Deep Learning Classification Models For Detection Of Covid Patients

Abstract

This study demonstrates how to evaluate deep Transfer Learning effectiveness in developing a classifier for identifying COVID-19-positive patients using CT scan images. The study found that deep learning (DL) effectively finds COVID-19 cases. A COVID-19 detection technique with high sensitivity and effectiveness is needed to stop it from spreading. This article offered a hybrid approach of image regrouping using ResNet and Densenet using the COVID-19 chest X-ray pictures as the basis for its research. The principal chest X-ray pictures were used to segment the lung area and split it into small portions. The very small components of the lung region were then randomly reconstructed into a normal image. In addition, the regrouped pictures were sent to a deep residual encoder block in order to have their features extracted. In order to prevent the model from becoming too specific and to improve its capacity for generalisation, the training dataset is improved by using a data augmentation technique. We looked at a set of pre-trained TL models DenseNet and ResNet for Convolutional Neural Networks, which improved performance, after pre-processing the data with Contrast Stretching, Histogram Equalization, and Log Transformation.