Vancouver experts are leading an international study to help predict the likelihood that a person has COVID-19, using CT scans, chest x-rays and AI technology to crunch the data.
The study is being lead by radiologists at Vancouver General Hospital (VGH) and University of British Columbia’s Department of Radiology.
Since January, teams have been collecting, labelling and coding CT and X-ray scans from around the globe, including data from the Middle East, Italy, South Korea, and Canada. More than 200 hours have already gone into the process, with a data set that represents the world’s largest global review of COVID CT scans and X-rays.
The data set is being used to create an AI-powered model that can predict the presence of COVID-19 on CT scans to assess the risk of lung disease severity leading to ICU admission, ventilation, lung fibrosis and death. The model will integrate clinical data to help support and supplement existing tools to improve patient care.
For example, it could help physicians determine whether individuals are best treated at home or whether they may require hospitalization and ventilation.
“The model will also assist in detecting similarities and differences in variations of patterns across different cultural and ethnic groups, and help us understand early and late stages of patterns of disease,” says Dr. Kendall Ho, VGH emergency physician and Academic Director, UBC Cloud Innovation Centre. It could also help flag those who may ultimately develop permanent lung damage/fibrosis.
“We know the lungs of COVID-19 patients are white and hazy, like a white-out or blizzard,” says Dr. Savvas Nicolaou, Director of Emergency and Trauma Radiology at VGH. “Currently, we can’t predict disease severity and its clinical impact in different patient populations. We’re confident this new tool will help us do that.” Dr. Nicolaou is leading the project with Dr. William Parker, a radiology resident at VGH/UBC.
Once developed, the new AI model will be piloted at VGH with an aim to embedding it in routine diagnostic procedures to improve the accuracy of COVID-19 diagnostics.
“We’ve seen patients present in the emergency department with non-typical symptoms such as severe abdominal pain, stroke and acute chest pain, and upon reviewing their CT scans for those conditions, we see the tell-tale haziness of COVID-19 in their lungs,” says Dr. Nicolaou.
This project is partly funded by the UBC Community Health and Wellbeing Cloud Innovation Center (UBC-CIC), which is powered by Amazon Web Services (AWS) and funding from the AWS Diagnostic Development Initiative (DDI).
Opened in January, the CIC is the first of its kind in Canada, and one of seven other AWS CICs around the world.