Comparison of Total X-Ray Energy Absorbed by Fibroglandular Tissue and its Mass Derived from Full Field Digital Mammography and Digital Breast Tomosynthesis

Abstract

This study investigated whether total x-ray energy absorbed by fibroglandular tissue and fibroglandular tissue mass derived from raw and clinical full field digital mammography (FFDM) images corresponded with the respective quantities derived from digital breast tomosynthesis (DBT). A total of 840 paired FFDM and DBT observations were analyzed separately for the craniocaudal (CC) and mediolateral oblique (MLO) orientations. Total absorbed x-ray energy and fibroglandular tissue mass were calculated using an automated breast density detection method in conjunction with the recorded average glandular dose. Relationships between FFDM and DBT quantities were evaluated in the log-log domain using linear and quadratic models, with model performance assessed by R2, adjusted R2, mean squared error (MSE), chi-square goodness-of-fit tests, and residual analysis. Strong associations were observed between FFDM and respective DBT quantities. For raw FFDM, linear energy models produced R2 values of 0.734 and 0.901 for the CC and MLO views, respectively, while the corresponding mass models produced R2 values of 0.634 and 0.870. Inclusion of a quadratic term improved model performance modestly. For clinical FFDM, linear energy models produced R2 values of 0.854 and 0.919, while the corresponding mass models produced R2 values of 0.721 and 0.812. The first-order coefficients for the clinical-image energy models were approximately unity (0.99-1.05), and inclusion of the quadratic term produced negligible changes in model performance and residual distributions. The goodness-of-fit tests did not reject adequate fit for any model. Clinical FFDM images can therefore provide estimates of total absorbed x-ray energy and fibroglandular tissue mass that are strongly associated with corresponding quantities derived from DBT. The approximately unit first-order coefficients and minimal benefit from quadratic modeling support a simple linear log-log calibration for clinical FFDM. This approach may facilitate combining FFDM and DBT based data in larger retrospective and longitudinal studies of cumulative mammographic x-ray exposure. These findings demonstrate proof-of-concept internal consistency but do not establish absolute dosimetric accuracy.