Detect COVID-19 from Chest X-Ray images using Deep Learning

Abstract

The COVID-19 pandemic has been causing devastating impacts on the well-being of people around the world as well as the global economy. Motivated by the effort of the open source community on collecting the COVID-19 dataset and the success of Deep Learning on previous studies with chest radiography, this thesis builds a Deep Convolutional Neural Network in order to detect COVID-19 using only chest X-Ray images. This project uses modern Deep Learning techniques such as using pretrained networks and fine-tuning as well as regularizations such as data augmentation and dropout to fight overfitting. The re-sulting model achieves an overall accuracy of 93% on the most realistic task of detecting COVID-19 patients among healthy normal people despite being trained on a dataset of only 115 images for each class. Out of 100 patients who do have COVID-19, the model accurately identifies 96 patients and misses out 4 patients. Out of 100 normal patients who do not have COVID-19, the model accurately identifies 91 patients as healthy and misclassifies 9 patients as COVID-19 positive. Overall, the model did a decent job as a COVID-19 detector but still has limitations and is far from being production-ready.