromodulation, and external and robotic device control are then covered in detail. Finally, the key challenges and future opportunities are summarized, including gelation control, long-term in vivo stability, integration with robotic hardware, and clinical translation. As the field advances, functional hydrogels are poised to become a core platform for next-generation neural interfaces. Full article

We present a novel mapping approach for lower-limb exoskeletons that generates real-time, robot-centric bird’s-eye view (BEV) occupancy maps to support safe and efficient local navigation. This work focuses on a self-balancing wearable humanoid exoskeleton, where BEV mapping is essential for enabling autonomous balance control, footstep planning, and adaptive navigation in complex, real-world environments. The proposed method explicitly incorporates camera motion alongside RGB-D observations to improve mapping accuracy under the dynamic conditions introduced by leg-mounted sensors. To meet the computational constraints of embedded platforms, the model is optimized for real-time operation and can effectively track dynamic elements such as moving pedestrians. We further introduce a semi-supervised framework that combines simulation-based supervised training with unsupervised learning on real-world data, enabling robust generalization despite limited ground-truth labels. Experiments in both simulated and real environments confirm that the model achieves fast inference (17 ms) with low memory consumption (600 MB). Moreover, the model remains robust to the exoskeleton’s motion as well as various sources of environmental noise. Full article

Automatic sign language recognition has become an important research area for improving communication between deaf individuals and the hearing community. While vision-based approaches achieve high recognition performance, their dependence on cameras and environmental conditions limits their suitability for portable real-time applications. This paper presents a fully embedded Moroccan Sign Language (MSL) recognition system based on a dual smart glove equipped with ten MPU6050 inertial measurement units (IMUs). A dedicated dataset of approximately 8000 gesture sequences representing 20 common MSL gesture classes was collected from 80 participants. The acquired multivariate inertial signals were preprocessed and used to train a lightweight Long Short-Term Memory (LSTM) network, which was quantized and deployed on a Raspberry Pi Pico microcontroller using TensorFlow Lite Micro. Experimental results achieved an overall recognition accuracy of approximately 98%, with high precision, recall, and F1-score, with an average inference latency of 27.4 ± 1.0 ms on the embedded platform. The proposed platform demonstrates the feasibility of accurate and low-latency MSL recognition on resource-constrained embedded hardware under controlled acquisition conditions, representing an initial proof of concept toward future wearable assistive communication systems. Full article

Type 2 diabetes mellitus (T2DM) affects hundreds of millions of people worldwide, and nearly half of all cases remain undiagnosed. This technical review, conducted under conducted under the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines and structured around the Population, Intervention, Comparison, Outcome and Context (PICOC) framework, identifies and synthesises the evidence on artificial intelligence (AI) algorithms for the early detection of T2DM and the risk stratification of its complications, evaluating their performance against conventional diagnostic methods and characterising their clinical implementation contexts. A systematic search in Scopus with a four-stage screening process, synthesised across five research dimensions, yielded 255 records, of which 47 studies met the inclusion criteria (2025–2026). Ensemble methods based on gradient boosting, particularly particularly extreme gradient boosting (XGBoost), emerged as the dominant primary modelling approach, while logistic regression was the most frequently reported algorithm, predominantly as a baseline comparator, with area under the receiver operating characteristic curve (AUC-ROC) values ranging from 0.63 to 0.97. AI models consistently outperformed glycated haemoglobin (HbA1c), whose reported sensitivity for prediabetes was only 49%, reaching accuracies of up to 96.7%, although that upper figure comes from a narrowly defined cohort with a high baseline prevalence and does not represent general screening performance. Multimodal approaches achieved the highest discrimination, combining wearables, continuous glucose monitoring, and genomics (AUC 0.96 internal, 0.90 external, for the separation of self-reported glycaemic states), and a real-world clinical deployment integrated into electronic health records improved glycaemic control outcomes among high-risk patients. An assessment of the eight anchor studies with thePrediction model Risk Of Bias ASsessment Tool updated for AI (PROBAST + AI) found high concern regarding quality in six of them, driven mainly by cross-sectional designs and self-reported outcomes. Overall, AI stands out as a viable and scalable complement to conventional screening, although gaps remain in multicentre prospective validation, reporting standardisation, and algorithmic equity across populations. Full article