Adaptive crossed reconstructed (ACR) K-mean clustering segmentation for computer-aided bone age assessment system

Abstract

The development of computer-aided design (CAD) system for clinical usage has been given excessive attention in recent years. Nonetheless, many problems still remain unsolved in the CAD field especially the segmentation problem in digital image processing. In order to increase the accuracy and efficiency in Bone age assessment (BAA), CAD system has been developed to assist the doctor and radiologist. The crucial step in the system is the bone segmentation before proceeding to the subsequent analysis and comparison with atlas. Therefore, in this paper, a method proposed to solve the problem based on grey-level co-occurrence matrix (GLCM) and k-means clustering, namely adaptive crossing reconstruction (ACR) k-mean clustering method. The method begins with bands separations into vertical and horizontal direction. Next, the pixels of each section are clustered and performed with GLCM texture analysis. At last, all the sections will be reconstructed based on the texture analysis. The resulting outcome shows that this method could segment the bone from the soft-tissue region and background effectively compared to global clustering method.