Knee osteoarthritis automatic detection using U-Net

Abstract

Knee osteoarthritis or OA is one of the most common diseases that can affect the elderly and overweight people. OA is occurred as the result of wear, tear, and progressive loss of articular cartilage. Kellgren-Lawrence system is a common method of classifying the severity of osteoarthritis depends on knee joint width. According to Kellgren-Lawrence, knee OA is divided into five classes; one class represents a normal knee and the others represent four levels of knee OA. In this work, we aim to automatically detect knee OA according to the Kellgren-Lawrence classification. The proposed system uses the U-Net architecture. The overall system yielded an accuracy of 96.3% during training.