Abstract Neural electrodes serve as the bidirectional bridge between the nervous system and external devices and play an irreplaceable role in the diagnosis and treatment of neurological disorders. However, conventional electrodes often suffer from high impedance, limited charge‐storage capacity, and poor stability, which constrain their durability. Herein, we coupled a machine learning‐assisted strategy with a multipotential‐step method to optimize the deposition of MoTe 2 quantum dots (QDs) on bare electrodes. After preliminary experimental screening to define the parameter ranges, the surrogate model mapping deposition parameters to overall electrochemical performance score was established and cross‐validated. Prediction over the full parameter space yielded an optimal combination (0.5 mg mL −1 , 0.2 V, 2 s, and 3 layers) that coincided with the experimentally screened optimum, and uncertainty analysis indicated its robustness. Relative to bare electrodes, the resulting MoTe 2 electrodes showed a 6.8‐fold lower impedance at 1 kHz together with 5.8‐ and 9.2‐fold higher charge‐storage capacity and double‐layer capacitance. The coating retained stable electrochemical behavior after 180° bending, repeated cycling, ultrasonic agitation, and 7‐week immersion, indicating good adhesion and durability. Together, these improved electrochemical and mechanical‐durability properties highlight MoTe 2 QDs as a promising coating material for next‐generation neural interfaces.
Abstract Menstrual cycle tracking remains challenged by the limited physiological scope and inconsistent generalizability of existing wearable‐based approaches, which often rely on individual sensing modalities that incompletely capture the complexity of menstrual physiology. This study presents a personalized multimodal machine learning framework for menstrual phase classification by integrating endocrine, autonomic, metabolic, and respiratory physiology through hormonal, glucose, heart rate variability, and respiration‐derived signals. Using the open‐source mcPHASES dataset, 126 menstrual cycles from 42 participants were analyzed over fixed 90‐day monitoring periods. Daily physiological measurements were transformed into statistical, spectral, and nonlinear descriptors and modeled using an Extra Trees ensemble classifier with leakage‐controlled five‐fold cross‐validation and fold‐specific Synthetic Minority Oversampling Technique balancing. Hormone‐derived features consistently provided the strongest individual predictive performance, reflecting the central role of endocrine regulation in menstrual physiology. However, multimodal integration produced the highest overall classification performance, yielding a consistent median improvement of +0.0156 macro F1 over the strongest unimodal models (Wilcoxon signed‐rank test, p = 8.0 × 10 −6 ; effect size r = 0.87). Although these improvements were modest in absolute magnitude, they were remarkably consistent across participants, indicating that multimodal integration primarily enhanced prediction robustness through complementary physiological information rather than large performance gains. The substantial variability in participant‐specific optimal sensing configurations (normalized entropy = 0.834) further demonstrated pronounced inter‐individual physiological heterogeneity. Collectively, these findings suggest that menstrual phase prediction is more appropriately viewed as a personalized multimodal inference problem than a conventional sensor optimisation task. Endocrine biomarkers provide the principal representation of ovarian activity, while autonomic, metabolic, and respiratory signals contribute complementary, context‐dependent information that refines participant‐specific prediction. This adaptive multimodal framework provides a foundation for scalable precision menstrual health analytics, longitudinal reproductive monitoring, and next‐generation personalized wearable health technologies.
Abstract In mainland China, Health Information Technology (HIT) is usually considered a practical combination of informatics and automation for medical care. In practice, HIT has been gradually integrated into daily affairs in most hospitals in China over the past decades. In this work, based on 2016–2021 panel data collected from a nationwide online survey and public yearbook data, an overview of the HIT development in mainland China hospitals is portrayed, demonstrating the features of such a centralized system. The HIT construction network demonstrates the mechanisms and effectiveness of informatization construction in China from 3 levels. First, the top‐level institutional design is established as a driving force which can be represented by the degree increase in the HIT construction network. Second, Grade III hospitals (with abundant medical resources) play a central role in the construction of HIT, exerting a significant correlated development on Grade II hospitals (with fewer medical resources than Grade III hospitals). Third, central and eastern regions display significant impacts on western regions, which is consistent with the policy support by Chinese development. The development of HIT in China shows its positive impacts on hospital efficiency but there are still shortcomings and challenges to address in the near future.
