Journal of NeuroEngineering and Rehabilitation

BackgroundChronic stroke-related gait impairment remains a major source of disability. InTandem is an autonomous neurorehabilitation system delivering individualized, progressive rhythmic auditory stimulation for home-based gait rehabilitation. ObjectivesTo evaluate: (1) engagement during a 12-week autonomous, home-based intervention, (2) changes in walking endurance and functional mobility, and (3) outcome differences across pre-defined engagement and baseline speed subgroups. MethodsThis pragmatic, decentralized trial enrolled adults [&ge;]6 months post-stroke with residual gait deficits. Participants were asked to complete 30-minute sessions 3x/week for up to 12 weeks. Engagement was primarily assessed as the proportion achieving moderate-to-high weekly usage (> 4 weeks; benchmark p1 = 0.60). Changes in 6-Minute Walk Test (6MWT) distances and Timed Up and Go (TUG) times were analyzed using linear mixed-effects models. ResultsOf the 204 who initiated the intervention, 81.9% (95% CI [0.76-0.87]) engaged at least 4 weeks, meeting the primary endpoint (p < 0.001). Overall, 58.1% achieved high engagement (> 9 weeks), 23.9% moderate engagement (4-8 weeks), and 18.1% low engagement ([&le;]3 weeks). Significant improvements in 6MWT distance (+ 26.1 {+/-}5.6 m; 95% CI [14.99, 37.22]) and TUG times (-1.45{+/-}0.31 s; 95% CI [-2.06, -0.84]) (p < 0.001) were observed. Engagement influenced effectiveness: each additional week engaged predicted a 5.82 m greater gain in the 6MWT (SE = 2.05; 95% CI [1.77, 9.87], p < 0.005). ConclusionsAutonomous home-based delivery of music-based rhythmic auditory stimulation achieved moderate-to-high engagement and improved walking endurance and functional mobility, supporting InTandem as a scalable approach to chronic stroke gait rehabilitation. Trial registrationTrial registration: Clinicaltrials.gov NCT06051539. Registered on 20 September 2023. https://clinicaltrials.gov/study/NCT06051539

A passive device that attaches to the feet, called an exotendon, can reduce the energetic cost of running at moderate speeds, but its efficacy and optimal design parameters at higher speeds are unknown. Identifying optimal parameters at new speeds experimentally would require many experimental trials with different exotendon designs, which is challenging for participants at higher running speeds. We developed a muscle-driven simulation framework to predict the effect of various exotendon designs on the energetic cost of running at an experimentally untested speed (4 m/s). We used these predictions to select four designs, which we evaluated experimentally as users ran at this speed. The framework correctly predicted that an exotendon that reduced energetic cost at 2.7 m/s would also reduce energetic cost at 4 m/s (10% predicted vs. 5.7% measured) and that a short, stiff exotendon and a long, compliant exotendon would not significantly reduce energetic cost. However, exotendon parameters predicted by the simulation to maximize energetic savings did not significantly reduce energetic cost when evaluated experimentally. There was variability between participants in both the magnitude of maximum energy savings and the exotendon condition associated with those savings. In a 5-km time trial performed with and without the exotendon condition that elicited the largest energy savings for each participant during the experiment, we observed a lower average heart rate (-3.9 {+/-} 3.8 beats/min; P=0.03; mean {+/-} standard deviation) and increased cadence (15.9 {+/-} 9.6 steps/min; P=0.002) when participants ran with the exotendon but did not observe a statistically significant difference in finishing time (-13.5 {+/-} 24.6 sec; P=0.3). These results demonstrate exotendons can reduce energetic cost across multiple running speeds and that predictive simulations provide a framework for guiding experiments to evaluate assistive device designs. Author summaryDesigning assistive devices that help people move more efficiently usually requires many experimental trials. These studies can be time-consuming and physically demanding, especially when testing multiple device designs. In this study, we explored whether computer simulations could help guide the design of an assistive device for running called an exotendon. The exotendon is a simple elastic band that connects the feet and can help runners use less energy. Previous experiments showed that the device reduces the energy needed to run at moderate speeds, but it was unclear whether it would also work at faster speeds or which design would lead to energetic savings. We first used simulations of human running to test many possible exotendon designs at a faster speed. These simulations allowed us to identify promising designs before conducting experiments. We then tested a small number of these designs with runners. The experiments confirmed that the exotendon can reduce the energy required to run at faster speeds, although the efficacy of different designs varied between individuals. Our results show that computer simulations can help researchers rapidly evaluate a variety of assistive device ideas and focus experimental testing on the most promising designs.

