Automatic Joint Position Estimation Method for Diagnosis Support System in Rheumatoid Arthritis

Rheumatoid arthritis is a chronic inflammatory disease characterized by inflammation of the synovial membrane of the joints. The destruction of the joints results in a narrowing of the interarticular space, distortion of the joints, and deformity of the fingers. Currently, the mainstream method of diagnosing such symptoms is by direct visual inspection of X-ray images by a physician. However, the diagnosis of the progression of joint destruction requires a lot of time and effort because the number of joints to be observed is large and it is difficult to accurately read minor pathological changes. In addition, the subjective evaluation by the physician is subject to large intra- and inter-examiner variability and lacks reproducibility, which is a major problem. As a means to solve these problems, the goal of this paper has been to develop a diagnosis support system by image processing that is quantitative and labor-saving. In this paper, we propose a method to improve the accuracy of joint position estimation in X-ray images in order to obtain the temporal variation between two images, and confirm the effectiveness of our proposed method by experiments.