Detection of COVID-19 from CT Lung Scans Using Transfer Learning

Sahil Lawton; Serestina Viriri

doi:10.1155/2021/5527923

Detection of COVID-19 from CT Lung Scans Using Transfer Learning

Comput Intell Neurosci. 2021 Apr 8:2021:5527923. doi: 10.1155/2021/5527923. eCollection 2021.

Authors

Sahil Lawton¹, Serestina Viriri¹

Affiliation

¹ School of Mathematics, Statistics and Computer Science University of KwaZulu-Natal, Durban, South Africa.

Abstract

This paper aims to investigate the use of transfer learning architectures in the detection of COVID-19 from CT lung scans. The study evaluates the performances of various transfer learning architectures, as well as the effects of the standard Histogram Equalization and Contrast Limited Adaptive Histogram Equalization. The findings of this study suggest that transfer learning-based frameworks are an alternative to the contemporary methods used to detect the presence of the virus in patients. The highest performing model, the VGG-19 implemented with the Contrast Limited Adaptive Histogram Equalization, on a SARS-CoV-2 dataset, achieved an accuracy and recall of 95.75% and 97.13%, respectively.

MeSH terms

Area Under Curve
COVID-19 / diagnostic imaging*
Datasets as Topic
Diagnosis, Computer-Assisted
Humans
Image Processing, Computer-Assisted / methods*
Lung / diagnostic imaging*
Lung / virology
Machine Learning*
ROC Curve
SARS-CoV-2*
Tomography, X-Ray Computed / methods*