PurposeQuantification of walking function, including joint motions, ground reactions, and joint loads, outside the lab is a growing research area. Because only joint motions can currently be measured outside the lab, researchers are utilizing tracking optimizations of walking to estimate associated ground reactions and inverse dynamic joint loads. However, foot-ground contact models used in such optimizations have been generic rather than personalized, which may limit the accuracy of estimated ground reactions and joint loads. This study compares the predictive capabilities of generic versus personalized foot-ground contact models. MethodsGeneric and personalized foot-ground contact models were evaluated in calibration and tracking optimizations performed using experimental walking data collected from three subjects in varying states of health. Foot-only calibration optimizations evaluated how well both models could reproduce experimental ground reaction and foot motion data while tracking both types of data simultaneously, while whole-body tracking optimizations evaluated how well both models could reproduce experimental ground reactions, joint motion, and joint load data while tracking only experimental joint motion data and achieving dynamic consistency. ResultsFor all three subjects and both types of optimizations, personalized foot-ground contact models reproduced experimental ground reaction, joint motion, and joint load data more accurately than generic foot-ground contact models. ConclusionPersonalized foot-ground contact models can improve the accuracy with which ground reactions and joint loads can be estimated via tracking optimizations of walking using only experimental motion data as inputs. Personalized models require little time and effort to calibrate using freely available software tools and should improve the accuracy of predictive simulations of walking as well.

BackgroundFunctional recovery after chronic stroke remains limited, requiring intensive and engaging rehabilitation approaches. Non-immersive virtual reality (NIVR) provides task-oriented, feedback-driven training that may enhance motor recovery in this population. ObjectiveTo evaluate the clinical effectiveness of a NIVR-based intervention (MindMotion GO) on upper limb motor function in patients with chronic left middle cerebral artery ischaemic stroke (LMCA stroke). MethodsA single-blind randomized controlled trial was conducted in 26 patients with chronic middle cerebral artery stroke. Five participants were lost to follow-up, resulting in a final sample of 21 patients allocated to the non-immersive virtual reality group (NIVR, n = 9) and conventional occupational therapy group (n = 12). Both groups completed an 8-week intervention consisting of two 30-40-minute sessions per week. The primary outcome was upper limb motor function assessed using the Fugl-Meyer Assessment-Upper Extremity (FMA-UE). Secondary outcomes included health-related quality of life (SF-12v2), emotional status (Hospital Anxiety and Depression Scale), and caregiver burden (Zarit Burden Interview). Statistical analyses were performed using the intention-to-treat principle with non-parametric tests. ResultsThe NIVR group showed a clinically meaningful improvement in FMA-UE (median {Delta}21), exceeding the minimal clinically important difference (MCID = 7.35), whereas the control group showed smaller gains ({Delta}2.50) that did not reach clinical relevance. Both groups improved significantly over time; however, between-group differences were not statistically significant (P > 0.05). No significant changes were observed in quality of life, mood, or caregiver burden. ConclusionsNIVR using MindMotion GO is a safe and feasible intervention that can induce clinically meaningful improvements in upper limb motor function in chronic stroke patients. These findings support the incorporation of accessible, task-oriented virtual rehabilitation strategies in long-term stroke care.

Background: Occupational therapists can improve stroke survivors hand and arm movement and participation in daily activities through action observation (AO). AO involves watching another persons hand or arm complete a movement or task. While research generally supports the use of AO with stroke survivors, there are limited AO videos are available to occupational therapists which makes applying AO challenging. Objective: The purpose of this work is to develop structured and widely accessible tool to support access to AO for stroke survivors, occupational therapists, and researchers. Methods: To develop an AO video library for stroke rehabilitation, functional and non-functional upper limb task deficits were first identified through clinical observations and clinician interviews to establish a prioritized list of daily activities. In collaboration with media production specialists, healthy adult volunteers were recruited and filmed performing these tasks from both first- and third-person perspectives. The recorded videos were then systematically edited, enhanced with instructional title slides, and distributed via a public YouTube channel for clinical application and a categorized digital repository for research purposes. Results: Initial assessments revealed a complete lack of familiarity, awareness, and utilization of AO resources among local occupational therapists, despite high perceived clinical utility. To address this gap, a final library of 150 tasks was established, resulting in the production of 419 finalized, standardized videos featuring six healthy volunteers. For clinical application, these videos were hosted on a free, public YouTube channel organized into 18 functional playlists, while a parallel set was structured into distinct movement categories for research repository storage. Conclusion: By providing a structured and highly accessible tool, this repository enables clinicians, researchers, and caregivers to readily implement evidence-based action observation interventions in both clinical and home settings.

