Classification of COVID-19 infection using Deep learning and Radiomic features extracted from Computed Tomography Scans of Patients' lungs.

Author/Creator

Author/Creator ORCID

Date

2020-01-01

Department

Computer Science and Electrical Engineering

Program

Computer Science

Citation of Original Publication

Rights

Access limited to the UMBC community. Item may possibly be obtained via Interlibrary Loan through a local library, pending author/copyright holder's permission.
This item may be protected under Title 17 of the U.S. Copyright Law. It is made available by UMBC for non-commercial research and education. For permission to publish or reproduce, please see http://aok.lib.umbc.edu/specoll/repro.php or contact Special Collections at speccoll(at)umbc.edu

Subjects

Abstract

COVID-19 is an air-borne viral infection, which infects the respiratory system of ahuman body and it became a global pandemic in March, 2020. Nearly forty million people have been affected by this infection till date, and the numbers are increasing each day. While the scientists and the doctors were trying to find a vaccine and a cure for this infection, this theses discussed the role of an engineer in analyzing the images of the virus infected on a human lung using CT scans. The radiomic COVID features were extracted by extracting the radiomic features of the lung image using its infection mask. Those features are used by Radiologists to quantitatively analyze the presence and severity of abnormalities in the lungs such as: Ground-glass Opacity, Consolidation, and Crazy-paving patterns. In this work, Deep learning models using 2DCNN and 3D CNN architectures were developed and trained for the extracting Deep Learning features. Then, both COVID Radiomic and Deep learning features are used for COVID19 classification by Random forest model. There are two types of patients' CT scans that were used. A list of 3888 COVID 2D images from38 patients' CT scans obtained from a public dataset and a collection of hospital dataset have been used. The experiment results show that the 3D model achieves a highly true positive rate with (Area Under the Receiver Operating Characteristics) AUC-ROC curve of 0.997.