Deep Learning Models for Knee Osteoarthritis Diagnosis: An Advancing Healthcare Approach

Abstract

Knee Osteoarthritis (OA) is one of the most common forms of arthritis and is primarily a degenerative joint disease. It causes pain, stiffness, and loss of movement in the knee joint by affecting the surrounding tissues, bones, and cartilage. Manual diagnosis of this condition involves looking at knee region X-ray images and using the Kellgren-Lawrence (KL) system to categorize severity levels. Conventional techniques for identifying and classifying osteoarthritis, such as manual radiography evaluation, are frequently laborious and subjective. This chapter introduces an automated system for classifying knee OA using deep-learning models such as ResNet18, SqueezeNet, EfficientNetB0, and ResNeXt50. The performance of the deep learning models is evaluated using performance metrics like accuracy, precision, recall, and F1-score. Compared to other models, the ResNeXt50 model promises to reduce the subjectivity in OA diagnosis and enhance diagnostic accuracy, potentially leading to more timely and effective treatment strategies for patients with knee osteoarthritis.