Journal of NeuroEngineering and Rehabilitation最新文献

Background: Severe upper limb impairment (ULI) presents a significant challenge in the rehabilitation of chronic stroke survivors and affects their quality of life. Identifying biomarkers and understanding the neural mechanisms associated with severe ULI are essential for evaluating recovery potential and enhancing rehabilitation effectiveness. This study aimed to identify resting-state electroencephalography (EEG) functional connectivity features associated with severe ULI in chronic stroke survivors using machine learning (ML) methods.

Methods: EEG data were collected from 34 chronic stroke survivors. Participants were categorized into two groups based on their Fugl-Meyer Assessment for Upper Extremity (FMA-UE) scores: a mild/moderate ULI (FMA-UE ≥ 30; n = 19) and severe ULI (FMA-UE < 30; n = 15). We employed ML algorithms to classify severe ULI, including logistic regression with L1, elastic net regularization, stochastic gradient descent, and support vector machines, along with several feature selection methods. Coherence was evaluated across six frequency bands in both the ipsilesional (affected by the lesion) and contralesional (opposite side of the lesion) hemispheres.

Results: The logistic regression model with L1 and ReliefF feature selection methods was the most effective, achieving a balanced accuracy of 0.91 (sensitivity = 0.93; specificity = 0.90). This approach identified 14 significant features for distinguishing severe ULI from mild to moderate ULI, including delta interhemispheric and intrahemispheric connectivity in the frontal, parietal, and temporal regions. Additionally, interhemispheric and intrahemispheric theta connectivity was observed in the prefrontal, frontal, temporal, and parietal regions. Low-beta intrahemispheric connectivity was also observed in the contralesional parietal regions.

Conclusions: Our research highlights the association between alterations in connectivity within low-frequency bands and severe ULI across widespread brain regions, including areas outside the sensorimotor cortex and bilateral intrahemispheric and interhemispheric regions. Further research utilizing larger longitudinal datasets from early stroke survivors employing ML approaches could contribute to the development of more accurate predictive models for motor recovery and rehabilitation responses.

Background: Individuals who experience severe speech and physical impairment face significant challenges in communication and daily interaction. Visual brain-computer interfaces (BCIs) offer a potential assistive solution, but their usability is limited when facing restrictions in eye motor control, gaze redirection and fixation. This study investigates the feasibility of a gaze-independent visual oddball BCI for use as an augmentative and alternative communication (AAC) device in the presence of limited eye motor control.

Methods: Seven participants with varying degrees of eye motor control were recruited and their conditions were thoroughly assessed. Visual oddball BCI decoding accuracy was evaluated with multiple decoders in three visuospatial attention (VSA) conditions: overt VSA, with fixation cued on the target, covert VSA, with fixation cued on the center of the screen, and free VSA without gaze cue.

Results: covert VSA with central fixation leads to decreased accuracy, whereas free VSA is comparable to overt VSA for some participants. Furthermore, cross-condition decoder training and evaluation suggests that training with overt VSA may improve performance in BCI users experiencing gaze control difficulties.

Conclusions: These findings highlight the need for adaptive decoding strategies and further validation in applied settings in the presence of gaze impairment.

Unilateral spatial neglect (USN) is a failure to respond or orient to stimuli in contralesional space, not explained by primary sensory or motor deficits. It affects up to two-thirds of right hemisphere stroke survivors and significantly impacts rehabilitation and functional outcomes. Recent advances in three-dimensional (3D) technologies, such as virtual reality (VR) and robotics, offer promising tools for assessment and treatment, providing realistic scenarios and precise clinical stimulation. This systematic review explores the current use of 3D technologies in USN, focusing on their features, level of development, and reported outcomes. A structured search of four databases using the PICO format identified 37 relevant studies out of 2891. The most frequently employed technologies were immersive and non-immersive VR, augmented and mixed reality, and robotics. However, these tools are still in early experimental phases. Among the studies, 15 addressed assessment, 17 focused on treatment, and 5 were technical in nature. Key challenges include methodological variability and the lack of standardized protocols. Due to the heterogeneity of technologies and outcomes, a meta-analysis was not feasible. Future studies should adopt rigorous designs to validate these approaches and support their integration into clinical practice.

