Early detection of Knee Osteoarthritis using SVM Classifier

Abstract

Osteoarthritis of Knee (KOA) is the most commonly found arthritis types. It is distinguished by wearing away of cartilage which is meant to ensure smooth movement between bones in joints. As a result of the wearing away, bones slide and rub against each other and this leads to swelling, pain and eventual loss of motion. Early detection of KOA can help ensure retarded progression of the ailment. Estimation of the thickness of articular cartilage is important in determining the stage of development of KOA. The estimated cartilage thickness along with the four textural features such as contrast, correlation, energy, homogeneity and three statistical features such as mean, median, variance (totally eight) is fed as training features to SVM to automatically classify the images into KOA and non-KOA cases. The results obtained from SVM with RBF kernel, SVM with linear kernel is 95.45% and SVM with polynomial kernel is 87.8%.