Abstract With global population aging, age‐related hearing loss has become a major public health concern. Understanding the relationship between auditory functional decline and cochlear structural degeneration is essential for early detection and precise intervention. However, conventional morphological approaches are limited in achieving high‐resolution in vivo imaging of the cochlea. Optical coherence tomography enables high‐resolution visualization of cochlear microstructures, yet its in vivo and clinical applicability remains constrained by anatomical and optical limitations. In this study, auditory brainstem response, distortion product otoacoustic emission, and cochlear structural features were collected from mice aged 4–12 months. Five structural parameters related to cochlear development were quantified and correlated with 32 electrophysiological characteristics. For each structural parameter, the three most relevant electrophysiological features were selected to predict cochlear morphological changes based on linear regression and random forest models, both of which achieved moderate predictive performance. This work establishes a noninvasive and repeatable framework that infers potential cochlear structural changes using routine auditory electrophysiological features, thereby overcoming the inherent limitations of current inner ear imaging methods. It provides a potential tool for longitudinal hearing monitoring and early assessment of age‐related cochlear degeneration.
Abstract Inflammatory bowel disease (IBD), with rising global incidence and disease burden, is a multifactorial disorder characterized by chronic and relapsing gastrointestinal inflammation. Its pathogenesis involves immune dysregulation, reactive oxygen species bursts, and microbiota imbalance. Intestinal immune dysregulation drives IBD progression, making precise immune regulation essential for controlling inflammation and improving long‐term outcomes. Despite advances in clinical therapeutics such as corticosteroids and biologics, durable remission remains difficult due to limited targeting and side effects. Nanozymes have emerged as a promising treatment strategy for IBD owing to their structural stability, tunable catalytic performance, and capacity to regulate oxidative stress and inflammation. Deep learning (DL)‐assisted nanozyme engineering facilitates precise matching of catalytic activity and environmental responsiveness to the complex and dynamic inflammatory microenvironment of IBD, thereby enhancing immune regulation and therapeutic efficacy. In this review, we summarize recent advances in nanozyme‐based strategies for IBD, with an emphasis on DL‐assisted design and functional optimization. We discuss current challenges and future perspectives for nanozyme therapeutics, including biosafety evaluation and clinical translation toward precision management of IBD.
Abstract Evidence regarding the prognosis of thyroid carcinoma is heterogeneous, ranging from age effects and nodal burden metrics, such as lymph node ratio (LNR) and log odds of positive nodes (LODDS), to preoperative imaging models comprising ultrasound, CEUS, and radiomics. We conducted a systematic review in line with PRISMA 2020 and SWiM, tabulated under four domains: age relative to the American Joint Committee on Cancer (AJCC‐8) staging system, LNR and LODDS, and preoperative prediction models. Terminology and units were standardized through dual data extraction and Python‐based harmonization. Prognostic studies were evaluated by Quality in Prognosis Studies, and prediction models were assessed using PROBAST. The AJCC‐8 55‐year threshold remains pragmatically useful, yet continuous nonlinear modeling of age offers better support for individualized risk estimates. Supplementing anatomic N staging with LNR significantly enhances prognostication, with compartment‐specific ratios refining the N1 subgroup. LODDS should be coreported with LNR because it is less sensitive to lymph‐node yield, preserves information at extreme values, and often equals or outperforms LNR. Preoperative radiomics and nomograms are promising but often lack external validation and adequate calibration, limiting clinical readiness. Common limitations include endpoint heterogeneity, variable follow‐up, node‐yield dependency, and sparse reporting of calibration or decision‐curve analysis. Residual confounding in retrospective cohorts and reporting bias remain significant challenges.
