Masters Theses

Date of Award

5-2000

Degree Type

Thesis

Degree Name

Master of Science

Major

Computer Science

Major Professor

Jens Gregor

Committee Members

Michael G. Showasm

Abstract

A new system for x-ray cone-beam micro-tomography has been developed to screen mice for internal phenotypic abnormalities at the Oak Ridge National Laboratory Mammalian Genetics Facility. Currently this system uses an image reconstruction algorithm that is based on two-dimensional (fan-beam) reconstruction techniques. The disparity between the actual scanner geometry and that assumed for reconstruction purposes introduces artifacts into the reconstruction volume that become increasingly worse the further their axial distance from the midplane. In order to reconcile this disparity and reduce axial distortion artifacts, a volumetric reconstruction algorithm based on cone beam geometry was implemented. The volumetric algorithm is derived and its heuristic implementation is explained within the constraints of the system, which limit the arclength of the scanning trajectory. Reconstructions using the volumetric algorithm are analyzed and compared to reconstructions from the current method. We show that our implementation produces images of equivalent quality in the midplane, and a marked decrease in axial distortion elsewhere Volume reconstruction times are shown to be comparable to those currently achieved. The theoretical foundations are given for future work to optimize the implementation through parallelization and by overcoming the data sufficiency problem.

Download
Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS