“It may be that you really can’t do that with, essentially, a line that separates the two,” Siegel said. “You may need something much more complicated and by just looking at the observations in the National Lung Screening Trial data set, you could come up with something that is a much more complicated relationship that a machine learning model would be able to predict.”

FDA regulatory issues will also be a huge challenge for AI, though Siegel shared some promising news: that the FDA is planning to launch a pilot software pre-certification program, similar to TSA Pre-Check, to make approvals faster.


“With TSA Pre-Check, they get to know you, they check your background once and then once they get to know you they end up passing you through a lot more quickly than everybody else,” Siegel said. “The idea with the FDA, they get used to certain vendors, big or small, look at their operational excellence, culture of quality, commitment to monitoring real world performance, et cetera, and then once they know the vendor, actually making it fast and easy for the vendor who’s using the same type of software to get approved.”

During a panel discussion on AI, two vendors, Marwan Sati, development executive for medical imaging AI at IBM Watson Health, and Woojin Kim, chief medical information officer at Montage Healthcare, a company acquired by Nuance, discussed AI’s role in medical imaging.

Kim said there was “so much more low-hanging fruit” that companies can apply AI to, such as radiology workflow and no-show patients, which can be a big budget hit for facilities.

“You can almost predict who’s going to show up or not and use that data to help you with tomorrow or next week’s scheduling, to really reduce the no-show rates, that alone could bring a tremendous amount of benefits to radiologists,” Kim said.


Back to HCB News

Artificial Intelligence Homepage