Diagnosing and localizing Covid-19 in High resolution CT(HRCT) scans using Deep learning

Abstract

With the break-out of covid-19 as a world-wide pandemic that has a higher spread rate, it became a need to find a solution that would work in the favor of the patient as well as the radiologist. Since 2020, there have been many attempts to cater for the problem. Many researchers proposed detection and classification models in an attempt to automate some parts of the diagnostics process. The common methods found in the reported literature includes using models like VGG16, FCNN, Unet, ResUnet, Inception net and Alex net for the tasks of detection and classification of covid-19 benign or malignant. This thesis aims to explore the possibility of detecting and localizing covid-19. The covid lesions were segmented and then detected using Attention Res-Unet. The lungs were segmented into the major lobes using Unet and then an attempt was made to localize the detected lesions with respect to segmented Lung Lobes.