Modelling Clinical and Radiographic Data for Early Detection and Prediction of Osteoarthritis

Osteoarthritis (OA) is a prevalent degenerative disease in joint area that impacts the quality of life of people who face it. This chronic disease causes lifelong pain and affects mobility, thus causing lifelong suffering. The early detection and prediction of this disease can aid in timely treatment thus saving quality of life and reducing healthcare costs. Despite numerous researches, accurate early detection of this disease has always been a challenge in the medical field due to the complex and multifactorial nature of the disease. This research provides a comprehensive approach to the detection and prediction of OA by two different types of data: symptoms, clinical data embedded with radiological data and patient’s X-ray images of knee joints. By integrating diverse data, the model uses EfficientNetB3, a powerful Convolution Neural Network (CNN)-based algorithm for the X-ray image analysis and XGBoost and a Deep Learning approach for analyzing clinical data which is diverse and has a wide spectrum of clinical fields for analysis. The EfficientNetB3 achieves an accuracy of 95.2% for detecting the presence of OA and accuracy of 84% for grading the severity of OA. The XGBoost model achieves an accuracy of 85.35% for detecting the presence of OA and accuracy of 72.46% for grading the severity of OA. The proposed deep learning achieves validation accuracy of 84.26% for detecting OA and 70-72% for grading the severity of OA. The independent analysis of the data with various models has given us promising results and the research has shown us the advantage of using several types of data thus enabling further early prediction and improving clinical outputs for patients with OA.