Date of Award
2008
Degree Type
Thesis
Degree Name
Doctor of Philosophy
Program
Electrical and Computer Engineering
Supervisor
Dr. James Lacefield
Second Advisor
Dr. Hanif Ladak
Third Advisor
Dr. Vijay Parsa
Abstract
Computer-based models are increasingly used in biomedical imaging research to clarify links between anatomical structure, imaging physics, and the information content of medical images. A few three-dimensional breast tissue software models have been developed for mammography simulations to optimize current mammography systems or to test novel systems. It would be beneficial in the development of ultrasound breast imaging to have a similar computational model for simulation. A three-dimensional breast anatomy model with the lobular ducts, periductal and intralobular loose fibrous tissue, interlobular dense fibrous tissue, fat, and skin has been implemented. The parenchymal density of the model can be varied from about 20 to 75% to represent a range of clinically relevant densities. The anatomical model was used as a foundation for a three-dimensional breast tumour model. The tumour model was designed to mimic the ultrasound appearance of features used in tumour classification. Simulated two-dimensional ultrasound images were synthesized from the models using a first-order k-space propagation simulator. Similar to clinical ultrasound images, the simulated images of normal breast tissue exhibited non-Rayleigh speckle in regions of interest consisting of primarily fatty, primarily fibroglandular, and mixed tissue types. The simulated images of tumours reproduced several shape and margin features used in breast tumour diagnosis. The ultrasound wavefront distortion produced in simulations using the anatomical model was evaluated and a second method of modeling wavefront distortion was also proposed in which 10 to 12 irregularly shaped, strongly scattering inclusions were iii superimposed on multiple parallel time-shift screens to create the screen-inclusion model. Simulations of planar pulsed wave propagation through the two proposed models, a conventional parallel time-shift screen model, and digitized breast tissue specimens were compared. The anatomical model and screen-inclusion model were able to produce arrival-time fluctuation and energy-level fluctuation characteristics comparable to the digitized tissue specimens that the parallel-screen model was unable to reproduce. This software is expected to be valuable for imaging simulations that require accurate and detailed representation of the ultrasound characteristics of breast tumours.
Recommended Citation
Shen, Yi-Ting, "Software for Modeling Ultrasound Breast Cancer Imaging" (2008). Digitized Theses. 4708.
https://ir.lib.uwo.ca/digitizedtheses/4708