Abstract Visually induced motion sickness (VIMS) is characterized by symptoms such as nausea, disorientation, and oculomotor discomfort, arising from a visually induced illusionary sense of self‐motion. Although previous studies have produced inconsistent spectral findings, particularly in the alpha band, electroencephalogram (EEG) offers high temporal resolution for objectively assessing VIMS. This scoping review synthesizes 28 studies to investigate the sources of this heterogeneity, focusing on the impact of experimental settings (inducing scenarios and presentation equipment) on EEG outcomes. By categorizing studies into abstract motion cues, virtual scene videos, vehicle driving, and gaming scenes, we reveal distinct neurophysiological patterns. Results indicate that while an increase in delta and theta band activities serves as a consistent correlate of sensory conflict, alpha band activity exhibits context‐dependent divergence: predominantly increasing in passive viewing scenarios but decreasing in active tasks. Furthermore, head‐mounted displays were more frequently associated with alpha enhancement compared to monitors, likely due to higher immersion and spatial orientation demands. This review clarifies these method‐dependent variations and offers practical recommendations for selecting experimental scenarios and parameters to improve the reliability of VIMS assessment.
Abstract In recent years, there has been a notable rise in sepsis incidence leading to more multiple organ failure and higher mortality. The lack of effective treatments for sepsis highlights the importance of early prediction in preventing multiple organ failure. This study aimed to develop a model for the early prediction of multiple organ failure in sepsis patients facilitating timely intervention by healthcare professionals. We extracted data of 2720 sepsis patients from Medical Information Mart for Intensive Care III database and used support vector machine (SVM), logistic regression (LR), random forest (RF), and eXtreme Gradient Boosting algorithms to predict heart, kidney, liver, and respiratory failure. Our models demonstrated strong predictive performance. LR performed best in predicting heart failure and kidney failure with area under the curve (AUC) values of 0.95 and 0.87, respectively. SVM and RF showed good performance in predicting liver failure (AUC = 0.93) and respiratory failure (AUC = 0.87), respectively. Furthermore, we conducted an importance ranking to identify the most significant features for predicting organ failure. The results demonstrated all four models effectively predicted multiple organ failure in sepsis patients. The study highlights machine learning's value as an early prediction tool and clarifies links between organ failure types and physiological parameters.
Abstract Epidemiological and clinical studies have suggested possible associations between type 1 diabetes (T1D) and neurodevelopmental disorders (NDDs), but these relationships remain inconsistent across disorders and populations. To clarify whether such mixed findings, we investigated the genetic architecture linking T1D with autism spectrum disorder (ASD), attention deficit hyperactivity disorder (ADHD), obsessive‐compulsive disorder (OCD), Tourette syndrome (TS), and anorexia nervosa (AN) using genome‐wide association study summary statistics. Global genetic correlations were estimated using linkage disequilibrium score regression, and local genetic correlations across 1703 genomic regions were assessed using ρ‐HESS. Shared genes and pathways were further explored. Global analysis showed limited overall genetic overlap, with a significant negative correlation observed only between ASD and T1D ( p = 0.03). In contrast, local analysis identified 61 genomic regions with significant local genetic correlations between T1D and four disorders, with both positive and negative effect directions. Gene mapping identified two cross‐phenotype candidate genes, ERBB3 shared between ADHD and T1D and GABBR1 shared between ASD and T1D. ERBB3 localized to a region with significant local genetic correlation between ADHD and T1D, and pathway analysis highlighted MAPK signaling as a potential shared mechanism. These findings support limited global but locus‐specific shared genetic architecture between T1D and NDD traits.
Abstract In recent years, the clinical treatment and symptom management of neurological disorders have faced significant challenges due to the high complexity of the nervous system's structure and function. Against this backdrop, physical stimulation techniques have emerged as a vital complementary approach to traditional pharmacological treatments and surgical interventions. These techniques utilize physical modalities to specifically target and modulate neural function and tissue activity, thereby achieving therapeutic outcomes. They offer several advantages, including noninvasive or minimally invasive properties, high safety profiles, and strong adjustability, which enable precise intervention while effectively mitigating the risks associated with drug dependence and adverse effects. This article systematically reviews four physical stimulation modalities—ultrasound, electrical, magnetic, and optical stimulation—focusing on their mechanisms and research findings for neuromodulation and the treatment of neurological diseases. It also summarizes the latest research trends in the field of physical neuromodulation, including the application of novel experimental models (such as brain organoids), the development of closed‐loop modulation systems, and the integration of multimodal stimulation approaches. Furthermore, the article discusses the existing limitations and challenges in this field and outlines future directions and prospects, aiming to provide reference for the further development and clinical translation of physical neuromodulation technologies.
