A new study published in Radiology introduces an artificial intelligence system called AsymMirai, which can forecast breast cancer risk up to five years in advance. The model analyzes mammogram images by comparing the left and right breasts, looking for tiny asymmetries that may signal early biological changes linked to cancer risk.

According to researchers from the Massachusetts Institute of Technology (MIT) and Massachusetts General Hospital (MGH), AsymMirai is a simpler and more transparent version of earlier deep learning models. Instead of scanning all image details, it focuses only on the differences between the two breasts, an approach that makes it easier for doctors to understand why a particular patient is considered high risk.

Results from the study showed that AsymMirai could predict long-term breast cancer risk with an AUC (area under the curve) of about 0.66 for five-year forecasts. While this performance is modest compared to more complex systems, it proves that meaningful prediction is possible using interpretable image features alone.

Experts say this innovation could be a step toward more personalized breast cancer screening. By identifying women at higher risk earlier, health systems can schedule closer monitoring, preventive testing, or lifestyle interventions to catch cancer before it develops.

Every year, breast cancer affects about one in eight women, and approximately one in 39 will die from it. Traditional mammograms, while essential, can miss up to 20 percent of cases especially in women with dense breast tissue. Tools like AsymMirai may help close that gap by providing doctors with additional insight long before visible tumors form.

While the technology is still in the research phase and not yet used in clinics, its potential is enormous. AsymMirai shows that artificial intelligence can do more than detect existing disease, it can help predict and prevent it, saving lives and reducing healthcare costs in the future.

Sources:

Yala A. et al. (2024). Interpretable Artificial Intelligence for Breast Cancer Risk Prediction Using Bilateral Mammographic Asymmetry. Radiology. PubMed

Radiological Society of North America News Release. News-Medical.net

National Cancer Institute (U.S.). Breast Cancer Statistics

American Cancer Society. Limitations of Mammograms