Value and prognostic impact of a deep learning segmentation model of COVID-19 lung lesions on low-dose chest CT

Axel Bartoli; Joris Fournel; Arnaud Maurin; Baptiste Marchi; Paul Habert; Maxime Castelli; Jean-Yves Gaubert; Sebastien Cortaredona; Jean-Christophe Lagier; Matthieu Million; Didier Raoult; Badih Ghattas; Alexis Jacquier

doi:10.1016/j.redii.2022.100003

Value and prognostic impact of a deep learning segmentation model of COVID-19 lung lesions on low-dose chest CT

Res Diagn Interv Imaging. 2022 Mar:1:100003. doi: 10.1016/j.redii.2022.100003. Epub 2022 Mar 22.

Authors

Axel Bartoli^{1

2}, Joris Fournel^{1

2}, Arnaud Maurin¹, Baptiste Marchi¹, Paul Habert^{1

3

4}, Maxime Castelli¹, Jean-Yves Gaubert^{1

3

4}, Sebastien Cortaredona^{5

6}, Jean-Christophe Lagier^{5

7}, Matthieu Million^{5

7}, Didier Raoult^{5

7}, Badih Ghattas⁸, Alexis Jacquier^{1

2}

Affiliations

¹ Department of Radiology, Hôpital de la Timone Adultes, AP-HM. 264, rue Saint-Pierre, 13385 Marseille Cedex 05, France.
² CRMBM - UMR CNRS 7339, Medical Faculty, Aix-Marseille University, 27, Boulevard Jean Moulin, 13385 Marseille Cedex 05, France.
³ LIEE, Medical Faculty, Aix-Marseille University, 27, Boulevard Jean Moulin, 13385 Marseille Cedex 05, France.
⁴ CERIMED, Medical Faculty, Aix-Marseille University, 27, Boulevard Jean Moulin, 13385 Marseille Cedex 05, France.
⁵ Institut Hospitalo-Universitaire Méditerannée Infection, 19-21 boulevard Jean Moulin, 13005, Marseille, France.
⁶ IRD, VITROME, Institut Hospitalo-Universitaire Méditerannée Infection, 19-21 boulevard Jean Moulin, 13005, Marseille, France.
⁷ IRD, MEPHI, Institut Hospitalo-Universitaire Méditerannée Infection, 19-21 boulevard Jean Moulin, 13005, Marseille, France.
⁸ I2M - UMR CNRS 7373, Aix-Marseille University. CNRS, Centrale Marseille, 13453 Marseille, France.

Abstract

Objectives: 1) To develop a deep learning (DL) pipeline allowing quantification of COVID-19 pulmonary lesions on low-dose computed tomography (LDCT). 2) To assess the prognostic value of DL-driven lesion quantification.

Methods: This monocentric retrospective study included training and test datasets taken from 144 and 30 patients, respectively. The reference was the manual segmentation of 3 labels: normal lung, ground-glass opacity(GGO) and consolidation(Cons). Model performance was evaluated with technical metrics, disease volume and extent. Intra- and interobserver agreement were recorded. The prognostic value of DL-driven disease extent was assessed in 1621 distinct patients using C-statistics. The end point was a combined outcome defined as death, hospitalization>10 days, intensive care unit hospitalization or oxygen therapy.

Results: The Dice coefficients for lesion (GGO+Cons) segmentations were 0.75±0.08, exceeding the values for human interobserver (0.70±0.08; 0.70±0.10) and intraobserver measures (0.72±0.09). DL-driven lesion quantification had a stronger correlation with the reference than inter- or intraobserver measures. After stepwise selection and adjustment for clinical characteristics, quantification significantly increased the prognostic accuracy of the model (0.82 vs. 0.90; p<0.0001).

Conclusions: A DL-driven model can provide reproducible and accurate segmentation of COVID-19 lesions on LDCT. Automatic lesion quantification has independent prognostic value for the identification of high-risk patients.

Keywords: ACE, angiotensin-converting enzyme; Artificial intelligence; BMI, body mass index; CNN, convolutional neural network; COVID-19; COVID-19, coronavirus disease 2019; CT-SS, chest tomography severity score; Cons, consolidation; DL, deep learning; DSC, Dice similarity coefficient; Deep learning; Diagnostic imaging; GGO, ground-glass opacity; ICU, intensive care unit; LDCT, low-dose computed tomography; MAE, mean absolute error; MVSF, mean volume similarity fraction; Multidetector computed tomography; ROC, receiver operating characteristic.