Abstract Auditory attention decoding (AAD) aims to detect the target speaker from electroencephalography (EEG) signals in multi‐talker environments. Existing methods often insufficiently exploit joint spatial and temporal information, which limits decoding performance. This paper presents STHANet (spatiotemporal hybrid attention network), a dual‐branch model that integrates depth‐wise spatial filtering, log–variance temporal characterization, and transformer‐based spatiotemporal fusion. Experiments on the KUL and DTU datasets show that STHANet achieves competitive performance with a 1‐s decision window, reaching accuracies of 93.6% and 75.8% under within‐trial partitioning and 76.7% and 66.1% under strict cross‐trial partitioning, respectively. Further evaluation on the AV–GC–AAD dataset under moving‐target gaze‐incongruent conditions shows that all evaluated direct AAD models decrease to near‐chance performance when gaze‐related shortcuts are more strictly controlled, whereas only STHANet remains significantly above chance. These findings support the effectiveness of STHANet for spatiotemporal EEG feature extraction and highlight the importance of controlling both data‐partitioning bias and gaze‐related confounds in direct AAD.
Abstract As a prevalent non‐invasive screening technique, Wireless Capsule Endoscopy is often hindered by poor image quality, including under‐/overexposure and low light condition. While illumination correction based on diffusion modeling or frequency‐domain decomposition has shown effectiveness, existing methods often (1) underexploit structural information, and (2) lack adaptive strategies for varying illumination degradations, leading to suboptimal restoration and unnecessary computation. To this end, we propose Brownian Bridge Diffusion Transformer‐Mixture‐of‐Experts (BiT‐MoFE), a unified adaptive framework that integrates the merits of the two paradigms for endoscopic illumination correction. We adopt a Brownian Bridge Diffusion framework, in which an efficient Transformer serves as the backbone network, and design a frequency‐decomposed MoFEs module to explicitly handle illumination and image structure simultaneously. By dynamically selecting the most suitable experts conditioned on exposure cues and diffusion timesteps, our framework achieves a strong balance between restoration fidelity and computational efficiency. Extensive experiments on multiple public datasets demonstrate that BiT‐MoFE achieves state‐of‐the‐art performance on both exposure correction and low‐light enhancement tasks.
Abstract This study investigates how long‐distance running training affects daily walking gait, addressing a research gap in runners' everyday gait characteristics. To compare the variations in everyday gait between distance runners and nonrunners, we gathered 28 marathon runners (15 females and 13 men) with good Harris scores and a control group of 11 nonrunners. A wearable device was used to gather data on gait. Gait parameters from single‐task and dual‐task experiments were collected and assessed by the gait analysis system. In the single‐task experiment, runners demonstrated higher gait velocity and cadence than nonrunners; shorter strides, stance and swing phases; significantly lower toe‐off and heel strike angles; and significantly higher symmetry of stride time and heel strike angles. The toe‐off angle and heel strike angle of runners were significantly lower than those of nonrunners throughout the dual‐task trial, and the symmetry of their stride length was superior to that of the nonrunners. For runners, when performing dual tasks, gait velocity, cadence, stride time, and stance phase time were closely related to marathon running pace. Toe‐off angle and heel strike angle may distinguish runners from nonrunners. This study shows that sustained training for marathons can alter walking patterns and improve gait symmetry, decreasing the risk of falling. Comment on the translational aspect of the work presented in the paper and its potential clinical impact.
