Nine years ago, one of the main artificial intelligence scientists in the world has identified an extinction professional species.
“People should stop training the radiologists now,” said Geoffrey Hinton, adding that it was “completely obvious” that within five years the IA would overcome humans in that field.
Today, radiologists – the doctor's specialists in medical humaging who look inside the body to diagnose and treat diseases – are still in great demand. A recent study by the American College of Radiology has projected a constantly growing workforce through 2055.
Dr. Hinton, who was awarded a Nobel Prize in Physics last year for pioneering research in AI, was widely correct that technology would have had a significant impact, but not as a work killer.
This is true for the radiologists of the Mayo Clinic, one of the main medical systems of the nation, whose main campus is in Rochester, Minn. In recent years, they have started using the IA to refine the images, automate routine tasks, identify medical anomalies and predict diseases. Artificial intelligence can also serve “a second series of eyes”.
“But would replace the radiologists? We didn't think about it,” said dr. Matthew Callstrom, president of Radiology of the Mayo Clinic, remembering the 2016 forecast. “We knew how difficult it is and everything that is involved.”
IT scientists, work experts and political managers have long discussed how the IA will eventually take place in the workforce. Will it be intelligent help, improving human performance or a robotic surrogate, thinning millions of workers?
The debate intensified since the avant -garde technology behind the chatbots seems to improve faster than expected. The leaders of Openai, anthropics and other Silicon Valley companies now provide that the IA eclipse humans in most cognitive tasks within a few years. But many researchers provide a more gradual transformation in line with seismic inventions of the past, such as electricity or internet.
The expected extinction of radiologists provides a case of study. So far, the IA is showing that it is a powerful medical tool to increase efficiency and enlarge human skills, rather than taking the work of anyone.
When it comes to developing and distributing artificial intelligence in medicine, radiology was a primary objective. Of the over 1,000 artificial intelligence applications approved by the Food and Drug Administration for use in medicine, about three quarters are in radiology. Artificial intelligence generally excels to identify and measure a specific anomaly, such as a lung lesion or a breast lump.
“There have been incredible progress, but for the most part these artificial intelligence tools,” said dr. Charles E. Kahn Jr., Professor of Radiology at the Perelman School of Medicine and publisher of Radiology magazine: Artificent.
Radiologists do much more than studying images. They recommend other doctors and surgeons, talk to patients, write relationships and analyze medical records. After identifying a suspicious cluster of fabric in an organ, they interpret what could mean for a single patient with a particular medical history, touching years of experience.
The forecasts according to which the IA will steal the jobs often “underestimate the complexity of the work that people actually carry out – just as radiologists do much more than reading scans,” said David Autor, an economist of work at the Massachusetts Institute of Technology.
At the Mayo Clinic, artificial intelligence tools have been designed, developed and personalized to adapt to the work routines of the medical doctors. The staff grew up 55 percent from the prediction of Dr. Hinton's Doom, to over 400 radiologists.
In 2016, driven by the warning and progress in the recognition of images fed by the AI, the leaders of the Radiology Department brought together a group to evaluate the potential impact of technology.
“We thought that the first thing we should do was to use this technology to improve ourselves,” said Dr. Callstrom. “This was our first goal.”
They decided to invest. Today, the Radiology Department has an artificial intelligence team of 40 people including artificial intelligence scientists, radiology researchers, data analysts and software engineers. They developed a series of artificial intelligence tools, from tissue analyzers to the predictors of diseases.
That team works with specialists such as Dr. Theodora Pestzke, who focuses on kidneys, bladder and reproductive organs. It describes the role of the radiologist as “a doctor for other doctors”, clearly communicating the imaging results, assisting and recommending.
Dr. could collaborate in an artificial intelligence tool that measures the volume of the kidneys. Renal growth, if combined with cysts, can predict the decline of renal function before it occurs in the blood tests. In the past, it has measured the renal volume largely by hand, with the equivalent of a sovereign on the screen and conjectures. The results varied and the ungrateful work was requested by time.
Dr. Pestzke was a consultant, end user and tester while working with the AI team of the Department. He contributed to planning the software program, which has a color coding for different fabrics, and has checked the measurements.
Today, it makes an image appear on the screen of her computer and clicks on an icon and the renal volume measurement appears instantly. He saves them from 15 to 30 minutes every time he examines a renal image and is constantly accurate.
“It is a good example of something that I feel very comfortable to deliver to the AI for efficiency and precision,” said dr. Could. “It can increase, help and quantify, but they are not in a place where I give up interpretative conclusions to technology.”
At the bottom of the corridor, dr. Francis Baffour, a staff radiologist, explained the various ways in which the IA had been applied to the field, often in the background. Magnetic resonance scanner manufacturers and coach use artificial intelligence algorithms to accelerate images and to clean them, he said.
Artificial intelligence can also automatically identify the images that show the maximum probability of abnormal growth, essentially telling the radiologist, “look here before”. Another program scanning images for blood clots in the heart or lungs, even when the medical focus can be elsewhere.
“Artificial intelligence is everywhere in our work flow now,” said dr. Baffour.
Overall, the Mayo Clinic uses more than 250 AI models, both developed internally and authorized by suppliers. The departments of radiology and cardiology are the major consumers.
In some cases, the new technology opens a door to intuitions that go beyond human ability. An artificial intelligence model analyzes data from electrocardiograms to predict patients who are more likely to develop atrial fibrillation, a heart rate anomaly.
A research project in radiology uses an artificial intelligence algorithm to discern thin changes of form and consistency of the pancreas to detect cancer up to two years before conventional diagnoses. The Mayo Clinic team is working with other medical institutions to further test the algorithm on multiple data.
“Mathematics can see what the human eye cannot,” said dr. John Halamka, president of the Mayo Clinic platform, who supervises the digital initiatives of the health system.
Dr. Halamka, an optimist of artificial intelligence, believes that technology will transform medicine.
“In five years, it will be negligent not to use the IA,” he said. “But they will be human and to the work together.”
Dr. Hinton agrees. In retrospective, he believes he has spoken too wide in 2016, he said in one and -mail. He did not clarify that he spoke exclusively about the analysis of the images, and he was wrong in times but not on the direction, he added.
In a few years, most of the interpretation of medical images will be performed by “a combination of AI and a radiologist and will make radiologists much more efficient as well as improving accuracy”, said dr. Hinton.