Computational synthesis of ultrasound breast images from a three-dimensional anatomical model

Abstract

A three-dimensional breast anatomy model has been implemented using spline surfaces and fractal structures to represent the architecture of the lactiferous ducts, mammary fat lobules, skin, and supporting connective tissues. The model randomly varies user-specified structural parameters to provide an unlimited number of realizations of the gross anatomy. Cross-sectional views extracted by slicing through a realization of the model are input to a two-dimensional k-space (i.e., spatial frequency domain) ultrasound propagation simulator. The k-space simulator iterates pressure and particle velocity fields in 30-ns steps to compute scattering from the structures defined by the anatomical model and small random variations in compressibility that are added to generate speckle. A synthetic aperture method is employed to simulate B-mode imaging with a 5 MHz, 192-element linear array operated using multiple transmit focal zones and dynamic receive focusing. Simulated images of random-scattering phantoms possess approximately Rayleigh speckle statistics. The anatomical model is expected to yield images with speckle statistics comparable to clinical breast images. The long-term objectives of these simulations are to investigate sources of focus aberration in ultrasound breast imaging and the impact of aberration on cancer detection. [Work supported by an NSERC Discovery Grant.]