Doctors who use AI can “forget” certain skills: gunshots

Artificial intelligence is starting to help doctors detect patients for several routine diseases. But a new study raises concerns as to whether doctors could become too dependent on AI.

The study examining gastroenterologists in Poland revealed that they seemed to be around 20% of worse to identify polyps and other anomalies during colonoscopies by themselves, after having used to use an AI assisted system.

The results, published in the journal Lancet gastroenterology and hepatology, Suggest that even after a short period of use of AI, experts can become too dependent on AI to do certain aspects of their work.

“We were quite surprised,” said Marcin Romańczyk, an MD-PH.D. Gastroenterologist at HT Medical Center in Tychy, Poland, who led the study.

But not everyone is convinced that the document proves that doctors lose essential skills because of AI.

“I think three months seem to be a very short period to lose a skill that you have taken 26 years to accumulate,” explains Johan Hulleman, researcher at the University of Manchester in England who studied human dependence on artificial intelligence.

Hulleman thinks that statistical variations in patient data could be part of the explanation of the reasons why the figures seem to drop. Factors such as the average age of patients used in different sections of the study could explain the variation, he says.

You have in medicine

Artificial intelligence is increasingly common for a number of routine medical scans. The next time you get an eye scan, breast cancer or colon disease, there is decent chances that AI could analyze the images.

“AI is spreading everywhere,” said Romańczyk. At the same time, many doctors play a catch -up because learning to use technology was not part of their training.

“We have learned books and our teachers,” he said. “No one told us how to use AI.”

A few years ago, four clinics in Poland tried an AI system to detect polyps and other anomalies during colonoscopies. The AI works in real time, analyzing the video of a camera inside the colon.

If he sees something, he will highlight the area so that the clinician can see him.

“In this particular box, there is a green box, showing where the polyp could be,” he said.

In the box

The clinics collected data on the operation of the AI system. It turns out that this is the case, but when Romańczyk and his colleagues reanalyzed the data, they found something else: after the introduction of the system, doctors became clearly worse to detect possible polyps when the AI was extinguished.

According to their analysis,, After doctors obtained AI, possible polyps detection rates increased from 28.4% to 22.4% when their new AI system was extinguished. In other words, doctors seemed to become quickly dependent on AI systems catching polyps. Romańczyk says he doesn’t really know why it happens, but he has theories.

“We unconsciously wait for the green box to come out to show us the region where the polyp is located and we are not paying attention,” he said.

There are other examples that support this idea: a similar study has shown that non-experts are doing a worst work of scanning mammography if they know that they can get an AI system to help them push a button.

Johan Hulleman, who helped direct this mammography study, describes him as a “security network effect”. He says these last results could be interesting, but he is skeptical. The study of colonoscopies took place over three months and participating doctors had decades of experience. He thinks that statistical variations due to a number of factors, such as patient age, could be the cause of the apparent decline.

In addition, he says: “We do not know how many polyps there were really, so we do not know the truth on the ground.” By that, it means that it is not clear how many possible polyps that doctors have missed were really medically important.

The author of the study, Romańczyk, believes that the decline is real – although he admits that the study of AI in a clinical context like this can be delicate. There are many variables that researchers could not control.

It is not against the use of AI. He actually thinks that small green boxes help him make better colonoscopies. But he thinks that there should be more studies like these examining how AI could change the way doctors work in the real world.

“Because looking at what’s going on,” he said. “We have AI systems that are available, but we don’t have the data.”