IEEE Transactions on Neural Systems and Rehabilitation Engineering最新文献

Behavioral and psychological symptoms of dementia pose challenges to the safety and well-being of individuals in residential care. The integration of video surveillance in common areas of these settings presents a valuable opportunity for developing automated deep learning methods capable of identifying such behavior of risk. By issuing real-time alerts, these methods can support timely staff intervention and reduce the likelihood of incidents escalating. However, a persistent limitation is the considerable drop in performance when these methods are deployed in environments unseen during training. To address this issue, we propose an unsupervised scene-invariant fusion-based deep learning network. It combines language model-based captioning and scoring with video anomaly detection scoring to improve the generalization performance for unseen camera scenes. The video anomaly detection scoring uses a depth-weighted spatio-temporal autoencoder to reduce false positives, and the caption-based scoring uses a large language model to generate anomaly scores from captions of video frames. The study uses video data collected from nine individuals with dementia, recorded via three distinct hallway-mounted cameras in a dementia unit. The performance was investigated in both the same camera and cross-camera settings, where the proposed method performed consistently better than the existing methods. The proposed approach obtained the best area under receiver operating characteristic curve performance of 0.855, 0.84 and 0.805 for the three cameras. This work motivates further research to develop cross-camera behavior of risk detection systems for people with dementia in care environments.

Supernumerary robotic fingers (SRFs) are wearable assistive devices, which are increasingly incorporated into robotic rehabilitation programs aimed at restoring upper-limb function and promoting task-specific compensation. Despite growing evidence of SRF efficacy in improving motor performance, limited attention has been given to physiological adaptation and autonomic nervous system (ANS) integration during SRF use. This study investigated phase coherence (PC) and amplitude-weighted phase coherence (AWPC) of RR intervals derived from photoplethysmogram (PPG) as noninvasive biomarkers for autonomic nervous system adaptation during SRF-assisted activities of daily living. Thirty healthy participants completed a baseline (no SRF), pre-training SRF application and post-training SRF use, including rest periods protocol. Drinking water, driving and shape sorting were the functional activities of daily living (ADLs) that had to be completed. The results for PC and AWPC in the low (0.04-0.15) and high (0.15-0.4) frequency bands indicated an overall significant reduction in stress associated with SRF use (p <0.05). During the shape sorting task, post-training AWPC was significantly higher than in the pre-training phase (p = 0.037), and PC also increased significantly (p = 0.044), indicating enhanced vagal modulation. Driving task AWPC improved in the high-frequency band increasing from $0.68~pm ~0.12$ (no SRF) to $0.74~pm ~0.10$ (pre-training SRF) and $0.79~pm ~0.09$ (post-training SRF), while PC increased from $0.54~pm ~0.11$ to $0.62~pm ~0.08$ after training demonstrating significant task, phase, and frequency-specific alterations in autonomic coherence. This work provides an innovative perspective on physiological embodiment and how robotic compensation/augmentation improve both motor performance and physiological regulation. PD analysis indicated central autonomic adaptation. The current findings support the integration of coherence-based autonomic measures into assistive device evaluation frameworks to optimize training protocols and personalize robotic rehabilitation strategies.

Aphasia is a common condition following brain injury, traditionally assessed and treated by speech therapists through manual evaluations and conventional language rehabilitation. However, these methods are time-consuming, reliant on professionals, and subject to subjective biases. This study aims to develop a virtual speech therapy room with an automated assessment model to assist clinicians in evaluation. It provides immersive virtual reality (VR) language training modules, combining analysis of physiological data to achieve the goal of smart healthcare. Twenty individuals with aphasia (IWA) and ten healthy participants were involved, with aphasia subjects randomly assigned to the experimental and control group A, and healthy participants forming control group B. Clinical scales, VR tasks, and neurobehavioral data were measured as needed. Statistical analysis confirmed that using virtual reality can enhance the effectiveness of aphasia treatment interventions. Utilizing virtual reality and behavioral sensing technology, significant differences were observed in the left frontal and occipital regions between IWA and healthy participants, aligning with clinical observations of impaired language and visual processing areas. The assessment model, established through these data, achieved an average classification accuracy of 97% in distinguishing between individuals with aphasia and healthy participants using multimodal fusion with repeated cross-validation, indicating its potential as an auxiliary tool for physician assessment and treatment.

