SegJSW: A Hybrid Model for Knee Osteoarthritis Classification

Knee osteoarthritis is currently one of the most serious and prevalent health issues in society. Over the years, numerous research have investigated applying Artificial Intelligence for early disease detection. However, most current approaches have merely relied on image processing techniques and thus not incorporating domain medical knowledge, which somehow limits the ultimate results. Therefore, in this paper, we propose a new method called SegJSW. Our model carries out the prediction process based on two branches of information: (1) overall diagnostic outcomes obtained from image classification models; and (2) Joint Space Width Evalutaion, developed from our proposed method. Experimented on public dataset of actual patients, SegJSW yields significant accuracy improvements compared to the baseline method (81.09% vs. 77.65%), while also enjoying recall enhancement as well (77.38% vs. 68.73%).