Classification of Gait Type in Hip Osteoarthritis Patients using Principal Component Analysis and Gaussian Mixture Model

Abstract

This paper proposes and compares two approaches that can identify different gait types of patients with hip osteoarthritis (OA) quantitatively using machine learning techniques. One is a simple and intuitive method that does not need clustering steps, and the other is a more detailed classification method that subdivides the gait classification results of the first method. Force plate measurements of 22 patients with hip OA and 18 healthy subjects without surgical history were collected and analyzed using principal component analysis (PCA) and Gaussian Mixture Model (GMM) to identify different types of gait. The physical meanings of the identified gait types are explained using the latent features of gait obtained from PCA and muscle forces calculated using OpenSim. The approaches will not only be useful for understanding the gait patterns of patients with hip OA but also will be applicable in analyzing different types of gait other than those of patients with hip OA.