Background: Artificial sensory feedback can improve function and user experience in lower-limb prosthesis users. Non-invasive methods like vibrotactile stimulation are clinically convenient, as they require no surgery. Most studies evaluate single feedback approaches, typically under controlled conditions promoting reliance on feedback. This study presents a flexible framework to compare multiple feedback approaches using microprocessor-controlled prosthesis (MP) sensors during daily-life activities.

Methods: Ten able-bodied participants and one prosthesis user with transfemoral amputation (TFA) tested two feedback locations (waist "Belt", or thigh/residual limb "Socket") to investigate tradeoffs between perception quality and compactness, using Sensation Thresholds (ST), Weber Fraction (WF), Spatial Discrimination (SD), and comfort. TFA then completed an out-of-the-lab walking session with the Socket configuration to evaluate the impact of four feedback approaches on spatiotemporal parameters and kinematics symmetries, cognitive load, and user experience during overground walking and stair climbing. Three approaches used embedded MP sensors, conveying (1) knee angle, (2) hybrid (gait phases overground, knee angle during stairs), and (3) damping (velocity-dependent resistance to flexion/extension) feedback. The fourth method used a sensorized insole, providing (4) force feedback (plantar pressure under the prosthetic foot).

Results: Able-bodied participants perceived the Belt configuration better-lower ST (29.09 ± 0.60% vs. 33.19 ± 0.60%, p < 0.001), lower WF (14.49 ± 7.02% vs. 17.98 ± 5.72%, p < 0.01), better SD at higher task difficulty (four choices: 99.3 ± 2.0% vs. 91.5 ± 2.0%, p < 0.01; eight choices: 96.0 ± 2.0% vs. 78.1 ± 2.0%, p < 0.001)-and found it also more comfortable (9.17 ± 0.3 vs. 8.15 ± 0.3; p < 0.05). Similar trends were observed in TFA. Feedback did not impact the kinematics symmetry but slightly affected stance time/percentage symmetry, with force feedback demonstrating the most consistent benefits. These suggest that incidental feedback provided intrinsically by the prosthesis (e.g., motion, sound, socket pressure, vibration) may already support gait in experienced users. Nevertheless, TFA preferred having feedback, especially force and damping, which reduced cognitive load.

Conclusion: Embedded MP sensors enable flexible, compact feedback solutions, combining internal signals (e.g., damping feedback) with external sensing (e.g., omnidirectional force feedback). Belt-mounted vibromotors are effective for testing complex encoding schemes. Feedback should be co-developed with users, balancing objective performance and subjective experience.

Background: Stroke is a condition marked by considerable variability in lesions, recovery trajectories, and responses to therapy. Consequently, precision medicine in rehabilitation post-stroke, which aims to deliver the "right intervention, at the right time, in the right setting, for the right person," is essential for optimizing stroke recovery. Although artificial intelligence (AI) has been effectively utilized in other medical fields, no current AI system is designed to tailor and continuously refine rehabilitation plans post-stroke.

Methods: We propose a novel AI-based decision-support system for precision rehabilitation that uses reinforcement learning (RL) to personalize the treatment plan. Specifically, our system iteratively adjusts the sequential treatment plan-timing, dosage, and intensity-to maximize long-term outcomes based on a patient model that includes covariate data (the context). The system collaborates with clinicians and people with stroke to customize the recommended plan based on clinical judgment, constraints, and preferences. To achieve this goal, we propose a contextual Markov decision process (CMDP) framework and a novel hierarchical Bayesian model-based RL algorithm, named posterior sampling for contextual RL (PSCRL), that discovers and continuously adjusts near-optimal sequential treatments by efficiently balancing exploitation and exploration while respecting constraints and preferences.

Results: We implemented and validated our precision rehabilitation system in simulations with 150 diverse, synthetic patients. Simulation results showed the system's ability to continuously learn from both upcoming data from the current patient and a database of past patients via Bayesian hierarchical modeling. Specifically, the algorithm's sequential treatment recommendations became increasingly more effective in improving functional gains for each patient over time and across the synthetic patient population. As a result, the algorithm's treatments were superior to non-adaptive, "one-size-fits-all" dosing schedules (uniform, decreasing, and increasing).

