Radiology finds clinical decision support in AI-based tool

Radiologists are overworked and, as is the trend these days, some are turning to digital tools backed by artificial intelligence to help ease the pain.

One of those tools, Aidoc, provides decision support by helping detect abnormalities in imaging studies. As soon as a patient leaves the scanner, Aidoc's software starts analyzing the images with AI-based tools and flags any abnormalities.

“Our aim is to help them sift through the mountains of data to help them find what really matters,” said Elad Walach, co-founder and CEO of Tel Aviv, Israel-based Aidoc. “This is an added layer of confidence that, if there is something in the scan,” it would help radiologists make decisions about the next step.

That could be especially beneficial given the workload of most radiologists. On average, a radiologist who looks at CT or MRI scans must interpret one of the images every 3 to 4 seconds over the course of an eight-hour workday to keep up with the workload, according to a 2015 study in the journal Academic Radiology.

To meet that demand, radiologists have been getting more productive, according to the AMGA, with a 3.4% increase in relative value units in 2016 compared with the year before. The median increase across other specialties is much lower, at 1.54%.

The quality of service must remain high, given radiology is disproportionately lucrative for hospitals. It brings in one-third of outpatient contribution profit—the most of any outpatient service, according to the Advisory Board.

Radiologists tend to get paid more than other specialties, though salaries dipped in 2017, falling to $487,239, according to the AMGA. That's in part because of lower reimbursement per service, which means it'll take even greater productivity for radiologists to pull in the same revenue as before.

Increased productivity could come from the very technology that some say will eventually replace radiologists.

So far, Aidoc has deployed its software at 50 health centers worldwide, including more than 10 in the U.S., where it's paid for in the software-as-a-service model.

The software is currently limited to CT scans, on which it can detect intracranial hemorrhages, for instance, and pulmonary embolisms, among other problems (though it can't tell radiologists specifically what's wrong).

Because Aidoc's software can be applied to images from many parts of the body—including the head, abdomen, chest and spine—the company's approach differs from that of many other AI radiology companies, which tend to focus on specific pathologies, Walach said. “If you really want to impact the workflow, you cover a significant portion, which means radiologists can rely on you in their day-to-day operations and not just for specific cases.”

In the U.S., though, Aidoc is limited to just the brain, since only its brain tool has received federal clearance. Outside of the U.S., Aidoc's software has been commercialized since December 2017. The company is now working on getting approval for use on other parts of the body.

Though the software is becoming more widespread, getting early adopters to sign on was tough. “The most difficult is to install in the first few sites, because everyone wants to see someone else act first,” said Tom Valent, Aidoc's vice president of business development.

Generally speaking, AI tools have been around for awhile, though, and hospitals and health systems are attracted to them as ways to make their specialists more efficient and to ultimately lower costs, said Sarah Danielson, managing director at Deloitte. “Whole business models can be built around low-cost screening alternatives,” she said, adding that it's about getting services to a cost point that's manageable.

While the software is a boon to radiologists and also potentially to health systems' bottom lines, it also raises concerns—namely, staffing. But that concern is misguided, said Dr. Eric Weinberg, medical director of University Medical Imaging at the University of Rochester (N.Y.) Medical Center. “Anyone who thinks they're going to be replaced as a radiologist by this is mistaken.”

Instead, he said, it's better to think of these tools as add-ons. “These tools are aids to accuracy and efficiency, which is great,” he said. “This is going to be very important for the future of my field.”