Knee Osteoarthritis Detection Using Deep Feature Based on Convolutional Neural Network
Dilovan Asaad Zebari, Shereen Saleem Sadiq, Dawlat Mustafa Sulaiman
Abstract
Various diseases have recently wreaked influence on people's way of life. Several bone illnesses significantly influence the quality of life, including Knee Osteoarthritis. When the cartilage in the knee joint between the femur and tibia wears down, it causes Knee Osteoarthritis, which results in significant j oint pain, joint movement restrictions, gait abnormalities, and even effusion. This study presents a method based on deep features. We employed Convolutional Neural Network to extract deep features from Knee Osteoarthritis images. Then, the extracted features are fed to different machine learning classifiers, namely Support Vector Machine, K-Nearest Neighbour, and Naive Bayes. The classification of this work has been performed to differentiate between healthy and unhealthy Knee Osteoarthritis images. The experimental result uses different evaluation matrices to test this work by obtaining 90.01 % of accuracy, 90% recall, and 87.8% specificity. The obtained results showed that the K-Nearest Neighbour based deep features achieved better classification accuracy compared to Support Vector Machine and Naive Bayes.