Will radiologists meet the same fate from the advent of artificial intelligence (AI) as horses following the development of machine power? It's possible, according to an opinion article published online May 11 in the Journal of the American College of Radiology.
There were 25 million horses in the U.S. in 1920 but only 6 million by 1960. Similarly, far fewer than the current 34,000 radiologists in the U.S. will be needed unless they do more than just interpret imaging studies, according to Dr. Robert Schier of imaging services provider RadNet.
"This is a complex situation filled with unknowns, and events are moving fast," Schier wrote. "We need to figure how to deal with this coming change. And we need to do it in a hurry."
Assuming limits to AI
Some radiologists believe that AI will never be able to solve challenging problems such as finding lobular cancer in dense breasts or comparing multiple postoperative spine MRIs with patient motion from different machines, Schier said. But this ignores the fact that AI algorithms will improve in the coming years.
Systems such as early mammographic computer-aided detection (CAD) software may have underperformed and generally made the radiologist's life more difficult; however, AI software applications that possess real intelligence and can observe and learn from the world are already on the horizon, according to Schier. What's more, they will steadily become more capable.
"If computers can do something now, they will only get better at it," he wrote. "If computers cannot do something now, they will probably learn how to."
Eventually, diagnostic image interpretation systems will be developed that have read every textbook and journal article, as well as know a patient's history, records, and laboratory reports, he noted. It will also have memorized millions of imaging studies.
Cambrian explosion
Life on Earth evolved slowly until the Cambrian explosion 500 million years ago, which rapidly increased the types and complexity of life that exist today. We are likely now in the midst of a similar explosion in computer intelligence, according to Schier.
The advent of computers that can accurately interpret diagnostic imaging studies will upend radiology, but it remains to be seen how much upending will take place and how long it will take to happen.
"There are vastly differing opinions, from the apocalyptic claim that AI will make all radiologists extinct to the delusional assertion that computers will always merely assist -- and never replace -- radiologists," Schier wrote. "Both extremes are mistaken, but the truth is in the direction of the first."
