Search Results (204)

Background/Objectives: Wearable technology is increasingly used to provide biofeedback in physical rehabilitation; however, there is no consensus on which biofeedback parameter is most appropriate for clinical use, as most studies evaluate only one arbitrarily selected parameter. This study presents a wearable multimodal biofeedback system integrating multiple parameters selected based on the prior literature and evaluates its feasibility, usability, and implementation within a rehabilitation context through a longitudinal post-stroke case study. Methods: The system integrates inertial and electromyographic sensors to monitor centre of mass (CoM-B), joint angle (ANG-B), and muscle activity (EMG-B), delivering real-time sensory cues based on the monitored parameters. Feasibility was assessed in a post-stroke participant (male, 32 years, 29 months post-stroke, left hemiparesis, Fugl-Meyer Lower Extremity Score = 27) across 15 sessions involving stand-to-sit, split-stance weight shifting, and walking tasks. Each task was practiced with all three biofeedback parameters, with five sessions per parameter. Results: The motor performance varied across biofeedback parameters and tasks. CoM-B was associated with favourable trends in motor performance during stand-to-sit, showing improvements in medio-lateral displacement (0.03/session); ANG-B during walking, showing increased ankle dorsiflexion (1 deg/session); and EMG-B during split-stance weight shifting, showing increased tibialis anterior activation (5 µV/session). Conclusions: The findings generate the hypothesis that the ability of biofeedback to elicit favourable motor performance is task-dependent, suggesting that the choice of biofeedback parameters may need to be adapted to task demands. The system demonstrated high usability and feasibility, supporting its potential for post-stroke rehabilitation. Further studies are needed to test the generated hypothesis and evaluate the system efficacy. Full article

Objectives: Bilateral robotic arm training (BRT) may enhance poststroke motor recovery by reducing interhemispheric inhibition and promoting bilateral motor network engagement. However, previous reviews have often pooled bilateral and unilateral robotic approaches, potentially masking differential effects. This systematic review and meta-analysis compared the effects of BRT with those of unilateral robotic training (URT) and conventional rehabilitation on upper-limb motor function after stroke. Methods: Randomized controlled trials were identified through systematic searches of major electronic databases and trial registries in accordance with PRISMA guidelines. The risk of bias was assessed via the Cochrane Risk of Bias 2 tool. Random effects meta-analyses were performed using standardized mean differences (SMDs). Predefined subgroup and sensitivity analyses were used to examine the influence of participant characteristics, training dose, intervention duration, and robotic device type. Results: Fourteen randomized controlled trials involving 440 participants were included. Overall, compared with control interventions, BRT did not significantly improve upper-limb motor function, as measured using the Fugl–Meyer Assessment for Upper Extremity (SMD = 0.18, 95% CI −0.01–0.36). Significant effects were observed in participants younger than 60 years, with training doses > 15 h, intervention durations > 4 weeks, and use of Bi-Manu-Track systems. Conclusions: BRT did not demonstrate a significant overall advantage over URT or conventional rehabilitation. However, subgroup analyses suggest that treatment effects may vary according to patient characteristics, training dose, duration of the intervention, and device type. Full article

Objective: To systematically evaluate the rehabilitation effect of brain-computer interface (BCI)-controlled hand robot training on post-stroke motor functions, especially upper limb functions. Methods: PubMed, Embase, Web of Science, Cochrane Library, CNKI, SinoMed, WanFang Data, and VIP Database were searched from inception to 13 March 2026. Randomized controlled trials (RCTs) with dose-matched designs were included, where the test group underwent BCI-controlled hand robot training and the control group received either pure hand robot training or routine rehabilitation. Meta-analysis was performed on RevMan 5.4. Results: Totally 11 RCTs involving 380 patients were included. Compared with hand robot training alone, BCI-controlled hand robot training significantly improved Fugl-Meyer Assessment for Upper Extremity (FMA-UE) scores (MD = 4.87, 95% CI: 1.04 to 8.69) and FMA-UE proximal scores (MD = 4.44, 95% CI: 0.15 to 8.74), and significantly reduced finger flexor spasticity (MD = −0.44, 95% CI: −0.68 to −0.21), but showed no significant difference in distal upper limb motor function or Action Research Arm Test (ARAT) scores. Compared with routine rehabilitation, BCI-controlled hand robot training significantly improved FMA-UE scores (MD = 6.55, 95% CI: 3.49 to 9.61). Conclusions: BCI-controlled hand robot training can effectively improve overall upper limb and proximal motor function after stroke and alleviate finger flexor spasticity, but the evidence for distal hand function and long-term efficacy remains limited. Full article

