Exhibit Hall | Forum 8
Purpose: To generate and validate patient-derived, heterogeneous digital breast phantoms for advanced breast dosimetry in mammography and digital breast tomosynthesis.
Methods: The proposed phantoms were developed starting from realistic shapes of compressed breasts, representing the two standard views, cranio-caudal(CC), and medio-lateral-oblique(MLO). Internally, the breast phantoms were defined as consisting of an adipose/fibroglandular tissue mixture, with a non-spatially uniform relative concentration. For the MLO view, the pectoralis muscle was defined starting with no thickness at the line of the nipple, increasing in width and thickness towards the cranial edge of the image. Various instances of the phantom were created with different thicknesses, overall glandularity, and maximum pectoralis muscle thickness. Using quadratic fits of the normalized Dg deviations from the mean for the different phantoms, the most appropriate “average” geometric model for the muscle shape, from a dosimetric point of view, was determined. Using a validated Monte Carlo dosimetry simulation, the Dg and normalized glandular dose (DgN) were estimated for each breast phantom, against 88 compressed breast CT patient phantoms that reflect the real distribution of the fibroglandular tissue in the breast, and compared with homogenous models as currently used for breast dosimetry applications.
Results: For each breast thickness/glandularity combination, the estimated Dg showed little sensitivity to the pectoralis muscle thickness. The pectoralis muscle thicknesses at the image edge that minimized the deviation from the mean dose values were 70% of the total breast tissue thickness. The DgN absorbed by the proposed phantoms was concordant with that absorbed by the breast CT patient phantoms, with errors of 5%(CC) and 4%(MLO), and was over 30% lower than that absorbed by the homogenous models.
Conclusion: The pectoral thickness profile has a small impact on the resulting dose estimates. The developed phantoms can be used for dosimetry simulations to improve the accuracy of dose estimates.
Funding Support, Disclosures, and Conflict of Interest: This research was supported in part by grant R01CA181171 from the National Cancer Institute, National Institutes of Health, and grant IIR13262248 from the Susan G. Komen Foundation for the Cure. The authors would like to thank the Comision Sectorial de Investigacion Cientifica (CSIC) under project C681 in Uruguay.