Many women who are considered at high risk of developing breast cancer struggle to choose enhanced surveillance or risk-reducing interventions, which can be highly invasive and have negative side effects. Such decisions are highly personal, requiring accurate quantification of individual risk and response characteristics to risk-reducing interventions. Current breast cancer risk assessment in the clinic is imprecise at the individual level. A personalized risk assessment that incorporates a woman's particular risk profile, such as anatomical, functional, or biological characteristics of her breast, can help to individualize breast cancer risk management. Mammographic breast percentage density (MPD) has been an established independent risk factor. Recent advances in breast magnetic resonance imaging (MRI) provide exquisite and high-resolution capabilities to characterize in-vivo properties of breast tissue that are related to breast cancer risk. Recent studies, including our pilot quantitative studies, indicate that: 1) volumetric fibroglandular (i.e., dense) tissue (FGT), contrast enhancement of FGT (aka background parenchymal enhancement [BPE]), and enhancement kinetics computed on normal (not cancerous) breast tissue, all measured from dynamic contrast-enhanced MRI (DCE- MRI), are predictive of breast cancer risk; and 2) changes of BPE and FGT after risk-reducing interventions measure patients' responses to applied intervention. We propose to investigate the clinical utility of breast MRI- based quantitative measures as new non-invasive breast cancer risk factors. Our hypothesis is that objectively quantified BPE, kinetics, and FGT measured on breast DCE-MRI are biomarkers of breast cancer risk and response to risk-reducing interventions, providing predictive value independent of MPD. We will optimize our automated computer algorithms and retrospectively analyze the DCE-MRI scans of 600 women in a case- control setting, including analysis of longitudinal MRI scans acquired over an 8-year timeframe. We will assess the MRI-derived measures as a response biomarker to risk-reducing interventions (e.g., salpingo- oophorectomy, or tamoxifen/raloxifene). We have achieved strong preliminary results across all of the proposed aims. This project will combine the multi-disciplinary expertise of a computational imaging scientist, radiologists, a medical oncologist (breast cancer high-risk program director), and a biostatistician. This study is the first of its kind that uses fully automated computerized analysi to develop significant breast DCE-MRI- derived risk biomarkers. Quantitative DCE-MRI-based biomarkers will advance our understanding of intrinsic breast characteristics pertaining to individual risk profiles. This study will provide strong data and rationale for incorporating quantitative breast DCE-MRI-derived biomarkers to more accurately assess breast cancer risk and to aid in the decision-making regarding risk-reducing interventions, all at the individual leve. This study will optimize the use of a large volume of breast DCE-MRIs that are routinely performed in major medical centers; the outcome of this study is therefore highly translational to the clinic.

Public Health Relevance

This study will develop and evaluate significant breast MRI-derived breast cancer risk biomarkers independent of mammographic breast density for women at high risk of developing breast cancer. This work will provide high-risk women with a more personalized risk assessment, which could aid in their clinical decision-making regarding risk-reducing interventions. Our fully automated computer methods will enable reproducible and objective risk assessment. Large volumes of breast DCE-MRIs are routinely performed in major medical centers for screening and diagnostic purposes. This study will optimize the use of these breast MRIs and, as a result, the outcome of this study will be highly translational to the breast cancer clinic.

National Institute of Health (NIH)
National Cancer Institute (NCI)
Research Project (R01)
Project #
Application #
Study Section
Special Emphasis Panel (ZRG1)
Program Officer
Zhang, Huiming
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
University of Pittsburgh
Schools of Medicine
United States
Zip Code
Mohamed, Aly A; Luo, Yahong; Peng, Hong et al. (2018) Understanding Clinical Mammographic Breast Density Assessment: a Deep Learning Perspective. J Digit Imaging 31:387-392
Aboutalib, Sarah S; Mohamed, Aly A; Berg, Wendie A et al. (2018) Deep Learning to Distinguish Recalled but Benign Mammography Images in Breast Cancer Screening. Clin Cancer Res 24:5902-5909
Mohamed, Aly A; Berg, Wendie A; Peng, Hong et al. (2018) A deep learning method for classifying mammographic breast density categories. Med Phys 45:314-321
Ma, Wenjuan; Zhao, Yumei; Ji, Yu et al. (2018) Breast Cancer Molecular Subtype Prediction by Mammographic Radiomic Features. Acad Radiol :
Wu, Shandong; Zuley, Margarita L; Berg, Wendie A et al. (2017) DCE-MRI Background Parenchymal Enhancement Quantified from an Early versus Delayed Post-contrast Sequence: Association with Breast Cancer Presence. Sci Rep 7:2115
Wu, Shandong; Berg, Wendie A; Zuley, Margarita L et al. (2016) Breast MRI contrast enhancement kinetics of normal parenchyma correlate with presence of breast cancer. Breast Cancer Res 18:76