Measuring neuromotor control after stroke is crucial for identifying the mechanisms underlying asymmetrical walking and guiding rehabilitation. The lower extremity portion of the Fugl-Meyer (FM-LE) and the number of muscle synergies are commonly used measures, but have important limitations. The dynamic motor control index has emerged as a complementary metric, yet its relationship to established clinical measures (i.e., FM-LE), muscle synergy number, and gait biomechanics remains unclear. This study evaluated the ability of the dynamic motor control index to quantify post-stroke neuromotor impairment relative to FM-LE and muscle synergy number and examined its relationship with propulsion asymmetry. Electromyography data from 22 individuals post-stroke and 31 neurotypical controls were analyzed using non-negative matrix factorization. The dynamic motor control index and not the muscle synergy number differentiated paretic, non-paretic, and neurotypical limbs ({chi}2(2) = 27.57, p < .001). It also differed significantly between less and more impaired individuals classified by FM-LE (p = .05) and demonstrated good discriminative performance between these groups (AUC: 0.777, p = .017). The index also moderated the relationship between FM-LE and propulsion asymmetry ({Delta}R2 = 0.223, p = .007). These findings support the dynamic motor control index as a clinically relevant msarker of post-stroke neuromotor impairment and recovery.

BackgroundConventional evaluations of digital health interventions typically assess mean treatment effects, potentially masking heterogeneous impacts across the functional recovery distribution. Patients at the lower and upper tails of recovery trajectories may respond differently to AI-enhanced telerehabilitation, yet standard regression approaches cannot capture these distributional nuances. ObjectiveThis study applied Recentered Influence Function (RIF) quantile regression with Oaxaca-Blinder decomposition to examine how AI-enhanced telerehabilitation differentially affects functional recovery outcomes across the entire distribution, and to decompose observed disparities into explained (composition) and unexplained (structure) components. MethodsWe analyzed data from 486 post-stroke patients across three rehabilitation centres in Singapore (January 2023-December 2025). Patients received either AI-enhanced telerehabilitation (n=241) incorporating natural language processing-based progress monitoring and adaptive exercise prescription, or standard care (n=245). RIF-quantile regressions were estimated at the 10th, 25th, 50th, 75th, and 90th quantiles of the Functional Independence Measure (FIM) score distribution. Oaxaca-Blinder decomposition at each quantile partitioned group differences into composition effects (attributable to differences in observable characteristics) and structure effects (attributable to differential returns to those characteristics). ResultsThe AI-enhanced telerehabilitation group demonstrated significantly greater FIM improvements across all quantiles, with the largest effects at the 10th quantile ({beta} = 12.74, 95% CI: 8.92-16.56, p < 0.001) and 25th quantile ({beta} = 9.83, 95% CI: 6.71-12.95, p < 0.001), diminishing at the 90th quantile ({beta} = 3.21, 95% CI: 0.88-5.54, p = 0.007). RIF decomposition revealed that at the 10th quantile, 68.3% of the treatment-control gap was attributable to structure effects, indicating that AI-enhanced telerehabilitation fundamentally altered recovery mechanisms for lower-performing patients rather than merely leveraging differences in patient characteristics. ConclusionsAI-enhanced telerehabilitation produces its most pronounced benefits among patients at the lower end of the functional recovery distribution, suggesting a potential mechanism for reducing outcome inequality in stroke rehabilitation. RIF-quantile regression decomposition offers a methodologically rigorous framework for understanding distributional treatment effects that are invisible to conventional mean-focused analyses.

BACKGROUNDFreezing of gait (FOG) is a disabling feature of Parkinsons disease (PD). Although physical practice (PP) improves gait, maintaining gains remains challenging. Mental practice (MP), including Dynamic Neuro-Cognitive Imagery (DNI), may enhance gait control, but evidence on remote combined interventions is limited. PURPOSETo investigate whether adding MP grounded in DNI principles to remote physical practice supports greater and more sustained improvements than remote physical practice alone in people with PD and FOG. METHODSA prospective, single-blind, parallel-group randomized controlled trial was conducted. Forty-three participants with idiopathic PD and FOG were randomized to an experimental group (EG, n = 20) or control group (CG, n = 23), stratified by cognitive performance. Both groups received 10 remote sessions over 6 weeks. All performed structured physical practice targeting gait components; the EG additionally performed MP based on DNI, while the CG performed time-matched seated stretching. Assessments were conducted at baseline (BI), post-intervention (AI), and 30-day follow-up (FU). The primary outcome was Rapid Turns Test performance; secondary outcomes included FOG severity, motor aspects of daily living, mobility-related quality of life, and global cognition. RESULTSAll randomized participants were included in intention-to-treat analyses; 38 completed all assessments. Significant group x time interactions were found for Rapid Turns Test duration (p = 0.0019) and FOG time (p = 0.0108). Both groups improved short-term, but only the EG maintained gains at follow-up. Additional interactions favored the EG for mobility-related quality of life (p = 0.001) and global cognition (p = 0.0018). Self-reported FOG improved over time in both groups (p < 0.001) without between-group differences, while motor aspects of daily living showed a time effect only (p = 0.001). CONCLUSIONMP based on DNI principles may enhance retention of gains when combined with remote physical practice, supporting its use as an adjunct in FOG rehabilitation. Trial registrationThis trial is registered at ClinicalTrials.gov with trial registration number NCT06957405 (registered on April 25, 2025). Protocol and statistical analysis planThe full trial protocol and statistical analysis plan are available upon request from the corresponding author. Data sharingThe datasets generated, used and analyzed during the trial are or will be available from the corresponding author upon reasonable request. Funding and conflicts of interestThis article was produced as part of the activities of FAPESP Research, Innovation and Dissemination Center for Neuromathematics (grant #2013/07699-0, Sao Paulo Research Foundation). Co-author PRS received individual support from FAPESP (grant number 2025/14403-7). The authors declare no conflict of interest.

