Item – Theses Canada

​
OCLC number
275811352
Link(s) to full text
LAC copy
LAC copy
Author
Vu, Randy Hoang,1981-
Title
Strategies for three-dimensional segmentation of the primary tumor mass in computed tomographic images of neuroblastoma.
Degree
M. Sc. -- University of Calgary, 2006
Publisher
Ottawa : Library and Archives Canada = Bibliothèque et Archives Canada, [2007]
Description
2 microfiches
Notes
Includes bibliographical references.
Abstract
Segmentation of the primary tumor mass in neuroblastoma could aid the radiologist by facilitating reproducible, objective quantification of the tumor's tissue composition and volume. However, segmentation continues to be a difficult problem because the tumor is usually composed of heterogeneous tissue types. In the present thesis, procedures are proposed to segment and remove known structures prior to segmentation of the tumor. Several image processing tools, in particular, thresholding, fuzzy connectivity, mathematical morphology, and deformable contours, are deployed for this purpose. By incorporating opening-by-reconstruction, and by applying an improved procedure to remove the peripheral fat, peripheral muscle, and thoracic cavity, the average false-positive rate was reduced by 41% as compared to a previously proposed method; the computational requirements were also reduced by a factor of 35. The average sensitivity of the methods with 10 computed tomographic exams was 82.5%. Potential clinical applications of the methods are discussed in this thesis.
ISBN
9780494192672
0494192674