Background/Objectives: Recovery of upper limb function after stroke is highly heterogeneous, and accurate prediction of clinically meaningful functional transition remains a major challenge in rehabilitation medicine. We developed and temporally validated machine learning (ML)-based prognostic models for predicting transition from non-functional movement to functionally usable upper limb capacity in patients undergoing intensive inpatient rehabilitation during the early subacute phase of stroke. Methods: This retrospective cohort study included 960 patients with ischemic or hemorrhagic stroke admitted to a tertiary rehabilitation center between 2010 and 2025. Three functional recovery outcomes were defined: motor impairment recovery, defined as Fugl-Meyer Assessment for Upper Extremity score ≥ 32; gross manual dexterity recovery, defined as Box and Block Test score ≥ 2 blocks/min; and functional pinch strength recovery, defined as pinch strength ≥ 1.1 kgf. Multidimensional predictors spanning demographic, clinical, neurophysiological, neuroimaging, and rehabilitation-related domains were integrated. Four ML algorithms were evaluated using stratified 5-fold cross-validation and temporal validation in a chronologically independent cohort (2024–2025). Models were developed under two tracks: Track A, incorporating only baseline variables available at admission (primary prognostic model), and Track B, additionally incorporating cumulative rehabilitation-related variables (exploratory). Results: Random Forest demonstrated the best overall performance. During temporal validation, models achieved AUROC of 0.800 for motor impairment recovery, 0.958 for gross manual dexterity recovery, and 0.888 for functional strength recovery. Baseline motor severity and corticospinal tract integrity were the dominant biological determinants of recovery. Earlier rehabilitation initiation and greater upper-limb robot-assisted therapy exposure were also associated with improved outcomes; however, these findings should be interpreted as observational associations subject to treatment-selection bias rather than evidence of causal effects. Conclusions: Probabilistic ML prediction integrating neural reserve and rehabilitation-related exposure variables can support individualized precision rehabilitation planning and improve functional outcome stratification in early subacute stroke. Full article

Background/Objectives: Stroke remains a leading cause of global disability, with upper limb impairment affecting over 80% of patients. During the acute phase (first seven days), a critical neuroplastic window exists where interventions may significantly influence recovery. This systematic review and meta-analysis aimed to evaluate the effectiveness and safety of electrical stimulation—specifically Functional Electrical Stimulation (FES) and Neuromuscular Electrical Stimulation (NMES)—on upper limb functional recovery and complication prevention during the acute phase of stroke. Methods: A systematic search was conducted across eight databases (including Medline, PEDRo, and Cochrane) for randomized and non-randomized clinical trials published between 2016 and 2025. Methodological quality was assessed using the PEDRo scale. Quantitative synthesis was performed via meta-analysis using a random-effects model, focusing on the Fugl-Meyer Assessment (FMA-UE). Results: Eight randomized clinical trials were selected with a total of 384 participants. The meta-analysis results showed a positive and statistically significant effect in favor of the experimental group compared to the control group (Z = 2.39; p = 0.02), with a combined Standardized Mean Difference of 0.53 (95% CI: 0.10 to 0.96), indicating a moderate effect size on the Fugl-Meyer Assessment Upper Extremity scale. Although high heterogeneity was detected (I² = 74%), the analysis suggests that Functional Electrical Stimulation (FES) and Neuromuscular Electrical Stimulation (NMES) improve manual dexterity, prevent disuse atrophy, and reduce glenohumeral subluxation. Conclusions: Electrical stimulation shows a positive trend in early stroke recovery; however, it should be considered a promising adjunct rather than a definitive treatment. Further research into standardized protocols is required to confirm their clinical significance. Full article

