Why radiologists may not be goners
Apart perhaps from cashier and truck driver, radiologist is said to be the most imperiled job on the planet in the new age of automation. But artificial intelligence researchers say the challenge they're addressing at the moment is to make it easier for radiologists to work faster and with less fear of a bad call.
Elad Walach, CEO of a startup called Aidoc, tells Axios that he has trained an AI system to detect the most ordinary but urgent and pernicious maladies — ones that, if missed by a radiologist, could lead to "a nightmare" for the patient. He calls them "acute findings," which include stroke, hemorrhage and fractures.
- Once the AI detects an abnormality, it informs the radiologist, who can then decide — "Is it a tumor? Is it urgent?" Walach says.
- The objective, he says, is to "relieve the bottleneck. Today, there are enough scanners but not enough radiologists. Radiologists will be able to handle more scans."
Louis Rosenberg, founder of Unanimous.ai, is another AI expert working on better x-ray interpretation. He's collaborating with Stanford Medical School to create what he calls a "hive mind."
- Here's what happens: Generally speaking, a hive mind is created by putting a group of people in front of individual screens. In what Fast Company calls a "virtual Ouija board," they are asked a series of questions, and use a cursor to push a puck to the answer they favor.
- At the same time, they watch what everyone else is favoring. Ultimately — just as a flock of birds will decide where to fly from danger or to a feeding ground — some compromise and they arrive at a collective decision.
- It's not clear how this method would be optimized for x-rays, but "when people think together as a swarm, they converge on decisions that get the value of the group better than a survey," Rosenberg says. "A survey gets the most popular answer, but not the most representative one. With a swarm, you converge on the sentiment that reflects the optimal outcome."
- Radiologists "are terrified about the future of their profession," Rosenberg says, but the hive mind would work only to help them make fewer errors.