RGB-D Videos-Based Early Prediction of Infant Cerebral Palsy via General Movements Complexity

Early detection and intervention of cerebral palsy can promote neural remodeling in the process brain development, thus reducing negative effects palsy. In this paper, we proposed a novel method for early prediction infant based on General Movements Assessment (GMA) theory with RGB-D videos. Firstly, explored human pose recognition supine position Then further apply it to auto-GMA. Specifically, employ current estimation RGB images achieve full body 2D key points. By combining depth information, 3D movement be obtained. infant's complexity index is achieved by extracting whole-body characteristic. order verify effectiveness method, did some experiments public dataset consisting 12 real recorded infants' videos, 4 samples were diagnosed as abnormal infants GMA expert. We use expert ratings these movements gold standard. Our state-of-the-art sensitivity 100%, specificity 87.5%, accuracy 91.7%. The results show that has great potential assisting doctors diagnose