A Deep Learning-based Application for COVID-19 Diagnosis on CT: The Imaging COVID-19 AI Initiative

Laurens Topff, José Sánchez-García, Rafael López-González, Ana Jiménez Pastor, Jacob J. Visser, Merel Huisman, Julien Guiot, Regina G. H. Beets-Tan, Angel Alberich-Bayarri, Almudena Fuster-Matanzo, Erik R. Ranschaert, on behalf of the Imaging COVID-19 AI initiative.

Abstract

Recently, artificial intelligence (AI)-based applications for chest imaging have emerged as potential tools to assist clinicians in the diagnosis and management of patients with coronavirus disease 2019 (COVID-19).

Introduction:

Coronavirus disease 2019 (COVID-19), caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has become a global health emergency since its appearance by the end of 2019. Dyspnoea, fever, dry cough, and myalgia are common manifestations of COVID-19. However, its clinical presentation is variable, ranging from asymptomatic to severe and potentially fatal. As a result, COVID-19 continues to present challenges in diagnosis and patient monitoring.

Materials and Methods:

Study design

The Imaging COVID-19 AI initiative was a large-scale collaborative effort to develop a generalisable deep-learning model for automatic classification and disease segmentation of chest CTs in COVID-19 suspected patients.

Patient selection

Outpatient or hospitalised patients (≥18 years old) with suspected or known COVID-19 who underwent chest CT in secondary or tertiary referral centres were included in the study.

Imaging data

For this study, a dataset including CT scans routinely acquired from December 2019 through July 2020 was created.

Discussion:

We developed and evaluated an automated deep learning-based application for the diagnosis of COVID-19 on chest CT images. In addition, the tool performed segmentation of infectious lung opacities, enabling the calculation of the extent of lung involvement, as well as the prediction of COVID-19 disease severity. As a result of the image analysis pipeline using both models, a complete and visual report can be delivered that can assist clinicians in the decision-making process of suspected and known COVID-19 patients.

Acknowledgments:

The authors are very grateful to the team of Robovision for providing and supporting the data annotation platform, and to the team of AContrario.Law for their legal work to allow the sharing of data. Without their support this project would not have been possible.

Citation: Topff L, Sánchez-García J, López-González R, Pastor AJ, Visser JJ, Huisman M, et al. (2023) A deep learning-based application for COVID-19 diagnosis on CT: The Imaging COVID-19 AI initiative. PLoS ONE 18(5): e0285121. https://doi.org/10.1371/journal.pone.0285121

Editor: Calogero Casà, Fatebenefratelli Isola Tiberina - Gemelli Isola, ITALY

Received: January 26, 2023; Accepted: April 15, 2023; Published: May 2, 2023.

Copyright: © 2023 Topff et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: The study involves human research participant data containing potentially sensitive patient information. Legal and ethical restrictions, imposed by the participating institutions and by the Institutional Review Board of the Netherlands Cancer Institute, prohibit public sharing of this data. Requests for data can be made by contacting the corresponding author or the Institutional Review Board of the Netherlands Cancer Institute (via IRB@nki.nl). The code is protected by Intellectual Property right laws and therefore cannot be shared publicly. Requests for code can be made by contacting Angel Alberich Bayarri, CEO of Quibim (via angel@quibim.com). All other non-sensitive data are available within the paper and its Supporting Information files.

Funding: The authors received no specific funding for this work.

Competing interests: I have read the journal’s policy and the authors of this manuscript have the following competing interests: Laurens Topff, Merel Huisman, Regina G.H. Beets-Tan: no disclosures; Jacob J. Visser: medical advisor Noaber Foundation, medical advisor NLC, medical advisor Contextflow GmbH, medical advisor Quibim S.L.; Julien Guiot: personal fees for advisory board, work and lectures from Boehringer Ingelheim, Janssens, GSK, Roche and Chiesi, non-financial support for meeting attendance from Chiesi, Roche, Boehringer Ingelheim and Janssens, permanent SAB of Radiomics (Oncoradiomics SA) for the SALMON trial without any specific consultancy fee for this work, co-inventor of one issued patent on radiomics licensed to Radiomics (Oncoradiomics SA); Erik R. Ranschaert: medical advisor Quibim S.L, medical advisor Robovision BV; Ángel Alberich-Bayarri: CEO and a shareholder of Quibim S.L.; José Sánchez-García, Rafael López-González, Ana Jiménez Pastor, Almudena Fuster-Matanzo: current or past employees of Quibim S.L.