A new artificial intelligence (AI) model that looks only at mammogram images can predict a woman’s five-year risk of breast cancer more accurately than traditional methods, including breast density, according to research presented at the Radiological Society of North America’s annual meeting.
Dr. Constance Lehman, a professor of radiology at Harvard Medical School, explained that current ways to estimate breast cancer risk—like age, family history, genetics, and breast density—aren’t enough.
Most women diagnosed with breast cancer don’t have a strong family history, and only 5–10% of cases are hereditary. Breast density alone has also proven to be a weak predictor of risk.
The new AI model, called Clairity Breast, is the first image-only breast cancer risk tool to be authorized by the FDA. It was trained using over 421,000 mammograms collected from 27 sites around the world.
By comparing scans from women who did and didn’t develop cancer, the AI learned to detect very subtle changes in breast tissue that might indicate future cancer risk—changes that human eyes can’t see.
The AI uses a deep learning method called a convolutional neural network to produce personalized five-year risk scores. It was tested on over 245,000 mammograms from the U.S. and Europe, taken between 2011 and 2017. Researchers then compared these AI-generated risk scores to actual cancer cases recorded in medical records.
The risk scores were grouped into three levels using standard cancer risk guidelines: average (less than 1.7%), intermediate (1.7 to 3.0%), and high (more than 3.0%). Women in the high-risk group identified by the AI were more than four times as likely to develop breast cancer than those in the average-risk group (5.9% compared to 1.3%).
By comparison, breast density alone showed much smaller differences—3.2% for women with dense breasts vs. 2.7% for those without. This means the AI tool provided a much clearer and more accurate picture of future risk.
Dr. Christiane Kuhl, the study’s first author and presenter, emphasized that these results show how image-only AI tools can greatly improve how we assess a woman’s risk of breast cancer. This could lead to more personalized screening strategies.
Currently, the American Cancer Society recommends that women at average risk begin mammograms at age 40. But women under 40 are now the fastest-growing group being diagnosed with advanced breast cancer.
The researchers say that image-based AI risk tools could help spot high-risk women earlier—possibly in their 30s—so they can start screening sooner.
In fact, the team suggests that a baseline mammogram at age 30 could be used to generate an AI-based risk score. If the score is high, that woman could be offered earlier and more frequent screenings—even if she has no family history of cancer.
Today, many U.S. states require doctors to tell women whether they have dense breast tissue after a mammogram. But Dr. Lehman says we can go further. She believes women should also be told their AI-generated risk score, giving them a more complete understanding of their risk and helping guide future care decisions.
If you care about breast cancer, please read studies about a major cause of deadly breast cancer, and this daily vitamin is critical to cancer prevention.
For more information about cancer, please see recent studies that new cancer treatment could reawaken the immune system, and results showing vitamin D can cut cancer death risk.
Copyright © 2025 Knowridge Science Report. All rights reserved.
