A new screening solution: COVID-Net

0

Over the past several months, PCR has become the gold standard for testing COVID-19 cases. But is it the best type of test?

“[PCR tests] have very high specificity, but people have recently found that sensitivity isn’t so great, depending on when you measure it,” Dr. Alexander Wong, Associate Professor at UW, and the Canada Research Chair in AI and Medical Imaging, said.

“A recent study also found that the longer somebody has had COVID-19, the less likely they are to be detected,” Dr. Wong said.“But the most important thing is that while PCR test tells you if someone has COVID-19 or not, it doesn’t tell you about the progression of the disease.”

As a potential solution, Dr. Wong has been collaborating with the COVID-Net project, an open source project aiming to improve detection via X-ray imaging.

“COVID-Net is an open source open access initiative where the underlying goal is to use deep learning in the case of COVID-19 screening,” Dr. Wong said. “It’s based on medical imaging, with chest X-rays and CT Scans being the primary ones, and we’re expanding beyond that. These help us determine the level of progression of the disease, so the doctors and the clinicians can have a better idea on how to handle it.” Open source projects are collaborative by nature, and COVID-Net is no exception.

“A lot of my students have become inspired by this initiative and joined our team. We’ve also had clinical institutes from around the world collaborating with us as they’re interested in leveraging AI to help with the pandemic. We’ve also received a lot of support from a lot of corporations who believe in the cause and they’ve provided many resources to undertake this project,” Dr. Wong said.

When asked about the progress in the past six months, Dr. Wong said, “We have actually released COVID X, which is the largest dataset of its kind with lung chest X-rays and people from all around the world have been leveraging it for their own experimentation, testing and development. We have also continuously released COVID-Net models for the world to use and expand upon -and they have. We recently released a COVID-Net CT model for detecting COVID-19 based on chest CT scans.”

“We are treating COVID-Net as a complimentary technology. For COVID-Net, the actual acquisition process will be done in minutes and if it passes through AI and it is able to actually provide results in seconds. This way, doctors and clinicians can figure out atleast how to take care of a patient while they’re waiting for the more accurate PCR test,” Dr. Wong explained.

“The future for COVID-Net is quite interesting. We continue to release regular updates to our datasets as well as models so that the entire world can use it and build upon our own experiences,” Dr. Wong said.

“We’re also building an open source frontend platform so doctors and clinicians who are not experts on deep learning can use the platforms directly and visualise, organise their images and so on and so forth.” Dr. Wong said.

Dr. Wong is looking for collaborators to join his research team and contribute to the advancement of COVID-Net.