Wearable exoskeletons are a promising tool for augmenting balance and reducing fall risk. Recent work suggests that active ankle exoskeletons need to act faster than the human to improve reactive balance control. However, the magnitude of exoskeleton torque that is best for improving reactive balance remains unknown. Drawing from the optimal torque for minimizing metabolic expenditure, we hypothesized that reactive balance would improve with increased exoskeleton torque. Participants wearing bilateral ankle exoskeletons were instructed to maintain standing balance during 15cm backward support-surface perturbations. Three exoskeleton plantarflexion torque conditions were tested: NO (Off), LOW (15Nm), or HIGH (30Nm). LOW torque improved balance performance compared to NO torque (p<0.001), with a 7{+/-}3% decrease in peak center of mass (CoM) displacement. Although HIGH torque caused a 9{+/-}11% decrease in peak CoM displacement compared to NO torque (p=0.12), it was not significant due to high intersubject variability. Whereas LOW torque decreased peak CoM displacement in all (range: -0.2 to -1.6cm), HIGH torque only decreased it in some (range = 1.2 to -2.6cm). The change in CoM displacement from LOW to HIGH torque was associated with balance ability, quantified by the narrowing beam test (R2=0.29, p=0.06), while this relationship didnt meet conventional statistical significance, likely due to the small sample size, it suggests that higher levels of exoskeleton torque may hinder balance performance in individuals with better balance ability. Taken together, more exoskeleton torque is not always better for balance, highlighting a potential need to personalize exoskeleton torque for balance augmentation.

The opportunity to collect movement data from smartphones for prolonged periods has opened new perspectives in the field of clinical movement analysis. However, when monitoring peoples mobility in free-living conditions, smartphone placement can influence the validity of the extracted digital mobility outcome. This study aimed to develop and validate an automatic smartphone placement recognition classifier and to investigate potential critical factors that can influence performance. The classifier was trained on data from 15 healthy participants using inertial signals collected from smartphones placed at six body placements during free-living walking and externally validated on over 3,000 individuals from external datasets, including blind participants and patients with cardiovascular or Parkinsons disease. A decision-tree ensemble model was developed using feature subsets of increasing dimensionality, with the optimal subset comprising 50 features. Classification accuracy increased consistently when front and back pocket placements were aggregated (81.1%) and further improved when coat pocket was also included in the pocket class (88.5%), underscoring the challenge of distinguishing between fine-grained pocket placements. The best-recognized placements across the external datasets were lower back (precision: 100%, recall: 72.5%), hand (precision: 94.2%, recall: 94.5%), and the aggregated pocket class (precision: 86.7%, recall: 90.2%). Recognition accuracy changed across cohorts (0.73 - 0.85), activities (0.63 - 0.94) and speed (0.79 - 0.87), however it stayed consistent across various technological and environmental factors. Overall, this study demonstrates the feasibility of robust placement recognition in walking and underscores the importance of accounting for key influencing factors when designing frameworks intended for deployment in heterogeneous real-world or clinical contexts. HighlightsO_LIMachine learning accurately identifies smartphone placement during real-world gait C_LIO_LISix on-body placements recognized, including pockets, hand, bag, and lower-back C_LIO_LIFree-living data used for training, ensuring robust performance across conditions C_LIO_LIFeature selection and hyperparameter tuning optimize classification accuracy C_LIO_LIExternal validation confirms generalizability across >3,000 healthy and diseased adults C_LI

Brain-Computer interfaces (BCIs) offer a link between neural signals and external computation, enabling control of devices for the purposes of restoring function to motor-affected individuals and enhancing capabilities of a wider set of populations. Electroencephalography (EEG) offers a high temporal resolution for dynamic and potential real-time feedback for non-invasive systems. However, its practical efficacy remains limited due to low spatial resolution and poor signal-to-noise ratio, leading to insufficient decoding accuracy and unintuitive control paradigms that hinder reliable user interaction. In this study, we present a framework for an online EEG foundation model by creating a custom foundation model through spectrogram reconstruction of compact temporal windows and online constraints during pretraining. We evaluate the performance of the model in a challenging control paradigm of single-arm, directional motor imagery with dynamic movements for guided and free movement cursor control tasks. Our foundation model approach achieved a final average accuracy of 51.3% during a goal-oriented guided control task. This represents a 15.8% increase over a conventional deep learning framework and a 26.3% increase above chance level, evaluated in a cohort of 11 human participants. During the free movement task, the foundation model invoked a higher rate of completion and lower completion times. Furthermore, the custom EEG foundation model demonstrated superior adaptability from same-session finetuning and indicated an enhanced capability to assist subject learning. These findings highlight the potential of EEG foundation models to support more robust and intuitive non-invasive BCI systems, providing a promising modelling framework for future BCI development.

