Yiqiu Shen, Assistant Professor at NYU Grossman School of Medicine, shared on LinkedIn:
“Excited to share our new paper in the American Journal of Roentgenology on predicting long-term breast cancer risk using longitudinal digital breast tomosynthesis (DBT)!
DBT is now widely used for breast cancer screening. Beyond detecting existing cancers, these 3D images may contain valuable information about a woman’s future breast cancer risk.
The challenge
Most AI-based breast cancer risk models rely on 2D mammograms from a single screening visit, potentially missing important information from 3D breast tissue structure and changes over time.
In this study, we developed a deep learning model that combines current and prior DBT examinations with age and breast density to predict breast cancer risk over 2–5 years.
Highlights
- 313,335 DBT examinations from 161,077 women
- 5-year AUC of 0.721 on an independent test set of 34,570 examinations
- Longitudinal imaging improved prediction over a single examination (AUC 0.721 vs. 0.707, p < 0.001)
- 39.7% of examinations in women with extremely dense breasts were classified as average risk, with only 0.8% observed 5-year cancer incidence
- 14.8% of examinations in women with fatty breasts were classified as high risk, with 2.6% observed 5-year cancer incidence
Clinical impact
Our findings suggest that routine DBT examinations can support more personalized breast cancer risk assessment beyond breast density alone. With further validation, this approach could help guide supplemental screening and identify women who may benefit from closer surveillance.
Congratulations to authors and all collaborators across Center for Advanced Imaging Innovation and Research (CAI2R) at NYU Langone Health and NYU Center for Data Science!
Acknowledgments
We are grateful for the generous support from the National Institute of Biomedical Imaging and Bioengineering (NIBIB), the Milstein Pilot Project Fund, the Shifrin-Myers Breast Cancer Discovery Fund, and the MANHASSET WOMENS COALITION AGAINST BREAST CANCER INC. Their support has been instrumental in making this research possible.”
Title: Predicting 5-Year Breast Cancer Risk From Longitudinal Digital Breast Tomosynthesis: A Single-Center Retrospective Study
Authors: Yanqi Xu, Laura Heacock, Jungkyu Park, Felicia L. Pasadyn, Qi Lei, Alana Lewin, Krzysztof Geras, Linda Moy, Freya Schnabel, Yiqiu Shen.
Read the full article.
This also can be interesting:
