Year of Publication

1996

Paper Type

Master's Thesis

College

College of Computing, Engineering & Construction

Degree Name

Master of Science in Computer and Information Sciences (MS)

Department

Computing

First Advisor

Yap S. Chua

Second Advisor

Ralph M. Butler

Third Advisor

Behrooz Seyed-Abbassi

Abstract

In this thesis, I shall study and compare various methods for manipulating two- and three-dimensional image data produced with a nuclear magnetic resonance scanner. In particular, I will examine ways of focusing upon specific structures internal to the object under study (segmentation); and will explore means of rendering realistic images of these structures on a computer screen using depth-cueing, shading, and ray-casting techniques.
The 3DHEAD volumetric dataset used for this project was created with the Siemens Magnetom and was provided courtesy of Siemens Medical Systems, Inc., Iselin, NJ. This dataset consists of 109 slices of a human head, with each slice stored consecutively as a 256 x 256 array. Each pixel is represented by two consecutive bytes, which make one binary integer. (A similar dataset of a human knee is also available.) The 3DHEAD dataset requires about 14 Mb of disk space uncompressed. The programs which manipulate this data are MS-DOS-based and were written and compiled using Microsoft QuickC version 2.51. The 2-D programs were executed on a CompuAdd 486DXl2-50 with 8 Mb of RAM, running MS-DOS version 6.22; the 3-D programs were executed on a 133 MHz Pentium clone with 48 Mb of RAM, running the DOS shell of Microsoft Windows 95.
Our immediate objectives are to produce pleasing and informative 2-D and 3-D pictures of the internal structure of some component of the human head: for example, the brain.
We need to remove from the original dataset all of the data which do not represent the brain. Then, for the 3-D images, we need to render the remaining data in such a way that it possesses depth and realism.

Share

COinS