Abstract
Meniscus tears are serious knee abnormalities that can cause kneeosteoarthritis disorder. Therefore, early detection and treatment of meniscustears that may occur in the knee with computer-aided systems will prevent theprogression of these disorders. In this study, an approach which can detect themeniscus tears automatically by using and comparing two different featureextraction methods have been presented. With these methods, features of theknee MR images were obtained and automatic meniscus tear classification wasperformed by such features. Four different classifiers have been used to modelthe features in the classification phase. The most successful classificationresults were obtained from the support vector machines (SVM) with a successrate of 90.13% and the extreme learning machines (ELM) with a success rate of87.85% via the LBP feature extraction method. It is observed that betterresults are obtained than the ones in similar studies in the literature. It isaimed to improve the existing success with the use of deep feature extractionmethods in the future.