Classification of MRI Images in 2D Coronal View and Measurement of Articular Cartilage Thickness for Early Detection of Knee Osteoarthritis

Osteoarthritis (OA) is a degenerative joint disease which is most prevalent in the knee joint. It can be characterized by the gradual loss of articular cartilage. The knee OA-affected bones slide together due to degradation of cartilage, causing joint pain, swelling, stiffness and eventual loss of motion. Magnetic resonance imaging (MRI) is the most suitable non-invasive imaging modality to detect damages in cartilage, ligament and tendon which cannot be visualized using an x-ray. In the proposed work, the cartilage is segmented using pixel-based segmentation technique. Image processing techniques such as contrast enhancement, histogram equalization, thresholding and canny edge detection are implementedusing MATLAB R2013a (8.1) software on the MR images in 2D coronal view. Then a rough mask is created which undergoes morphological operations and the background noise is reduced. The segmented image undergoes GLCM feature extraction process. The texture features are calculated from the segmented image. The extracted GLCM features are given to the SVM classifier for classifying the image as normal and OA-affected. The accuracy was found to be 86.66% for the classification of the subject into normal and OA-affected. Articular cartilage thickness is measured using Euclidean distance formula and compared with the standard values for early detection of knee Osteoarthritis.