Open Access

ARTICLE

Automatic Detection and Classification of Human Knee Osteoarthritis Using Convolutional Neural Networks

by Mohamed Yacin Sikkandar^1,*, S. Sabarunisha Begum², Abdulaziz A. Alkathiry³, Mashhor Shlwan N. Alotaibi¹, Md Dilsad Manzar⁴

1 Department of Medical Equipment Technology, College of Applied Medical Sciences, Majmaah University, Al Majmaah, 11952, Saudi Arabia
2 Department of Chemical Engineering, Sethu Institute of Technology, Kariapatti, 626115, Tamilnadu, India
3 Department of Physical Therapy, College of Applied Medical Sciences, Majmaah University, Al Majmaah, 11952, Saudi Arabia
4 Department of Nursing, College of Applied Medical Sciences, Majmaah University, Al Majmaah, 11952, Saudi Arabia

* Corresponding Author: Mohamed Yacin Sikkandar. Email:

Computers, Materials & Continua 2022, 70(3), 4279-4291. https://doi.org/10.32604/cmc.2022.020571

Received 29 May 2021; Accepted 15 July 2021; Issue published 11 October 2021

Abstract

Knee Osteoarthritis (KOA) is a degenerative knee joint disease caused by ‘wear and tear’ of ligaments between the femur and tibial bones. Clinically, KOA is classified into four grades ranging from 1 to 4 based on the degradation of the ligament in between these two bones and causes suffering from impaired movement. Identifying this space between bones through the anterior view of a knee X-ray image is solely subjective and challenging. Automatic classification of this process helps in the selection of suitable treatment processes and customized knee implants. In this research, a new automatic classification of KOA images based on unsupervised local center of mass (LCM) segmentation method and deep Siamese Convolutional Neural Network (CNN) is presented. First-order statistics and the GLCM matrix are used to extract KOA anatomical Features from segmented images. The network is trained on our clinical data with 75 iterations with automatic weight updates to improve its validation accuracy. The assessment performed on the LCM segmented KOA images shows that our network can efficiently detect knee osteoarthritis, achieving about 93.2% accuracy along with multi-class classification accuracy of 72.01% and quadratic weighted Kappa of 0.86.

---

Keywords

---