Knee osteoarthritis detection based on the combination of empirical mode decomposition and wavelet analysis

Abstract

The early-stage of knee osteoarthritis (OA) is usually asymptomatic. However, timely detection of osteoarthritis can prevent further cartilage degeneration via appropriate exercise prescription and behavioral change. In this article, a noninvasive method to diagnose the OA of a knee recording the knee vibroarthrographic (VAG) signals over the mid-patella during the standing movement is proposed. A method that combines empirical mode decomposition (EMD) and wavelet transform is developed to analyze the nonstationary VAG signals. The least squares support vector machine algorithm (LSSVM) that is a type of support vector machine is used to classify the knee joint VAG signals (26 normal and 25 abnormal) collected from healthy subjects and patients suffering from the knee OA using the Kellgren and Lawrence grading system III and IV (KLGS III and IV). The LSSVM classifier achieves an accuracy of 86.67% in differentiating the normal and abnormal subjects that proves the effectiveness of the autocorrelation function features and continuous wavelet transform (CWT) features. Therefore, the VAG signals can be clinically significant for the classification of healthy and OA subjects.

Keywords

• osteoarthritis (oa)
• vibroarthrographic (vag)
• empirical mode decomposition (emd)
• wavelet transform
• least squares support vector machine algorithm (lssvm)