Hierarchical Matching of Cortical Features for Deformable Brain Image Registration

  • Authors:
  • Marc Vaillant;Christos Davatzikos

  • Affiliations:
  • -;-

  • Venue:
  • IPMI '99 Proceedings of the 16th International Conference on Information Processing in Medical Imaging
  • Year:
  • 1999

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper builds upon our previous work on elastic registration, using surface-to-surface mapping. In particular, a methodology for finding a smooth map from one cortical surface to another is presented, using constraints imposed by a number of sulcal and gyral curves. The outer cortical surface is represented by a map from the unit sphere to the surface which is obtained by a deformable surface algorithm. The sulcal and gyral constraints are defined as landmark curves on the outer cortical surface representation. The unit sphere is then elastically warped to itself in 3D using the predefined sulcal and gyral constraints, yielding a reparameterization of the original surface. This method is tested on MR images from 8 subjects, showing improved registration in the vicinity of the sulci used as constraints. We also describe a hierarchical framework for automating this procedure, by using conditional spatial probability distributions of cortical features on the spherical parametric domain, in order to automatically identify cortical features. This approach is demonstrated on the central and precentral sulci.