The medical software maker Lunit is stepping up its commitment to help healthcare providers combat COVID-19 with the release of an AI solution designed for X-ray analysis. The firm announced that it is freely releasing its software online to assist with the detection of coronavirus-related pneumonia, such as a “consolidation” radiologic finding seen on X-rays of the lungs. This finding is usually present when areas of the lung are filled with fluid instead of air.
“We hope our decision to publicly release a special version of our software can help ease the strain in the healthcare system in countries around the world,” said Brandon Suh, CEO of Lunit. "Our AI solution has been in use at coronavirus care centers in South Korea, which is one of the countries that are successfully 'flattening the curve' by applying diverse and innovative measures against the virus."
While effective screening tools for COVID-19 do exist, they are in short supply in many parts of the world. Genetic tests, called the RT-PCR, are highly accurate but are harder to obtain and slower to provide results than the widely available X-ray and CT machines. AI-powered software can provide efficient, reliable analysis of scans so that clinicians can quickly diagnose and treat the patients that need it the most.
But not all that glitters is AI gold, cautions the senior editor for artificial intelligence at MIT Technology Review. In a recent article about the launch of COVID-NET, an AI tool trained to identify signs of coronavirus, the editor points out that many of these diagnostic tools have not been made fully available to the public, so it's impossible to know how effective they are.
On the other hand, COVID-NET was developed by Linda Wang and Alexander Wong at the University of Waterloo and the Canadian-based firm DarwinAI. The algorithm was taught to identify specific radiologic findings by analyzing 5,941 chest X-ray images of 2,839 patients with varying lung conditions. It's not a finished product, they say, and is released to the general public so researchers from around the world can make their adjustments to the tool.
The biggest problem facing those in the COVID-19 fighting business is that big data doesn't deliver due to how fragmented and disparate the sources. Unless an organization has deep experience building AI-enabled imaging software and an established network like Lunit does, many will need to patchwork together their partnerships to get access to that information. Also, the coronavirus is a new disease, and it can be challenging to identify defining characteristics of the condition on a chest X-ray.
Lunit’s software release is limited only to those radiologic findings that are linked to a COVID-19 presentation but can find red flags in images for other conditions like breast cancer, heart problems, and lung conditions. It was founded in 2013 and has raised $46.5 million from venture investors.
