Clinical Report: An Advanced Deep Learning Model for Assessing Breast Cancer Risk

Overview

This study evaluates a deep learning (DL) model, Mirai, for predicting breast cancer risk compared to traditional breast density assessments. The findings suggest that the DL model provides superior risk stratification, potentially enhancing supplemental screening accuracy.

Background

Breast density is a significant factor in breast cancer risk and can obscure cancer detection on mammograms. Current policies based on binary density classification may lead to overuse of resources and increased false positives. Advanced risk prediction tools like the DL model could offer more personalized screening strategies, addressing these limitations.

Data Highlights

Key Findings

• The DL model outperformed traditional breast density assessments in predicting future breast cancer risk.
• It effectively stratified risk across a diverse population, reducing interreader variability.
• The study included a large cohort of 123,091 mammograms from 67,019 patients.
• Deep learning models capture a broader range of risk-relevant imaging features compared to density alone.
• Results indicate potential for improved access to supplemental imaging based on individualized risk rather than binary density classification.

Clinical Implications

The implementation of the DL model could lead to more accurate breast cancer risk assessments, allowing for tailored screening strategies. This may reduce unnecessary imaging and associated costs, particularly for underserved populations.

Conclusion

The advanced DL model presents a promising alternative to traditional breast density assessments, enhancing the precision of breast cancer risk stratification and potentially improving screening outcomes.

References

1. FDA, Important Information: Final Rule to Amend the Mammography Quality Standards Act (MQSA), 2024 -- FDA's new regulations on breast density notification
2. Artificial Intelligence Risk Model (Mirai) Delivers Robust Generalization and Outperforms Tyrer-Cuzick Guidelines in Breast Cancer Screening - PubMed, 2022 -- Study on DL model performance
3. European Radiology, Evaluating Breast Cancer Risk for Screening Using a Combined Artificial Intelligence Method, 2025 -- AI methods in breast cancer screening
4. asco ai in oncology — Breast Cancer Recurrence Risk Determined by Deep Learning Model Trained on Histopathologic Slides
5. the asco post — Breast Cancer Recurrence Risk Determined by Deep Learning Model Trained on Histopathologic Slides
6. the asco post — Large AI Breast Cancer Screening Trial Increases Detection Rate by 20%
7. ASCO AI in Oncology - Breast Cancer Recurrence Risk Determined by Deep Learning Model
8. The ASCO Post - Large AI Breast Cancer Screening Trial Increases Detection Rate by 20%
9. Important Information: Final Rule to Amend the Mammography Quality Standards Act (MQSA) | FDA
10. Cost-Effectiveness of Magnetic Resonance Imaging Screening for Women With Extremely Dense Breast Tissue | JNCI: Journal of the National Cancer Institute | Oxford Academic
11. Artificial Intelligence Risk Model (Mirai) Delivers Robust Generalization and Outperforms Tyrer-Cuzick Guidelines in Breast Cancer Screening - PubMed