AI could help identify early signs of breast cancer years before the disease is diagnosed through conventional screening, according to new research. In a large retrospective study conducted in Sweden, researchers found that commercially available AI systems were able to detect subtle abnormalities in mammograms up to six years before a formal breast cancer diagnosis.
The findings add to a growing body of evidence suggesting that AI could become a valuable tool in breast cancer screening, not only for detecting existing tumors but also for identifying women at increased risk of developing the disease.
Researchers from Karolinska University Hospital analyzed nearly 89,000 mammograms from more than 31,000 women collected over a ten-year period through the Validation of Artificial Intelligence for Breast Imaging (VAI-B) database. The study evaluated three commercially available AI-based computer-assisted detection (AI-CAD) systems and compared their assessments with later clinical outcomes.
AI scores years before diagnosis
The analysis showed that women who were eventually diagnosed with breast cancer had consistently higher AI-generated risk scores in earlier mammograms than women who remained cancer-free. According to senior author Fredrik Strand, approximately one in five breast cancer cases displayed mammographic features that AI systems could recognize as early as six years before diagnosis. The predictive performance increased as the time to diagnosis shortened.
At a specificity level of 90 percent, the AI systems identified potential future cancers in up to 19.7 percent of women six years before diagnosis. This rose to 25.2 percent four years before diagnosis and 39.3 percent two years before diagnosis. The results suggest that subtle imaging changes may be present long before radiologists can confidently identify a developing cancer during routine screening. The study was recently published in Radiology.
Personalized screening
The researchers emphasize that the technology is not intended to replace radiologists. Instead, AI-generated scores could provide an additional layer of information to support clinical decision-making and help identify women who may benefit from closer monitoring.
Sweden’s national breast cancer screening program invites women aged 40 to 74 for mammography every two years, with each examination traditionally reviewed by two radiologists. Integrating AI into this process could potentially enable a more personalized approach, in which screening intervals and follow-up strategies are adapted to an individual’s risk profile.
While further research is needed before such approaches can be implemented in clinical practice, the study highlights AI’s growing role in preventive oncology. Beyond detecting existing tumors, future AI systems may help clinicians understand how breast cancer develops over time and identify disease at an earlier, potentially more treatable stage.
According to the researchers, longitudinal analysis of AI scores could eventually offer valuable insights into the earliest detectable changes associated with breast cancer, opening new opportunities for risk-based screening and earlier intervention.
Breast cancer risk prediction
Last year, researchers at the University of Cambridge developed an AI-based risk prediction tool that can identify women at increased risk of interval breast cancer, a more aggressive form of cancer that emerges between routine mammography screenings. The study analyzed more than 134,000 mammograms from the U.K. breast screening program using the deep learning algorithm Mirai.
The AI generated personalized three-year risk scores based on breast density and subtle imaging features invisible to radiologists. Women in the highest-risk 20% accounted for up to 42% of all interval cancer cases. The model showed its strongest predictive performance within the first year after screening and outperformed traditional risk assessment methods, even among women with dense breast tissue. Researchers believe the technology could support more personalized screening strategies and help target additional imaging to women at greatest risk.
