Keywords:
Computer Applications-Detection, diagnosis, Imaging sequences, Mammography, Breast, Screening, Cancer, Quality assurance
Authors:
X. Lin, N. Sauber, R. Highnam; Wellington/NZ
DOI:
10.1594/ecr2013/C-1770
Results
Volpara™ generates consistent results across different x-ray vendors
We observed consistent volumetric breast density measurements,
as determined by Volpara™,
from mammograms taken on the same woman using different x-ray systems (see Figure 1).
A 0.940 and 0.974 correlation was observed using Volpara™ on the Hologic and GE,
and Fuji CR and Fuji DR datasets,
respectively.
Given that the images were taken at different timepoints,
some genuine variation should be expected.
Volpara™ generated density maps
The density maps generated by Volpara™ provided us with an excellent tool for the analysis of dense tissue changing over time,
as they essentially standardized mammograms taken on different x-ray machines.
“For Presentation” images are variable due to different processing algorithms used by each x-ray vendor.
A single case was selected to illustrate the difficulties in using the “For Presentation” mammograms for temporal studies.
We have shown the “For presentation” mammograms (Figure 2) and the Volpara™ density maps (Figure 3) from a woman who had yearly mammograms taken over eight years,
on a combination of GE,
Hologic and Siemens x-ray systems.
The accompanying graphs showing the volumetric breast densities over time clearly do not correspond well visually,
to the “For Presentation” mammograms.
In contrast,
using the Volpara™ density maps clearly showed the decrease in density as this particular woman aged.
Temporal movies
Using the same example case as above,
two movies were generated by appending either the eight density maps (Figure 4) or the eight “For Presentation” images (Figure 5).
The normalization and image mapping processes worked extremely well and is evident in the temporal movie of our density map.
The contrast levels in the mammograms are consistent,
even across multiple x-ray systems,
and the image mapping processing effectively aligned the appearance of each breast in the sequential mammograms.
One can also clearly see the impact of using “For Presentation” mammograms without any prior modification on assessing changes in density.
Three radiologists independently confirmed that our temporal density maps offer an easier and more accurate assessment of breast density changes.