Facial reconstruction

Search LJMU Research Online

Browse Repository | Browse E-Theses

Diagnosis of Covid-19 through Imaging Modalities using Deep Learning

Rehman, A, Sadad, T, Saba, T and Kolivand, H Diagnosis of Covid-19 through Imaging Modalities using Deep Learning. Journal of Ambient Intelligence and Humanized Computing. ISSN 1868-5137 (Accepted)

[img] Text
Diagnosis of Covid-19 through Imaging Modalities using Deep Learning.pdf - Accepted Version
Restricted to Repository staff only

Download (603kB)

Abstract

A novel coronavirus now known as COV19 has emerged in the latest SARS outbreak. In Wuhan, China, the pandemic was first reported, but then disperse throughout the world very quickly. Computed tomography (CT) or X-ray has become a useful tool for precise diagnostics. Therefore, it is necessary to acquire an effective computer-aided diagnosis (CAD) system to facilitate doctors in diagnosing infected cases of COVID-19 via image modality. We employed an open-source dataset containing 2482 CT images in which 1252 CT scans belong to COVID-19 infection and 1230 CT images belong to non-COVID-19 (dataset is accessible www.kaggle.com/plameneduardo/sarscov2-ctscan-dataset). A hybrid based deep learning model is proposed to diagnosis the patients with COVID-19 or not. In the proposed technique, four pre-trained Deep Convolutional Neural Network (DCNN) models were applied using transfer learning: ResNet50, DenseNet201, MobileNet V2 and Inception V3 and accomplished 95.98% through DensNet201 architecture. However, the model achieved good results, which could reduce the burden on radiologists and healthcare centers. Still, diverse and more extensive datasets are required to assess the techniques in a practical situation.

Item Type: Article
Uncontrolled Keywords: 0801 Artificial Intelligence and Image Processing, 0805 Distributed Computing
Subjects: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Q Science > QR Microbiology > QR355 Virology
R Medicine > RA Public aspects of medicine
Divisions: Computer Science & Mathematics
Publisher: Springer
Date Deposited: 01 Apr 2021 09:39
Last Modified: 01 Apr 2021 09:39
URI: https://researchonline.ljmu.ac.uk/id/eprint/14734

Actions (login required)

View Item View Item