Multi-classification of COVID-19 and pneumonia in chest X-ray images using convolutional neural network

In December 2019, Wuhan became the epicenter of the deadly novel coronavirus (COVID-19), which spread to every corner of the world. Many health systems have seen collapse due to the limited capacity of the system and an exponential increase of suspected COVID-19 cases. A dependable and cost-effective technique is sought to reduce the overloading of radiologists’ efforts to detect suspected cases early, which may reduce the patient’s death rate. For the detection of COVID-19, this paper has proposed a deep learning multi-classification model. The model is developed using feature extraction by a convolutional neural network (CNN) from chest X-ray images, which could detect COVID-19 and pneumonia correctly. The proposed model is then compared with a pre-trained model, Xception. The proposed model shows 95.53 % accuracy, 0.936 recall, 0.936 precision, 0.936 F1-score while the Xception pre-trained model shows 93.04 % accuracy, 0.936 recall, 0.906 precision, 0.910 F1 score.