Background/Objectives: Upper-limb motor dysfunction is common after stroke, and patients often have limited recovery during rehabilitation. In this study, we aimed to investigate the relationship between contralesional neurophysiological parameters and the effects of rehabilitation on upper-limb motor function in stroke patients with corticospinal tract damage. Methods: Forty patients with subacute stroke with an absent MEP response on the ipsilesional side before admission were included. Contralesional neurophysiological parameters, including resting motor threshold, contralesional MEP, contralesional short-interval intracortical inhibition (short-ICI), and contralesional long-interval intracortical inhibition (long-ICI), were assessed via transcranial magnetic stimulation (TMS) pre-admission. The coefficients of variation for MEP, short-ICI, and long-ICI were calculated to assess cortical stability. Rehabilitation effect was measured using changes in the Fugl–Meyer assessment score after 21 days of rehabilitation. Results: No single contralesional parameter significantly predicted rehabilitation effect. Further exploratory analysis revealed that a model combining contralesional neurophysiological parameters was associated with proximal limb motor function recovery. Short-ICI played a prominent role in this exploratory model. Conclusions: Contralesional neurophysiological markers demonstrated limited predictive value in patients with stroke with moderate-to-severe motor dysfunction and damaged corticospinal tract function on the ipsilesional side. However, a model combining multimodal contralesional TMS measures, particularly short-ICI, may offer incremental value in predicting proximal limb motor improvement following 21-day rehabilitation. Although this mechanism was not directly measured, the findings suggest a compensatory role of the cortico-reticulo-spinal pathway. These exploratory results should be interpreted with caution regarding their clinical applicability and are premature as a predictive tool, pending rigorous external validation. Full article

Gait dysfunction is a common and disabling consequence of stroke, frequently associated with impaired lower-limb torque generation and reduced balance. Robot-assisted gait training (RAGT) has emerged as a promising intervention; however, its phase-specific biomechanical effects remain incompletely characterized. This pilot mechanistic study investigated the effects of Walkbot-assisted gait training on phase-specific lower-limb torque and clinical outcomes in individuals with unilateral hemiplegia. Fifteen patients with hemiplegia underwent Walkbot-assisted gait training. Joint torque values were normalized to body mass (Nm/kg). Phase-specific torque was analyzed during the swing and stance phases for the affected and unaffected limbs. Pre–post differences were evaluated using the Wilcoxon signed-rank test. Functional balance and motor impairment were assessed using the Berg Balance Scale (BBS) and the Fugl–Meyer Assessment—Lower Extremity (FMA-LE). Significant torque increases were observed in both gait phases. Median swing-phase torque increased from 0.261 to 0.361 Nm/kg in the affected limb and from 0.254 to 0.334 Nm/kg in the unaffected limb (p ≤ 0.017). Stance-phase torque increased from 0.197 to 0.454 Nm/kg in the affected limb and from 0.158 to 0.471 Nm/kg in the unaffected limb. Clinical outcomes improved significantly, with median BBS scores increasing from 22.0 to 34.0 and FMA-LE scores from 14.0 to 24.0 (p = 0.001). Walkbot-assisted gait training was associated with significant phase-specific torque gains, accompanied by improvements in balance and lower-limb motor recovery. Full article

Background/Objectives: Robot-assisted therapy (RAT) can deliver repetitive, feedback-enriched upper-limb practice after stroke, but evidence comparing RAT with dose-matched conventional occupational therapy (COT) under routine inpatient conditions—and concurrent neurocognitive data—remains limited. We compared motor recovery between an end-effector RAT-based program (30 min RAT plus 30 min COT) and dose-matched COT alone in subacute stroke survivors, with neurocognitive outcomes prespecified as exploratory endpoints. Methods: In this single-center retrospective non-randomized cohort study, adults with first-ever ischemic or hemorrhagic stroke who completed routine baseline and week−4 assessments received 4 weeks of upper-limb rehabilitation: combined RAT plus COT (60 min daily) or COT alone (60 min daily). The primary outcome was the week-4 Fugl–Meyer Assessment–Upper Extremity (FMA-UE) motor score adjusted for baseline. Primary inference used covariate-adjusted linear regression on outcome-specific complete cases, with a prespecified stabilized inverse probability of treatment weighting (IPTW) average treatment effect analysis as the sensitivity test. Secondary and exploratory endpoints were interpreted descriptively; Benjamini–Hochberg false discovery rate (FDR) control and multiple imputation were applied as supportive analyses. Results: The analytic cohort comprised 65 patients (RAT, n = 33; COT alone, n = 32). Both groups improved over 4 weeks, but the RAT group had worse baseline upper-limb motor status. The adjusted between-group difference for the week-4 FMA-UE motor score was non-significant (adjusted mean difference, 4.39; 95% confidence interval (CI), −2.43 to 11.21; p = 0.203), and the stabilized IPTW estimate was concordant (β = 2.17; 95% CI, −3.63 to 7.98; p = 0.464). In unadjusted analyses, the FMA-UE motor gain was larger after RAT than COT alone (14.70 ± 15.53 vs. 7.91 ± 9.42), and only the RAT group exceeded the prespecified 12.4-point clinically important threshold; this signal attenuated after adjustment. No secondary or exploratory endpoint remained significant after FDR control. Multiple imputation for the primary endpoint was concordant with the complete-case result (pooled β = 4.52; 95% CI, −1.91 to 10.94; p = 0.168). Conclusions: End-effector RAT did not demonstrate adjusted superiority over dose-matched COT alone for upper-limb motor recovery. The larger unadjusted FMA-UE gain should be interpreted as a descriptive impairment-level signal rather than as evidence of comparative efficacy. Neurocognitive results were exploratory; the retrospective non-randomized design, baseline imbalance, differential missingness, and unavailable confounder data require cautious interpretation. Full article