BackgroundModerate- to high-intensity walking training (M-HIT) is an established intervention for improving walking capacity in chronic stroke. Musculoskeletal (MSK) adverse events commonly occur during M-HIT, yet tools to identify individuals at higher risk are limited. Baseline clinical characteristics may provide insight into susceptibility to training-related MSK adverse events during M-HIT. Thus, this study aimed to develop and internally validate a model for predicting MSK adverse events during a 12-week M-HIT program in chronic stroke using baseline clinical characteristics. MethodsParticipants (n=100) from HIT-Stroke Trials 1 and 2 were included. Baseline clinical characteristics included measures of orthopedic history, pre-existing pain, motor function, recent exercise history, demographics and health characteristics, stroke chronicity, and psychological health. Logistic regression models evaluated all possible combinations of baseline characteristics with up to three predictors. Leave-one-out cross-validation was used for internal validation to mitigate overfitting. Predictive performance was quantified using the C-statistic, and the candidate model with the highest cross-validated C-statistic was selected as the final model. ResultsMSK adverse events occurred in 32.0% of participants. The optimal three-variable model included prior orthopedic condition (Odds ratio [OR] 3.02 [95% CI 1.14-8.64]), Fugl-Meyer lower extremity motor score (OR 1.14 [95% CI 1.02-1.28]), and self-reported participation in regular walking exercise (OR 0.17 [95% CI 0.05-0.49]) at baseline. This model demonstrated moderate discrimination (cross-validated C-statistic = 0.74; apparent C-statistic = 0.78). ConclusionsParticipants reporting at least one pre-existing lower extremity or lumbar spine orthopedic condition and those with better lower-extremity motor function exhibited greater odds of experiencing MSK adverse events during M-HIT, while participants reporting participation in regular walking exercise had lower odds. These findings suggest that baseline clinical characteristics may help identify individuals at elevated risk for MSK adverse events during M-HIT who may warrant closer monitoring or risk-reduction strategies. Future studies are needed for external validation. Clinical Trial Registrationhttps://ClinicalTrials.gov; Unique identifiers: NCT03760016, NCT06268041

Purpose: To identify, through iterative user-centered design, the auditory biofeedback requirements and sound preferences supporting gait training in children with cerebral palsy (CP), and to determine which feedback variables, sound mappings, and sound types yield clinically viable and movement-interpretable paradigms. Methods: The iterative process spanned two prototype phases. Prototype A comprised seven paradigms demonstrated to two experienced physiotherapists (Workshop 1A). Two of these were subsequently discarded owing to poor sound-movement interpretability and two were modified. Six paradigms were added to Prototype B, demonstrated to four children, five parents, and one therapist (Workshop 1B) and two therapists (Workshop 2B). Data were analyzed using systematic text condensation. Results: Within-child sound preferences varied with energy level and sensory state on a given day. Sound-movement interpretability tended to suffer for paradigms with greater acoustic complexity (e.g. computer-generated music). Therapists endorsed a repertoire spanning both movement quality and movement quantity targets. Participants independently proposed paradigms rewarding restrained and controlled movement, a feedback category absent from the current prototype. Conclusions: Session-level calibration is preferable to fixed sound profiles, requiring real-time interface support for paradigm adjustment. Acoustic complexity must remain subordinate to movement-sound interpretability. Paradigms targeting movement restraint are a development priority unaddressed in the literature.

