Model Based Tomography: A Comprehensive Approach to Iterative Image Reconstruction

点击翻译按钮获取中文摘要

Low dosage transmission medical imaging, emission medical imaging, and nondestructive testing of materials are all examples of tomographic reconstruction problems which can benefit greatly from improved reconstruction techniques. This research introduces a strategy for the development of computationally efficient reconstruction algorithms which directly search for the statistically optimal fit to measured data. This approach exploits the wide availability of digital computation and storage to substantially improve reconstruction in demanding applications. The research has three essential components: 1) A general measurement system model is being proposed which characterizes the physical measurement apparatus for both the transmission and emission problems, and can be extended to include nonlinear effects such as scattering. 2) A new class of computationally tractable image cross section models is being proposed which preserves image detail while suppressing noise artifacts. 3) A fast numerical algorithm, known as Gauss-Seidel, is proposed as a basis for efficient and accurate multiscale reconstruction methods. Algorithmic techniques are being evaluated using data collected from the Nondestructive Evaluation Section at the Lawrence Livermore National Laboratories.