A PSEUDO-SPECTRAL APPROACH TO THE FORWARD PROBLEM OF 2D ULTRASOUND COMPUTER TOMOGRAPHY

Open Access
Author:
Cairnie, Mark A
Area of Honors:
Electrical Engineering
Degree:
Bachelor of Science
Document Type:
Thesis
Thesis Supervisors:
  • Mohamed Almekkawy, Thesis Supervisor
  • Julio Urbina, Honors Advisor
Keywords:
  • Ultrasound
  • Medical Imaging
  • USCT
  • Tomography
  • Inverse Problems
  • Acoustics
  • Simulation
Abstract:
While medical imaging technologies such as X-Ray and MRI have seen wide spread use, they are not without significant shortcomings. To meet the evolving needs of medical professionals and satisfy the shortcomings in these technologies, there has been a recent surge the pursuit of realizable ultrasound computer tomography (USCT) imaging systems. USCT offers comparable image quality, with reduced overall cost and simple design that aids medical professionals, specifically in the area of breast cancer screening. Studies from Karmanos Cancer Institute have shown the ability of USCT to resolve cancerous tissue by recording sound-speed, acoustic attenuation, and acoustic reflectivity data from the imaging medium. The largest constraint on a realizable USCT is the image reconstruction problem due to its immense computational load. The inherent ill-posedness of the problem, in conjunction with its non-convex nature, requires the problem to be solved iteratively with both a forward and an inverse solver. The computational complexity of the problem is the focus of the majority of current research; however, the cost of data collection has severely limited this progress. While several research groups have managed to construct full-scale operational prototypes of a USCT system, access to the data is limited. As a result, the purpose of this thesis is to develop a forward model that can be used to simulate data in place of a full scale prototype. The forward solver can be used for data generation and as a component of the image reconstruction algorithm. The solver is built in MATLAB and, as a result, is both compact and portable. A pseudo-spectral method is used to implement the solver because it is more computationally efficient then more typical finite element and finite difference methods and can therefore be run on standard desktop hardware.