Three-dimensional in silico breast phantoms for multimodal image simulations.

Anatomic simulators have provided researchers with the realistic objects needed to develop and evaluate medical imaging approaches. Today we have new insights into the cellular biology of breast tissues that is driving many new targeted diagnostic and therapeutic approaches, including molecular imaging. We report on our initial efforts to build a scalable framework for the construction of realistic 3-D in silico breast phantoms (ISBP) capable of leveraging existing knowledge and yet adaptable to fully integrate future discoveries. The ISBP frames are developed with scalable anatomical shapes and morphologic features as adapted from a rich literature on this topic. Frames are populated with tissue subtypes essential for imaging and object contrast functions are assigned. These data can be resampled to match the intrinsics scales of various imaging modalities; we explore mammography, sonography and computed tomography. Initial comparisons between simulated and clinical images demonstrate reasonable agreement and provides guidance for future development of a more realistic ISBP. An end-to-end simulation of breast images is described to demonstrate techniques for including stochastic variability and deterministic physical principles on which image formation is based.