Nine years ago, one of the world's leading artificial intelligence scientists chose an endangered professional species.
“People now have to stop training radiologists,” said Geoffrey Hinton, adding that it was “just completely clear” that AI would surpass the people in that area within five years.
Nowadays, radiologists – the physician specialists in medical imaging look in the body to diagnose and treat disease – are still a lot of demand. A recent study by the American College of Radiology projected a steadily growing workforce until 2055.
Dr. Hinton, who received a Nobel Prize in Physics last year for groundbreaking research in AI, was broadly correct that the technology would have a significant impact – just not as a court murderer.
This applies to radiologists in the Mayo Clinic, one of the most important medical systems in the country, whose main campus is in Rochester, Minn. There they have started using AI in recent years to grind images, to automate routine tasks, identify medical abnormalities and predict diseases. AI can also serve as 'a second set of eyes'.
“But would it replace radiologists? We didn't think so,” said Dr. Matthew Callstrom, the chairman of the Radiology of the Mayo Clinic, who remembered the 2016 prediction. “We knew how difficult it is and everything involved.”
Computer scientists, labor experts and policymakers have long debited how AI will ultimately play in the labor force. Will it be a smart helper, improving human performance or a robot -like surrogate, which moves millions of employees?
The debate is intensified because the leading technology behind chatbots seems to improve faster than expected. Leaders at OpenAI, anthropic and other companies in Silicon Valley now predict that AI will darken people in most cognitive tasks within a few years. But many researchers provide a more gradual transformation in accordance with seismic inventions of the past, such as electricity or internet.
The predicted extinction of radiologists offers a meaningful case study. Until now, AI appears to be a powerful medical device to increase efficiency and increase human skills, instead of taking someone's job.
When it comes to the development and use of AI in medicine, radiology has been an excellent target. Of the more than 1,000 AI requests approved by the Food and Drug Administration for use in medicine, there are about three-quarters in radiology. AI usually excels in identifying and measuring a specific deviation, such as a long lesion or a chest chock.
“Great progress has been made, but these AI tools are looking for one thing for the most part,” said Dr. Charles E. Kahn Jr., a professor of radiology at the Perelman School of Medicine of the University of Pennsylvania and editor of the magazine Radiology: Artificial Intelligence.
Radiologists do much more than study images. They advise other doctors and surgeons, talk to patients, write reports and analyze medical records. After identifying a suspect of tissue in an organ, they interpret what it could mean for an individual patient with a certain medical history, where they tick years of experience.
Predictions that AI will steal jobs often underestimate “the complexity of the work that people actually do – just like radiologists do much more than reading scans,” said David Autor, a labor economist at the Massachusetts Institute of Technology.
In the Mayo Clinic, AI tools were investigated, developed and adapted to the work routines of busy doctors. The staff has grown 55 percent since Dr. Hinton's prediction of Doom, up to more than 400 radiologists.
In 2016, stimulated by the warning and progress in AI-driven image recognition, the leaders of the radiology department compiled a group to assess the potential impact of technology.
“We thought the first thing we had to do, would use this technology to make us better,” Dr. Callstrom. “That was our first goal.”
They decided to invest. Nowadays, the Radiology Department has an AI team of 40 people, including AI scientists, radiology researchers, data analysts and software -engineers. They have developed a series of AI tools, from tissue analyzers to illness forelators.
That team works with specialists such as Dr. Theodora Potretzke, who focuses on the kidneys, bladder and reproductive organs. She describes the role of the radiologist as 'a doctor for other doctors', who clearly communicates, assist and advise the imaging results.
Dr. Potretzke has collaborated on an AI tool that measures the volume of the kidneys. The growth of the kidney, in combination with cysts, can predict the deterioration of kidney function before it is displayed in blood tests. In the past they mostly measured the kidney volume by hand, with the equivalent of a ruler on the screen and guesswork. The results varied and the job was a time -consuming.
Dr. Potretzke served as a consultant, end user and tester while working with the AI team of the department. She helped with the design of the software program, which has color coding for different tissues, and checked the measurements.
Today she brings an image on her computer screen and clicks on a icon and the kidney volume measurement appears immediately. It saves her 15 to 30 minutes every time she examines a kidney image and it is consistently accurate.
“It is a good example of something that I feel very comfortable with AI for efficiency and accuracy,” said Dr. Potretzke. “It can expand, help and quantify, but I am not in a place where I give up interpretative conclusions to technology.”
In the hall, Dr. Francis Baffour, a radiologist of the staff, the varied ways to which AI was applied to the field, often in the background. The makers of MRI and CT scanners use AI algorithms to accelerate and clean up the taking images, he said.
AI can also automatically identify images that show the highest probability of an abnormal growth, essentially tell the radiologist: “Look here first.” Another program scans images on blood clots in the heart or lungs, even when the medical focus can be elsewhere.
“Ai is now everywhere in our workflow,” said Dr. Baffour.
In general, the Mayo clinic uses more than 250 AI models, both internally developed and license from suppliers. The radiology and cardiology departments are the greatest consumers.
In some cases, the new technology opens a door to insights that have to do with people. One AI model analyzes data from electrocardiograms to predict patients who previously develop atrial fibrillation, an abnormality of the heart rhythm.
A Radiology research project uses an AI algorithm to distinguish subtle changes in shape and texture of the pancreas to detect cancer for two years before conventional diagnoses. The Mayo Clinic team works together with other medical institutions to further test the algorithm for more data.
“Mathematics can see what the human eye cannot do,” said Dr. John Halamka, president of the Mayo Clinic Platform, who supervises the digital initiatives of the health system.
Dr. Halamka, an AI optimist, believes that technology will transform medicine.
“In five years it will be malpractice not to use AI,” he said. “But it will be people and AI who work together.”
Dr. Hinton agrees. In retrospect, he believes that he spoke too broadly in 2016, he said in an e -mail. He did not make it clear that he spoke purely about image analysis and was wrong with timing, but not about the direction, he added.
In a few years, most medical image interpretation will be done by “a combination of AI and a radiologist, and it will make radiologists a lot more efficient in addition to improving accuracy,” Dr. Hinton.
