Convolutional Neural Network Based Feature Hybridization For Ct Scan Images: Covid-19 Detection Perspective

ABSTRACT

Coronavirus disease (COVID-19) is an infectious disease caused by the SARS-CoV-2 virus. Early diagnosis is only the proactive process to resist against the unwanted death. However, machine vision-based diagnosis systems show unparalleled success with higher accuracy and low false diagnosis rate. Working with the proposed method, this research has found that Computed Tomography (CT) provides more satisfactory outcomes regarding all the performance metrics. The proposed method uses a feature hybridization technique of concatenating the textural features with neural features. The literature review suggests that medical experts recommended chest CT in covid diagnosis rather than chest X-ray as well as RT-PCR. It is found that chest CT is more effective in diagnosis for being low false-negative rate. Moreover, the proposed method has used segmentation technique to dig the potential region of interest and obtain accurate features. Compared with different CNN classifier, such as, VGG-16, AlexNet, VGG-19 or ResNet50 and scratch model also. To obtain the satisfactory performance VGG-19 was used in this study. The Proposed machine learning based fusion technique achieves superior performance according to COVID-19 positive or negative with the accuracy of 98.63%, specificity of 99.08% and sensitivity of 98.18%. © 2022 IEEE.