“AI has no soul, it is not sensitive to satisfaction from abnormality” according to a radiologist

This summer, franceinfo is interviewing employees, freelancers, and business leaders about their relationship with artificial intelligence. How do they use it, how do it change their professional practices? Today, Marie-Pierre Revel, head of the radiology department at Cochin Hospital, in charge of screening for the French Society of Radiology.

Published


Reading time: 2 min

Artificial intelligence in hospitals and clinics: At Cochin Hospital in Paris, AI has long been used in the radiology department. (VITHUN KHAMSONG / MOMENT RF / GETTY IMAGES)

What is the relationship between employees, business leaders and the self-employed and artificial intelligence? Sarah Lemoine met Marie-Pierre Revel, head of the radiology department at Cochin Hospital, in charge of screening for the French Society of Radiology.

Marie-Pierre Revel has been using artificial intelligence for a long time in her work as a radiologist. She began experimenting with it in 2002, first through voice recognition, well before this technology was widely deployed in the hospital world. “I had the chance to participate in the opening of the Pompidou hospital in Paris. At the time, in terms of artificial intelligence, we tested the use of voice recognition to generate reports. That was 22 years ago!

Before that, the classic method was to dictate on a cassette, to give the recording to a secretary who would type the report. Voice recognition has allowed us to make very significant time savings. The reports are available much more quickly to patients, they are also more relevant. We see the text displayed on the screen and we immediately self-criticize.”

In the radiology department that she now heads at Cochin Hospital, Marie-Pierre Revel uses two artificial intelligence systems on a daily basis. One to help detect fractures, the other nodules in the lung.

“It’s as if there were two readers. A human reader and a reader who is artificial intelligence. We always start by analyzing without the AI, and then we look at what the AI ​​has found. The latter has the advantage of not having any qualms, it is not sensitive to fatigue, nor to what is called ‘satisfaction of search (SOS)’ in English.”

Marie-Pierre Revel, radiologist

to franceinfo

“In fact, we radiologists, continues Marie-Pierre Revelwhen we have found a first anomaly, our vigilance goes out a little, we have the impression of having done the work, whereas there may be something more discreet, but much more important for the patient. Artificial intelligence, it does not care, whether it has already found something before or not, it does not influence it!

When Professor Revel is asked whether artificial intelligence in the service of medicine raises questions, she points out the limits of prediction algorithms. “I don’t have a crystal ball! Once I’ve analyzed the scan, if the algorithm tells me that a patient is not going to respond to treatment, or that they have a five-year survival rate of 12%, that poses a problem for me in terms of liability, because I have no way of verifying it.”

“This also poses an ethical problem for me, because I don’t know if we should predict everything. If we knew that in the year 2047, at 6:02 p.m., we were going to die, I’m not sure that would be helpful, and that it would be easy to live with. This is what I call ‘cumbersome knowledge’.”

Marie-Pierre Revel, radiologist

to franceinfo

If Marie-Pierre Revel considers artificial intelligences to be essential today in radiology, particularly in terms of detection, the latter require permanent human validation, because contrary to their name, she says, they are not very intelligent.


source site-14

Latest