IEEE Transactions on Neural Systems and Rehabilitation Engineering最新文献

Understanding the frequency dependent alterations in brain-muscle communication after stroke is crucial for advancing targeted neurorehabilitation strategies. In this study, we propose a novel multilayer corticomuscular network (MCMN) model based on functional corticomuscular coupling characteristics. Using multi-channel electrophysiological recordings acquired during a multi-joint motor task, we constructed a super-connectivity matrix by combining phase synchronization and phase-amplitude coupling across frequency bands. We then examined both local (single-layer) and global (multilayer) network properties by comparing nodal metrics between stroke patients and healthy controls in terms of functional connectivity and topological organization. The results revealed that stroke patients exhibited enhanced theta band within-frequency subnetwork relative to controls, but significantly reduced beta and gamma band subnetworks. Cross-frequency subnetworks in patients showed diminished integrative capacity compared to controls, with the exception of proximal muscle nodes in the beta-gamma subnetwork, which displayed pronounced hub properties. At the global level, patients demonstrated contralateral compensatory reorganization, whereas the contralateral hemisphere exhibited impaired cross-layer integration. The MCMN of stroke patients showed reduced algebraic connectivity, reflecting lower network robustness and information transfer efficiency. Finally, we found that node degree of gamma band and multiplex clustering coefficient of ipsilateral exhibited a linear correlation with FMA-UE scores in stroke patients. This multilayer network approach reveals frequency-specific and topological reorganization of corticomuscular interactions following stroke, providing a novel systems level framework for exploring motor network plasticity and informing precision neurorehabilitation.

The surface electromyogram (sEMG) based motion intent recognition has been considered as a promising approach for prosthetic hand control. However, the variations of EMG signals in daily applications and the insufficient residual muscles of amputees make the clinical application of EMG-based prosthesis remain a challenge. Tactile signals provide critical haptic information for human beings during object manipulation, which is expected to serve as a signal source for motion intent recognition. In this study, a novel tactile-triggered method of motion intent recognition and prosthetic hand control was proposed. This method is implemented by a finite state machine (FSM) using a multi-axis tactile sensor and is deployed in an integrated system. Validation consisted of three evaluation tests, the last involving both able-bodied and transradial-amputee participants. For comparison, the EMG pattern recognition (EMG-PR) control method was also implemented in all three tests. The experimental results demonstrate that the tactile-triggered control strategy provides an effective approach for prosthetic hand opening/closing control, enabling stable grasping of diverse objects of varying shapes and sizes. Compared to EMG-PR, this method substantially eliminates control performance degradation caused by arm posture changes and exhibits superior grasping stability. Additionally, it avoids muscle fatigue in users. This approach offers a promising prosthetic control strategy that can either function as a non-EMG control paradigm or serve as a complementary modality to myoelectric control.

The loss of an upper limb significantly affects daily activities, making advanced prosthesis control crucial for improving the quality of life. Pattern recognition applied to electromyographic signals has emerged as a leading solution for controlling prosthetic hands; yet, most studies focus solely on steady-state muscle activity, neglecting the transient phase of contraction, thereby limiting real-world applicability. To address this limitation, this study introduces a hierarchical approach that combines an Onset Detection Algorithm, a 9-class steady-state gesture classifier, and a three-level force classifier. Additionally, it investigates Self-selected contraction levels across three grasp types, corresponding to subjectively perceived low, medium, and high forces, chosen according to each participant's preference or perceived exertion Results demonstrate improved classification accuracy and responsiveness, particularly during early muscle contraction, outperforming state-of-the-art methods. Moreover, optimal contraction levels were found to be grasp-dependent and significantly lower than those commonly used in the literature, emphasizing the need to adjust reference values to reduce fatigue and enhance comfort.

