Feedback-Enhanced Virtual Reality Upper-Limb Training With Body Position Measurement in Healthy Adults: Development...

Virtual reality (VR) systems are increasingly used in rehabilitation to facilitate motor learning by providing visual feedback. However, few studies have validated the motion tracking accuracy of VR devices compared with gold-standard motion capture systems. In particular, validation evidence for upper-limb reaching with commercially available VR tracking setups remains limited.

This study aimed to evaluate the validity of a custom VR-based rehabilitation system (VRactice) by comparing its motion tracking data with that of a Vicon motion capture system during a goal-directed reaching task in healthy adults.

This laboratory-based validation study was conducted at Tokyo Kasei University, Sayama Campus, Japan (August-December 2023). Participants were recruited via posted announcements on campus (convenience sampling) and received a 1000 JPY gift card (US $7.00; JPY 142.79=US $1 as of August 1, 2023). A total of 16 healthy participants (n=6, 37.5% male and n=10, 62.5% female participants; mean age 25.3, SD 4.56 years; all right-handed) performed reaching tasks in a VR environment while being tracked simultaneously by both the VRactice system and a Vicon system. Trackers and reflective markers were attached to the hand and elbow to capture 3D coordinates. Each participant performed 10 reaching trials at a frequency of 1 Hz. Data were upsampled to 100 Hz, synchronized, and normalized to the initial position. Valid cycles were identified, and distance time series from the initial position were extracted for the 500-millisecond interval preceding the peak displacement. For each participant, all valid cycles were pooled, and the coefficient of determination (R2) between VRactice and Vicon trajectories was calculated. Of 160 planned trials (16 participants×10 trials), 4 (2.5%) trials were not recorded; the remaining 156 (97.5%) trials were analyzed without imputation. Statistical significance was evaluated at a 2-sided α level of .05.

Strong agreement between VRactice and Vicon was observed at both the individual and group levels. The R2 ranged from 0.75 to 0.99 across participants, and all comparisons were statistically significant (P<.001). Deviations between the 2 systems remained minimal, confirming that VRactice reliably reproduced the temporal and spatial characteristics of reaching trajectories. At peak displacement, the participant-level mean absolute difference (mean of 10 trials per participant) was 36.5 (SD 29.3) mm (95% CI 20.9-52.1), suggesting spatial agreement that may be acceptable for upper-limb reaching measurements in this experimental context.

The findings support the validity of VRactice in capturing reaching movements with high spatial accuracy compared with a motion capture system. By providing reliable motion data, VRactice may serve as a useful platform for delivering real-time visual feedback and supporting motor training applications in rehabilitation settings. This study is innovative in that it provides formative validity evidence for a VR-based system that integrates real-time trajectory monitoring with adaptive visual guidance, supporting subsequent clinical implementation and evaluation.