Due to the high prevalence of breast cancer among women, much is being done to detect breast cancer
earlier and more accurately. In current clinical practice, the most widely-used mode of breast imaging is
mammography. Its main advantages are high sensitivity and low patient dose, although it is still merely a two-dimensional
projection of a three-dimensional object. In digital breast tomosynthesis, a three-dimensional image of
the breast can be reconstructed, but x-ray projection images of the breast are taken over a limited angular span.
However, the breast tomosynthesis device itself is more similar to a digital mammography system and thus is a
feasible replacement for mammography. Because of the angular undersampling in breast tomosynthesis, the
reconstructed images are not considered quantitative, so a worthwhile question to answer would be whether the
voxel values (VVs) in breast tomosynthesis images can be made to indicate tissue type as Hounsfield units do in CT.
through some image processing scheme. To investigate this, simple phantoms were imaged consisting of layers of
uniform, tissue-equivalent plastic for the background sandwiching a layer of interest containing multiple, small
cuboids of tissue-equivalent plastic. After analyzing the reconstructed tomosynthesis images, it was found that the
VV in each lesion increases linearly with tissue glandularity. However, for the two different x-ray tube energies and
for the two different beam exposure levels tested, the trend-lines all have different slopes and y-intercepts. Thus,
breast tomosynthesis has a definite potential to be quantitative, and it would be worthwhile to study other possible
dependent parameters (phantom thickness, overall density, etc.) as well as alternative reconstruction algorithms.