Interpretable Temporal Facial-Region Motion Analysis for In-the-Wild Parkinson's Disease Video Classification
Researchers have developed a method to classify Parkinson's disease from in-the-wild videos using facial motion analysis. The study focused on temporal motion descriptors extracted from facial keypoints, finding that normalized velocity descriptors combined with a Random Forest classifier achieved a balanced accuracy of 0.826. This approach is highlighted as lightweight and interpretable for classifying Parkinson's-related videos, though it does not claim to assess clinical severity. AI
IMPACT Offers a new, interpretable method for AI-assisted diagnosis of neurological conditions from video data.