IEEE ... International Conference on Rehabilitation Robotics : [proceedings]最新文献

The research paper examines the challenges of remote control delays in exoskeleton-aided physiotherapy. It focuses on the implications of telerehabilitation requirements for safe treatment. As the global share of disabled and elderly people increases, the demand for effective teleoperated physiotherapy solutions rises, necessitating precise control over robotic systems. The study utilizes a mixed-methods approach to assess delays in communication between a rehabilitation exoskeleton and the control system's based on the digital twin in VR (virtual reality) components at different stages. Moreover, additional delays coming from distant operations are assessed. Key findings reveal that the average latency for joint control on shorter distances remains below 300 ms-considered acceptable for comfortable manual teleoperation. The delays between setting the device's configuration and reaching it by a physical exoskeleton differ for particular joints and depend mainly on the torque-to-moment of inertia ratio. The total communication cycle should not exceed 400 ms, which can be further reduced by using a 6G network in the future or ADS (Automation Device Specification) communication protocol instead of the database. Ultimately, the study provides a foundation for developing algorithms that can mitigate delays and improve the efficacy of telehealth solutions with physically supportive devices.

Upper limb differences present tremendous challenges for autonomy in daily living, and current prostheses often face high abandonment rates due to complexity and lack of functionality. This study investigates fully unsupervised incremental myocontrol using higher-density surface EMG. Utilizing incremental sparse non-negative matrix factorization (ISNMF), we employed two 32-channel sEMG bracelets to incrementally extract muscle synergies from EMG signals in real-time. Eight able-bodied participants underwent this unsupervised training paradigm with an increasing number of target synergies and were evaluated with a virtual target achievement control (TAC) test. Participants demonstrated up to six independently controllable synergies in full-intensity tasks, exceeding the current state of the art. However, proportional control remained challenging, reflected in a median success rate of 10% for half-intensity targets. Subjective feedback across the number of synergies showed only small variations in cognitive and physical workload despite increased complexity. This approach shows promise for enabling fully unsupervised myocontrol, but further refinement of training protocol and hyperparameters, as well as testing on users with limb differences, are necessary to validate and improve this approach.

Stroke affects the neuromuscular control of humans, influencing muscle coordination and consequently gait patterns. Robotic devices can be useful to improve walking function and kinematics after stroke, but there is not much research related to muscle coordination through muscle synergies. This is the first study in which muscle synergies have been studied in functionally-dependent subacute post-stroke patients (FAC <2) in a robotic rehabilitation context. We analyzed muscle synergies during gait of three post- stroke patients before and after 2 months of robotic rehabilitation and correlated the results with functional improvements and kinematic changes. After the rehabilitation, patients walked faster, more independently, with normalized gait patterns and different muscle coordination. However, the results obtained underline the importance of a patient- specific therapeutic approach guided by the combination of multi-level metrics.

Stroke often causes sensorimotor deficits, impairing hand dexterity and disrupting independence for millions worldwide. While rehabilitation devices leveraging visual and haptic feedback show promise, their effectiveness is limited by a lack of perceptual equity, which is necessary to ensure fair comparisons between sensory modalities. This study refines cross-modal matching protocols to address this gap, enabling unbiased evaluation of multimodal feedback. Using the Hand Articulation and Neurotraining Device (HAND), 12 healthy participants matched visual and haptic stimuli in a structured task. A streamlined protocol, requiring just $2-3$ blocks and 3 reference intensities, reduced experimental time fivefold while preserving data integrity. Data were analyzed using linear and exponential models applied to both full and reduced datasets. The results demonstrated consistent participant performance across trials, with higher matching errors at greater stimulus intensities, likely attributable to sensory saturation effects. Furthermore, the study offered practical methodological insights, including the use of reduced data sampling paradigms to enhance experimental efficiency significantly while preserving data integrity. This work advances perceptual equity in multisensory feedback systems, addressing sensory encoding variability to support scalable, personalized therapeutic strategies for stroke recovery.

In recent years, a novel human-machine interface for prosthetic control has been developed: the myokinetic interface, which decodes the user intent by monitoring the displacement of implanted magnets in the muscles. Simulations and the first in-human demonstration of this interface indicate that the placement of the magnets is crucial for the range and stability of the control signals. Therefore, a preoperative estimation of individual muscle displacement is necessary to identify optimal implantation regions and generate synthetic datasets of magnet displacement. In this study, we developed a finite element model of pennate muscles, calibrated and validated using the geometries of healthy muscles and in vivo measurements from healthy subjects. The performance of the model was further assessed on three amputated muscles by comparing simulations with in vivo data from a limb-impaired individual. Overall, the simulation results aligned well with experimental data, with average errors below 0.7 mm for the healthy muscles and 1.7 mm for the amputated ones. These results suggest that this model could serve as a valuable tool for optimizing surgical procedures and control strategies prior to clinical implementation. This framework could be expanded to investigate muscle behavior in different amputee populations or individuals with neuromuscular diseases, to enhance understanding of muscle biomechanics and advance the design of personalized rehabilitation devices.

