AI model predicts breast cancer risk

A new artificial intelligence model trained solely on mammogram images is proving far more accurate than traditional breast density assessments in predicting a woman’s five-year risk of breast cancer (Radiological Society of North America).

For years, breast cancer risk stratification has relied on factors like age, family history, genetics, and breast density. But these indicators only tell part of the story: the vast majority of women diagnosed each year have no strong hereditary markers, and breast density alone separates risk only slightly.

The new model: Clarity Breast, the first FDA-authorized image-only AI risk tool, was developed using more than 421,000 mammograms from facilities in Europe, South America, and the United States. By analysing both scans from women who went on to develop cancer and those who did not, the system learned subtle, complex patterns within breast tissue that humans cannot detect.

A new layer of risk detection

Unlike tools focused on spotting visible abnormalities, this AI model assesses the underlying texture and biological signals within the breast. The output is a five-year risk probability that categorizes women into average, intermediate, or high-risk groups using standard National Comprehensive Cancer Network thresholds.

When applied to more than 245,000 mammograms from six screening sites, the model sharply distinguished differences in cancer incidence:

High-risk AI group: 5.9% incidence

Average-risk AI group: 1.3% incidence

By comparison, breast density, long treated as a key risk marker, made only a modest difference (3.2% in dense breasts vs. 2.7% in non-dense). The findings show that image-based AI uncovers meaningful risk signals that density measurements overlook, offering a far more granular, individualized picture of future cancer risk.

Toward more personalized screening

The results come at a pivotal moment. While guidelines recommend starting routine mammograms at age 40, younger women are the fastest-growing group diagnosed with breast cancer and often at more advanced stages.

With an AI-derived risk score, clinicians could identify high-risk women earlier, including those in their 30s who would otherwise be considered average risk. A baseline mammogram at 30 paired with image-based risk assessment could help channel the right patients into earlier and more frequent screening pathways.

Breast density notification laws in 32 U.S. states already require providers to inform women of their breast density after screening. The study’s authors argue that pairing density information with an AI-derived risk score could offer women a far clearer understanding of their personal risk, moving beyond a simple “dense or not dense” classification.

The model does not replace traditional risk markers, but it adds a powerful new layer. And as AI tools evolve, they may redefine how women worldwide are screened, counselled, and monitored long before cancer develops.