Compositional Mammography for Breast Cancer Detection

Institution: University of California, San Francisco
Investigator(s): John Shepherd, Ph.D. -
Award Cycle: 2012 (Cycle 18) Grant #: 18IB-0042 Award: $149,895
Award Type: IDEA
Research Priorities
Detection, Prognosis and Treatment>Imaging, Biomarkers, and Molecular Pathology: improving detection and diagnosis

Initial Award Abstract (2012)

Breast cancer early detection screening and diagnosis, primary via mammography, focuses on maximizing the benefits of early detection while minimizing harms. However, one-third of women who are screened with mammography over 10 years have abnormal results, even though no breast cancer is present, thus increasing the costs due re-screening an follow-up procedures. Enhanced imaging techniques are needed to reduce this false-positive rate. Current X-ray mammography measures low energy X-ray attenuation, which cannot classify tissue beyond “dense” or “fatty.” Breast lesions are subjectively classified according to shape and/or the presence of calcifications. Full-field digital mammography (FFDM) and digital tomosynthesis (i.e., producing a series of slices at different depths) systems have advanced detector and electronics hardware with sophisticated image processing methods. However, both techniques use a fixed X-ray energy spectrum for a given breast type and do not achieve potential improvements in image contrast that are afforded by multi-spectral techniques.

Thus, the diagnostic imaging field has stalled on the analysis of lesion shape and microcalcifications. The central hypothesis of this grant project is that the water, lipid, and protein content of breast lesions varies sufficiently according to lesion type to provide clinically useful diagnostic information. We will utilize a novel dual-energy imaging technique, 3-component breast imaging (3CB), to analyze breast lesions of various types. These include invasive ductal carcinoma (IDC), ductal carcinoma in situ (DCIS), fibroadenoma and benign breast tissue. First, we will prospectively measure compositional 3-component breast components in a BIRADS 4 and 5 (i.e., suspicious breast abnormalities that indicate the need for biopsy) cohort of 60 women to characterize invasive breast cancer compared to DCIS and benign findings. Next, our goal is to produce a new “risk model” to exclude some women from biopsy and specifically identify women whose lesions are most likely to be cancer.

We envision this 3CB technology will be used to augment standard mammogram interpretation by a radiologist to estimate the composition of suspicious breast lesions. Because tissue composition is expected to be associated with vasculature and malignant status, we believe that this technique will provide additional information regarding lesion malignant/benign status that is not currently considered during mammography. This imaging technique does not require contrast agents and is expected to be applicable even to dense breasts where sensitivity and diagnostic accuracy, in current practice, are at their lowest.

Final Report (2014)

We had proposed to research a novel mammography technique to measure the composition (amount of water, lipid, and protein) of suspicious breast lesions and compare these measurements to the lesion type. We call this technique “3CB Mammography”. The expected result as that different lesion types contain different cell types resulting in unique compositions. For example, magnetic resonance imaging diagnoses invasive cancer primarily through variations in the number of blood vessels in malignant lesions; blood vessels, for instance, are expected to contain additional water content. Our research is highly relevant to breast cancer because we expect it to improve the sensitivity and specificity of diagnostic mammography, thereby reducing the number of women who are unnecessarily biopsied and subjected to its associated physical and mental harms.

Central hypotheses of the research:
The water, lipid, and protein content of different types of breast lesions are different enough that they can be used to improve the accuracy of mammography.

The general methodology:
We imaged women with suspicious mammographic lesions (BI-RADS 4 or 5) using a using a special mammography method we developed to estimate water, lipid, and protein content. We estimated the water, lipid, and protein content of the lesion and compare that estimate to the lesion type.

Outcome of the project:
Our goal was to recruit 60 women. We achieved this goal. All women received a biopsy to determine lesion type. X-ray mammography relies on qualitative, shape-based information to identify cancer; this includes the presence of an attenuating mass, the shape of the mass edges, and the presence of calcifications. Tumors are often identifiable using this information. Unfortunately, some benign and malignant lesions look qualitatively similar. The result is that only 1 in 4 biopsies actually find cancer. Our study sample contained 9 invasive breast cancers,17 ductal carcinoma in situ (DCIS), 9 fibroadenomas, and 75 benign findings (note that some women had more than one biopsied lesion.) We found that the 3CB information was able to identify the difference between the two types of lesions that should be biopsied (invasive and DCIS) versus the benign lesions (fibroadenomas and other benigns). Furthermore, we found that the 3CB information was largely independent of the shape-based information normally used to identify cancerous lesions. The key innovation of this project was the development of a method to estimate quantitative compositional information of breast tissue using data obtainable from an X-ray mammography system. This was achieved and has been since applied to a larger NIH study of 600 women. 16 unique measures of lipid, water, and protein were measures for a total of 48 measures in and around each lesion. Our findings suggests that this approach can change the fundamental paradigm of diagnostic breast imaging; breast lesions would be imaged using the 3CB protocol to provide a quantitative assessment of the probability of malignancy before a biopsy. This would result in a more accurate determination of who would benefit from an immediate biopsy and who may opt to either forego or delay the biopsy of the lesion. This new approach is expected to circumvent the intrinsic limitations of diagnosing lesions based on shape alone and result in significantly increased utility for mammography.

Mammogrphic quantitative image analysis and biologic image composition for breast lesion characterization and classification.
Periodical:Medical Physics
Index Medicus: Med Phys
Authors: Drukker K, Duewer F, Giger ML, Malkov S. Flowers CI et al.
Yr: 2014 Vol: 41 Nbr: 3 Abs: Pg:0310915