Evaluation of markerless hand-tracking accuracy in VR application designed for hand range-of-motion measurement and rehabilitation

Background Virtual reality offers new opportunities for the objective assessment and training of hand function during rehabilitation. However, the reliability of performance metrics depends on the accuracy and stability of hand tracking. This study evaluates thumb-finger distance measurement in a virtual reality application designed for hand range-of-motion measurement and rehabilitation by comparing native Meta Quest hand tracking with MediaPipe. The comparison was designed as a technical pilot assessment rather than a population-level validation. Methods The protocol comprised a static head-to-head comparison between Meta Quest and MediaPipe and a separate dynamic orientation analysis conducted using Meta Quest only. The core comparison involved one healthy right-handed male participant aged 25 years. Static tests for both tracking approaches were performed at reference distances of 5, 10, and 15 cm verified with a digital caliper, whereas the Meta Quest dynamic tests used a 10 cm reference during full rotations and half rotations in palm-facing and dorsum-facing sectors. Results Differential-error analysis showed that the reproduction of reference-distance changes depended on hand-finger configuration. In the static cross-system comparison, Meta Quest showed lower MAE in 23 of the 24 matched static configurations and higher median tolerance-band coverage than the tested MediaPipe pipeline, while MediaPipe was more susceptible to occlusion and landmark-assignment errors. A separate Meta Quest dynamic analysis demonstrated orientation-dependent changes in tracking performance. An exploratory extension involving five additional healthy male students aged 18–24 years provided descriptive multi-user context and revealed between-subject variability, particularly in dynamic conditions. Conclusions Native Meta Quest tracking appears more suitable than MediaPipe for distance-based VR hand rehabilitation applications under the tested conditions. The findings represent technical validation of relative thumb–fingertip distance measurement and should not be interpreted as clinical validation or absolute 3D positional validation. These findings support further technical development and evaluation of native headset-based hand tracking for adaptive VR systems and provide practical guidance for reducing tracking errors and standardizing the reporting of performance metrics in rehabilitation-oriented VR environments.