Abstract Acute kidney injury (AKI) is a common and severe complication of rhabdomyolysis (RM), and early risk stratification remains challenging because of its multifactorial and heterogeneous nature. We developed and externally validated an interpretable machine learning (ML) model for early prediction of AKI in RM across traumatic and non‐traumatic etiologies. Data were obtained from four public critical care databases and a multicenter cohort from tertiary hospitals in China. A total of 1569 patients were included in the derivation cohort and 401 in the external validation cohort. Eighteen variables within 24 h of admission were used to train 12 ML models. Performance was assessed by area under the receiver operating characteristic curve (AUC), and interpretability was evaluated using SHapley Additive exPlanations. The random forest model achieved the best performance (AUC = 0.940) and was simplified into a five‐variable model including lactate dehydrogenase, serum creatinine, Alb, prothrombin time, and activated partial thromboplastin time. The final model achieved AUCs of 0.919 and 0.900 in internal and external validation, respectively, with consistent performance in non‐traumatic (0.911) and traumatic (0.882) subgroups. This interpretable model may support early risk stratification in patients with RM across different etiologies.
Abstract Attentional control theory posits that anxiety disrupts goal‐directed attentional system; however, its neurophysiological mechanisms in generalized anxiety disorder (GAD) remain unclear. This study investigated two core operations of the goal‐directed attentional system in GAD—inhibition and shifting—by characterizing their behavioral and neurophysiological correlates. Twenty‐four patients with GAD and 28 healthy controls completed the Attentional Control Scale (ACS) and performed Go/No‐Go and more‐odd shifting tasks whereas behavioral performance and 64‐channel EEG were recorded. ACS scores indicated significantly lower total and subscale scores in GAD, whereas behavioral performance revealed higher inverse efficiency, reflecting reduced attentional control and processing efficiency. During the Go/No‐Go task, GAD patients showed attenuated NoGo‐N2 and NoGo‐P3 amplitudes, indicating impaired neural subprocesses underlying response inhibition. During the more‐odd shifting task, GAD patients exhibited reduced theta power and weakened theta–gamma coupling (TGC), reflecting disrupted neural coordination supporting cognitive flexibility. Furthermore, anxiety severity was significantly correlated with behavioral performance, NoGo‐N2 and NoGo‐P3 amplitudes, and TGC. These results indicate that greater anxiety severity is associated with impaired inhibitory control, reduced attentional resources, and disrupted neural oscillatory dynamics. Elucidating these deficits may inform the development of targeted interventions aimed at enhancing cognitive control in GAD.
Abstract Postural instability and gait disorder (PIGD) subtype of Parkinson's disease (PD) is marked by heterogeneous motor and cognitive impairments, making rehabilitation response difficult to predict. Identifying robust multimodal predictors is essential for precision rehabilitation. This study aimed to identify key multimodal features associated with response to motor‐cognitive interactive rehabilitation and to develop a generalizable prediction framework. Twenty‐one PD patients with PIGD completed a motor‐cognitive interactive rehabilitation program. Multimodal data, including demographics, clinical scales, gait parameters, magnetic resonance imaging (MRI), and EEG, were collected across 14 feature modalities. A multimodal sequential forward selection framework based on mutual information (MSFSF‐MI) was proposed where predictive stability of selected feature sets was assessed across five machine learning models (support vector machine, RBF, random forest, stochastic gradient boosting, and XGB). Multimodal feature subsets derived by the proposed framework consistently outperformed unimodal models across classifiers. Cross‐model analyses highlighted functional connectivity, cortical thickness, low‐frequency power spectral density, and phase–amplitude coupling as reproducible predictors, forming a key feature set mainly from MRI and EEG domains. This study identified predictive and potentially robust multimodal neural features of PD rehabilitation response. The introduced nested prediction framework demonstrates strong potential for future generalization, providing a methodological foundation for personalized neurorehabilitation strategies.
