This poster is published under an
open license. Please read the
disclaimer for further details.
Keywords:
Breast, Mammography, Dosimetry, Dosimetric comparison
Authors:
C. Tromans, A. Chan, R. Highnam; Wellington/NZ
DOI:
10.1594/ecr2014/C-0360
Methods and materials
Raw digital mammograms from 102 women with a wide-range of breast densities with standard four-view temporal mammograms were analyzed (i.e.
left and right craniocaudal and mediolateral oblique views).
96 women were imaged one year apart on GE Healthcare (GE) or Hologic Selenia (Hologic) x-ray systems and 6 women were imaged over several years on a mix of GE,
Hologic and Siemens Novation systems.
All images were run through automated breast density assessment software (VolparaDensity™),
to obtain the volumetric breast density (VBD),
breast volume (BV) and,
at each pixel in the image,
an estimate of the volume of dense tissue between that pixel and the x-ray source (a “density map”).
Our intent was to use Dance’s algorithm for dose estimation [6-9] but,
rather than use his rough estimate of glandularity from the breast thickness,
we sought to use the woman’s actual glandularity,
which we can compute from the density map.
The VBD for a woman is the percentage of fibroglandular tissue within the entire breast,
this is typically a low number (0-35% when the skin is excluded),
whereas Dance defines glandularity as being the percentage of glandular tissue inside the glandular disk.
To generate “Dance’s glandularity” we focused only on a central portion of the breast,
and we excluded the top and bottom layers of subcutaneous fat.
Figure 1 shows the relationship between our estimates of Dance’s glandularity and breast thickness for a large US dataset (blue dots),
compared to the relationship reported by Dance et al [6-9] for two separate age ranges.
The figure shows that the values used by Dance appear to represent mid-range values for all the potential glandularities which might exist for each breast thickness.
We use average dose estimates and Pearson Correlation Coefficients (PCCs) to compare manufacturer reported doses to patient-specific doses,
and to compare dose estimates over time for consistency.
Unless otherwise stated,
“average” refers to mean values.