Video-based action recognition for neural rehabilitation—spanning stroke recovery, Parkinsonian gait assessment, and cerebral palsy monitoring—faces critical challenges, including temporal ambiguity, non-causal motion correlations, and the absence of causally grounded dynamics modeling. While transformer-based architectures achieve strong performance, they often exploit spurious temporal and environmental cues, limiting reliability in safety-critical clinical settings. We propose NeuroPrisma, a neuro-prismatic video framework that integrates frequency-domain spectral decomposition with causal intervention under Structural Causal Models (SCMs) via the backdoor criterion. NeuroPrisma introduces (i) a Prismatic Spectral Attention (PSA) module, which applies discrete Fourier transforms to decompose temporal features into multi-scale frequency bands, disentangling slow postural dynamics from rapid corrective movements, and (ii) a Causal Intervention Layer (CIL), which performs do-calculus-based backdoor adjustment to remove confounding influences and produce causally invariant representations. PSA preconditions representations prior to intervention, improving confounder estimation and causal robustness. Extensive evaluation against seven state-of-the-art models (I3D, SlowFast, TimeSformer, ViViT, Video Swin Transformer, UniFormerV2, and VideoMAE) demonstrates that NeuroPrisma achieves 98.7% Top-1 accuracy on UCF101, 82.4% on HMDB51, 71.2% on Something-Something V2, and 91.5%/95.8% on NTU RGB+D (Cross-Subject/Cross-View), consistently outperforming prior methods. It further reduces the Causal Confusion Score (CCS) by 42.3%, indicating substantially lower reliance on spurious correlations, while maintaining real-time performance with 23.4 ms latency per 16-frame clip on an NVIDIA A100 GPU. All improvements are statistically significant (p < 0.001, Cohen’s d = 0.72–1.24). Evaluation was conducted exclusively on benchmark datasets (UCF101, HMDB51, Something-Something V2, and NTU RGB+D) under controlled conditions, without direct clinical validation on neurological patient cohorts. Overfitting was mitigated using three random seeds (42, 123, 456), RandAugment, Mixup (α = 0.8), weight decay (0.05), and early stopping. Cross-dataset generalization from UCF101 to HMDB51 without fine-tuning achieved 76.2% Top-1 accuracy. Future work will focus on prospective clinical validation across stroke, Parkinson’s disease, and cerebral palsy populations, including correlation with standardized clinical assessment scales such as Fugl–Meyer, UPDRS, and GMFCS. These results establish NeuroPrisma as a causally grounded and computationally efficient framework for reliable, real-time movement assessment in clinical rehabilitation systems. Full article

