Automating General Movements Assessment with quantitative deep learning to facilitate early screening of cerebral palsy

The Prechtl General Movements Assessment (GMA) is crucial for evaluating infant nervous system integrity and predicting motor dysfunction, especially cerebral palsy (CP). However, GMA's reliance on highly trained professionals limits its widespread use. This study introduces a deep learning-based motor assessment model (MAM) automating GMA at the fidgety movements (FMs) stage. MAM achieves an AUC of 0.967 during validation, accurately identifying FMs and exhibiting strong agreement with expert assessments. A quantitative GMA method using predicted FMs frequency achieves an AUC of 0.956, improving GMA beginners' diagnostic accuracy by 11.0%. MAM streamlines early CP screening and advances video-based quantitative medical diagnostics.

Qiang Gao, Siqiong Yao, Yuan Tian, Chuncao Zhang, Tingting Zhao, Dan Wu, Guangjun Yu, Hui Lu