Regularized GAN-Augmented CNNs for Enhanced Rheumatoid Arthritis Detection

Abstract

Rheumatoid arthritis (RA), a chronic joint disease with significant implications for patient health, demands early detection and management to mitigate its potentially severe consequences. The advancement of artificial intelligence, the computer-based automatic diagnosing systems has significantly reduced human involvement in diagnosing the disease. Numerous research endeavors have focused on leveraging machine learning (ML) and deep learning (DL) algorithms to automate RA detection methods. These studies concentrate on various joints affected by RA, such as the hands, knees, and feet, providing a rapid and precise means of identification. Early detection facilitated by these advanced technologies is key to effective intervention and improved patient outcomes. This paper presents the enhanced deep learning model that can automatically detect RA. In this, we used a Kaggle knee x-ray data set; due to the less number of samples, we first augmented the samples with regularized GAN and balanced the images. Finally, these samples are trained on an optimized neural network model. In this, we got an accuracy of 0.97. Moreover, we tested all dimensions to prove the consistency of the model, and our model consistently performed with other prescribed models.