Background/Objectives: Abnormal muscle tone and impaired motor control commonly limit gait recovery after stroke. Robot-assisted gait training has been introduced to augment conventional rehabilitation; however, its effects on stage-based motor recovery, functional ambulation, and muscle tone during the subacute phase remain unclear. Methods: This prospective, single-center, randomized controlled trial enrolled 30 patients with subacute stroke who received robot-assisted gait training plus conventional rehabilitation (R-BoT Plus group, n = 15) or conventional rehabilitation alone (control group, n = 15) over 4 weeks. The primary outcome was the change in Brunnstrom recovery stage of the lower extremities (BRS-LE). Secondary outcomes included Functional Ambulation Category (FAC), Fugl–Meyer Assessment for the Lower Extremity (FMA-LE), clinical spasticity measures (Modified Ashworth Scale and Modified Tardieu Scale), and muscle mechanical properties (MyotonPRO). Exploratory analyses were conducted to examine the associations between changes in stage-based motor recovery (ΔBRS-LE), functional ambulation (ΔFAC), and MyotonPRO parameters. Within-group changes were assessed using the Wilcoxon signed-rank test. Between-group effects were primarily evaluated using baseline-adjusted ANCOVA with HC3 robust standard errors, with Wilcoxon rank-sum tests on change scores as sensitivity analyses. Associations between changes in clinical outcomes and MyotonPRO parameters were evaluated using Spearman’s rank correlation coefficient (ρ). Results: BRS-LE (p = 0.014) and functional ambulation (p = 0.041) were significantly improved in the R-BoT Plus group. Changes in FMA-LE and clinical spasticity measures did not differ significantly between groups. Quantitative myotonometry revealed selective muscle- and parameter-specific changes. No robust correlations were observed between MyotonPRO parameters and changes in BRS-LE. Conclusions: The addition of robot-assisted gait training to conventional rehabilitation was associated with greater improvements in stage-based lower-limb motor recovery and functional ambulation in patients with subacute stroke. In contrast, cumulative impairment scores and conventional clinical spasticity measures demonstrated limited changes between groups. Quantitative muscle mechanical assessment revealed selective muscle-specific adaptations, supporting its role as a complementary tool for mechanistic characterization rather than as a surrogate marker of motor recovery. Future studies incorporating dose-matched designs and longer follow-up periods are warranted to clarify the independent and long-term effects of robot-assisted gait training. Full article

Stroke remains a leading cause of death and disability worldwide, imposing significant burdens on patients, families, and healthcare systems. Despite advances in acute management and rehabilitation, effective interventions to promote neural recovery remain limited. Hyperbaric oxygen therapy (HBOT) has emerged as a potential adjunctive treatment, but its effects on cortical functional activity—particularly the neurophysiological mechanisms underlying clinical improvements—remain insufficiently understood. This study aimed to investigate the effects of hyperbaric oxygen therapy (HBOT) on cerebral activation in stroke patients using functional near-infrared spectroscopy (fNIRS) and to evaluate its therapeutic efficacy. A total of 23 patients with intracerebral hemorrhage and 20 with cerebral infarction were enrolled. fNIRS data were collected before HBOT and within 10–30 min after treatment completion. During data acquisition, participants performed an alternating left- and right-hand grip task while wearing the fNIRS device throughout the procedure. Changes in near-infrared light intensity were monitored to objectively reflect cortical activity. The results showed that after HBOT, activation patterns in relevant brain regions during the grip task were significantly altered: activation channels during the bilateral grip task changed in cerebral infarction patients, with some brain regions overlapping with those observed in intracerebral hemorrhage patients. In intracerebral hemorrhage patients, the number of significantly activated channels decreased during the left-hand grip task but increased notably during the right-hand grip task, which may be related to cerebral functional compensation and right-hand dominance. Clinical assessments revealed significant post-treatment improvements in Brunnstrom stage, Fugl-Meyer scores, and activities of daily living. These findings suggest that HBOT may contribute to multifaceted recovery of brain function in stroke patients, not only by enhancing cerebral blood flow and oxygenation but also by facilitating neural repair and regeneration, as well as optimizing cerebral activation and functional connectivity. Thus, this study provides an objective basis for understanding the mechanisms and efficacy of HBOT in stroke rehabilitation. Full article

