Multi-operator Variant of Differential Evolution and its Application in Classification of COVID-19 CT-scan Images

Novel coronavirus disease (COVID-19), caused by the virus (SARS-CoV-2), has drastically impacted human beings' lives since early 2020. The virus is constantly changing, and with mutations, it becomes diverse and spreads more easily. Several automatic COVID-19 diag-nostic tools are proposed that emphasize feature extraction mechanism from radiographical images using modern deep learning technology. The general idea is to leverage smart solutions of pre-trained networks for deep-feature processing. However, all the extracted features may not essentially contribute to the performance of the COVID-19 diag-nostic model, and hence an optimal subset of features must be discov-ered. Motivated by this, we propose a novel feature selection method based on multi-operator differential evolution (MODE), which helps to acquire optimal feature-subset. To show the efficacy of the proposed algorithm, we focus on applying the COVID-19 classification model through medical imaging. Eight advanced pre-trained architectures have been selected for COVID-19 feature extraction from CT-scan medi-cal imaging. After that, the proposed feature selection technique based on MODE is applied. A customized SVM kernel is implemented that supports feature classification. The result analysis shows the perfor-mance of the existing COVID-19 designs with the proposed feature selection technique, MODE, integrated with a customized SVM kernel. It even beats the existing state-of-the-art frameworks carried forward for COVID-19 diagnosis. We have observed that MODE feature selec-tion is suitable for fast COVID-19 detection, having overall accuracy of 85.27%.