Deep Expert: Domain-specific features extraction for knee osteoarthritis using deep learning.

Abstract

Early diagnosis of knee osteoarthritis (KOA) is essential to prevent joint degeneration, which significantly impacts quality of life and imposes an in- creasing burden on healthcare systems worldwide. However, the traditional diagnostic approach is cumbersome and difficult due to the narrow space between the tibia and the femur and the difficulty in identifying bone ero- sions and even osteophytes at the joint margins and the generalization of cross data sets. In our suggested method, we introduce a mixed deep sys- tem that analyzes X-ray images of the front of the knee to determine how severe the disease is, using region of interest, gap extraction and classi- fier. This approach eliminates the need for manual annotations and has real-world flexibility. This work contributes to computer-aided diagnosis by providing efficient, interpretable, and scalable osteoarthritis diagnosis mechanisms, ultimately supporting clinicians in early intervention and in- dividualized treatment planning.