Surface electromyography (EMG) can drive assistive training systems in neurorehabilitation. This systematic review and meta-analysis evaluated whether EMG-driven device-assisted rehabilitation improves upper-limb (UL) and lower-limb (LL) outcomes versus conventional therapy (CT). The review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) and was registered in PROSPERO (CRD420251029642). We searched databases for randomized controlled trials in adults with neurological disorders; three reviewers screened records, extracted data, and assessed risk of bias using the Revised Cochrane risk-of-bias tool (RoB 2). Seven trials (n = 160) were included, all in post-stroke populations (UL: 3; LL: 4). UL trials showed mixed findings, and pooled effects were imprecise and not statistically significant for activities of daily living (ADL) (standardized mean difference, SMD −0.55; p = 0.09; I² = 0%). LL pooled estimates showed no significant differences in motor function (Fugl-Meyer Assessment, lower extremity, FMA-LE) (mean difference, MD −1.69; p = 0.40), walking independence (Functional Ambulation Categories, FAC) (MD −0.24; p = 0.61), balance (SMD 0.12; p = 0.61), mobility (Timed Up and Go, TUG) (MD −3.24; p = 0.71), or endurance (SMD −0.19; p = 0.43). Current evidence does not demonstrate clinical superiority over CT. EMG-driven systems may be used as an adjunct, but larger trials with standardized protocols, implementation outcomes, and neurological pathologies beyond stroke are needed. Full article

Stroke frequently results in persistent upper limb impairments, which are often accompanied by compensatory movement strategies that are not fully captured by conventional clinical assessment scales. Quantitative kinematic analyses may provide more objective and sensitive measures of motor dysfunction. In this study, we propose a probabilistic, distribution-based analysis of upper limb kinematics to quantify motor disability in post-stroke patients. We analyzed reaching movement data acquired with a markerless Kinect V2 system from 36 post-stroke patients and age-matched healthy controls. Wrist velocity profiles were characterized using distribution metrics, including variance, skewness, kurtosis, and entropy, and divergence measures (Hellinger distance, Kullback–Leibler divergence, and Jensen–Shannon divergence). Group differences between patients and controls, as well as across impairment levels stratified by the Fugl-Meyer (FM) score, were evaluated. Several distribution metrics significantly discriminated patients from controls and scaled with motor impairment severity. In particular, divergence-based measures showed a strong association with FM scores, indicating increasing deviation from normative movement patterns with greater impairment. These findings demonstrate that distribution-based metrics focusing on kinematic analysis provide a clinically meaningful, objective descriptor of motor dysfunction and complement conventional biomechanical assessments, offering a sensitive framework for quantifying motor disability after stroke. Full article

This study proposes a video-based approach for quantitatively evaluating upper-limb joint abnormalities in individuals with brain lesions during resistance exercises. While the Fugl–Meyer Assessment (FMA) is a reliable clinical tool, its use is limited by the need for expert involvement and repeated assessments. To address this issue, skeletal joint data were extracted from RGB exercise videos using OpenPose, and joint abnormalities were identified by learning normal movement patterns from non-disabled participants. A total of 26 non-disabled individuals and 12 individuals with brain lesions performed chest press, shoulder press, and arm curl exercises. Joint movement patterns were analyzed using correlation analysis and a long short-term memory (LSTM) autoencoder. Only joints relevant to each exercise were evaluated, and joint-level results were integrated to compute arm-level abnormality rates. The correlation-based abnormality rate showed a significant negative correlation with FMA scores (r = −0.7789, p = 2.83 × 10⁻³), while the LSTM autoencoder-based abnormality rate exhibited a stronger correlation(r = −0.8454, p = 5.33 × 10⁻⁴). In addition, affected-side classification accuracy reached 78.0% and 83.3% for correlation analysis and the LSTM autoencoder, respectively. These results indicate that the proposed method is consistent with clinical assessments and can serve as a non-invasive, cost-effective tool for video-based rehabilitation evaluation. Full article

Accurate upper-limb motor assessment is critical for post-stroke rehabilitation but relies on subjective clinical scales. This study proposes the Action-Aware Multimodal Wavelet Fusion Network (AMWFNet), integrating surface electromyography (sEMG) and robotic kinematics for automated Fugl-Meyer Assessment (FMA-UE)-aligned quantification. Continuous Wavelet Transform (CWT) converts heterogeneous signals into unified time-frequency scalograms. A learnable modality gating mechanism dynamically weights physiological and kinematic features, while action embeddings encode task contexts across 18 standardized reaching tasks. Validated on 40 participants (20 post-stroke, 20 healthy), AMWFNet achieved 94.68% accuracy in six-class classification, outperforming baselines by 9.17% (Random Forest: 85.51%, SVM: 85.30%, 1D-CNN: 91.21%). The lightweight architecture (1.27 M parameters, 922 ms inference) enables real-time assessment-training integration in rehabilitation robots, providing an objective, efficient solution. Full article