(A) study on automatic bone segmentation in dual x-ray absorptiometry image

Abstract

Active shape model (ASM) has been widely adopted in automated segmentation of bone from radiographs. ASM is an iterative algorithm that tries to fit the shape model to the contours of object in the image. ASM has the advantage that it makes use of prior knowledge of object shape and appearance and therefore, do not need to rely completely on the information in the images. However, in conventional ASM, the multiple edges resulting from the bone, soft tissue and air area in near vicinity in ASM cause segmentation errors when segmenting bone in digital radiographs. To avoid the segmentation errors, in this paper, we propose the modified ASM which emphasizes bone edge and downplays soft tissue edge by taking Dual X-ray Absorptiometry (DXA) decomposed bone image into account. For automatic procedure, the proposed bone segmentation system consists of mean shape training, chamfer matching, DXA decomposition, and modified ASM process. Our experimental results show that the proposed method accurately segment distal radius. For the test images, our segmentation method reduces in average 39.5% segmentation error measured by point-to-line distance error.