In today’s medical practice, various types of medical images are increasingly used, and X-ray images are used as one of the diagnostic methods in rheumatoid arthritis. Rheumatoid arthritis causes deformity of joints, especially in the fingers, and narrowing of joint space distances (JSD) as the disease progresses; however, an accurate method for measuring the JSD on X-ray images has not been established, and diagnosis is based on subjective evaluation by skilled medical specialists. In a previous study, a diagnostic application was developed for the early detection of rheumatoid arthritis; however, it was unable to correctly measure the JSD because the images were unclear and reflected parts of the bone that were not the original bone edges. In this paper, we propose a segmentation method using UNet to accurately capture the outlines of finger nodes from blurred joint images and verify the effectiveness of our proposed method through experimental results.