OsteoGA: An Explainable AI Framework for Knee Osteoarthritis Severity Assessment
Hieu Phan Trung, Su Nguyen Thiet, Tuan Nguyen Trung, Loc Le Tan, Minh–Triet Tran, Tho Quan · 2023
Knee osteoarthritis is among the most common joint disorders. Recent studies have investigated the application of Artificial Intelligence (AI) in automated diagnosis using knee joint X-ray images. However, these studies have primarily focused on diagnosing the severity of osteoarthritis without providing explanations for the underlying reasons that led to those results. In this paper, we present OsteoGA, an AI framework that focuses on the interpretability of the model in order to assist in the diagnosis of knee osteoarthritis. OsteoGA introduces a novel generative adversarial autoencoder model, called GAE, to reconstruct an assumed healthy knee joint image from the original ones. The reconstruction process in OsteoGA combines image reconstruction, data imputation, and adversarial learning to generate high-quality images that are consistent and coherent with the patient’s image. Apart from effectively diagnosing the severity of knee osteoarthritis, the OsteoGA framework also produces an anomaly map. This map highlights valuable information about the abnormal regions in X-ray images, offering additional insights to medical experts during the diagnosis process.