Background: Accurate evaluation of fine motor abilities is a key aspect of neurological rehabilitation. However, conventional approaches like goniometry are limited by variations among raters and their difficulty in detecting active movement. On the other hand, computer vision-based software delivers non-invasive and quantitative analysis of hand movements. An innovative computer-vision-based software tool, F.A.I.R. Chance(C), was developed to track and analyze individual finger joint movements on a camera-equipped laptop and give real-time numerical feedback. However, its metrics require validation in a healthy population before the tool can be used for clinical purposes. Objective: To assess the reliability and validity of finger movement assessment by the F.A.I.R. Chance computer vision-based tool in healthy adult participants. Methods: An observational cross-sectional study was done at MGM School of Physiotherapy, comprising 30 healthy participants between 18 and 60 years of age. Finger movements like flexion, extension, abduction, and adduction were measured with a standard handheld goniometer. These same finger movements were then measured with the tool at two time points separated by a 30-minute interval to determine the test-retest reliability. The tool's measurements were compared with the goniometric measurements to determine its concurrent validity. Test retest reliability was checked by the Intra-class Correlation Coefficient ICC (2,1), while concurrent validity was tested through Pearson's correlation coefficients. Results: Metacarpophalangeal and proximal interphalangeal joint motions demonstrated moderate to good test-retest reliability (ICC: 0.716-0.953) for the F.A.I.R. Chance tool. However, distal interphalangeal joint movements had lower consistency. Good reliability (ICC: 0.754-0.908) was seen for movements of abduction and adduction in the fingers. Strong concurrent validity for extension movements of the metacarpophalangeal joints (r=0.760-0.914) and moderate concurrent validity for flexion movements of the metacarpophalangeal joints (r=0.427-0.604) was demonstrated for all fingers for the F.A.I.R. Chance tool. Concurrent validity for adduction and abduction movements demonstrated a low to fair correlation with goniometric measurements (r=0.210-0.440). This is consistent with previous research showing poor agreement between goniometry and adduction-abduction movements of the fingers. Conclusion: The F.A.I.R. Chance tool shows good reliability and acceptable concurrent validity to assess fine motor movements in the healthy adult population. This sets a basis for further clinical study of the tool in the target population with fine motor impairments. Keywords: artificial intelligence; assistive technology; computer vision; fine motor evaluation; hand function;

Background and PurposeWalking speed is the dominant clinical metric used to classify post-stroke hemiparetic gait severity. However, speed does not describe how mechanical energy is generated and redistributed. We tested whether whole-body center-of-mass (COM) work patterns provide a biomechanically grounded supplement to speed-based severity classification. MethodsLimb-specific COM power and work were computed from ground reaction forces using the individual-limbs method across five walking speeds (0.2-0.7 m/s). We quantified net COM work index of asymmetry (IA_Wnet), positive COM work asymmetry (IA_Wpos), and the Propulsion-Support Ratio (PSR = impFy/impFz). Piecewise and quadratic regressions were used to assess speed-dependent trends. ResultsIA_Wnet remained elevated across speeds and showed no significant high-speed association. IA_Wpos demonstrated a significant quadratic relationship with speed (p=0.023, R{superscript 2}=0.23), decreasing near 0.5 m/s before rising again. Paretic limb PSR remained constrained and exhibited a quadratic association (p=0.012, R{superscript 2}=0.14), while unaffected limb PSR declined significantly at higher speeds (p=0.019, R{superscript 2}=0.38). Below 0.5 m/s, COM power profiles collapsed to a two-phase pattern without paretic limb push-off; at [&ge;]0.5 m/s, a four-phase structure emerged. ConclusionIncreasing walking speed did not normalize interlimb mechanical imbalance. COM work organization revealed a biomechanical transition near 0.5 m/s and distinguished compensation from recovery-based restoration. Supplementing speed with COM work and propulsion-support metrics may refine severity stratification and guide mechanism-targeted rehabilitation.

Background and Objectives Patients with peripheral neuropathies (PN) commonly exhibit balance impairment. In clinical practice, balance is typically assessed using the Rombergs test and ataxia scales, which rely on examiner interpretation, while objective biomarkers for quantifying balance remain lacking. Wearable sensors are valuable tools for objectively quantifying gait abnormalities in PN patients and may capture clinically meaningful changes over time. By integrating these parameters, artificial intelligence (AI) can assist in generating a digital score that enables easy, objective, and reproducible monitoring of patients postural balance. This study aims to generate and assess an AI-generated digital Rombergs test to quantify balance impairments in a cohort of PN patients. Methods PN patients were assessed in a longitudinal study using a wearable system composed of inertial sensors placed on the trunk and plantar pressure sensors integrated in insoles. Patients performed the Rombergs test under both eyes-open and eyes-closed conditions and were classified according to ataxia severity (mild, moderate, or severe) following the score obtained in item 1 of MICARS and SARA scales. Results We included 97 patients with PN (including autoimmune and hereditary polyneuropathies), and 117 healthy controls (HC). Significant differences in trunk sway and center of pressure (COP) were observed between groups, particularly with eyes closed. Using wearable sensor parameters, we developed an AI digital Rombergs test, which correlated with clinician-rated Rombergs test performance and distinguished patients with and without ataxia (AUC=0.632) and across different PN pathologies. Longitudinally, digital Rombergs test and iRODS showed concordant trajectories. Also, changes [&ge;]25% in the score were associated with clinical changes in ataxia severity measured by an increase in MICARS-SARA score (+1.42 points), whereas improvement was associated with a decrease (-0.20 points) in the scale. Discussion This study demonstrates that wearable sensors are useful to detect and quantify balance impairment. The AI-generated Rombergs test is an objective and reproducible tool for postural balance assessment, with robust discriminatory performance across clinical ataxia severity in PN. Scores longitudinal changes aligned with clinical severity, supporting its potential for monitoring disease progression and treatment response. Its strong association with balance measures reinforces its role as a quantitative biomarker of postural control in ataxia patients.

