Evaluation Method of Rheumatoid Arthritis by the X-ray Photograph using Deep Learning

Rheumatoid arthritis (RA) is evaluated using the Sharp/van der Heijde scoring system that is currently the most widely used X-ray photograph scoring system. However, it is also difficult for the doctor who is not major in RA cannot evaluate it exactly since this diagnosis relies on implicit knowledge which is difficult to transfer to another person by the text or figures. Therefore, it is desirable to develop a system that enables the condition of RA to be evaluated automatically. Our objective is to develop an automatic evaluation system that is improved by the evaluation and the modification by the doctor in diagnosis in the hospital. This paper proposes the method for the realization of both recognition and classification at the same time using deep learning. And the experiments are performed using X-ray photographs. As a result, we confirm that the effectiveness of the system is verified.