An image-only artificial intelligence (AI) model demonstrated stronger and more precise risk stratification than breast density assessment for predicting 5-year risk for developing breast cancer, according to findings from a study presented at the 2025 Annual Meeting of the Radiological Society of North America (RSNA) (Abstract S5-SSBR02-1). The model, Clairity Breast, is the first U.S. Food and Drug Administration–authorized image-only AI breast cancer risk model.
“The results of this large-scale analysis demonstrate that AI risk models provide far stronger and more precise risk stratification for 5-year cancer prediction than breast density alone,” said study first author and presenter Christiane Kuhl, MD, PhD, Director, Department of Diagnostic and Interventional Radiology at University Hospital RWTH Aachen, Germany. “Our findings support the use of image-only AI as a complement to traditional markers supporting a more personalized approach to screening.”
Study Methods
Clairity Breast was trained on 421,499 mammograms from 27 facilities across Europe, South America, and the United States. The model consisted of a deep convolutional neural network that was calibrated on an independent test cohort to predict 5-year risk probabilities.
The model was applied to a cohort of 245,344 bilateral 2D screening mammograms taken between 2011 and 2017. The AI model categorized the womens' mammograms into thresholds by National Comprehensive Cancer Network (NCCN) criteria of average (< 1.7%), intermediate (> 1.7%–3.0%) and high risk (> 3.0%). Univariable and multivariable time-to-event models were also established to determine hazard ratios (HRs) for AI risk and breast density.
Key Findings
Breast density was associated with a slight increase in breast cancer probability (HR = 1.16; 95% confidence interval [CI] = 1.11–1.22). Intermediate AI risk alone showed an HR of 2.06 (95% CI = 1.93–2.20) and high AI risk had an HR of 4.49 (95% CI = 4.24–4.76), regardless of breast density.
When factored in together, the HR was 1.12 (95% CI = 1.07–1.17) for breast density–adjusted AI risk, 2.04 (95% CI = 1.92–2.18) for breast density–adjusted intermediate risk, and 4.47 (95% CI = 4.22–4.73) for adjusted high risk.
The study authors suggested that the image-only AI model offers a more precise risk stratification for 5-year risk of breast cancer, and that the findings of the study support the complementary use of Clairity Breast in addition to traditional markers for guiding personalized breast cancer screening.
“An AI image-based risk score can help us identify high-risk women more accurately than traditional methods and determine who may need screening at an earlier age,” said senior author Constance D. Lehman, MD, PhD, Professor of Radiology, Harvard Medical School.
Disclosure: For full disclosures of the study authors, visit rsna.org.
