A Novel Breast Phantom Created Using GPU-Based Voxelization for X-Ray Breast Imaging Research

Anthropomorphic x-ray medical phantoms are models of the human tissue, organ, system, whole body, created with different purposes, such as optimization of imaging techniques, optimization of dose to the patient, testing reconstruction images for newly developed modalities, as well as performing quality control of x-ray medical devices. In the field of breast imaging, significant efforts have been devoted to developing heterogeneous breast phantoms for use with 3D imaging techniques. It is crucial that both, physical and computational phantoms share identical properties. Anthropomorphic computational phantoms can be derived from patient imaging data or constructed using combinations of geometric primitives. In the latter case, the use of large number of primitives can significantly increase computational complexity, making ray-tracing techniques time-consuming and inefficient, while also complicating the insertion of lesions. This study aims to develop a voxel-based version of a breast mathematical phantom, originally constructed from geometric primitives, to facilitate the efficient generation of x-ray images. The mathematical phantom consists of 15863 geometric primitives, including two semicylinders, one parallelepiped, and the rest spheres of varying diameters. The voxelization process was carried out using a CPU programming approach. X-ray images of both, the mathematical and voxel-based phantoms were generated and compared, with an observed maximum difference of 0.6%, attributed to the voxelization algorithm. Improvements to this algorithm are also planned. The model is suitable for use with various lesions inserted for virtual experiments.