All mammograms performed at Johns Hopkins locations are reviewed using artificial intelligence (AI) software as a secondary tool, providing dedicated breast radiologists additional support in detecting early-stage breast cancer AI software can also help to reduce false positive readings.

A health system policy enacted in March 2024 makes the additional AI review a standard part of cancer screenings and diagnosis, delivered at no additional cost to patients and requiring no additional insurance approval.

This is a departure from industry practice, says Pouneh Razavi, who directs breast imaging in the Greater Washington area.

A Johns Hopkins committee that included Razavi, along with oncologists, administrators and other radiologists, reviewed the AI software, chose a vendor and set the no-fee policy.

“While most places charge the patient for this as an out-of-pocket add-on for their mammogram, we decided that we would make it complimentary,” she says. “We don’t want anyone to miss out on the technology because they can’t afford it or they don’t think it’s worthwhile.”

National data shows that facilities that use AI as a secondary screening tool for mammograms are finding more tiny, invasive cancers, she says.

A Second Opinion, Not a Replacement

AI software is developed by reviewing hundreds of thousands to millions of mammograms to create a model for determining what a normal mammogram looks like and what a mammogram showing cancer looks like.

The screening software serves as a second opinion and does not replace the experience and judgement of Hopkins’ fellowship-trained breast radiologists who review mammograms across Johns Hopkins.

When a patient gets a mammogram, a Johns Hopkins breast radiologist and the AI software separately interpret the images, she explains.

Razavi notes that the technology is particularly effective at detecting small and subtle distortions, which could prompt clinicians to recommend a biopsy. However, it is not yet able to analyze images from year to year. As a result, the software may flag findings that the breast radiologist knows has been stable for years.

If AI identifies an area of concern, the radiologist reassesses the finding. If there’s still a question about what was flagged, a second breast radiologist is consulted. “AI complements our reading, but the final interpretation is by the breast radiologist,” Razavi says. “If there’s a discrepancy where I'm not sure if I agree with AI, we always consult with another breast radiologist.”

Patients can be referred to two convenient breast imaging locations in the Greater Washington area:

Sullivan Breast Center at Sibley Memorial Hospital

Washington, D.C.

202-537-4545

Johns Hopkins Medical Imaging

Bethesda, Md.

301-530-5151