Abstract
Breast cancer has a significant impact on the well-being of the female population both nationwide and worldwide. This has motivated the establishment of national breast screening programmes in most western countries, in order to reduce the mortality associated with this disease via early detection. X-ray mammography is considered the current gold standard technique for breast cancer detection in such screening programmes. However, this suffers from performance limitations due to tissue superposition which can either mimic or obscure malignant pathology. Therefore, alternative X-ray modalities, such as digital breast tomosynthesis (DBT), which employs a series of X-ray projections at different (but limited) angles, are being explored in order to improve breast cancer detection rates. In all such X-ray based imaging methods, scattered photons produced deleterious effects on image quality to varying degrees. In order to model such scatter distribution, Monte Carlo (MC) simulations is often chosen as the default approach, and as such is used in this thesis to quantify its effects in X-ray mammography and DBT scenarios. Following validation on the use of the GEANT4 MC package for use in mammography, three commercially available full-field digital mammography (FFDM) systems were simulated with their corresponding anti-scatter grids using a CDMAM geometry. It was observed that, for the particular geometry studied, the scattered radiation recorded at the detector was 17% using a linear anti-scatter grid design. However, this figure was reduced by a factor of three when employing a cellular anti-scatter grid geometry. In DBT geometries, scattered radiation is larger than in FFDM and, spatially, may vary more rapidly due to the absence of an anti-scatter grid. The excessively long times needed to run MC simulations (8-10 hours) for such analysis motivates the need for an alternative approach. A non-stationary kernel-based approach has thus been developed. It was found that using kernels based on breast thickness-only, can overestimate scatter radiation by more than 60% (compared to MC simulations) at the breast edge region. Simulation work presented here shows that this overestimation in scatter is largely due to the air gap between the lower curved breast edge and the image receptor. In this thesis, a more accurate scatter field estimator is proposed for use in DBT which not only considers the breast thickness and primary incidence angle, but also accounts for scatter exiting the breast edge region and traversing an air gap prior to absorption in the image receptor. This proposed approach has reduced such errors to an average error of 10% in scatter, and a maximum of 20% across the projected breast phantom, and has decreased the run-time ten-fold. Such an approach has potential applications in scatter correction methods in DBT, and as an efficient modelling tool in imaging system development and in evaluation of virtual clinical trials.