Women Will Pay for AI-Enhanced Breast Cancer Screening

According to research presented at the annual meeting of the Radiological Society of North America (RSNA), more than a third of women across 10 health care practices chose to enroll in a self-pay, artificial intelligence (AI)-enhanced breast cancer screening program, and the women who enrolled were 21% more likely to have cancer detected,

AI has shown great promise in mammography as a “second set of eyes” for radiologists providing decision support, risk prediction and other benefits. Despite its promise, AI is not yet reimbursed by insurance, which likely is slowing its adoption in the clinic. Some practices have elected to offer enhanced workflows enabled by AI at additional cost, much like what was done when digital breast tomosynthesis was originally deployed.

For the study, researchers investigated the impact of AI—including a safeguard review—as a self-pay option in screening mammography. A self-pay, AI-powered screening mammography program was offered to patients across 10 clinical practices, ranging from a few sites up to 64 sites at the largest practice. Women who enrolled had FDA-compliant AI software applied to their mammograms. An expert breast radiologist provided a third, safeguard review in cases where there was discordance between the first reviewer and the AI.

Out of the 747,604 women who underwent screening mammography over an initial 12-month period, the overall cancer detection rate was on average 43% higher for enrolled women than for unenrolled women. The pattern of a substantially higher cancer detection rate in enrolled women was observed at all 10 practices.

Further analysis attributed 21% of the increase in cancer detection to the AI program. The researchers credited the remaining 22% increase in detection to the fact that higher-risk patients chose to enroll more frequently.

“These data indicate that many women are eager to utilize AI to enhance their screening mammogram, and when AI is coupled with a safeguard review, more cancers are found,” said study senior author Gregory Sorensen, MD, from DeepHealth Inc. in Somerville, Massachusetts.

The recall rate—the rate at which women were called back for additional imaging—was 21% higher for enrolled versus unenrolled women. Relatedly, the positive predictive value for cancer was 15% higher for the enrolled women, indicating that each recall resulted in more cancer diagnoses in the enrolled population.

“This is the first report on results from a program that provides an AI-powered enhanced review that patients can elect to enroll in,” said study lead author Bryan Haslam, PhD, from DeepHealth. “The AI-driven enhanced review program leverages AI in a novel workflow to ensure women with suspicious findings get expert level care that could help detect many more breast cancers early. The number of women electing

for this program is now at 36% and growing, and the rate of cancer detection continues to be substantially higher for those women.”

In the future, the researchers hope to better quantify the benefit of the AI-driven safeguard review with prospective randomized controlled trials that would eliminate the self-selection bias and provide the highest level of evidence.