This poster is published under an
open license. Please read the
disclaimer for further details.
Keywords:
Breast, Mammography, Technology assessment, Cancer
Authors:
C. Tromans1, R. Highnam1, O. Morrish2, R. Black2, L. Tucker2, F. J. Gilbert2; 1Wellington/NZ, 2Cambridge/UK
DOI:
10.1594/ecr2014/C-0363
Methods and materials
Digital breast tomosynthesis (DBT) uses a series of low-dose x-ray projections at various angles,
to reconstruct a pseudo-3D image of the breast.
There are several DBT systems available on the global market,
but only one with FDA approval (at time of writing): Hologic’s Selenia Dimensions 3D.
Initial FDA approval of the Dimensions System,
for screening,
required it to function in “combo” mode,
where conventional 2D mammograms are performed at the same time as the series of low-dose x-ray projections.
FDA approval has now been granted to allow the use of synthesized or “C-View” images rather than the conventional 2D mammograms.
In this study,
raw images (all anonymized) from 20 women imaged in combo-mode on a Hologic DBT unit were selected,
which represent a range of BIRADS density categories as visually assessed on the conventional images.
All images were acquired in combo-mode,
so that the 2D and 3D images were acquired under the same breast compression and at (approximately) the same time.
Volpara™ was used to automatically estimate breast volume,
volumetric breast density (VBD) and BIRADS density category from the raw conventional mammograms and from the raw DBT projections.
As noted,
DBT is simply a series of low-dose x-ray projections and those low-dose projections can be exported from a DBT system in “raw” format which means that the manufacturer has not overly manipulated the data for visualization purposes.
To run Volpara over DBT images,
software was written to simply extract a raw projection image and populate it with the appropriate DICOM tags before running Volpara.
To compare the density results from the 2D and 3D images,
Pearson Correlation Coefficients (PCCs) for breast volume (BV),
fibroglandular volume (FGV) and VBD were calculated.
Volumetric parameters were calculated as the mean across a typical four-view study.
Student’s t-test was used to compare whether there were any significant differences between measuring VBD from conventional mammograms or DBT projections.
A confusion matrix showing the agreement between BI-RADS categories as determined from the conventional mammograms versus the DBT images,
was also generated.