A Bayesian cost function applied to model-based registration of sub-cortical brain structures

Brian Patenaude, Stephen Smith, Mark Jenkinson

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Morphometric analysis and anatomical correspondence across MR images is important in understanding neurological diseases as well as brain function. By registering shape models to unseen data, we will be able to segment the brain into its sub-cortical regions. A Bayesian cost function was derived for this purpose and serves to minimize the residuals to a planar intensity model. The aim of this paper is to explore the properties and justify the use of the cost function. In addition to apure residual term (similar to correlation ratio) there are three additional terms, one of which is a growth term. We show the benefit of incorporating an additional growth term into a purely residual cost function. The growth term minimizes the size of the structure in areas of high residual variance. We further show the cost function's dependence on the local intensity contrast estimate for a given structure.

Original languageEnglish
Title of host publicationBiomedical Image Registration - Third International Workshop, WBIR 2006, Proceedings
PublisherSpringer Verlag
Pages9-17
Number of pages9
ISBN (Print)3540356487, 9783540356486
DOIs
Publication statusPublished - 1 Jan 2006
Externally publishedYes
Event3rd International Workshop on Biomedical Image Registration, WBIR 2006 - Utrecht, Netherlands
Duration: 9 Jul 200611 Jul 2006

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume4057 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other3rd International Workshop on Biomedical Image Registration, WBIR 2006
CountryNetherlands
CityUtrecht
Period9/07/0611/07/06

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this