Validation of a breast density quantization software with 3D printed breast phantoms

Breast density has been shown to be a significant and independent estimator of breast cancer risk and can mask lesions from being visible in mammograms. Therefore, the accuracy of estimates of breast density are important, both for evaluating the risk of breast cancer development and for estimating the potential efficacy of imaging. One commercial breast density estimation software assesses the volume of dense tissue based on modeling of the image acquisition process using the information present in the raw ("for processing") mammogram. Although this software has been extensively validated for precision and to a limited extent for accuracy using other imaging modalities (e.g., breast MRI), direct validation of its accuracy on mammograms has not been reported. The aim of this work is to evaluate the accuracy of the breast density estimates of this commercial density quantization software. For this, five different 3D printed phantoms were used, each obtained from an actual patient mammogram. Mammograms of these phantoms were acquired and analyzed using the density software. In parallel, the equivalent breast density of the phantoms was estimated independently using the known design of the phantoms used for the 3D printing. The spectra used by the automatic exposure control for each mammogram was accurately characterized and modeled using a previously published spectral model. It was found that the estimates of the amount of dense tissue replacing the estimated fully adipose areas of the breast are accurate to within an infra-phantom relative mean of 10% (std. dev. <4.4%), with negligible bias.