A Comparison among Distinct Deep Learning Techniques for Real-Time Testing of Covid-19 Infected Patient using Chest Radiography
Siddharth Gupta, Avnish Panwar, Devyani Rawat
Abstract
The people across the world in late 2019 came in contact with the harmful virus which affected many lives, destroyed the economy of many countries, and brings down the best medical facilities across the world. In the beginning, the etiology of this virus was not known but with the passing time WHO investigated and named it Coronavirus or SARS-COV2. Covid-19 is highly communicable and may spread from one human being to several others. To stop the transmission of Covid-19, patients suffering from this virus need to be separated from normal people. Several techniques such as RTPCR were opted for the analysis of Covid-19 infection. However, high cost and large time in revelation can be observed using RT-PCR. Radioscopic approaches including Computed Tomography (CT) and the chest x-ray scans were adopted to overcome the dis-advantages of RT-PCR. In the current work, different Deep Learning (DL) models were fed with Covid-19 positive x-ray scans, pneumonia x-ray scans, and healthy person chest x-ray scans for feature extraction. After, extracting the features these images were passed to different machine learning classifiers for the analysis of covid19 positive scans and healthful person chest x-ray scans. The obtained result concludes that the VGG19 model with the Logistic Regression (LR) classifier gives an accuracy of 96.8%.