Researchers create Artificial Intelligence models to detect Covid-19 from X-Rays

Researchers from the University of Waterloo Vision and Image Processing Lab and the startup DarwinAI partnered to create COVID-Net - a convolutional neural network for COVID-19 detection from chest radiographies.

They started working on this project with the objective of using their expertise to develop an AI-based tool that could help in the screening and diagnosis of patients for Covid-19.

The COVID-Net tool was trained and tested through the COVIDx dataset – an open resource that hold more than 16.000 chest x-rays from over 13.000 patient cases. These test determined that COVID-Net was able to detect Covid-19 infections with a PPV (Positive Predictive Value) of 88,9%.

The project is on-going and the models provided are meant to be used as a reference and improved upon as new data is made available.

The COVID-19 project is completely open source, with the codes, documentation, dataset, and scientific papers available at the GitHub website.

Adapted from:https://syncedreview.com/2020/04/02/darwinai-open-sources-covid-net-as-m...

Project Page at Github for more information: https://github.com/lindawangg/COVID-Net