Volumetric breast composition analysis: reproducibility of breast percent density and fibroglandular tissue volume measurements in serial mammograms

Abstract

Volumetric breast composition analysis represents a useful tool for assessing changes in breast composition over time. However, no data exist on the reproducibility of this method in serial mammograms. To assess the reproducibility of two volumetric breast composition parameters, breast percent density (PD) and fibroglandular tissue volume (FTV), in consecutive mammograms. Volumetric breast composition analysis to determine PD and FTV was performed in two consecutive unilateral mammograms of 211 patients. All mammograms were obtained on the same digital mammography unit within a maximum interval of 24 months. Volumetric data for analysis for both examinations were available for 174 patients. Thirty-two patients had successful volumetric analysis of additional consecutive examinations on a second digital mammography unit. Inter-examination correlation of measurements and absolute differences were analyzed. Bland-Altman analysis was performed to compare readings from different mammography units. Mean FTV remained constant over the study period. A reduction in PD of 0.5% and a mean increase in breast volume (BV) of 3% were observed. FTV measurements obtained on the same mammography unit were significantly more reproducible than PD measurements (Pearson correlation coefficients of 0.947 and 0.920, respectively; P < 0.05). A 15% difference between mean absolute volume measurements (FTV and BV) obtained on different mammography units was observed (P ≤ 0.001), while mean PD was close to the expected value. Volumetric breast composition analysis is highly reproducible in serial mammograms in normal women. FTV is a more reproducible parameter than PD, indicating that absolute quantification of breast parenchyma may be preferable to the measurement of relative parameters such as PD. However, a disadvantage of using FTV is that it is susceptible to systematic differences when measurements are obtained on different imaging platforms.