Skip to Main Content
Rheumatoid arthritis (RA) is a chronic autoimmune disease which affects multiple joints and causes bone erosions and joint damage. Among the available imaging modalities like X-ray, Computed Tomography, Ultrasonograph and Magnetic Resonance Imaging (MRI) for diagnosing RA, Conventional radiographs have been considered to be the gold standard method for evaluating the progression of bone and joint damage in RA. The aim and objective of this proposed approach is as follows: i) to automatically segment the bone regions in hand radiographs of rheumatoid arthritis patients using dual tree complex wavelet based watershed algorithm i) to extract the features using Gray level Co-occurrence matrix (GLCM) and to classify the arthritis using back propagation neural networks. Hand radiographs of ten RA patients and five normal persons were used in this study. First, the hand radiographs are pre-processed and segmented using watershed algorithm. Then ten features are extracted from the segmented image using gray level co-occurrence matrix (GLCM). The feature vector extracted from segmented image is given as input to the back propagation network. The BPN network classifies and produce the output as arthritis (abnormal) or normal. The performance of classification was evaluated using various statistical measures such as sensitivity obtained as 78%, specificity as 75% and accuracy was 77%. In this proposed approach, the combination of dual tree wavelet with watershed algorithm and BPN network are very efficient in segmentation and disease classification.
Date of Conference: Oct. 30 2012-Nov. 2 2012