Anthropomorphic breast phantoms are used for preclinical testing or optimization of imaging systems or image analysis methods. We discuss mathematical issues of recursive partitioning algorithms for simulation of software breast phantoms. The considered aspects include modeling of tissues and anatomical details (e.g., Cooper’s ligaments, milk ducts) and control and validation of model parameters. We demonstrate the asymptotic time complexity of the algorithm. We provide theoretical evidence of possible improvement in projected image quality without reducing voxel size. We propose characterization of multi-tissue voxels through calculation of subvolumes occupied by each tissue type. To calculate subvolumes we utilize local planar approximations of the boundary surfaces and decomposition into simple geometric shapes. Statistical methodology is introduced to validate the implementation of the algorithm and evaluate the quantitative error of the approximation.