Restoring sensory function post amputation remains a major challenge. Peripheral nerve stimulation and targeted reinnervation may partially restore somatotopic feedback, but their need for surgery hinders widespread adoption. Here, we investigate the feasibility of transcutaneous spinal cord stimulation (tSCS) as a non-invasive approach for sensory restoration in upper-limb amputees. In a study involving seventeen able-bodied participants and five individuals with upper-limb amputation, we show that tSCS can evoke a range of sensations, including touch, tapping, vibration, and movement, perceived as originating from the missing limb. Notably, these perceptions were primarily isolated to the missing limb and absent in the residual limb in 98% of trials. Participants with amputations found tSCS tolerable, with some reporting increased comfort during stimulation. tSCS evoked sensations in the fingertips of 93% of able-bodied participants, though these were mainly paraesthetic. We further characterised how stimulation parameters, including electrode placement, carrier frequency, and burst frequency, modulated the quality and type of perceived sensations. Additionally, we show that tSCS maintained force proprioception necessary for effective prosthesis control. These findings support the potential of tSCS as a non-invasive sensory feedback approach for upper-limb prosthesis users.

Previously, we proposed a power-free isokinetic training robot designed to provide resistive isokinetic training for knee injury patients during advanced-stage rehabilitation. However, patients in the early stage often lack sufficient muscle strength and necessitate assistive support. To address this limitation, this study introduces a hybrid assistive-resistive isokinetic training robot that integrates active assistance for early-stage knee rehabilitation and power-free resistive training for advanced stages. The system features a compact mechanical design and a reconfigurable control circuit capable of dynamically switching among three modes: active, passive (regeneration), and passive (consumption). Ten healthy subjects and ten knee-injury patients participated in the experimental validation. The results confirmed the adaptability of the system across multiple rehabilitation stages. These findings demonstrate the feasibility of the hybrid assistive-resistive isokinetic training robot and highlight the potential of the system for both clinical application and home-based rehabilitation. Future work will focus on extending the system to multi-joint training and enhancing control algorithms for broader patient populations.

Robotic devices can expand the repertoire of rehabilitation methods by enabling actions that cannot be replicated by a physical therapist. We previously developed a technique, we term 'rapid assistance,' that can assist movements beginning within the electromechanical delay between muscle activation and muscle contraction. Here, we evaluated the effects of repeated arm extension training with rapid assistance in older adults (n = 18) during a single session. We compared training with rapid assistance to a control group that performed unassisted arm extension training. Participants positively adapted to rapid assistance indicated by quickening reaction times (15.69%, t = -1.79, p = 0.089, d = 0.36) and greater extension angular velocities (47.93%, t = 3.47, p = 0.002, d = 0.56) compared to the control group following training. These motor performance improvements following rapid assistance training may be due to reducing Golgi-tendon inhibition during muscle contraction thereby, introducing an alternate strategy to improve motor performance. This specialized assistive timing may address a trade-off present in rehabilitative practice between assisting a patient or sufficiently challenging them to facilitate functional recovery.

Stroke is an emergency cerebrovascular event, resulting in damage to cranial nerves, and the subsequent rehabilitation is necessary to restore neurological function and improve patient outcomes. In clinical settings, the rehabilitation effectiveness is assessed subjectively by experienced physicians or through the use of performance assessment scales. Although several techniques, such as electroencephalography (EEG), cardiopulmonary exercise testing (CPET), and functional magnetic resonance imaging (fMRI), may also be utilized in the evaluation of rehabilitation effectiveness, they require higher costs and the expertise of professional medical staff for experienced operation and post-analysis. Based on the technique of near-infrared spectroscopy (NIRS), a field programmable gate array (FPGA)-based rehabilitation effect assessment headband was proposed. The designed headband could monitor the change in cerebral blood flow non-invasively and continuously under exercise. From the changes in cerebral blood flow, eight perfusion indexes were also extracted in real time and utilized to assess the cardiopulmonary function status of middle-aged and older adults before and after rehabilitation. The analysis algorithm would be completed in the wireless and wearable headband to greatly improve the convenience of use. The experimental results showed that the cardiopulmonary function status could be effectively classified from defined perfusion indexes, and the differences between defined perfusion indexes and the neural network output before and after rehabilitation were also significant.

