Johns Hopkins radiologists have repurposed a deep learning algorithm designed to detect tuberculosis on chest x-rays to, instead, help identify COVID-19.
The only caveat is that the AI model would likely come back with a COVID-positive diagnosis when faced with lung findings common to TB but rare in COVID, the authors note.
Still, the algorithm correctly classified 78 of 88 COVID-19 chest x-rays as positive. That’s a success rate of 89%, and the positive cases were confirmed in lab tests.
The proof-of-concept study further suggests that visual heatmaps generated by the model could be useful in several scenarios.
“Diagnostic tools to aid nonexpert radiologists may become particularly relevant as the pandemic overwhelms hospitals in both the developed world, where health care staffing is decreasing by mandates for social distancing and quarantining of workers who contract COVID-19, and the developing world, where few dedicated radiologists work at a baseline,” write lead author Paul Yi, MD, and colleagues.
Additionally, they note, a deep learning model such as theirs could be used as a triage tool to quickly isolate likely COVID-positive patients in ER waiting areas.
In the Hub coverage, Yi says the idea for the repurposed model was sparked by the newness of the novel coronavirus.
“Our goal was to demonstrate the ability of a deep learning model that had never ‘seen’ a case of COVID-19 to identify these cases,” he says. “Because COVID-19 is a new infection, large datasets are not currently available to train deep learning models. We hypothesized that images of other infections with similar appearances to COVID-19 could be used to train models capable of identifying this new disease.”