Abstract MscL‐G22S‐mediated sonogenetics is a promising strategy for precision, cell‐targeted ultrasound neuromodulation of deep brain structures and for next‐generation noninvasive ultrasound brain–computer interfaces. However, how ultrasound fundamental frequency (FF) and pulse repetition frequency (PRF) individually and jointly shape MscL‐G22S‐mediated hippocampal responses remains unclear, and their molecular correlates are largely unknown. Here, we recorded hippocampal local field potentials (LFPs) and ultrasound‐evoked potentials (UEPs) in anesthetized rats expressing MscL‐G22S under factorial combinations of FF (0.5, 1, and 5 MHz) and PRF (40, 1000, and 4000 Hz), followed by transcriptomic profiling and integrative bioinformatics. FF and PRF exerted independent additive effects on LFP total power, with a significant PRF main effect and a nonsignificant but large‐effect‐size FF trend. UEPs revealed pronounced FF‐dependent PRF selectivity: 0.5 MHz stimulation supported UEP induction across PRFs, whereas 5 MHz stimulation required matched high PRF for robust phase‐locked responses. Exploratory transcriptomic analysis identified nonlinear, frequency‐specific gene expression programs enriched in mechanotransduction‐related pathways, together with candidate hub genes and exploratory LASSO‐prioritized candidate molecular features. These findings define a compact electrophysiological–transcriptomic parameter–response map for MscL‐G22S‐mediated hippocampal sonogenetic neuromodulation and provide reference data for future model‐informed optimization of ultrasound parameters.
Abstract Stroke is a leading cause of global mortality and disability, often resulting in severe motor dysfunction. Acupuncture has shown promise in stroke rehabilitation, but its neural mechanisms remain unclear. This study investigated the immediate effects of acupuncture at GV26, PC6, and SP6 on cortical and corticomuscular functional connectivity in stroke patients with motor dysfunction. Fifteen stroke patients were recruited and their resting‐state electroencephalography (EEG), as well as EEG and electromyography (EMG) during unilateral static ankle dorsiflexion were recorded under five conditions: no acupuncture, acupuncture at GV26, PC6, SP6, and all above three acupoints. Then, we analyzed resting‐state brain networks and corticomuscular coherence (CMC) during ankle dorsiflexion tasks. Acupuncture at PC6 and all three acupoints significantly enhanced topological parameters of brain networks in the theta band, indicating improved cortical functional integration. Additionally, acupuncture at PC6, SP6, and all three acupoints increased CMC in beta and gamma bands, suggesting strengthened corticomuscular coupling. These results demonstrate that acupuncture acutely modulates both central and peripheral neural pathways, with acupoint‐specific immediate effects. The study provides preliminary neurophysiological evidence supporting the acupoint specificity effects in stroke patients and highlights the potential for developing personalized acupuncture prescriptions pending further long‐term validation.
Abstract Brain development in preterm infants shows marked heterogeneity, often obscured by group‐level analyses. Between the group‐level and the individual difference, subgroup can model the heterogeneity of early developmental trajectories. To characterize this, we analyzed longitudinal functional connectome data from 90 preterm infants (scanned at birth and term‐equivalent age) and 521 full‐term controls from the developing Human Connectome Project. A machine learning model predicted individual brain‐age gap (BAG), quantifying maturational deviation. Clustering of longitudinal BAG trajectories revealed two distinct preterm subgroups with divergent developmental pathways. These subgroups exhibited significantly different functional network architectures and, at 19‐month follow‐up, distinct behavioral outcomes in cognitive, language, and motor domains. Our findings establish that early preterm brain maturation follows identifiable, heterogeneous trajectories, providing a data‐driven framework for early risk stratification.
Abstract Spatial cognition is a key ability of human cognition and intelligence. In this study, we validated the feasibility and effectiveness of origami training in enhancing spatial cognition and elucidated the underlying neural mechanisms. We assigned participants to either an origami group or a control group, with the origami group completing a training program. We collected electroencephalography (EEG) signals and eye movement data during the spatial tasks pre‐, during‐, and post‐training. A cognitive questionnaire was also collected. We then compared event‐related synchronization and event‐related desynchronization in different bands and constructed weighted Phase Lag Index brain network maps. We also analyzed eye‐tracking metrics. Origami training enhanced cognitive performance, improving accuracy and reducing response time. The origami training increased the frontal midline θ power and decreased the parietal α power. The origami training modulated brain connectivity differently across tasks. Eye‐tracking data revealed a reduction in cognitive load, increased focus, and more efficient cognitive processing following the training. The frontal, parieto‐occipital, and frontal‐occipital regions actively contribute to spatial cognition. Origami training enhances spatial cognition by re‐shaping the brain networks and functional connectivity. These findings support the development of a portable and cost‐effective digital therapy for neurodegenerative disorders.
