City Research Online

Lightweight Model for the Prediction of COVID-19 Through the Detection and Segmentation of Lesions in Chest CT Scans

Ter-Sarkisov, A. ORCID: 0000-0002-1300-6132 (2021). Lightweight Model for the Prediction of COVID-19 Through the Detection and Segmentation of Lesions in Chest CT Scans. International Journal of Automation, Artificial Intelligence and Machine Learning, 2(1), pp. 1-15.

Abstract

We introduce a lightweight model that segments areas with the Ground Glass Opacity and Consolidation and predicts COVID-19 from chest CT scans. The model uses truncated ResNet18 and ResNet34 as a backbone net, and Mask R-CNN functionality for lesion segmentation. Without any class balancing and data manipulations, and using only a small fraction of the training data, COVID-CT-Mask-Net classification model with 6.12M total and 600K trainable parameters, achieves 91.35% COVID-19 sensitivity, 91.63% Common Pneumonia sensitivity, 96.98% true negative rate and 93.95% overall accuracy on COVIDx-CT dataset (21191 images). The full source code, models and pre-trained weights are available on https://github.com/AlexTS1980/COVID-CT-Mask-Net.

Publication Type: Article
Additional Information: © 2021 by the authors; licensee Research Lake International Inc., Canada. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution License.
Publisher Keywords: Convolutional neural networks, COVID-19, Lesion segmentation, Lesion detection
Subjects: H Social Sciences > HM Sociology
Q Science > QA Mathematics > QA75 Electronic computers. Computer science
R Medicine > RA Public aspects of medicine > RA0421 Public health. Hygiene. Preventive Medicine
Departments: School of Mathematics, Computer Science & Engineering > Computer Science
[img]
Preview
Text - Published Version
Available under License Creative Commons: Attribution International Public License 4.0.

Download (564kB) | Preview

Export

Downloads

Downloads per month over past year

View more statistics

Actions (login required)

Admin Login Admin Login