AI models developed in Waterloo can identify COVID-19 patients at gravest risk

WATERLOO New artificial intelligence models developed at the University of Waterloo can help doctors prioritize care by predicting which COVID-19 patients are at greatest risk of death or injury.

The computer software identifies vulnerable patients by learning from the outcomes of previous cases of COVID-19, to allow doctors to quickly intervene and treat patients before conditions worsen.

“Hospitals are already extremely overburdened by the pandemic,” said Alexander Wong, a professor of systems design engineering. “Having AI models to help health-care workers identify who needs care in an efficient and effective manner can significantly reduce the burden, as well as the costs of health care.”

But it doesn’t just give predictions. “It’s able to also explain to you why and how it made that decision,” by explaining which indicators were relied on, to allow doctors to confidently make decisions based on the findings, Wong said.

“Physicians don’t want to work with a black box,” said Wong, a Canada Research Chair in Artificial Intelligence and Medical Imaging.

Instead, explainable AI makes it “trustworthy and transparent.”

The goal is to have human and machine work together, learning from each other to improve decisions. The AI is continuously honed through feedback.

“In our lab, we always view AI as a collaborator, not just a pure tool,” Wong said.

The models analyze clinical and biochemical markers such as heart rate and blood pressure, blood ferritin levels and use of therapeutic heparin to automatically discover patterns that can predict the likelihood of a patient dying or suffering kidney damage. COVID-19 patients have been found to develop kidney injuries.

The research team worked with a Montreal emergency doctor and plans to test the models in clinical settings to improve their accuracy, as well as talk to more doctors and run expanded studies with more data.

The new models are part of an open-source project called COVID-Net that has produced other innovations since the pandemic’s start. The work and results, as well as previous research in the project, are available to researchers and scientists around the world.

The researchers are already looking into adapting the models for other conditions and diseases beyond COVID-19.

Hossein Aboutalebi and Maya Pavlova, both students in the Vision and Image Processing Lab, contributed to the project along with doctoral student Andrew Hryniowski and Mohammad Javad Shafiee, also a systems design engineering professor at Waterloo.

Having AI models to help healthcare workers identify who needs care in an efficient and effective manner can significantly reduce the burden, as well as the costs of health care. ALEXANDER WONG PROFESSOR OF SYSTEMS DESIGN ENGINEERING