Data-driven analysis of Armeo Spring performance across neurological disorders: implications for personalized upper...

Robotic-assisted therapy (RAT) has emerged as an effective approach to upper limb neurorehabilitation. Among available systems, the Armeo®Spring enables task-oriented, customizable training supported by virtual reality (VR), fostering motivation and neuroplasticity. This retrospective observational study aimed to evaluate longitudinal changes in performance across different VR exercises using Armeo®Spring session data from patients with diverse neurological conditions and to identify tasks exhibiting significant improvement within particular diagnoses, thereby supporting personalized robotic rehabilitation.

The dataset included adults (≥18) with common neurological disorders who completed ≥20 Armeo®Spring sessions using frequent integrated VR exercises. Performance across the first 20 sessions was analyzed using linear mixed-effects models with fixed effects for session, disease, age, sex, difficulty, and mechanical support, and random patient intercepts and slopes. False discovery rate (FDR) correction was applied to identify disease- and task-specific improvement trajectories.

After sequential filtering, the final cohort included 71 patients (30 with ischemic stroke, 15 with hemorrhagic stroke, 15 with multiple sclerosis, and 11 with Parkinson's disease) who underwent rehabilitation using five different VR exercises: Balloons, Roll the Ball, Fly High-Elbow, The Goalkeeper, and Pirate Adventure. A significant improvement in Roll the Ball scores was detected for MS (slope = +9.41 points/session, FDR = 0.0015), IS (+9.18 points/session, FDR = 0.0001), and HS (+7.28 points/session, FDR = 0.023). In Fly High (Elbow), MS patients demonstrated a significant improvement (+6.84 points/session, FDR <0.001) as for IS patients (+5.00 points/session, FDR <0.001). Task difficulty was consistently correlated with lower scores across all games (FDR <0.05), whereas age and sex were not significant predictors in the adjusted models.

Disease-specific recovery profiles suggest that proximal, multi-joint VR exercises, such as Roll the Ball and Fly High (Elbow), may be particularly effective for patients with multiple sclerosis and ischemic stroke, whereas other exercises show smaller or non-significant improvements. These findings support tailoring VR-based rehabilitation to the patient's neurological condition, enabling targeted, condition-specific exercise selection and progression, which may enhance the effectiveness and efficiency of upper-limb recovery.