A Deep Convolutional Neural Network Approach to Classify COVID-19 and Pneumonia From Chest X-ray Images

Research Supervisor
Journal Title
Journal Title
Journal ISSN
Volume Title
COVID-19 pandemic has frightened all the peoples of the world. It is a contagious disease caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), originated from Wuhan in December 2019 and spread quickly all over the world. The number of infections and deaths due to this disease are increasing day by day. Not only humans, but the global education system and economy of many countries are affected due to this viral disease. To combat against COVID-19 urgent diagnosis is important. However, Convolutional Neural Network (CNN) can be used as an alternative to diagnosing COVID-19 and reduce burden on doctors. COVID-19 detection from chest X-ray is suitable considering all aspects in comparison to RT-PCR and CT scan. In this paper we propose a deep CNN based approach to detect COVID+, pneumonia and normal cases, analyzing chest X-ray images. Several deep CNN models- VGG16, InceptionV3, DenseNet121, DenseNet201 and InceptionResNetV2 have been adopted in this proposed work. They have been trained individually to make particular predictions. After training the models we found DenseNet121 provides efficient results with an accuracy of 96.65%. The accuracy is evaluated in terms of precision, recall and F-1 score. All of the studies were carried out using a publicly available CXR image dataset. The final model has been deployed in a GUI based application for publically use. Users having an android device can upload their chest X-ray image and the system will return the result in a few seconds analyzing the particular image with reliable accuracy.
MEDICINE::Social medicine::Public health medicine research areas::Public health science
Department Name
Electrical and Computer Engineering
North South University
Printed Thesis