Identification of patellofemoral pain syndrome using a Support Vector Machine approach

Abstract

Patellofemoral pain syndrome (PFPS) is a common disorder resulting in varying degrees of knee pain. The diagnosis of PFPS is difficult since the exact biomechanical factors involved in the disorder are still unknown. Recent research has reported significant differences in rearfoot kinematics and foot ground reaction forces (GRFs) which could be indicative of PFPS, but the inter relationship between these measures and the pathology have been inconclusive so far. In this paper, we investigate Support Vector Machines (SVM)'s potential to classify gait patterns with PFPS using 14 GRF and 16 rearfoot kinematic features. Test results indicated that using GRF features alone resulted in a leave one out (LOO) classification accuracy of 85.15% compared to 74.07% using kinematic features. A hill climbing feature selection algorithm was subsequently applied to determine the subset of features which improved classifier performance. This reduced subset consisted of 6 features from a combination of GRFs and kinematic features, and provided a maximum LOO accuracy of 88.89% in detecting the PFPS gait.