Filenews 14 December 2021
A new high-tech system can help doctors do the best possible management of beds in COVID-19 Intensive Care Units, which are valuable and scarce in the midst of the pandemic. This is because it predicts with great accuracy which patients with the novel coronavirus will need admission to ICU.
More than 200 clinical parameters
The system, developed by researchers from the University of Waterloo and the start-up Company DarwinAI, uses artificial morbidity to predict the need for admission to an ICU based on more than 200 clinical parameters including the results of blood tests, vital functions and the medical history of each patient. "This is a very important step in terms of the clinical decisions that doctors have to make regarding the triage of patients and the development of therapeutic plans for each of them," said Alexander Wong, professor of Systems Design Engineering at the University of Waterloo.
How the AIS software was "trained"
The researchers' new AI software was "trained" based on data on about 400 cases of patients treated at the Sirio-Libanes Hospital in Sau Paulo, Brazil, by the hospital's doctors who were asked to decide which COVID-19 patients should be admitted to ICU.
Accuracy over 95%
Based on the data and training received, the neural network was able to predict the need to enter an ICU of a new case of COVID-19, with an accuracy of more than 95%. The system could also identify the key factors that guide its predictions in order to help doctors understand the... his way of thinking and trust him.
New useful tool
The new technology is in no way intended to replace doctors but to offer them a new useful tool to make faster and better decisions about the need to introduce patients to ICU. "Our goal is to help doctors make faster and more correct decisions based on previous patient cases and the outcome they had," noted Dr. Wong, who is also director of the VIP (Vision and Image Processing Lab) at the University of Waterloo. "We want to empower specialists to make the best use of medical resources and provide personalized treatment to patients."
Free access to all scientists
Researchers have made the technology they developed available to all scientists around the world to help optimize it. In fact, they are currently integrating it into a larger system aimed at strengthening better clinical decisions and developed as part of their ongoing COVID-Net Open Source Initiative – this initiative also helps doctors detect COVID-19 cases and determine the severity of their disease using AI analysis of imaging tests.
It is noted that the presentation of the new system is expected to take place on December 10th during the world's largest conference on artificial intelligence (2021 Conference on Neural Information Processing Systems).