Conclusions: Our novel AI-based precision rehabilitation system, based on contextual model-based RL, has the potential to play a key role in novel learning health systems in rehabilitation.

Background: Sensorimotor impairments following stroke frequently result in diminished voluntary control of the ankle, contributing to deficits in balance and gait. Robotic training paradigms targeting ankle motor control often use an assist-as-needed strategy, where compliant guidance is provided to assist movements towards a target trajectory. However, interaction with "perfect" reference trajectories may overly constrain movements during training and has been shown to limit learning in many upper-limb contexts; alternatives to robotic assistance have rarely been explored for post-stroke ankle training. Inspired by human-robot-human interaction studies, we investigated whether physical interaction with a therapist-termed human interaction-offers advantages over traditional trajectory guidance regarding short-term learning.

Methods: In a within-subject design, nine individuals with chronic stroke (61.6 ± 14.3 years) performed a 1-DoF visuomotor tracking task while wearing ankle robots designed to train dorsiflexion and plantarflexion movements. Two robotic training methods were evaluated in separate visits: (1) compliant connection to a sinusoidal target trajectory (i.e., trajectory guidance) and (2) compliant connection to a physical therapist who tracked the same target trajectory (i.e., human interaction). In each visit, tracking performance (i.e., errors, movement smoothness) and muscle activation were evaluated during and immediately after training.

Results: Both training types improved tracking accuracy and movement smoothness during training, however random error was more significantly suppressed with trajectory guidance. Immediately after training, we found no significant difference in tracking accuracy or movement smoothness across training types. However, participants demonstrated significantly higher dorsiflexor activation after training with human interaction compared to trajectory guidance.

Conclusion: Our results suggest that human interaction is a viable strategy for training ankle movements in chronic stroke participants, likely by providing assistance without over-constraining an individual's movement smoothness or variability. Training while physically interacting with a partner could serve as an effective alternative to conventional robot-guided therapy for post-stroke ankle rehabilitation, though further studies with larger cohorts are needed to assess the generalization of this approach regarding long-term retention and functional improvement. Registry: clinicaltrials.gov, TRN: NCT04578665, Registration date: 8 October 2020.

Background: Motor imagery (MI) is a widely used paradigm in brain-computer interface (BCI) research due to its potential applications in areas such as motor rehabilitation. As a purely cognitive process, MI produces low-amplitude, non-stationary EEG. Despite improving accuracies, cross-subject variability and limited generalization continue to motivate approaches that strengthen MI representations and enhance system robustness.

Methods: We designed a task-guided preparatory movement state-based motor imagery (PMS-MI) paradigm that elicits a brief motor preparatory state before MI and captures EEG features from both the preparation and imagery phases. To decode the features effectively, we introduced a multilayer energy decoder (MLED) that integrates graph signal processing (GSP): EEG is modeled as intra- and cross-frequency multilayer brain networks, and a graph Fourier transform (GFT) projects the signals into network energy features before classification. We benchmarked the PMS-MI paradigm and the MLED method across multiple time window lengths using a panel of classical and deep-learning classifiers.

Results: The PMS-MI paradigm elicited significant energy variations during the movement preparation phase and induced earlier event-related desynchronization (ERD) with broader frequency band activation during MI, compared to traditional MI paradigms. Classification performance using CSP in the PMS-MI paradigm surpassed that of the traditional paradigm at all time windows. Further accuracy improvements were achieved with the MLED method. Brain network analysis revealed distinct neural representations between the preparation and MI phases, and MLED effectively captured these differences. Feature fusion of preparation and MI stages resulted in classification accuracies exceeding 85% for both 1 s and 4 s windows. The results demonstrate that both algorithmic design and paradigm choice play important roles in MI EEG decoding, with their relative contributions varying across temporal windows and experimental conditions.

Conclusions: Integration of preparatory movement states into the movement imagery process can generate distinguishable features at different stages and improve the classification performance of BCI systems. The proposed PMS-MI paradigm, combined with the MLED decoding method, provides a promising direction for developing more accurate and robust BCIs, particularly in the context of neurorehabilitation.