Coronavirus Classification based on Enhanced X-ray Images and Deep Learning
Fallah H. Najjar, Safa Riyadh Waheed, Duha Amer Mahdi
Abstract
In light of the fact that the global pandemic of Coronavirus Disease 2019 (COVID-19) is still having a significant impact on the health of people all over the world, there is a growing need for testing diagnosis and treatment that can be completed quickly. The primary imaging modalities used in the respiratory disease diagnostic process are the Chest X-ray (CXR) and the computed tomography scan. In this context, this paper aims to design a new Convolutional Neural Network (CNN) to diagnose COVID-19 in patients based on CXR images and determine whether they are COVID or healthy. We have tested the performance of our CNN on the COVID-19 Radiography Database with three classes (COVID, Pneumonia, and Normal). Also, we proposed a new enhancement technique to enhance the CXR image using the Laplacian kernel with Delta Function and Contrast-Limited Adaptive Histogram Equalization. The proposed CNN has been trained and tested on 15153 enhanced and original images, COVID (3616), Pneumonia (1345), and Normal (10192). Our enhancement technique increased the performance metrics scores of the proposed CNN. Hence, the proposed method obtained better results than the state-of-the-art methods in accuracy, sensitivity, precision, specificity, and F measure.