Abstract
Osteoarthritis (OA) is a common and devastating joint disease that requires a precise diagnosis to be effectively treated. Knee OA (K-OA), a common form, results in cartilage degeneration, exposing underlying bone and affecting connective tissues, leading to knee pain and discomfort. In this research, we look into how machine learning (ML) and deep learning (DL) methods may be used on 3D-Magnetic Resonance Imaging (MRI) data to diagnose K-OA. Four machine learning algorithms, support vector machine (SVM), Naive Bayes (NB), convolutional neural network (CNN), and knearest neighbors (KNN), are tested for their diagnostic effectiveness. When it comes to data classification, the straightforward KNN technique uses similarity measurements, while SVM uses ensemble learning to include multiple decision tree forecasts. NB, a probabilistic model, assumes feature independence, and CNN excels in extracting hierarchical image features using deep learning. The dataset comprises 3D knee MRI scans from individuals with and without OA. Normalization, feature extraction, and dimensionality reduction are key aspects of preprocessing that improve algorithm performance. One way to measure the performance of these algorithms is by examining their Area under Curve-Region of Curve (AUC-RoC), which is a measure of their diagnostic ability. Other metrics include their sensitivity, specificity, and accuracy. Results reveal high diagnostic accuracy across all methods, with CNN outperforming the others due to its ability to extract features from raw MRI data autonomously. This study highlights CNN’s potential to significantly improve K-OA diagnosis through advanced analysis of 3D-MRI data, offering promising insights for clinical applications.
Keywords: 3D Magnetic Resonance Imaging, Convolutional Neural Network, K-nearest Neighbors, Knee Osteoarthritis, Naive Bayes, Support Vector Machine.