Knee Osteoarthritis Detection Using Deep Learning Algorithms

Abstract

Osteoarthritis is one of the most common form of arthritis that affects middle-aged and elderly people, and osteoarthritis (OA) usually affects knees and small finger joints, as well as thumb. The conventional prevalent practice to diagnose such as osteoarthritis diseases relies on human interpretation of medical images. This manual diagnosis approach is prone to the medical skills of individuals, time consuming, and error-prone task. However, the Artificial Intelligence (AI) has emerged in the recent years as a new diagnosis approach with the potential to overcome these limitations, and it shown outstanding success in processing and analysis of medical images. This paper presents a convolutional neural network based model and a mobile application that can help healthcare specialists and nonprofessional individuals in diagnosing knee OA from 2D (X-ray) images. The proposed model was trained and was tested using images from three different sources: an online reputed medical source, local diagnostic centers, and from local hospitals. The user-friendly mobile application was designed to take an X-Ray image as an input to this model and then displays the level of severity of the knee OA.

The obtained results were very satisfactory the model yields up to 92% accuracy to predict the presence of osteoarthritis or not. The model also achieved about 86% accuracy to predict the Knee OA severity level (five grades) based on the KL system.