Figure 1.
Best linear regression fit line with 95% confidence interval bands for percentage fibroglandular density (top), log fibroglandular volume (middle), and total breast volume (bottom) for MRI versus either SXA (left), Quantra (center), or Volpara (right) measures.
Solid points correspond to example images in Figure 4.
Table 1.
Description of population characteristics.
Table 2.
Density measures for study participants by method (n = 99).
Table 3.
Percent fibroglandular density quartile ranges by method (n = 99).
Figure 2.
Comparison of quartiles classification for percent fibroglandular density (top) and log fibroglandular volume (middle) for MRI versus SXA (left), Quantra (center), and Volpara (right).
The bottom row of plots show quartiles comparisons between mammographic density measures. Legend at right defines categories of agreement, where either the two compared method's agree completely (black) or are off by one or two quartiles up or down in comparison with the other method.
Figure 3.
Validation of SXA model using breast biology and adipose volume estimates from MRI.
The amount of water volume in the MRI adipose volume was estimated to be 15% of the volume, which is consistent with previous work estimating it to be between 8% [34] and 20% [35]. The MRI model does not include adipose density in the fibroglandular volume while SXA does. Subtracting out the adipose water volume from the SXA fibroglandular volume improved the agreement between SXA and MRI from R2 = 0.78 to 0.83 and removed most of the bias between the measures.
Figure 4.
Six comparisons of the LCC mammograms to their respective left central breast axial-slice MR images on five different women (c and d are the same woman).
The white line connecting points in the MR images define the total breast volume. The MRI fibroglandular volume is shown delineated with white lines without points. Solid data points 4a–4f in Figure 1 correspond to the image labels a–f. Compared to the mammographically-derived SXA values, a) MRI percent density is higher, b) MRI percent density is lower, c) MRI breast volume is higher, d) of the same woman as c (this mammogram not part of analyses, only here and measures plotted in Figure 1 to illustrate one reason for discrepancy between methods' results), MRI breast volume is better segmented due to the breast being extended more into the mammographic image field, e) MRI breast volume is lower, f) all MRI measures of density and volume were in substantial agreement.