Human Activity Recognition (HAR) is a valuable tool for healthcare and rehabilitation, enabling applications like remote patient monitoring and rehabilitation progress assessment. This paper introduces TIFEX-Py, a comprehensive Python toolbox designed for time series feature extraction in HAR. TIFEX-Py offers a rich set of 195 feature extraction methods across statistical, amplitude, spectral, and time-frequency domains. To evaluate its effectiveness, TIFEX-Py was applied to 11 publicly available HAR datasets: DSADS, HHAR, MHEALTH, MotionSense, PAMAP2, REALDISP, RealWorld, UniMiBSHAR, USC-HAD, WARD, and WISDM. Machine learning pipelines utilizing TIFEX-Py features, evaluated under both random and subject-stratified cross-validation settings, consistently achieved performance that is competitive with or superior to state-of-theart (SOTA) benchmark performances available for the datasets. In 11 out of 11 random split cross-validation scenarios, our pipeline surpassed or matched SOTA performance. For stratified by subject cross-validation, this was the case for more than half of the datasets. These results highlight the power of TIFEX-Py's feature space in representing time series data. TIFEX-Py is opensource and publicly available for researchers in rehabilitation and movement analysis fields.

We present a case study investigating the impact of integrating a novel rehabilitation device (Andago) into physiotherapy services of a public hospital in Australia. A 3month baseline of usual therapy was compared to a 3-month trial which involved an implementation team consisting of a physiotherapist and a rehabilitation engineer supporting a new rehabilitation device being used in clinical practice. Therapists were free to use the Andago for therapy. The patient progressed from non-ambulatory during baseline to walking with single-person assistance by trial end. There was a higher proportion of therapy dedicated to walking when comparing baseline to trial phases, and when comparing Andago and usual therapy within the trial phase. However, therapists reported increased perceived strain and fatigue during Andago-assisted sessions. Therapist engagement with the Andago was moderate, with the device utilized in approximately one out of five sessions per week. The implementation team enabled safe and supported uptake of the novel device which was valued by the case study participant and their therapist. Findings suggest the Andago's potential to enhance gait training but highlight the need to address usability challenges and staff burden to optimize integration and adoption in clinical practice. We believe this can be addressed by better integration of clinicallyoriented engineers in the delivery of rehabilitation services.

The thumb has been called a "hallmark of humanity", and yet, in stroke rehabilitation, there are few studies that have examined thumb function and even fewer that have focused specifically on thumb sensory function. Here we describe a novel robotic technique to assess thumb proprioception and apply it to quantify the relationship between thumb proprioception and hand function after stroke. Based on a video game, "SomatoCircleJump" challenges participants to "jump" to radial targets presented on a screen as their thumb rotates around a circle, driven by a robot. Proprioceptive ability is quantified by radial jump error. We evaluated 35 individuals in the chronic phase post-stroke as part of a randomized controlled trial of robotic finger training. Compared to an agematched control group, people with stroke had significantly increased jump error ($mathrm{p}

Assistive lower-body exoskeletons aim to improve quality of life for broad populations including older adults and people in physically exhausting manual jobs. By applying torque to augment human motion with backdrivable actuators, these devices can reduce human joint effort without restricting volitional motion. However, these backdrivable actuators are coupled by mechanical interfaces to soft tissues of the human body that together introduce resonator dynamics that can delay or diminish the torque assistance. Low interface stiffness and uncompensated dynamics can cause inefficient power delivery to the user, alter their perceived assistance and comfort, and destabilize feedback controllers. We hypothesize that the low stiffness in lateral strap interfaces, like those in the opensource M-BLUE exoskeleton, can be improved by mechanical redesign. Building on the open-source M-BLUE exoskeleton, this paper introduces an alternative interface design that loads the leg through anterior and posterior pads (normal loading) and straps, in which the pads provide extension assistance and the straps provide flexion assistance. We compare the interface dynamics of lateral and normal loading interfaces on N = 10 human subjects using both quasi-static spring measurements and frequency response methods, finding the new design to be 85.7% stiffer $(p<0.01)$ for a range of leg poses and in both flexion and extension loading.

While prototype prostheses and control algorithms have demonstrated compelling clinical benefits in research laboratories, studies with commercially-available robotic prostheses have often failed to demonstrate similar benefits for users, limiting their adoption into mainstream clinical practice. This work is a step towards addressing this shortcoming by presenting the implementation of a phase-based variable impedance controller on the commercially-available Össur Power Knee ${ }^{text {TM }}$ for walking and sit/stand tasks. We show that, through preliminary experiments with $mathrm{N}=4$ high-mobility above-knee prosthesis users, the Power Knee under our controller can produce clear clinical benefits compared to the users' prescribed prostheses. In sitting and standing, users demonstrated generally increased leg-loading symmetry and speed with the Power Knee, indicating easier motions with less over-use of the sound limb. In walking, users demonstrated improved gait with the Power Knee, including increases in toe clearance and early-stance knee flexion. These positive results are similar to our previous work on prototype hardware, demonstrating our controller's hardware generalization and its potential for generating clinical benefits with commercial prostheses. These results are a step towards a promising future in which commercially-available robotic prostheses provide users with concrete clinical benefits.