Knee osteoarthritis classification using social wolf swarm-based deep convolutional neural network

Abstract

ABSTRACT Osteoarthritis is an agonising disease that affects millions of people every year and the occurrence rate is also increasing promptly. The prompting leads to irrelevant lifestyle among the individuals and early detection and diagnosis help to reduce the influence of osteoarthritis to a greater extent. In this research, knee osteoarthritis is identified using the optimised deep convolutional neural network (deep CNN) classifier from the necessary features extracted. The significant steps in this research are feature extraction using a hybrid pretrained model and classification using optimised deep CNN. The hybrid pretrained model is developed using the Visual Geometry Group (VGG), Resnet 101, and Alexnet architectures and the composed characteristics provide effortless training with reduced error and loss function. The deep CNN classifier is optimised using the social wolf swarm-based optimisation that effectively tunes the classifier by its fitness function and finally, the knee osteoarthritis is determined. The proposed method achieved an efficient accuracy rate of 98%, f1 measure of 94.03%, precision of 98%, and recall of 97.173%, which is more efficient.