Efficiency Enhancement of Knee Osteoarthritis Classification Using Optimization Technique

Patients with knee osteoarthritis (KOA) have a significant reduction in their quality of life. Since the mid-twentieth century, the prevalence of KOA, a degenerative joint disease, has increased. The importance of early detection of longitudinal KOA grading for efficient monitoring and remediation has grown in recent years. Detecting and tracking the progression of KOA at an early stage is essential for preventative therapy. Kellgren and Lawrence (KL) developed a grading system used in clinical settings to categorize the severity of KOA; grades range from 0 (no) to 4 (severe). The human expert's evaluation of low-resolution images (i.e., X-ray images) leads to the wrong identification of disease. To solve this issue, in this research, we employed an optimized feature selection method to extract critical information from X-rays and built a deep learning (DL) model to accurately determine the degree of KOA from the radiometric images alone. There are four primary sections in the suggested approach. First, we developed certain processing methods for making X-ray images free from noise. Second, to retrieve the features from the images, we create texture- and color-based feature extraction methods. Third, an optimization method called the firefly approach is used to pick the most correlated features. Finally, to categorize the severity of KOA, we develop a convolutional neural network (CNN). The results of feature extraction with and without feature selection are used to train and validate the CNN. The performance of CNN by applying two different inputs is validated using the metrics. The results of the experiments demonstrate that the accuracy of the CNN model is improved by 2.5% thanks to the optimization technique in feature selection.