RSNA 2019: AI can do a lot for radiologists—but not everything

The rise of AI is one of the most popular topics in all of radiology, but there are still clear limits to its potential, according to a presentation given by Soonmee Cha, MD, at RSNA 2019 in Chicago.

Cha, a veteran neuroradiologist, has served as a program director at the University of California San Francisco since 2012 and currently oversees 100 radiology trainees. She began her presentation Monday, Dec. 2, by noting that she’s no expert when it comes to AI—but radiology education is a subject she knows quite well.

Trainees, Cha noted, are one of those things that AI simply can’t replace.

“AI will not replace radiologists or replace radiology trainees,” Cha said. “And it won’t cut costs. Maybe in the long run, but initially, it will actually drive up costs.”

AI also won’t be able to sign final reports or attend meetings, two things that are absolutely necessary in today’s radiology landscape.

“I have hundreds of meetings, just like all of you in this room,” Cha said. “If a machine could attend, that would be great!” (If machines could be trained to bring radiologists coffee, she also joked, it would certainly be appreciated.)

What AI can do

Of course, there is still a lot AI can do for radiologists. Cha said she can see a future where AI is improving image quality, decreasing acquisition times, eliminating artifacts, improving patient communication and even decreasing radiation dose.

“If AI can detect when machines are being set up incorrectly and alert us, it’s a win for us and for patients,” she said.

Interest in AI—both in terms of funding and academic research—are at all-time highs, Cha noted, showing that AI also has the ability to help jumpstart the careers of talented researchers.

But perhaps most importantly, AI technologies have shown the potential to help radiologists battle one of the biggest issues they face on a daily basis: increasing workloads. As radiologists get more and more sedentary, it could help them fight exhaustion and remain much happier on a daily basis if they could occasionally leave the reading room and speak with the very patients they are tasked with treating.

“I’m really waiting for that breakthrough where AI is going to help get me out of my chair so I can do some walking,” Cha said.

Around the web

The Palo Alto giant used exams from nearly 250,000 patients to upgrade its already robust algorithm.

Exams performed using the deep learning-based reconstruction tool also maintained high image quality, experts reported Wednesday.

Stratifying exams according to risk can reduce unnecessary imaging and downstream costs of care, Hawaiian researchers reported in Radiology.

Trimed Popup
Trimed Popup