Refining and expanding breast dosimetry for the Sacramento Area Breast Imaging Registry

Abstract The SABIR (Sacramento Area Breast Imaging Registry) was created to provide insights regarding the best approaches for breast cancer screening, surveillance, and to compare the interpreƟve performance of specific faciliƟes relaƟve to naƟonal benchmarks. The resulƟng predicƟve models are someƟmes

weighed by the cancer risk associated with the ionizing radiaƟon used in breast x‐ray imaging modaliƟes. To date, the SABIR database contains average glandular dose (𝐷𝑔) values that are extracted directly from the DICOM header of that exam. The models employed by the vendors to esƟmate 𝐷𝑔 are out of date

and mistakenly assume that the breast is composed of a homogenous mixture of 50% glandular and 50%

adipose Ɵssue (hereaŌer referred to as “𝐷𝑔,ℎ𝑜𝑚𝑜”). Over the last decade, advancements in 3D breast imaging have greatly advanced our understanding of breast anatomy, parƟcularly the amount and distribuƟon of glandular Ɵssue, which is vital for accurate dosimetry. The findings revealed that prior

esƟmaƟons of radiaƟon dose for digital mammography (DM) and digital breast tomosynthesis (DBT) were overesƟmated by about 30% due to not accounƟng for the realisƟc heterogenous distribuƟon of glandular Ɵssue. If we use the linear no threshold (LNT) model to predict risk as a funcƟon of dose, these

findings suggest that the radiaƟon risks of breast cancer screening and diagnosis using DM and DBT are lowered by this same factor (30%). We therefore propose to generate new 𝐷𝑔values for the SABIR

database using the now universally accepted, realisƟc heterogenous breast model “𝐷𝑔,ℎ𝑒𝑡𝑒𝑟𝑜” and

explore characterisƟcs of 𝐷𝑔,ℎ𝑒𝑡𝑒𝑟𝑜 in digital breast tomosynthesis DM and DBT across breast density, size, and paƟent BMI. Specific Aim 1: Develop new esƟmaƟons of mean glandular dose using the heterogenous breast dosimetry model (𝑫𝒈,𝒉𝒆𝒕𝒆𝒓𝒐) in both DM and DBT for the SABIR database: The breast density and

DICOM metadata for DM and DBT imaging examinaƟons will be used to generate new 𝐷𝑔,ℎ𝑒𝑡𝑒𝑟𝑜 using the universal heterogeneous breast model published by EFOMP/AAPM Joint Trask Group 282.

𝐷𝑔,ℎ𝑒𝑡𝑒𝑟𝑜values will be compared against exisƟng legacy 𝐷𝑔,ℎ𝑜𝑚𝑜 values across the range of acquisiƟon characterisƟcs and breast composiƟons found in the SABIR database.

Specific Aim 2: Explore how 𝑫𝒈,𝒉𝒆𝒕𝒆𝒓𝒐 values in DM and DBT vary with breast density, compressed breast thickness, and paƟent BMI found in the SABIR database: Effects of breast density, compressed breast thickness, and BMI will be explored for both DM and DBT examinaƟons.