Identifier

etd-10272005-155140

Degree

Doctor of Philosophy (PhD)

Department

Computer Science

Document Type

Dissertation

Abstract

In the recent years, the demand for automated processing techniques for digital medical image volumes has increased substantially. Existing algorithms, however, still often require manual interaction, and newly developed automated techniques are often intended for a narrow segment of processing needs. The goal of this research was to develop algorithms suitable for fast and effective correction and advanced visualization of digital MR image volumes with minimal human operator interaction. This research has resulted in a number of techniques for automated processing of MR image volumes, including a novel MR inhomogeneity correction algorithm derivative surface fitting (dsf), automatic tissue detection algorithm (atd), and a new fast technique for interactive 3D visualization of segmented volumes called gravitational shading (gs). These newly developed algorithms provided a foundation for the automated MR processing pipeline incorporated into the UniViewer medical imaging software developed in our group and available to the public. This allowed the extensive testing and evaluation of the proposed techniques. Dsf was compared with two previously published methods on 17 digital image volumes. Dsf demonstrated faster correction speeds and uniform image quality improvement in this comparison. Dsf was the only algorithm that did not remove anatomic detail. Gs was compared with the previously published algorithm fsvr and produced rendering quality improvement while preserving real-time frame-rates. These results show that the automated pipeline design principles used in this dissertation provide necessary tools for development of a fast and effective system for the automated correction and visualization of digital MR image volumes.

Date

2005

Document Availability at the Time of Submission

Release the entire work immediately for access worldwide.

Committee Chair

John Tyler

DOI

10.31390/gradschool_dissertations.1461

Share

COinS