Prediction of Knee-Replacement using Deep-Learning Approach

Knee Osteoarthritis (OA) is seen in both young and older people. It is a musculoskeletal disorder. If diagnosed at an early stage of development, the lifestyle of the patient can be altered in order to slow down the progression of osteoarthritis by adapting measures such as weight loss and exercise. When diagnosed at later stage, the only option available is Knee Replacement (KR). Only about 66.67% of patients have the possibility of operation being successful and problems can occur later that may require treatment. A representation is needed to recognize people having more possibility of KR, so the required treatment can be provided in order to slow down the progression of OA at early stages. This paper exhibits a deep learning predictive model which uses X-ray pictures with analytical along with statistical particulars to foretell KR having AUC 0.86 (p < 0.05).In addition, this predictive model is developed using both Xception and Inception models of CNN, where the graphs for Training and validation accuracy for both Training and Testing datasets and also the confusion Matrix for both the models used in CNN are compared, thereby identifying accurate KR predicting models for clinical utility.