Geant4 physics list comparison for the simulation of phase-contrast mammography (XPulse project)

Highlights

• Geant4 Standard EM option 4 and Livermore Physics Lists are in good agreement with EGSnrc in the energy range 20–100 keV.

• We confirmed that breast dose in mammography screening can be minimised if a beam energy around 60 keV is used.

• We confirmed that phase-contrast imaging is a promising technique for routine mammography screening.

Abstract

Purpose

Breast cancer is the most frequent cancer in women. Early and accurate detection of the disease is a major factor in patient survival. To this end, phase-contrast imaging has gained significant interest in recent years. The aim of this work was to validate the physics models of a Geant4 mammography imaging simulation (in the context of the XPulse project) by comparing to EGSnrc results.

Methods

We used three Geant4 electromagnetic physics lists of the version 10.4 of the toolkit: Standard, Livermore and Penelope. We calculated energy distributions in homogeneous and inhomogeneous phantoms and breast doses in DICOM images. The simulations used photon beams of energies 20–100 keV. The Geant4 calculations were compared with EGSnrc/DOSXYZnrc simulations.

Results

We found a very good agreement between the Standard Electromagnetic option 4 and Livermore Physics Lists (within 1% for all beam energies). Larger differences were found between Standard Electromagnetic option 4 and Penelope Physics Lists (about 4%). The agreement of longitudinal energy distributions between Geant4 Standard Electromagnetic option 4 and EGSnrc was good in water and light biological materials, but important discrepancies were found in heavy elements. We confirmed with both codes that dose to the breast is minimal at beam energy around 60 keV.

Conclusions

Overall, we found good agreement between the option 4 of the Standard Electromagnetic physics list and Livermore physics lists of Geant4, as well as EGSnrc for materials relevant to mammography screening. Further investigations are needed for the case of heavier materials.

Introduction

Breast cancer is the most frequent form of cancer in women. The implementation of mammography systems and organised breast screening has allowed for a better control of the disease since the 1970s due to early detection of the disease. Currently, the reference technology for breast cancer screening is digital mammography. The dose deposited in breast tissue during a diagnostic procedure (mean glandular dose, MGD) remains a concern in popular press even though the benefit of screening largely outweighs the risk of radiation-induced cancer [1]. Radiologists constantly try to reduce breast dose to the lowest amount technically achievable to alleviate the concern, in particular for women undergoing regular examinations in the age interval of systematic screening (40–74 years of age).

MGD estimation is done routinely as a part of quality control procedures [2]. Large variations of MGD have been reported in the literature depending on the technology and protocol used. For example, Bosmans et al. have reported a 60% higher breast dose delivered by computed radiography systems as compared to digital radiography systems [3]. At the same time as trying to reduce breast dose, radiologists strive for an increased image quality to reduce false positives/negatives in breast screening [4]. To address this issue, several novel techniques have been developed, including 3D imaging such as digital breast tomosynthesis (DBT) [5], [6] and breast computed tomography (BCT) [7].

Currently, the standard method for breast cancer screening is digital x-ray mammography. It relies on the selective absorption of radiation by tissues of different density in the patient. Detection efficiency therefore depends on the composition of the breast tissue: detecting a lesion in a dense breast is much more difficult than in a breast composed mostly of adipose tissue. This poses a problem of diagnostic accuracy in the subgroup of women with dense breasts, which overlaps significantly with the group of women of less than 50 years of age. The question has been raised whether photon absorption is the only physical process that can be exploited for the acquisition of diagnostic breast images. Several groups have advocated the use of phase contrast imaging as an alternative. Phase contrast 3D imaging would be of particular interest, as it could potentially maximize detection efficiency while providing patient comfort by removing the need for breast compression [8], [9].

The challenge in constructing a novel mammography system consists in the reduction of breast dose while maintaining an image quality equivalent or superior to this of current techniques. In this context, the need of estimating MGD accurately is no longer restricted to the estimation of cancer-induction risk. It is also a pivotal parameter for the design and optimisation of the new mammography system.

Breast dose cannot be obtained with direct measurements of dose in patients. MDG estimation is instead largely based on the use of Monte Carlo calculated conversion factors that translate measurements of incident air kerma to breast dose [2]. While at the design phase of a novel mammography system based on phase-contrast imaging, it is important to conduct a detailed study of the dose dependence on several parameters. These include not only the geometrical specifications of the system, but also the method of primary photon generation and their energy spectrum, the image reconstruction algorithms and the implementation of clinical protocols. To this end, Monte Carlo calculations are an indispensable tool. Geant4 is a Monte Carlo code [10], [11], [12] that has been frequently used for mammography applications [13], [14], [15]. In this work, we validate the physics options of Geant4 v.10.4 with the intent of using the code for the design and optimisation of a mammography system based on phase contrast imaging (XPulse project). We have used EGSnrc [16] as a reference, as it is a Monte Carlo code broadly used in medical physics applications. We compared dose distributions calculated for various Geant4 physics implementations and various geometries with EGSnrc results. We performed comparisons of doses in homogeneous phantoms and in patient CT images. In addition, we utilized both Monte Carlo codes to study the dependence of breast dose on primary photon energy.