KNEE OSTEOARTHRITIS PREDICTION DRIVEN BY DEEP LEARNING AND THE KELLGREN-LAWRENCE GRADING

Abstract

Degenerative osteoarthritis of the knee (KOA) affects the knee compartments and worsens over 10–15 years. Knee osteoarthritis is the major cause of activity restrictions and impairment in older persons. Clinicians' expertise affects visual examination interpretation. Hence, achieving early detection requires fast, accurate, and affordable methods. Deep learning (DL) convolutional neural networks (CNN) are the most accurate knee osteoarthritis diagnosis approach. CNNs require a significant amount of training data. Knee X-rays can be analyzed by models that use deep learning to extract the features and reduce number of training cycles. This study suggests the usage of DL system that is based on a trained network on five-class knee X-rays with VGG16, SoftMax (Normal, Doubtful, Mild, Moderate, Severe). Two deep CNNs are used to grade knee OA instantly using the Kellgren-Lawrence (KL) methodology. The experimental analysis makes use of two sets of 1650 different knee X-ray images. Each set consists of 514 normal, 477 doubtful, 232 mild, 221 moderate, and 206 severe cases of osteoarthritis of the knee. The suggested model for knee osteoarthritis (OA) identification and severity prediction using knee X-ray radiographs has a classification accuracy of more than 95%, with training and validation accuracy of 95% and 87%, respectively.

Keywords

• x-rays
• osteoarthritis (oa)
• cnn
• deep learning
• vgg16