• monte carlo simulation
• medical physics
• medical imaging
• geant4
• digital breast tomosynthesis
• breast dosimetry

In this thesis work are introduced a new method and a new formalism to calculate, via Monte Carlo (MC) simulations, dose conversion factors for asserting the dose delivered to breast in a Digital Breast Tomosynthesis (DBT) exam. The simulations were carried out through a C++ Geant4 based software that I've developed and validated. This work has been developed in the framework of a collaboration between University of Pisa, University of Naples Federico II and INFN (Istituto Nazionale di Fisica Nucleare) on breast dosimetry in DBT. Breast cancer is one of the most common cancer among the population, accounting for 25% of total cancers per year in the female population. Since screening programmes are based on the use of X-ray radiation, there is always the risk of radiation induced cancer, therefore an accurate estimation of the radiation dose absorbed by the patients breast is essential. The DBT, a pseudo 3D imaging technique, has been introduced and, in recent years, many instruments able to perform also DBT have been successfully employed in breast imaging. The current DBT dosimetric protocols were formulated as an extension to the DM ones. Various correction factors were calculated to reuse the dose coefficients already calculated for DM. Furthermore, these coefficients were calculated via Monte Carlo simulations employing a standard geometry for the DBT system model, disregarding some differences of geometrical features of the different instruments. Moreover, recent studies conducted on breast composition based on CT scans showed that the skin thickness, used in the breast model currently employed in protocols, was too large. This issues suggested the possibility of a general recalculation of dose coefficients. The formalism that I've introduced in this work defines DBT system specific Normalized Glandular Dose coefficients DGNDBT, calculated by directly simulating a DBT exams, thus being independent of DM coefficients. The developed Geant4 MC software is based on a model previously proposed for Digital Mammography, integrated with all the necessary geometric features to perform a simulation of a DBT exam. Also the breast model used has been optimized taking into account the latest findings on skin thickness in order to obtain a more accurate dosimetry. I've toughly validated the MC code vs. data of the AAPM TG 195 protocol Case I and III and against the data provided by Dance. Furthermore, I've conducted an experimental validation with data acquired with radiochromic films at the Senology Department of the 'Azienda Ospedaliero-Universitaria Pisana' (AOUP), taking dose measurements for different beam qualities in a breast PMMA phantom and comparing the results with MC simulations recreating the acquisition setup. The results of the simulations were two datasets of DGNDBT coefficients relative to two commercial DBT machines: the Hologic Selenia Dimensions and the GE Senoclaire. The datasets explore different X-ray beams qualities and breast characteristic. I've developed a MATLAB application that, by means of opportune interpolations, gives the end user the coefficients for the desired exam setup.
I've developed the MC code to take advantage of the multi-threading capabilities offered by the Geant4 kernel, obtaining a quasi-linear speedup with respect to the number of utilized core. Simulations were run on a i7-9700 @ 3.00 GHz CPU, and took a total of one month for validation purposes and two months for the elaboration of the DGNDBT coefficients. I've also developed several Python scripts to automatize the dataset creation and data analysis.