Motor Function Assessment of Children with Cerebral Palsy using Monocular Video

The assessment of movement abilities in individuals with neurological disorders is a critical task in clinical practice. Currently, clinical assessments are time-consuming and rely on qualitative scales typically conducted by trained clinicians. Moreover, these assessments offer only coarse snapshots of a person's abilities, failing to track the minutiae of recovery over time. To overcome these limitations, we propose a machine learning approach based on spatial-temporal graph convolutional network (STGCN) to extract movement features from pose data obtained from monocular videos collected with mobile devices (e.g., smartphones, tablets). Our proposed method achieves an accuracy of 76% in evaluating children with Cerebral Palsy (CP) using the Gross Motor Function Classification System (GMFCS), a 10% improvement in accuracy compared to current state-of-the-art methods, and shows substantial agreement with professional assessments based on the weighted Cohen's Kappa (kappa(lw) = 0:74). Furthermore, the proposed method can be efficiently implemented on a wide range of mobile devices in real-time or near real-time.