This thesis describes a novel approach to partial volume segmentation of cerebrospinal fluid (CSF) in sulci of 3D magnetic resonance (MR) brain images that overcomes a key limitation of other techniques. The method is based upon previous work in that it uses a Markov random field (MRF) to encourage adjacent voxels to have the same tissue label in most brain regions, but differs in that it adaptively reverses its behaviour to preserve tissue boundaries in sulci. Image curvature and medial surface information, both independent of MR pulse sequence, are used to identify regions where sulci have a high probability of occurring, in order to modulate MRF behaviour. The method is evaluated using both real and simulated data, showing a significant improvement in terms of sensitivity to voxels containing CSF. It is also shown qualitatively to be useful in improving a deformable model based cortical surface extraction procedure.
Identifer | oai:union.ndltd.org:LACETR/oai:collectionscanada.gc.ca:QMM.82635 |
Date | January 2005 |
Creators | Singh, Vivek |
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 Engineering (Department of Electrical and Computer Engineering.) |
Rights | All items in eScholarship@McGill are protected by copyright with all rights reserved unless otherwise indicated. |
Relation | alephsysno: 002210751, proquestno: AAIMR12649, Theses scanned by UMI/ProQuest. |
Page generated in 0.0014 seconds