In recent years, extended reality-based myoelectric training has emerged as a promising approach to prepare users for advanced prosthesis control. This study (1) identified User Needs for an ideal training tool through qualitative interviews with occupational therapists, (2) developed the Myoelectric Training in Extended Reality (MyoTrainXR) system, and (3) evaluated its usability using an advanced postural control strategy. Six individuals with intact limbs and two with trans-radial upper limb loss underwent four 45-minute training sessions with the Block Builder module. The Pasta Box Task was used during training and evaluation, and the Cup Transfer Task was used only during evaluation. In the Pasta Box Task, participants with intact limbs maintained a 100% completion rate, while their success rate increased from 86.1±5.8% to 98.5±3.7%. Participants with upper limb loss began with completion rates between 0% and 40%, improving to 100%, with success rates between 90.9% and 100% by the final evaluation. Iteration completion times showed significant reduction across all participants (p-value < 0.05, linear mixed-effects model), with the median decreasing from 19.4 to 15.8 seconds. The Cup Transfer Task showed a similar trend of significant improvement, demonstrating that the acquired skills generalized to an untrained task. The system also demonstrated excellent usability, with an average System Usability Scale score of 81.9±10.0. These findings indicate that our user-centered extended reality training tool holds promise for enhancing myoelectric control proficiency.

Robot-aided rehabilitation effectively supports treatment of upper-limb disorders and enhances outcomes when combined with traditional therapy. Artificial intelligence enables behavioral cloning of physiotherapists' expertise to autonomously modulate robot assistance from real-time multimodal patient data. Therefore, this paper aims to propose and validate a behavioral cloning strategy, namely Physiotherapist-Supervised Parameter Adaptation (PSPA), for online tuning the robot assistance level replicating the physiotherapists' decision-making. The experimental validation was conducted in a clinical setting involving ten post-surgical orthopedic patients who participated in a robot-aided rehabilitation session using the KUKA LWR 4+ robot. The sessions were supervised by physiotherapists who could adjust the level of robotic assistance as needed, thus labelling the collected patient multimodal data. The validation aimed at i) identifying the best-performing input modality, feature set, and classifier, and ii) comparing the capability of the approach in tailoring the assistance level with respect to the established performance-based (PB) one. Combining biomechanical and physiological features significantly improved the classification performance across all classifiers, with the highest performance observed for the Multi-layer Perceptron on the present dataset. Moreover, using the optimized feature set, the proposed PSPA methodology achieved an even greater alignment with the physiotherapists' decisions with respect to the PB approach (ΔF1-score = 15.40 ± 30.33%, ρ = 0.56 ± 0.21 for PSPA, ρ = -0.12 ± 0.43 for PB).

As skeletal data can be collected non-invasively while preserving patient privacy, it is widely used in public medical datasets to document patient behavior. Autism Spectrum Disorder (ASD) is characterized by significant behavioral heterogeneity, reflected in the topological structure and dynamic evolution of skeletal movements. This complexity poses substantial challenges for skeleton-based behavioral analysis. Existing methods struggle to effectively utilize behavioral evolution for subject-specific reasoning, leading to suboptimal representations that lack diagnostic relevance for autism. To address this limitation, we propose a Behavioral Evolution-based Edge Reconstruction (BER) Strategy for learning autism-related behavioral representations. By reconstructing a high-granularity adjacency matrix that spans both spatial and temporal dimensions, utilizing dynamic evolution and spatial location information, BERGCN enhances behavioral reasoning. Specifically, we first compute channel-level spatial and temporal edge reconstruction parameters by performing feature compression and targeted convolution operations on the differences between neighboring frames. Based on these, the spatial edge reconstruction module is designed by combining a generic attention map with two personalized attention maps, while the temporal edge reconstruction module is implemented using flexible frame replace ment and weighted aggregation. Finally, we investigate both single-modal and multimodal network architectures under various fusion strategies. We evaluate BERGCN on three autism clinical tasks and a benchmark action recognition dataset. Experimental results demonstrate competitive performance, showing improved sensitivity to subject-specific behavioral patterns while maintaining computational efficiency.