This thesis addresses several brain MRI segmentation methods including three methods of using normal distributions, similar methods using t-distributions and a new method using MCLUST software packages. The main purpose of this thesis is to try to improve current brain image segmentation methods and reduce the long computing time with satisfactory results. The methodology consists in applying histogram analysis for the initialization of parameters, using Simpson's rule for the approximation of numerical integration and describing methods of skull stripping and neighborhood information encoding. The two most significant contributions are suggestions for speeding up current methods and a newly proposed method that ignores spatial information for parameter estimation. Potential future work is proposed at the end.
| Identifer | oai:union.ndltd.org:LACETR/oai:collectionscanada.gc.ca:QMM.81458 |
| Date | January 2004 |
| Creators | Xu, Peiheng, 1965- |
| Publisher | McGill University |
| Source Sets | Library and Archives Canada ETDs Repository / Centre d'archives des thèses électroniques de Bibliothèque et Archives Canada |
| Language | English |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Format | application/pdf |
| Coverage | Master of Science (Department of Mathematics and Statistics.) |
| Rights | All items in eScholarship@McGill are protected by copyright with all rights reserved unless otherwise indicated. |
| Relation | alephsysno: 002186692, proquestno: AAIMR06475, Theses scanned by UMI/ProQuest. |
Page generated in 0.0011 seconds