Diagnosis of Covid-19 in X-ray Images Based on Convolutional Neural Network (CNN)

Abstract

A number of studies focus on the early diagnosis of COVID-19 to reduce the spreading of this virus in the communities in order to support the health system and economy. This paper proposes a Convolutional Neural Network (CNN) model based on row X-ray chest images to detect the COVID-19 disease. In addition, an augmentation technique was employed on these images to increas the dataset and reduce the overfitting inside the CNN. This system bases on x-ray images of the chest. The proposed system contains three stages, the first stage is the pre-processing that starts by resizing the x-ray images into equal size (224 x 224), converting X-ray images into Grayscale images, and enhances the resulting image using Histogram Equalization(HE) technique. The second stage features extraction using CNN after applying augmentation on the dataset. The classification is the last stage for detecting the test sample only if is infected with Covid-19 or not, where the SoftMax function were used to classify patients. The results showed high accuracy in the classification process of the test images Furthermore, specificity, sensitivity, accuracy, and F1-score are used as criteria to estimate the classification efficiency of the proposed CNN model, where the accuracy of the model is 100% in the test dataset (220 X-ray images).