The researchers note that scans don’t always show lung diseases when a patient first presents symptoms and lab tests can take days to come back. The AI helps address both of those problems.
“The high sensitivity of our AI model can provide a ‘second opinion’ to physicians in cases where CT is either negative (in the early course of infection) or shows nonspecific findings, which can be common,” Fayad said in a press release. “It’s something that should be considered on a wider scale, especially in the United States, where currently we have more spare capacity for CT scanning than in labs for genetic tests.”
The researchers trained the algorithm on over 900 scans from medical centers in China. The scans included 419 confirmed COVID-19 cases and 486 negative cases. Researchers also had access to clinical information, like blood test results showing abnormalities in white blood cell counts or lymphocyte counts.
The algorithm they created mimics a workflow a physician would use to diagnose COVID-19. It produces separate probabilities of being COVID-19 positive based on CT images, clinical data and both combined. Next, the researchers hope to find clues about how well patients will do based on subtleties in their CT data and clinical information.
“This is an early proof concept that we can apply to our own patient data to further develop algorithms that are more specific to our region and diverse populations,” said Matthew Levin, director of the Mount Sinai Health System’s Clinical Data Science Team.
While CT scans aren’t widely used to diagnose COVID-19 in the US, the team believes they have the potential to play an important role. Eventually, the AI could be used in hospitals around the world.