Detection and Classification Knee Osteoarthritis Algorithm using YOLOv3 and VGG-16 Models

Abstract

The paper proposes an algorithm to help doctors identify Knee Osteoarthritis automatically on X-ray images by deep learning method with the YOLOv3 model and VGG-16. We collect data including 2874 knee X-ray images taken from the OAI dataset which is a large data set of Knees and joints. In this paper, all data are labeled and tested by chiropractors at Bach Mai Hospital which will be preprocessed by applying the CLAHE algorithm to improve image quality. Next, the YOLOv3 modelis trained with a preprocessed dataset as input. It then predicts the location of the knee joint on the X-ray image automatically. Finally, we use VGG-16 to classify the image. Besides, we alsouse preprocessing methods to enhance the image quality before training, and thus the accuracy of the problem is also significantly improved. The results display that the proposed method achievesan accuracy of up to 89%