A robust deep learning approach for knee osteoarthritis disease classification

Abstract

Knee osteoarthritis (KOA) is a primary cause of chronic pain and impairment that greatly reduces the quality of life for those who have it. Precise categorization of KOA is crucial for efficient diagnosis, formulation of treatment strategies, and tracking of the disease's advancement. leaf diseases can significantly impact tea crop productivity and quality, necessitating timely and accurate detection for effective management. KOA is a common ailment that progresses gradually and causes noticeable changes to the bone in X-ray images. Because they are affordable and simple to use, X-rays are the recommended diagnostic method. Physicians assess the severity of each KOA patient's disease using the Kellgren and Lawrence (KL) grading system. This method classifies the illness into stages ranging from mild to severe. Treatment can slow down knee degradation by using this method for early diagnosis. In this work, we combined two datasets to create a raw picture collection of 2042 images. To enhance image quality and generate a substantial dataset, we used image pre-processing techniques, such as resizing images and CLAHE. We employed a number of well-known algorithms (ResNet50, VGG19, InceptionV3, VGG16, and the suggested model) in our work to identify five different KOA illness classes (normal, doubtful, mild, moderate, and severe). In the end, the suggested model has the best accuracy (96.56%) when compared to other algorithms.