DarwinAI developed the COVID-19 program with the University of Waterloo to improve COVID-19 detection and risk stratification on chest radiography. In March, a Canadian team published research showing the COVID-Net deep convolutional neural network achieved promising early results, including 100% sensitivity, 80% positive predictive value, and 83% accuracy for identifying COVID-19 on a small test set of chest radiography studies.
Now, Red Hat and DarwinAI are gearing the tool up for clinical and research use with the help of underlying technology from a computational research group at the Boston Children's Hospital. The collaboration aims to make COVID-Net easier for clinicians to use, including through a web-based graphical user interface designed to work with Boston Children's open-source ChRIS framework.
Rudolph Pienaar, PhD, the lead ChRIS technical architect and assistant professor in radiology at Harvard Medical School, said COVID-Net will help screen many cases at a large scale to focus on applying healthcare resources where they are most needed.
Copyright © 2020 AuntMinnie.com