• geant4
• breast dosimetry
• monte carlo

Breast cancer is the leading cause of cancer deaths in female subjects. Screening Digital Mammography (DM) is the principal technique used to detect tumours even in an early stage, allowing greater possibilities for treatments. Since DM is performed by using ionizing radiation, Mean Glandular Dose (MGD, or Average Glandular Dose AGD) is adopted for cancer risk estimates and it must be assessed with an accurate dosimetry. Since dose to the gland can’t be measured directly, Monte Carlo (MC) simulations are performed to evaluate MGD, multiplying incident air kerma on the upper surface of the breast by suitable conversion factors computed by Monte Carlo codes. The same principle regards dose estimates for the Digital Breast Tomosynthesis (DBT) technique, which allows to reduce the tissue superimposition effect and making it easier for radiologists to distinguish normal from cancerous tissues. For dosimetry purposes, in both DM and DBT, geometric assumptions are followed, by involving a simple mathematical phantom for reproducing the breast geometry, and by using a homogeneous compound of adipose and glandular tissues surrounded by 5 mm thick adipose skin to reproduce the breast anatomy. Research in breast dosimetry leads to improve accuracy in MC calculations and the improvement on the models adopted, whether anatomical, geometric, or physical assumptions are routinely investigated for even better MC estimates.
This thesis voluntarily deviates from the usual setting of scientific papers and a didactic description is followed; it is entirely inspired by works in literature published by the research group of which the author belongs.
In paragraphs 1.1-1.2 a brief review of breast screening procedures and formalisms followed by international dosimetry protocols are described. In this thesis a GEANT4-based MC code, developed by our research group is described and validated (par. 1.3-1.4). It has been developed for MGD estimates for commercial DBT units, taking into account specific operating and geometrical conditions for each one. GE SenoClaire, Hologic Selenia Dimensions, GE Senographe Pristina, Fujifilm Amulet Innovality, Siemens Mammomat Inspiration and IMS Giotto Class are reproduced in simulations and normalized glandular dose (DgN) coefficients are obtained. For gaining in MC accuracy, a new skin envelope is included in the phantom model, on the basis of new founding in literature regarding the breast anatomy (par. 2.1).
The characterization of DgN_DBT numbers versus breast density, compressed breast thickness beam quality is investigated, interpolation conditions explored, and a spreadsheet *.xlsm is presented which produces dose coefficients for arbitrary inputs of DBT units (from those simulated in this work), kV, HVL, compressed breast thickness and glandularity. The novelty consists not just in the DBT-specific simulations, but in the skin model adopted of 1.45 mm (instead of the adipose 5 mm thick skin model), increasingly presented in literature as the correct model to adopt. Computed coefficients may led to MGD estimates which differ at maximum by (-17.6%, +6.2%) from those provided by the adoption of the EUREF protocol (2018), due mainly to the different proposed breast models (skin thickness and composition and scanner specific calculation).
In the paragraph 2.2, a new and innovative digital phantom model is proposed and evaluated, which overcomes the drastic assumption of the homogeneous mixture of adipose and glandular tissues, involving a voxelized breast phantom approach. Clinical results based on Breast Computed Tomography (bCT) investigation performed in the United States and published in literature suggested a new method for representing the gland, basically far from the current uniform distribution of gland within the breast; the proposed model here presented involves a gaussian distribution of glandular tissues, by means of glandular voxels distributed according to the position in the breast volume. MGD estimates are evaluated and compared with the current dosimetry method, and a proposal for involving the phantom model for dosimetry routines is presented.
The new heterogeneous phantom model is used for investigating MGD estimates and differences with dose estimates obtained with the use of homogeneous phantom have been reported. Wide variations have been confirmed, mostly for low breast densities, the most common characteristic among women, and the new trend curve of 4th order polynomial fit is presented regarding the MGD values versus the glandularity. The underestimates of MGD values with the adoption of the voxelized phantom are in line with literature results. Nevertheless, a justification for today's adoption of the old protocols based on homogeneous phantoms can be given by confirming a conservative approach of the actual method related to glandular dose.
This thesis is not limited in considering only digital breast phantoms, but discussions regarding physical breast phantoms are introduced with the support of the MC code. In the last years, thanks to the spread of 3D printers, the state of the art in different fields of medical physics, whether is radiology or radiotherapy, is the production of built-in-house phantoms presenting attenuation characteristics and/or digital images similar to those of the real tissues. For DM and DBT, test physical phantoms represent fundamental tools used to perform quality assurance (QA) procedures and allow the calculation of useful parameters for imaging and radiation dosimetry. QA procedures and research are usually performed using polymethyl-metacrilate phantoms (PMMA) blocks simulating the breast composition, In par. 3.1 low density 3D-printing materials are evaluated by using MC calculations for built-in-house phantoms for applications in breast dosimetry. A physical breast phantom is introduced, presenting dedicating homogeneous compounds for both the skin envelope and the inner part of the breast phantom, according to the actual reference dosimetry methodology.
Results deriving from MC simulation runs over monoenergetic and polychromatic beams lead to consider PCABS material as a well substitute for the 5 mm thick skin layer and PLA material as substitute for the inner breast tissue. A physical test phantom, which design relies on the geometry adopted in the MC code, is proposed and validated with experimental measurements using GAFchromic films, which results confirms an agreement with transmission estimations in MC results for both PCABS layer and for the inner PLA material, supporting our method, which uses relatively low-cost equipment and procedures.