> Research & Grants
> Grants Program
> Research Grants
> Research Grants Awarded
Research Grants Awarded
Volumetric Estimation of Breast Percent Density from Tomosynthesis Images
Detection, Diagnosis and Prognosis
Studies have shown a 4- to 6-fold increase in breast cancer relative risk (RR) between patients with the highest and lowest percent breast density (PD) measured mammographically. The RR estimates vary in part due to the projective nature of mammograms. Tomosynthesis is a novel radiographic method of producing tomographic images. Breast tomosynthesis provides superior tissue delineation and has improved sensitivity and specificity as compared to mammography. Our long-term goal is to test the hypothesis that PD obtained from tomosynthesis is superior in estimating RR compared with other modalities. As a necessary step, here we propose to develop and evaluate two methods to calculate PD using tomosynthesis; one based on image thresholding similar to existing mammographic methods; the other, a probability-based assessment of breast composition similar to methods developed within our group for MRI PD assessment. Initially, we propose to use physical phantoms for development and evaluation of the PD estimation methods. We have extensive experience with modeling breast tissue anatomy and simulating image acquisition procedures corresponding to various clinical modalities. Accuracy of PD estimation will be assessed by comparison with the ground truth known from our phantom designs. After this phantom-based validation, we will evaluate the developed methods using retrospective analysis of breast images from an ongoing clinical study of high-risk patients at our department. Volunteers in that study have mammograms, breast MRI, and tomosynthesis taken on the same day. Out of the available cases, we will select breast images with adequate visualization of the tissue and without obvious dominant abnormalities, and compute PD using existing methods (for mammograms and MR images) and the new methods developed for breast tomosynthesis. Tomosynthetic PD estimates will be evaluated by testing the hypothesis that the PD estimates from tomosynthesis are equivalent with the PD values obtained using images from other modalities of the same breast. We plan to use the methods developed in this grant application for a future clinical study of breast cancer risk assessed via breast tomosynthesis.
Women with dense breasts have an increased risk of breast cancer. Breast density is measured as the percentage of non-fatty (i.e., dense) tissue in the breast, called percent density (PD). The ability to estimate breast cancer risk is of great importance since it may allow customization of breast cancer detection and treatment, especially for patients at high risk of breast cancer. There are existing methods to calculate PD from mammograms and magnetic resonance images (MRI). PD values obtained using mammography vary widely; in mammography the overlapping 2-dimensional projections of nonadjacent tissue can cause errors in calculating PD. MRI offers 3-dimensional (volumetric) imaging of the breast anatomy and, thus, provides a more accurate estimation of PD. However, breast MRI has lower spatial resolution than mammography, is not widely available, and is not cost effective for cancer screening. Based on these facts, there is a clear need to combine the ability of mammography to capture tissue detail at low cost, with the 3-dimensional clarity of MRI. One such approach is breast tomosynthesis, a novel x-ray technique in which images of parallel slices of the breast are reconstructed from a small number of projection (source) images. Such an approach offers improved volumetric tissue visualization and, thus, improved cancer detection and diagnosis. Tomosynthesis can be performed with simple modifications of standard mammography equipment, and therefore has the potential for use in breast screening, i.e., imaging of asymptomatic women for early detection of cancer. Developing a method to estimate PD from tomosynthesis images is important to better assess risk of breast cancer. Moreover, a successful method to estimate PD is a necessary capability of any candidate breast cancer screening technology. We propose to develop and evaluate two methods for estimating PD from tomosynthesis images; currently no such methods exist. This proposal would also result in the formation of a database of paired multimodality images to allow rapid assessment of future PD estimate methods.