Deep learning model for the estimation of Rheumatoid arthritis in hand radiographs

Rheumatoid Arthritis (RA) is an autoimmune illness that affects the hands and feet’s joints. The detection of RA is done using the scoring of radiographs and is a difficult task that requires trained clinicians. An automated system for diagnosing RA will eventually help physicians in better patient care and early detection. The study aims to implement an automated tool for the evaluation of RA disease. One hundred hand radiographs of RA and 100 normal hand radiographs were acquired for the study. ResNet50 and SqueezeNet models using the transfer learning approaches are used in this work. The accuracy obtained by the ResNet50 model is 89.8% and the SqueeNet model is 77.5% for the assessment of RA disease.