Estimating the severity of knee osteoarthritis using Deep Convolutional Neural Network based on Contrast Limited Adaptive Histogram Equalization technique

Abstract

Osteoarthritis is a degenerative joint disease that affects larger joints, including the knee, foot, hip, and spine by infecting the cartilage, which causes bones to rub against each other in extreme pain. Knee osteoarthritis (KOA) manual inspections demand a skilled physician to examine the x-ray image. In this paper, a method for detecting the severity of knee osteoarthritis using a Deep Convolutional Neural Network (3D CNN) is proposed in order to benefit physicians by examining the X-ray images. Here, the Contrast Limited Adaptive Histogram Equalization (CLAHE) technique is used to enhance contrast in images, and this strengthens the current approach. This study presents two models: one for identifying and categorizing normal (KL 0,1) and osteoarthritic results (KL 2, 3, 4). The second model is to assist in distinguishing between severe grades (KL 3, 4) and non-severe grades (KL 2) or normal findings (KL 0,1). For examination, X-ray images from the Knee Osteoarthritis severity grading dataset and the Osteoarthritis Initiative (OAI) dataset are used. Experimental results have demonstrated that the proposed technique achieves accuracy rate of 85.50%, outperforming a number of existing approaches.