AI Could Be Saving Lives Right Now

AI could be saving women’s lives today. How? By helping doctors read mammograms. A study called the MASAI Trial, recently published in The Lancet, found that one radiologist working with AI outperforms two radiologists working together. AI is consistent and methodical in ways humans simply aren't.

The most trustworthy authority on preventive health, the US Preventive Services Task Force, recommends that women get mammograms every two years starting at age 40. But the doctors who read mammograms aren't perfect — the NIH estimates they miss 20% of breast cancers. Another 10% come back as false positives, triggering unnecessary additional testing. And in rural areas, even getting a mammogram can be difficult due to a shortage of radiologists skilled in reading them.

To understand how AI works, I visited with Dr. Charles McDonnell, a radiologist at Sutter Medical Center in Roseville, California. He showed me a standard mammogram — two breast images side by side. Then he showed me the same images where AI added red marks demarcating where there was a suspicion of breast cancer.

"So it looks at the mammogram, and it puts a mark on something that it thinks could be a cancer, so that you pay extra attention to it,” says McDonnell.

Studies show AI catches those 20% of cancers radiologists miss — without increasing false positives. We continued to look at the mammograms.

"Maybe that patient would not have been called back at all,” he says. “But there were those marks there, so the patient was called back for bilateral diagnostic workups. It showed that the patient had bilateral breast cancer."

Dr. McDonnell says the AI identified what humans might miss.

"If 10 radiologists reviewed this screening study, maybe nine would've identified the abnormality on the right, maybe four would've identified the left, without marks."

AI also addresses the nationwide shortage of radiologists.

"The idea would be, you have 100 mammograms, let's say 50 or 60 of those have no marks on them. And if they have no marks on them, a human being doesn't need to look at it. So that would be a way to leverage your lack of manpower."

And, it's not just breast cancer. AI shows similar promise for detecting lung cancers that humans miss.

The case for AI-assisted screening is compelling across the board. So why isn't AI part of routine screening? Significant barriers remain: doctors don’t know how to bill for using it, insurers don’t know how to pay for it, some radiologists are reluctant to integrate AI into their workflow, and the FDA is slow to approve it. These delays are harming patients who aren't getting the most accurate diagnosis possible.

The technology exists. The evidence is clear. Every month of delay means cancers are going undiagnosed.

— Dr. Michael Wilkes with a Second Opinion

AI Could Be Saving Lives Right Now

Download the KCRW App