An externally validated fully automated deep learning algorithm to classify COVID-19 and other pneumonias on chest computed tomography

Akshayaa Vaidyanathan; Julien Guiot; Fadila Zerka; Flore Belmans; Ingrid Van Peufflik; Louis Deprez; Denis Danthine; Gregory Canivet; Philippe Lambin; Sean Walsh; Mariaelena Occhipinti; Paul Meunier; Wim Vos; Pierre Lovinfosse; Ralph T H Leijenaar

doi:10.1183/23120541.00579-2021

An externally validated fully automated deep learning algorithm to classify COVID-19 and other pneumonias on chest computed tomography

ERJ Open Res. 2022 May 3;8(2):00579-2021. doi: 10.1183/23120541.00579-2021. eCollection 2022 Apr.

Authors

Akshayaa Vaidyanathan^{1

2

3}, Julien Guiot^{4

3}, Fadila Zerka^{1

2}, Flore Belmans¹, Ingrid Van Peufflik¹, Louis Deprez⁵, Denis Danthine⁵, Gregory Canivet⁶, Philippe Lambin², Sean Walsh¹, Mariaelena Occhipinti¹, Paul Meunier⁵, Wim Vos¹, Pierre Lovinfosse⁷, Ralph T H Leijenaar¹

Affiliations

¹ Radiomics (Oncoradiomics SA), Liège, Belgium.
² The D-Lab, Depts of Precision Medicine, and Radiology and Nuclear Medicine, GROW-School for Oncology, Maastricht University, Maastricht, The Netherlands.
³ These authors have contributed equally to this work and share first authorship.
⁴ Dept of Pneumology, University Hospital of Liège, Liège, Belgium.
⁵ Dept of Radiology, University Hospital of Liège, Liège, Belgium.
⁶ Dept of Computer Applications, University Hospital of Liège, Liège, Belgium.
⁷ Dept of Nuclear Medicine and Oncological Imaging, University Hospital of Liège, Liège, Belgium.

Abstract

Purpose: In this study, we propose an artificial intelligence (AI) framework based on three-dimensional convolutional neural networks to classify computed tomography (CT) scans of patients with coronavirus disease 2019 (COVID-19), influenza/community-acquired pneumonia (CAP), and no infection, after automatic segmentation of the lungs and lung abnormalities.

Methods: The AI classification model is based on inflated three-dimensional Inception architecture and was trained and validated on retrospective data of CT images of 667 adult patients (no infection n=188, COVID-19 n=230, influenza/CAP n=249) and 210 adult patients (no infection n=70, COVID-19 n=70, influenza/CAP n=70), respectively. The model's performance was independently evaluated on an internal test set of 273 adult patients (no infection n=55, COVID-19 n= 94, influenza/CAP n=124) and an external validation set from a different centre (305 adult patients: COVID-19 n=169, no infection n=76, influenza/CAP n=60).

Results: The model showed excellent performance in the external validation set with area under the curve of 0.90, 0.92 and 0.92 for COVID-19, influenza/CAP and no infection, respectively. The selection of the input slices based on automatic segmentation of the abnormalities in the lung reduces analysis time (56 s per scan) and computational burden of the model. The Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) score of the proposed model is 47% (15 out of 32 TRIPOD items).

Conclusion: This AI solution provides rapid and accurate diagnosis in patients suspected of COVID-19 infection and influenza.