Evaluation of osteoarthritis in plain radiographs using machine learning algorithm

• Main Authors: Anbarasi, A., Snehalatha, U.
• Format: Conference Proceeding
• Language: English
• Subjects: Abnormalities; Algorithms; Arthritis; Biomechanics; Bones; Classifiers; Image segmentation; Knee; Machine learning; Multilayer perceptrons; Osteoarthritis; Radiographs; X ray imagery
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0126374

Osteoarthritis (OA) is the most common kind of arthritis which refers to biomechanical changes within a knee joint region. In clinical diagnosis, X-ray imaging is the best tool for detecting abnormalities in bone and joints. The purpose of this study is to use machine learning approaches to assess osteoarthritis in plain radiographs. The evaluation of knee OA is analyzed automatically based on a dataset that is collected from the Kaggle. To select the required region of interest from the knee X-ray images segmenting bones process was carried out. Following the segmentation, three classifiers, Random Forest, Naive bye, and multilayer perceptron, are used to extract features and determine the existence of OA. This machine learning system is learned and tested with 30 images, 15 of which are normal and 15 of which are abnormal. The system’s distinguishing performance yields an accuracy of 86.66% and 96.66% in detecting OA, respectively. This level of precision may be sufficient for it to serve a clinical purpose in an OA analysis system. Hence, the proposed automated system for detecting the OA region in the X-ray images is detected and classified with different classifiers by using machine learning techniques.