Lower-limb prosthesis users often overuse their intact joints due to the lack of positive work generated by their devices. This overreliance has been shown to increase joint loading, degeneration, and pain. While powered prostheses can generate positive work and therefore reduce this burden, clinical studies of commercialized single-joint devices have demonstrated inconsistent results. Recently, prototype powered knee and ankle prostheses have shown more consistent advantages over passive devices in laboratory settings. Most of the studies, however, focus on the biomechanics of the prosthesis rather than its impact on the user's joints, study isolated activities, and/or do not replicate the demands of continuous real-world use. This case series analyzes the intact joint moments and work for N=3 above-knee amputee subjects using a powered knee-ankle prosthesis vs. their prescribed passive device during a continuous, sustained sequence of the primary activities of daily life. The powered prosthesis decreased peak hip flexion moment (but increased peak extension moment) during level walking, and decreased peak knee extension moment for all other activities. For at least two of the three subjects, the powered prosthesis decreased total positive work across the intact joints during ascent activities (stair ascent, sit-to-stand) and decreased negative total work for descent activities (stair descent, stand-to-sit). This case series suggests that powered knee-ankle prostheses have the potential to reduce overuse of intact joints in emulated real-world conditions.

Previous work suggests that people with mobility impairments move differently in the traditional gait laboratory than in natural environments, but the extent to which this possible divergence may have limited understanding of gait following anterior cruciate ligament reconstruction (ACLR) remains unknown. We hypothesized that 1) patients following ACLR walk more asymmetrically in daily life than in the laboratory and 2) differences between patients following ACLR and individuals with no gait pathologies would be more pronounced in daily life than in the laboratory. Twelve participants (six post-ACLR, six healthy) completed gait assessments in the laboratory with optical motion capture and inertial measurement units (IMUs) and in daily life with IMUs. Patients walked similarly to healthy participants in the laboratory, but in daily life they walked with a longer gait-cycle time ( $1.22~pm ~0.03$ s vs. $1.08~pm ~0.06$ s), longer double-support phase ( $21.8~pm ~1.3$ % vs. $17.9~pm ~2.7$ % Gait Cycle Time), and greater single-support asymmetry ( $8.4~pm ~0.9$ % vs $4.1~pm ~0.7$ %). While the reasons behind the observed differences were not studied, these results suggest that gait-analysis studies to date may have not realistically captured natural-environment behavior. Wearable sensors now offer a path toward deeper understanding of post-surgical gait and the specific patterns that may place certain patients at risk for post-traumatic osteoarthritis.

Hemiplegic shoulder pain (HSP) is a common complication following stroke, significantly affecting upper limb recovery and quality of life. However, the underlying pathophysiological mechanisms of HSP remain poorly understood, which poses a major obstacle to the development of effective therapeutic strategies. This study aims to investigate the underlying mechanism of HSP by evaluating neuro-musculo-vascular activities using high-density surface electromyography (HD-sEMG), musculoskeletal ultrasound, and laser speckle contrast imaging (LSCI). A total of 12 HSP patients and 12 hemiplegic controls without shoulder pain (HNSP) participated in this study. Their neuro-musculo-vascular data in the affected shoulder were collected using a 64-channel HD-sEMG electrode array, a musculoskeletal ultrasonic probe, and a LSCI sensor. Muscle activity was quantified by the root mean square (RMS) of HD-sEMG signals, while neural firing activity was characterized by the discharge rate and coefficient of variation (CoV), decomposed from the HD-sEMG. Meanwhile, structural characteristics were measured by the shoulder subluxation distance (SSD) from ultrasonic image, and blood perfusion was evaluated by perfusion units (PU) from LSCI. Results showed significantly lower RMS and CoV in HSP group versus HNSP group (p<0.05), both strongly correlated with pain intensity (RMS: r=-0.792, ${p}={0}.{002}$ ; CoV: r=-0.698, p= 0.012 ). Pain intensity also linked to greater SSD (p <0.001) but not PU value (p >0.05), while SSD negatively correlated with both RMS and CoV (p <0.05). These findings suggest that HSP is more closely related to neuromuscular control abnormalities and shoulder joint instability than to microcirculatory dysfunction, emphasizing the need for targeted neuromuscular rehabilitation in treating HSP.

