Using Deep Learning for Efficient Diagnoses of COVID-19, Viral Illnesses (Other than COVID-19), and Bacterial Illnesses

Abstract

According to the World Health Organization the COVID-19 pandemic has killed more than 3.3 million people worldwide. Efficiently and accurately diagnosing people with COVID-19 is essential to help slow down the spread of the virus. Although swab tests do exist, they are not easily accessible in underdeveloped areas, whereas Chest X-Ray scanning has been available before the pandemic. However, there is a lack of radiologists to analyze and diagnose illnesses from Chest X-Rays. This is why this research aimed to use deep learning for efficient and automated diagnoses of COVID-19, Viral (Other than COVID-19), and Bacterial illnesses via Chest X-Ray images. Since the deep learning models had to analyze images, a CNN (Convolutional Neural Network) was built. There were three different CNN architectures fine-tuned and trained on real-time patient data. Out of all three fine-tuned CNN models, the VGG-16 fine-tuned CNN model received the highest testing accuracy of 92.34% when tested on 977 images. This means that when given a new image, the model was able to correctly classify it in one of the four different classes 92.43% of the time. Further improvements will be made to this project in order to make it into an actual usable platform.