Foundation-model-based multi-label phenotyping of combined hyperkinetic movement disorders
AI that spots multiple movement disorders in videos, without markers or retaining
Researchers combined two AI foundation models to identify eight different hyperkinetic movement disorders from video alone, without requiring special markers or sensors. Trained on just 25 people, the system transferred directly to new pediatric and adult datasets with zero false positives for dystonia and near-perfect detection of tremor, outperforming standard keypoint-tracking methods on high-confidence diagnoses.
Movement disorders are notoriously hard to diagnose consistently because even expert clinicians disagree on what they're seeing. A reliable, automated video-based system that works across ages and different clinical settings could make diagnosis faster, cheaper, and more consistent worldwide — eliminating the need for specialized equipment or travel to expert centers, while reducing the guesswork that currently leads to delayed or missed diagnoses.