Application of supervised machine learning algorithms in the classification of sagittal gait patterns of cerebral palsy children with spastic diplegia

Abstract

Gait classification has been widely used for children with cerebral palsy (CP) to assist with clinical decision making and to evaluate different treatment outcomes. The aim of this study was to evaluate supervised machine learning algorithms in the classification of sagittal gait patterns for CP children with spastic diplegia. Gait parameters were extracted from gait data obtained from two hundred children with spastic diplegia CP, and were used to represent the key kinematic features of each individual's gait. Seven supervised machine learning algorithms including an artificial neural network (ANN), discriminant analysis, naive Bayes, decision tree, k-nearest neighbors (KNN), support vector machine (SVM), and random forest were compared by constructing a gait classification system based on the same gait data. The performance of these algorithms was then evaluated using a standard 10-fold cross-validation procedure. The results show that the ANN has the best prediction accuracy (93.5%) with a low resubstitution error (5.8%), high specificity (>0.93) and high sensitivity (>0.92). The decision tree algorithm, SVM, and random forest approaches also have high prediction accuracy (>77.9%) with low resubstitution error (<14.3%), moderate specificity (>0.5) and moderate sensitivity (>0.2). The discriminant analysis, naive Bayes and KNN methods have relatively poor classification performance. Given these results for classification performance and prediction accuracy, the ANN is a good candidate for gait classifications for CP children with spastic diplegia. The decision tree is also attractive for clinical applications due to its transparency. Supervised machine learning algorithms can potentially be integrated into an expert gait analysis system that can interpret gait data and automatically generate high-quality analyses.