IntroductionDespite the proven benefits of reperfusion therapies in acute ischemic stroke, treatment decisions in the hyperacute phase remain complex and are rarely supported by individualized outcome predictions. Artificial intelligence (AI)-based clinical decision support systems (CDSS) offer potential real-time prognostic estimates, but prospective evidence of their feasibility and performance in routine clinical workflows is limited. Our aim is to prospectively evaluate real-time feasibility, usability, and predictive performance of an AI-based CDSS (VALIDATE-CDSS) for individualized outcome prediction in acute stroke care. Methods and analysisProspective, multicenter, observational study enrolling consecutive patients with acute ischemic stroke presenting to three tertiary stroke centers. Clinical management will follow standard practice at the discretion of treating physicians. In parallel, a dedicated researcher will collect patient data in real time and input them into the VALIDATE-CDSS using a mobile application, operating in shadow mode without influencing clinical decisions. The system will generate individualized predictions of 3-month functional outcome (modified Rankin Scale) for four treatment strategies (intravenous thrombolysis, endovascular thrombectomy, combined therapy, or no reperfusion) at three sequential time points: baseline clinical data, non-contrast CT, and CT angiography. The primary outcome is the real-world feasibility and usability of the VALIDATE-CDSS in the hyperacute stroke workflow. Secondary outcomes include predictive performance, agreement between model-suggested and actual treatments, incremental value with increasing data availability, and assessment of potential bias across predefined subgroups. This study will provide prospective real-world evidence on the implementation and clinical potential of AI-based decision support for personalized treatment selection in acute ischemic stroke Ethics and disseminationPatient enrollment began after approval from the ethics committees of all participating centers. Results will be disseminated through peer-reviewed open-access journals and conference presentations. Following open science principles, anonymized data and metadata will be made publicly available in the Zenodo repository upon study completion. Trial registrationClinicalTrials.gov (NCT05622539). Strengths and limitations of this study- First study to assess the feasibility of integrating an outcome-predictor CDSS into real-life hyperacute stroke workflows, addressing a critical gap between AI model development and clinical implementation - Multicenter, prospective observational time-motion shadow-mode design, which minimizes interference with standard care while capturing real-world operational data - Validation of a locked AI model developed from independent retrospective multicenter datasets across different populations, reducing the risk of overfitting to local case-mix - Real-time data acquisition in the emergency department poses a significant operational challenge, with potential for missing or delayed inputs that may affect model performance in practice - Risk of bias cannot be excluded, including spectrum bias from non-anticipated subgroups, temporal drift in clinical practice or patient populations, and centre-level variation in workflow and data quality

Wearable digital health technologies may complement traditional gait assessments in amyotrophic lateral sclerosis (ALS) by sensitively capturing real-world mobility changes. In this study, we validated six digital gait metrics derived from ankle-worn sensors in a natural history cohort of 182 individuals with ALS. Investigated metrics correspond to various aspects of gait, including volume, speed, intensity, similarity, variability, and fragmentation. Longitudinal analyses showed significant declines in step count, peak cadence, stride intensity, and stride similarity, with increasing stride duration variability and walking fragmentation over 52 weeks. Many participants exhibited greater relative change in the gait metrics than the self-reported ALS Functional Rating Scale-Revised (ALSFRS-RSE). Stratified analyses revealed that digital metrics captured significant functional decline even in participants with stable walking scores on the ALSFRS-RSE. These findings support the potential utility of these metrics for disease monitoring in ALS clinical care and trials.

BackgroundLower-limb EEG is a rehabilitation-facing control signal for stroke neurorehabilitation and future non-invasive brain-spine interfaces, but a public external benchmark that jointly audits source construction, minimal adaptation burden, and confound sensitivity is lacking. We therefore tested whether lower-limb effort-versus-rest decoders trained on healthy public EEG transport to a stroke target domain. MethodsWe conducted a retrospective public-data external benchmark using three public EEG datasets harmonised to a common lower-limb effort-versus-rest target. Classical and deep models were compared under zero-shot transport, 10-shot temperature calibration, and 10-shot fine-tuning. For few-shot analyses, each target participant contributed a trial-disjoint subject-internal support set of 10 labelled trials per class and a held-out remainder test set. Prespecified analyses audited source construction, support-resampling sensitivity, and montage controls. Uncertainty was summarised with participant-level bootstrap confidence intervals. ResultsWithin this benchmark, healthy-to-stroke zero-shot transport was weak. The best zero-shot result was classical rather than deep, with CSP+LDA reaching area under the receiver operating characteristic curve (AUROC) 0.603, whereas EEGNet remained near chance (AUROC 0.527). Ten-shot calibration improved operating behaviour more than discrimination: for CSP+LDA, expected calibration error fell from 0.267 to 0.035 and specificity increased from 0.180 to 0.485, whereas AUROC remained essentially unchanged (0.603 to 0.604). Ten-shot fine-tuning produced only modest gains; the best overall AUROC was 0.605 for pooled dataset-balanced CSP+LDA, numerically tied with pooled raw CSP+LDA (0.605). MILimbEEG-only source training was consistently weak, exploratory deep domain-generalisation variants did not rescue transport, and frontal and temporal montage controls remained relatively competitive. ConclusionsWithin this public benchmark, source construction and minimal adaptation burden mattered more than model novelty, and retrospective montage controls limited motor-specific interpretation. The results support harmonised prospective validation of lower-limb EEG transport over further retrospective model iteration.

