Cerebral Palsy classification based on multi-feature analysis using machine learning

Cerebral Palsy (CP) is an umbrella name for disorders caused by abnormal brain development or damage to the developing brain. It affects the child's ability to move and maintain balance and posture. CP is the most common cause of motor disability in childhood. CP has various classification systems based on body structure and function, such as the topographic classification into quadriplegia, diplegia, and hemiplegia, or based on activities such as the gross motor function classification system (GMFCS). Classifying children with CP is challenging for clinicians due to the homogeneous nature of the clinical presentations of children with CP. Classifications of CP can guide the planning of services that enhance the quality of life of children and their families. There are a few studies about children with CP in Jordan. Data about children with CP in Jordan are also lacking. This study, therefore, aims to compile a comprehensive, concise, and fully annotated national dataset for Jordanian children with CP and to provide a benchmark for related studies in the field of pediatric rehabilitation medicine (PRM). This work presents the methodology implemented to compile and analyze the dataset and the experiments conducted on this dataset. We evaluated the dataset using a set of five commonly used machine learning classification algorithms, namely, K-Star, Multilayer Perceptron (MLP), Naïve Bayes (NB), Random Tree (RT), and Support Vector Machine (SVM). The MLP classifier successfully classified CP-type cases with an accuracy rate of 84% and GMFCS cases with an accuracy rate of 53%. The obtained results were promising and encouraging to put the compiled CP dataset into practice for clinicians, researchers, and policymakers working in the field of PRM.