Marcela Bohm-Velez MD, Abstract Co-Author: Consultant, Koninklijke Philips NV Consultant, Matakina Technology Limited

Ralph Philip Highnam PhD, Presenter: CEO, Matakina Technology Limited CEO, Volpara Solutions Limited

Ariane Chan PhD, Abstract Co-Author: Employee, Matakina Technology Limited

Thomas S. Chang MD, Abstract Co-Author: Nothing to Disclose

Interpreting large volumes of mammographic site data for quality assurance is complicated by differences in imaging systems and population characteristics. In this study, we used a novel system-wide approach to determine whether GE and Hologic mammographic units  under- or over-estimated mean glandular dose (MGD) compared to a personalized estimate.

Mammographic images (11,254 images; 2864 studies) from December 2006 to March 2014 were retrospectively analyzed in our practice using automated quality assurance software (VolpraAnalyticsTM). Volumetric breast density (VBD) characteristics and a personalized estimate of MGD (P-MGD) were obtained from the raw images. Facility and image data were automatically extracted from the image headers (e.g. manufacturer-reported MGD (M-MGD), detector ID and vendor). Average MGD values were compiled across individual mammography units (3 GE and 2 Hologic). Differences between P-MGD and M-MGD were assessed by mammography unit, vendor and patients's VBD characteristics.

Overall, M-MGD significantly underestimated dose compared to the P-MGD (1.47 and 1.58 mGy, respectively; p=0.014). When stratified by vendor, the difference between M-MGD versus P-MGD estimates were 0.03 mGy (p<0.05) and 0.24 mGy (p<0.001) for the GE and Hologic units respectively. Subanalysis of one GE and one Hologic unit found that M-MGD and P-MGD estimates were very similar for the GE unit (1.44 and 1.46 mGY; p=0.381), but significantly different for the Hologic unit (1.60 and 1.84 mGy, respectively; p<0.001) despite similar VBD and breast volumes for both patient groups. The differences in vendor dose algorithms effectively masked some of the variation in dose between mammography units, albeit other influencing factors, such as compression, were also present.

Stratification of radiation dose data by mammography unit highlighted significant differences in MGD estimates between GE and Hologic units which can be attributed, in part, to the different dose algorithms used by each manufacturer. P-MGD estimates, which incorporate both individual breast densities and a manufacturer-independent dose algorithm, are essential for standardizing the assessment of radiation dose.

In monitoring radiation dose, a system-wide approach incorporating breast density characteristics allows standardization of MGD between mammographic units, for improved breast imaging safety.

Bohm-Velez, M, Highnam, R, Chan, A, Chang, T, System-Wide Monitoring of Mammographic Radiation Dose for Quality Assurance.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14008100.html