BreastSimulator: A software platform for breast x-ray imaging research

Abstract

Objective: It is well established that computer based models of x-ray imaging systems are basic and very important tools for developing and evaluating new emerging x-ray imaging techniques, optimizing technical parameters, and performing feasibility studies prior to implementation in clinical practice. Such models are essential for the development and the establishment of new breast x-ray imaging modalities that aim to detect and better characterize breast lesions in their early stage. This work presents a complete software package, called BreastSimulator, dedicated for breast x-ray imaging research.

Methods: The package consists of four modules used to create three-dimensional breast models in compressed and uncompressed state, simulate x-ray mammographic images and visualize the results of the simulations. The module that is used to generate breast models, Breast Modeling Module, consists of several sub-modules that are utilized to model the different breast components: external shape, glandular and adipose tissue, breast lesion, skin, pectoralis and lymphatics. The Compression Module is dedicated to simulate the mechanical compression of the breasts. Mammographic projection images are obtained with simulation of x-ray photon transport starting from the x-ray source, passing through the breast model and reaching the detector. This is accomplished in the Image Generation Module. Finally, the results of the simulations, i.e. breast models and mammographic images can be seen with the Visualization Module.

Results: Here, we demonstrate the application of the software package in conventional and dual-energy mammography as well as compression studies, as examples to highlight basic functions and applications of Breast Simulator. The first study aimed to define the optimal pair of ‘low’ and ‘high’ monochromatic x-ray energies for dual-energy mammography. It involved the synthesis of 225 dual-energy images obtained from combinations of ‘low’ and ‘high’ energy images acquired in the energy range 14 to 28 keV. Images were generated from a medium sized dense breast model that contained one calcification. The study showed that 17/28 keV incident monoenergetic beams are optimal to obtain maximal calcification detectability for this breast. The second study demonstrated the effect of breast compression on the quality of the obtained mammograms. It included a breast model based on breast CT slices subjected to simulated compression and generation of mammographic images. Increased image quality is observed for mammograms obtained from breasts with reduced thickness. The characteristics of the x-ray beams that exit a small dense breast model were investigated in the third study. For two mammographic spectra used in mammography imaging, the mean energy of the transmitted x-rays and the mean exit angle of the scattered radiation increase as the incident x-ray energy increases.

Conclusions: We believe that this tool and its functionalities will speed up the development, testing and optimization of new breast imaging modalities such as breast tomosynthesis, cone-beam CT and advanced two-dimensional techniques like dual-energy as well as specific parts of imaging chain, such as x-ray source, detector and acquisition geometry.