CoDE : AIR - COVID-19 Dilemma Evaluation : Artificial Intelligence and Radiology

The current outbreak of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and subsequent pandemic of COVID-19 is imposing a substantial stress on healthcare systems around the world. When severe COVID-19 cases require admission to intensive care units for respiratory distress and hypoxaemia, non-invasive approaches have been shunned in order to limit infections, especially in medical personnel, leaving endotracheal intubation and mechanical ventilation as the main remaining option.

A first problem faced by physicians is to decide on the need for invasive ventilation. This assessment is complicated by the aggressive nature of COVID-19, whereby earlier ventilation may increase survival rates and therefore patients who might have been only monitored should rather be intubated right away. In some locations, intensive care unit have been overwhelmed, and faced scenarios where there were insufficient resources for the number of critically sick COVID-19 patients requiring ventilation.

The MSSS has released guidelines to direct the triage operation in intensive care units in the case of surging demand and under-capacity in ventilation resources. These guidelines use clinical scores to determine the need of ventilation, however no score exist to predict the outcome of the ventilation. This information would be useful to further refine the triage decision, as well as guide transfer decisions from one region to another.

We hypothesize that radiomics features extracted from chest X-ray taken at the patient’s bedside while in intensive care unit, when combined with clinical and laboratory data, would be predictive of the need for, and the outcome of, invasive ventilation in the context of COVID-19. The objective of this project is to build an algorithm to predict which patients will require ventilatory support within 24 to 48 hours after admission to the intensive care unit, what the response to ventilation will be and its eventual outcome, as well as the time required to achieve this outcome. These tools can assist caregivers in the delicate but necessary choice of medical resources.

This study is conducted by the research team of Professor Simon Duchesne of the Faculty of Medecine at Université Laval, together with a number of co-investigators at the seven biggest hospital groups in the province of Quebec.