Autism Spectrum Disorder and Normal Gait Classification Using Machine Learning Approach

Abstract

Previous research has reported that children with autism spectrum disorder (ASD) exhibit unusual movement and atypical gait patterns. Automated classification of abnormal gait from normal gait can serve as a potential tool for early and objective diagnosis as well as post-treatment monitoring. The aim of this study is to employ machine learning approaches to differentiate between children with ASD and healthy controls by utilizing gait features extracted from three-dimensional (3D) gait analysis data. The gait data of 30 children with ASD and 30 healthy controls were obtained using 3D gait analysis during walking at a normal pace. Time-series parameterization techniques were applied to the kinematic and kinetic waveforms to extract useful gait features. Further, the dominant gait features were selected using statistical feature selection techniques. To highlight the efficacy of different machine learning classifiers towards devising an accurate gait classification, four machine learning classifiers were trained to classify ASD and control gait based on the selected dominant gait features. The classifiers are Artificial Neural Networks (ANN), Support Vector Machines (SVMs), K-Nearest Neighbor (KNN), and Linear Discriminant Analysis (LDA). The 10-fold cross-validation test results indicate that the ANN-SCG classifier with six dominant gait features was able to produce the optimum classification performance with 98.3% accuracy, 96.7% sensitivity, and 100% specificity. The findings indicate that the ANN classifier has the potential to serve as a valuable tool for assisting in the diagnosis of ASD gait and evaluating treatment programs.