Automatic joint space measurement on the hand radiographs for rheumatoid arthritis(International Forum on Medical Imaging in Asia 2009 (IFMIA 2009))

概要

論文の詳細を見る
The measurement of joint distance is an important task in the quantification of joint space narrowing of rheumatoid arthritis. In this paper, we used 66 study samples which include 28 patients and 36 normal cases. We attempted curvature based algorithm to find joint distance from phalangeal region of interest (ROI). Firstly, the input data was extracted phalangeal region from both hand bone images. We segmented each finger bone out of background using k-means clustering. The center information which is half point of binary image was used to apply polynomial curve fit algorithm. Pixel average was applied on normal direction of center curve to overcome the fluctuation by variation of pixel value on the bone region. In other to find minima of the average profile, we proposed the curvature index which is angle difference among reference position of center curve and forward, backward position on the profile. The curvature minima were identified the distal inter-phalangeal joint (DIP) and proximal inter-phalangeal joint (PIP) on the center index profile. As a result, a t-test between the mean joint distance of normal and abnormal subjects in case of DIP was statistically significant (P<0.05) and systemic success rate was 93.8%.
社団法人電子情報通信学会の論文
2009-01-12