Brain functional network analysis models the brain as a graph of regions of interest (ROIs) and quantifies the correlations across different regions derived from functional magnetic resonance imaging (fMRI). Recently, artificial intelligence-based brain functional network analysis methods have demonstrated exceptional performance in diagnosing related neurological disorders. These approaches primarily focus on extracting relevant information from global connectivity patterns to analyze functional brain networks. However, medical research indicates that the impact of brain disorders predominantly manifests in localized functional connections among disease-relevant regions. Treating all connections equally risks introducing interference from irrelevant brain regions, thereby compromising diagnostic accuracy. To address this issue, we propose a novel sub-connection learning method that effectively identifies diagnostically specific connections while suppressing ineffective redundant connections. Specifically, we begin by employing a dynamic functional connectivity construction strategy to generate a functional connectivity matrix encapsulating dynamic features. Subsequently, we design a sub-connection Mask Learning strategy, which employs a multi-head self-attention mechanism to adaptively learn connection masks from functional connectivity matrices, enabling the capture of disease-specific connections and the suppression of noise connections. Additionally, we introduce a Multi-mask Fusion strategy and a Mask Iterative Optimization strategy to further enhance mask quality. Experimental results demonstrate that our model outperforms state-of-the-art methods on the ABIDE I and ADNI datasets, achieving accuracies (ACC) of 72.30% and 80.99%.

Cervical spondylotic myelopathy (CSM) and parkinsonian syndromes (PS) present similar motor symptoms, often causing misdiagnosis due to current clinical diagnostic limitations. Misdiagnosis can exacerbate patient conditions or result in unnecessary surgical interventions, thereby increasing surgical risks and the likelihood of serious postoperative complications. This study aims to develop a mixed dual-branch network for classifying CSM patients, PS patients, and healthy individuals using gait data. This study recruits 51 CSM patients, 49 PS patients, and 33 healthy controls. The kinematic data are collected and used to calculate the time series of angle, angular velocity, and angular acceleration for the hip, knee, and ankle joints. From each time series, 20 features are extracted, including the time domain, frequency domain, time-frequency domain, and nonlinear features. A dual-branch model named DCDM-Net is proposed to classify subjects through collaborative decision making (CDM) method, with one branch using ResNet with convolutional block attention module (CBAM) and evidential deep learning (EDL) loss for analyzing time series, and the other employing multilayer perceptron (MLP) for dealing with multi-domain features. DCDM-Net achieves an ACC of 92.35% $pm ~0.76$ % and an AUC of 96.70% $pm ~0.47$ % in the three-class classification task. Additionally, in binary classification scenarios, the model demonstrates robust performance with an average ACC of 93.13% and AUC of 98.34%. Furthermore, comparative evaluations show that the integrated EDL module surpasses Softmax, MC-Dropout, and Deep Ensembles in uncertainty estimation, yielding the lowest Expected Calibration Error (ECE of 0.0304) and lower Brier score (0.1074), indicating superior reliability. However, cross-dataset OOD validation yielded an AUROC of $0.4022~pm ~0.2481$ and an AUPR of $0.9699~pm ~0.0162$ , revealing that restricting features to joint angles leads to significant distribution overlap; this conversely validates that angular velocity and acceleration are indispensable for preventing model overconfidence. Interpretable results obtained through the SHapley Additive exPlanations (SHAP) method and the integrated gradients (IG) method are confirmed by clinical findings. Our method provides a promising tool for diagnosing CSM and PS, with the potential to reduce misdiagnosis. The code implementation of this study is available at https://github.com/AImedcinesdu212/DCDM-Net.

Transcranial Focused Ultrasound Stimulation (tFUS) is a promising non-invasive technique capable of modulating brain activity with high spatial precision. However, its efficacy for seizure suppression requires further exploration. This study aims to address whether tFUS of white matter can suppress seizures non-invasively. Repeated injections of a 4-Aminopyridine (4-AP) cocktail into the right somatosensory cortex (S1) induced cortical seizures in male rats under anesthesia with recording of both EEG and intracranial signals. Approximately one hour of tFUS was applied to the corpus callosum (CC) using a 128-element random-array transducer with 20 ms pulse duration, 1 Hz pulse repetition frequency, 2% duty cycle, and ~127 kPa pressure. Another 2-3 hours were used to assess post-stimulation effects. Seizure duration, seizure count, percent time in seizure, and inter-seizure interval were compared to a sham control for quantifying efficacy. The absolute frequency power, asymmetry index (AI), and phase lag index (PLI) were calculated to analyze brain activity changes induced by tFUS. CC tFUS can significantly reduce percent time in seizure, seizure duration, and seizure count, as well as increase inter-seizure interval. These effects extended up to 2 hours post-stimulation. We also observed a decrease in absolute power of the beta band and changes in the brain network, as evidenced by a decrease in synchronization and an improvement in interhemispheric balance. Our study is the first to show that white matter tFUS can significantly suppress seizures with a lasting post stimulation effect, potentially providing a safer alternative for drug-resistant epilepsy patients.