COVIDNET-CT: a new detection model for CT scans

Dear Colleague,

The response to our release of COVID-Net three months ago continues to inspire us.

Here are the latest updates with the project:

1.) Today we are proud to announce the release of COVIDNet-CT, a family of neural networks to detect COVID-19 by means of Computed Tomography (CT) scans. Over 100,000 CT images from ~1500 patients were leveraged to build the model. The source code, trained models, data preparation scripts, training/inference/evaluation scripts, and instructions can be accessed at this GitHub repository, which will be regularly updated with better models and additions to data.

In keeping with the spirit of this initiative, our hope is for COVIDNet-CT to serve as a reference baseline to the research community to advance deep learning as a tool for assisting clinicians and healthcare workers in their fight against the COVID-19 pandemic.

2.) As with the original COVID-Net, our GenSynth platform was integral in accelerating the development of COVIDNet-CT using a trusted human-machine collaborative design approach.

In addition to generating highly efficient and accurate models, this design methodology results in robust and trustworthy networks that have been validated by means of our Explainable AI (XAI) technology to ensure the right decisions are being made for the right reasons. The figure below, for example, illustrates the critical factors the neural network is leveraging during the decision-making process to detect COVID-19. This type of trust validation made possible by GenSynth constitutes a powerful way to build trustworthy models in a transparent manner.

A more comprehensive treatment of this design approach can be found here.

A special thanks to NVidia for providing computing resources to accelerate the development of COVIDNet-CT.

Finally, if you’re a researcher or clinician and would like access to our explainability platform to assist with this project and gain transparency or have data to share, please email us at info@darwinai.ca.

In Solidarity,

The DarwinAI team