MARY LOUISE KELLY, HOST:
Artificial intelligence is showing up at the doctor's office. AI algorithms are beginning to help screen for several routine diseases. But as NPR's Geoff Brumfiel reports, a new study is raising concerns about whether doctors might become too reliant on AI.
GEOFF BRUMFIEL, BYLINE: The next time you get a scan for your eyes, breast cancer or colon disease, there's a decent chance that artificial intelligence might be analyzing the images.
MARCIN ROMANCZYK: AI is spreading everywhere.
BRUMFIEL: Marcin Romanczyk is a doctor at H-T. Medical Center in Tychy, Poland. He says although AI is moving rapidly into medicine, many doctors are playing catch-up. It simply wasn't part of their training.
ROMANCZYK: All the doctors - we've been taught from books and from our teachers. No one told us how to use AI.
BRUMFIEL: Romanczyk is a gastroenterologist. A few years ago, several clinics in Poland tried out an AI system to detect polyps during colonoscopies. The AI works in real time, analyzing video from a camera inside the colon. If it spots something...
ROMANCZYK: We see some kind of marking. In this particular one, there's a green box showing where the polyp might be.
BRUMFIEL: The clinics were collecting data on whether the AI worked. It turns out it did. But when Romanczyk and his colleagues reanalyzed the data, they found something else. After the system was introduced, doctors became significantly worse at detecting possible polyps when the AI was switched off.
ROMANCZYK: We were quite actually surprised.
BRUMFIEL: According to a new analysis published in the journal, Lancet Gastroenterology And Hepatology, after doctors got AI, detection rates without it fell by roughly 20%. In other words, the doctors seem to become quickly dependent on AI systems catching the polyps. Romanczyk says he's not quite sure why it's happening, but he has some theories.
ROMANCZYK: We are subconsciously waiting for the green box to come out and show us the region where the polyp is, and we are not paying so much attention.
BRUMFIEL: There are some other examples that support that idea. A similar study has shown that people do a worse job scanning mammograms if they know they can get an AI system to help them with the push of a button. Johan Hulleman is a researcher at Manchester University in the U.K. who helped lead that mammogram study. He says these latest results could be interesting, but he's a little skeptical. The study took place over three months and the doctors participating had decades of experience.
JOHAN HULLEMAN: I think three months kind of - it seems a very short period to lose the skill that you took, like, 26 years to build up.
BRUMFIEL: He thinks statistical variations due to a number of factors might be making the drop look larger than it is.
HULLEMAN: We don't know how many polyps there really were. And so we don't know the grand truth.
BRUMFIEL: By that, he means it's unclear how many of the possible polyps were medically important. The study's author, Romanczyk, does believe the drop is real, though he admits studying AI in a clinical setting like this can be tricky. There are a lot of variables the researchers couldn't control. And that's why Romanczyk thinks that there should be more studies like these, looking at how AI might be changing the way doctors work in the real world.
ROMANCZYK: Because look what's happening. We have AI systems that they are available, but we don't have the data.
BRUMFIEL: About how doctors interact with those systems. He's not against using AI. He actually thinks the little green boxes help him do better colonoscopies. He just wants to make sure this rapid rollout of a new technology is supported by facts, not just feels.
Geoff Brumfiel, NPR News.
(SOUNDBITE OF MUSIC) Transcript provided by NPR, Copyright NPR.
NPR transcripts are created on a rush deadline by an NPR contractor. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.
