Deep Learning Pose Estimation for Multi-Label Recognition of Combined Hyperkinetic Movement Disorders
Researchers have developed a novel framework using markerless pose estimation and a tabular foundation model to identify multiple hyperkinetic movement disorders from routine videos. The system was initially trained on adult patients and then tested on a pediatric cohort, demonstrating improved accuracy after a lightweight calibration. This approach aims to provide an objective and scalable method for diagnosing and monitoring conditions like dystonia, tremor, and tics, which are often challenging to assess due to their subjective and variable nature. AI
IMPACT Provides a more objective and scalable method for diagnosing and monitoring complex movement disorders, potentially improving patient care.