Advanced Bi-CNN for Detection of Knee Osteoarthritis using Joint Space Narrowing Analysis

Abstract

The prevalence of knee osteoarthritis is significantly increasing due to the expanding global ageing population and the rising incidence of obesity. Many researchers use artificial intelligence analytics for knee osteoarthritis (KOA) prediction and treatment. The majority of research is restricted to particular patient groups or attributes, such MRI, X-ray, or questionnaire groups. In our research we propose the use of advanced ortho bilinear convolutional neural network (CNN) classifier to enhance the precision of knee osteoarthritis detection through joint space narrowing analysis. Recognizing the critical need for accurate and early diagnosis in osteoarthritis, this study introduces a sophisticated approach leveraging the unique capabilities of bilinear CNNs (BiCNN). By integrating bilinear interactions within the CNN architecture, the model aims to capture convoluted spatial and channel-wise dependencies in knee radiographic images, thereby improving the capability to understated changes in osteoarthritis progression, particularly within the joint space. The proposed bilinear CNN classifier technique promises to refine the precision of knee osteoarthritis detection, providing clinicians with a powerful tool for identifying joint space narrowing with improved accuracy. Based on the experiment over unseen images, the recall was 93.04%, precision 96.33%, F1 Score was 95.46% and overall accuracy was 94.28%. Results show the superiority of the proposed method compared to other state-of-the-art methods. Hence the proposed method can be used for KOA diagnosis and KL grading in real time scenarios.