Deep learning of chest X-rays can predict mechanical ventilation outcome in ICU-admitted COVID-19 patients.

Autor: Gourdeau D; CERVO Brain Research Center, Québec, Québec, Canada. daniel.gourdeau.1@ulaval.ca.; Physics Department, Université Laval, Québec, Québec, Canada. daniel.gourdeau.1@ulaval.ca., Potvin O; CERVO Brain Research Center, Québec, Québec, Canada., Biem JH; Department of Radiology and Nuclear Medicine, Université Laval, Québec, Québec, Canada., Cloutier F; Department of Family and Emergency Medicine, Université Laval, Québec, Québec, Canada.; Centre de recherche intégrée pour un système apprenant en santé et services sociaux, Lévis, Québec, Canada., Abrougui L; Centre de recherche intégrée pour un système apprenant en santé et services sociaux, Lévis, Québec, Canada., Archambault P; Department of Family and Emergency Medicine, Université Laval, Québec, Québec, Canada.; Centre de recherche intégrée pour un système apprenant en santé et services sociaux, Lévis, Québec, Canada.; Centre de recherche sur les soins et les services de première ligne de l'Université Laval, Québec, Québec, Canada., Chartrand-Lefebvre C; Centre hospitalier de l'Université de Montréal, Montréal, Canada., Dieumegarde L; CERVO Brain Research Center, Québec, Québec, Canada., Gagné C; Electrical and Computer Engineering Department, Université Laval, Québec, Canada., Gagnon L; Department of Radiology and Nuclear Medicine, Université Laval, Québec, Québec, Canada., Giguère R; Centre de recherche intégrée pour un système apprenant en santé et services sociaux, Lévis, Québec, Canada., Hains A; Electrical and Computer Engineering Department, Université Laval, Québec, Canada., Le H; Jewish General Hospital, Montréal, Canada.; Department of Diagnostic Radiology, McGill University, Montréal, Canada., Lemieux S; Department of Radiology and Nuclear Medicine, Université Laval, Québec, Québec, Canada.; Institut universitaire de cardiologie et de pneumologie de Québec, Québec, Canada., Lévesque MH; Department of Radiology and Nuclear Medicine, Université Laval, Québec, Québec, Canada.; Institut universitaire de cardiologie et de pneumologie de Québec, Québec, Canada., Nepveu S; Centre hospitalier de l'Université de Montréal, Montréal, Canada., Rosenbloom L; Jewish General Hospital, Montréal, Canada.; Department of Diagnostic Radiology, McGill University, Montréal, Canada., Tang A; Centre hospitalier de l'Université de Montréal, Montréal, Canada., Yang I; Jewish General Hospital, Montréal, Canada.; Department of Diagnostic Radiology, McGill University, Montréal, Canada., Duchesne N; Department of Radiology and Nuclear Medicine, Université Laval, Québec, Québec, Canada.; Public Health Directory, Centre intégré universitaire santé et services sociaux de la Capitale Nationale, Québec, Québec, Canada., Duchesne S; CERVO Brain Research Center, Québec, Québec, Canada.; Department of Radiology and Nuclear Medicine, Université Laval, Québec, Québec, Canada.
Jazyk: angličtina
Zdroj: Scientific reports [Sci Rep] 2022 Apr 13; Vol. 12 (1), pp. 6193. Date of Electronic Publication: 2022 Apr 13.
DOI: 10.1038/s41598-022-10136-9
Abstrakt: The COVID-19 pandemic repeatedly overwhelms healthcare systems capacity and forced the development and implementation of triage guidelines in ICU for scarce resources (e.g. mechanical ventilation). These guidelines were often based on known risk factors for COVID-19. It is proposed that image data, specifically bedside computed X-ray (CXR), provide additional predictive information on mortality following mechanical ventilation that can be incorporated in the guidelines. Deep transfer learning was used to extract convolutional features from a systematically collected, multi-institutional dataset of COVID-19 ICU patients. A model predicting outcome of mechanical ventilation (remission or mortality) was trained on the extracted features and compared to a model based on known, aggregated risk factors. The model reached a 0.702 area under the curve (95% CI 0.707-0.694) at predicting mechanical ventilation outcome from pre-intubation CXRs, higher than the risk factor model. Combining imaging data and risk factors increased model performance to 0.743 AUC (95% CI 0.746-0.732). Additionally, a post-hoc analysis showed an increase performance on high-quality than low-quality CXRs, suggesting that using only high-quality images would result in an even stronger model.
(© 2022. The Author(s).)
Databáze: MEDLINE
Externí odkaz:
Nepřihlášeným uživatelům se plný text nezobrazuje