Related Experiment Video
Updated: Jun 5, 2025
15:48
Tracking the Mammary Architectural Features and Detecting Breast Cancer with Magnetic Resonance Diffusion Tensor Imaging
Published on: December 15, 2014
22.4K
Volumetric Breast Density Estimation From Three-Dimensional Reconstructed Digital Breast Tomosynthesis Images Using
Vinayak S Ahluwalia1,2,3, Nehal Doiphode4, Walter C Mankowski4
1Center for Biomedical Image Computing and Analytics (CBICA), University of Pennsylvania, Philadelphia, PA.
JCO Clinical Cancer Informatics
|December 9, 2024
Summary
A new deep learning model can estimate breast density from 3D digital breast tomosynthesis images, identifying women at higher risk for breast cancer diagnosis.
Area of Science:
- Radiology and Medical Imaging
- Artificial Intelligence in Healthcare
- Oncology
Background:
- Breast density is a known risk factor for breast cancer.
- Digital breast tomosynthesis (DBT) is increasingly used for screening.
- Current methods for estimating volumetric breast density (VBD) from DBT are limited by data requirements.
Purpose of the Study:
- To develop and validate a deep learning (DL) model for estimating VBD from 3D reconstructed DBT images.
- To assess the association between DL-derived VBD measures and breast cancer diagnosis.
Main Methods:
- Retrospective analysis of 1,080 DBT screening examinations.
- Development of a DL model to segment dense and fatty breast tissue.
- Estimation of %VBD and absolute dense volume (ADV) using the DL model.
- Logistic regression to correlate VBD measures with contralateral breast cancer diagnosis.
Main Results:
- The DL model demonstrated good agreement with reference segmentations (Dice scores 0.88 and 0.76).
- DL-derived VBD (%VBD and ADV) was significantly associated with increased odds of breast cancer diagnosis (ORs 1.41 and 1.45).
- Area under the curve (AUC) for breast cancer diagnosis ranged from 0.65 to 0.67.
Conclusions:
- Deep learning-derived breast density measures from 3D DBT images are a viable tool for risk assessment.
- This approach can help identify individuals at higher risk for breast cancer.
- The findings support the routine estimation of VBD in breast cancer screening using DBT.