BackgroundWalk n Watch (WnW) is a structured, progressive walking exercise intervention developed for Canadian inpatient stroke rehabilitation. Although its mechanisms align with UK guidance for intensive walking therapy, stroke rehabilitation in the UK is delivered predominantly in the community. This change in service context has implications for safety, feasibility, and fidelity, necessitating structured pre-implementation intervention adaptation to support delivery. MethodsA prospective adaptation process used ADAPT guidance. A multidisciplinary coalition and learning collaborative (UK clinicians, clinical- academics, people with lived experience, and Canadian WnW developers) participated in stakeholder co-production activities. Informed by ADAPT steps 1-2, co-production focused on rationale, core components, contextual mapping and planning adaptations. Discussions were analysed through rapid deductive mapping using Consolidated Framework for Implementation Research (CFIR) domains. Candidate fidelity-consistent adaptations were refined by the learning collaborative. Conceptual outputs of the process were synthesised. ResultsThree intervention core components were confirmed: 1) prioritised, high-volume, weight-bearing walking-related activities at moderate effort; 2) structured progression of steps based on performance on a walking test (e.g. Six-Minute Walk Test); 3) objective monitoring of steps and cardiovascular intensity. Several contextual determinants across CFIR domains were likely to influence UK community implementation. Fidelity-consistent modifications to the adaptable periphery were specified across four areas: 1) therapy & practice, 2) environment & safety, 3) monitoring & feedback, and 4) workflow & documentation. Adaptations included hybrid supervision, planned out-of-session practice, and monitoring using validated proxies. A WnW Adaptation Model was produced. ConclusionsThis paper provides a transparent pre-implementation adaptation of WnW for delivery within UK community stroke rehabilitation. Anchoring adaptations to intervention mechanisms and principles through co-production and implementation science frameworks, this work establishes a foundation for piloting and hybrid effectiveness-implementation evaluation. The WnW Adaptation Model offers support for future implementation efforts. Discussion positions adaptation as a pragmatic means for applying optimisation principles. PLAIN LANGUAGE TITLEAdapting the Walk n Watch walking exercise programme for home-based stroke rehabilitation in the UK: A structured step-by-step process PLAIN LANGUAGE SUMMARYO_ST_ABSBackgroundC_ST_ABSWalk n Watch (WnW) is a structured exercise programme that helps people improve their walking. It was originally developed for people recovering from stroke in hospital in Canada. While the approach fits well with United Kingdom (UK) recommendations for intensive therapy, stroke rehabilitation in the UK often takes place at home. Because of this difference, WnW needs careful adaptation for safe and effective delivery. MethodsPublished ADAPT guidance was used to adapt WnW. UK therapists, researchers, people with stroke, and Canadian WnW developers undertook adaptation activities. Together, they identified which parts of WnW were essential, explored differences between the Canadian and UK settings, and planned changes. Discussions were reviewed using an established framework to develop adaptations that kept the most important parts of WnW intact (fidelity-consistent adaptations). The adaptation process was summarised. ResultsThree essential intervention parts were confirmed: 1) prioritised, high-volume, weight-bearing walking-related activities at moderate effort; 2) structured progression of steps based on performance on a walking test; 3) objective monitoring of steps and cardiovascular intensity. Several factors were likely to influence delivery in the UK community. Changes focused on four areas: 1) therapy & practice, 2) environment & safety, 3) monitoring & feedback, and 4) workflow & documentation. They included using both in-person and online sessions, planning safe between session practice, and using non-digital monitoring. A WnW Adaptation Model was produced. ConclusionsThis paper clearly describes the steps taken to adapt WnW for delivery in UK community stroke rehabilitation. By working closely with stroke experts and using established research frameworks, the adapted programme keeps the most important parts of WnW while allowing it to fit into real-life. The WnW Adaptation Model offers support for further testing and may assist others looking to adapt WnW. Discussion offers perspective on how adaptation aligns with optimising interventions.