AI-based analysis of CT images for rapid triage of COVID-19 patients

Qinmei Xu,Changsheng Zhou,Guangming Lu,Zhen Zhou,Xiuli Li,Xianghao Zhan,Peiyi Xie,Longjiang Zhang,Yizhou Yu,Yiheng Li,Olivier Gevaert,Shu Zhang

doi:10.1038/s41746-021-00446-z

Abstract

The COVID-19 pandemic overwhelms the medical resources in the stressed intensive care unit (ICU) capacity and the shortage of mechanical ventilation (MV). We performed CT-based analysis combined with electronic health records and clinical laboratory results on Cohort 1 (n = 1662 from 17 hospitals) with prognostic estimation for the rapid stratification of PCR confirmed COVID-19 patients. These models, validated on Cohort 2 (n = 700) and Cohort 3 (n = 662) constructed from nine external hospitals, achieved satisfying performance for predicting ICU, MV, and death of COVID-19 patients (AUROC 0.916, 0.919, and 0.853), even on events happened two days later after admission (AUROC 0.919, 0.943, and 0.856). Both clinical and image features showed complementary roles in prediction and provided accurate estimates to the time of progression (p < 0.001). Our findings are valuable for optimizing the use of medical resources in the COVID-19 pandemic. The models are available here: https://github.com/terryli710/COVID_19_Rapid_Triage_Risk_Predictor.

Highlights

From 30 December to 11 October, the ongoing severe acute respiratory syndrome–coronavirus 2 (SARS-CoV-2) pandemic has caused over 37 million coronavirus disease 2019 (COVID-19) confirmed cases and 1 million deaths globally[1]
We provided risk stratification based on computed tomography (CT)-based radiomics features and clinical data for COVID-19 patients in terms of stable or severe disease on admission
We built a specific subset of Cohort 2 (Cohort 3, n = 662) for patients from the nine medical centers whose time intervals between admission and progression to critical outcomes (ICU/mechanical ventilation (MV)/death) were more than two days, aiming to evaluate the performance of our models on predicting events happening at least two days after admission

Summary

Introduction

From 30 December to 11 October, the ongoing severe acute respiratory syndrome–coronavirus 2 (SARS-CoV-2) pandemic has caused over 37 million coronavirus disease 2019 (COVID-19) confirmed cases and 1 million deaths globally[1]. To enable sufficient supply of medical resources, rapid triage method for COVID-effected patients with potentially serious outcomes has become an urgent priority for reallocating medical resources as well as distributing patients to balance ICU loads across affected regions so as to deliver timely treatment[5,6,7,8]. Evaluating the severity of patients with infectious pneumonia has been applied in clinics such as measuring the acute physiology and chronic health evaluation II (APACHE-II) score and laboratory indicators including neutrophil-to-lymphocyte ratio (NLR)[9,10,11,12]. The scoring systems of APACHE-II are highly subjective and time-consuming while laboratory indicators are not comprehensive enough to predict the adverse outcomes of the newly emerged COVID-19. Better ways to utilize multi-modal data for grouping hospitalized COVID-19 patients according to their potential clinical outcomes remain to be developed to deliver specific treatment timely

Methods

Results

Conclusion