Whether clinicians with varying levels of general movement assessment (GMA) expertise could accurately detect fidgety movements from skeleton (privacy-preserving) videos, and whether performance improved with exposure across expertise levels.
GMA tutors achieved high accuracy in detecting fidgety movements from skeleton videos; certified non-tutors and unqualified individuals performed heterogeneously, though overall performance improved with repeated exposure across the full sample.
The study does not report specific accuracy metrics or sample sizes for each expertise group, limiting quantitative interpretation; generalizability is uncertain given the likely small, specialized participant pool; real-world clinical implementation of skeleton-video-based GMA has not been tested.
Skeleton videos—which strip identifying visual detail—may offer a viable, privacy-preserving format for sharing GMA data at scale, enabling both expert human review and AI training without compromising patient privacy.
Explore related topics
