Accurate diagnosis of COVID-19 based on deep neural networks and chest X-ray images

Abstract EN

The present study aims at preventing spread out of COVID-19 by early detection of infected cases using chest X-ray images and convolutional neural networks.

Covid-19 chest X-ray dataset were collected from public sources as well as through agreements with hospitals and physicians with the consent of their patients.

A deep learning algorithm based on convolutional neural networks (CNN) was implemented utilizing X-ray images to diagnose COVID-19.

ResNet50, short for Residual Networks, is a classic neural network that was used as a backbone for the classification task.

It accelerates the speed of training of the deep networks and reduces the effect of vanishing gradient problems.

Images were first resized and then pre-processed to increase sharpness, contrast, and clarity.

Images were fed into a deep neural network to predict the probability of COVID-19 infectious.

The deep learning calculation acquired an area under the curve (AUC) of the receiver operating characteristics (ROC) of 0.9888, 96.2 % sensitivity, 98 % accuracy, and 100 % specificity.

Moreover, the algorithm can be easily modified to add extra images (normal and COVID-19) to improve performance.

The proposed system introduces a great help to all nations to screen and diagnose COVID-19 as a faster alternative compared with conventional method that uses PCR