Skip to Main content Skip to Navigation
Conference papers

Hippocampus-Specific fMRI Group Activation Analysis with Continuous M-Reps

Abstract : Continuous M-Reps. In earlier work [4], we developed a deformable modeling framework called the continuous medial representation (cm-rep), which, in simplest terms, is a continuous analog of the m-rep approach by Pizer et al . [3]. Cm-reps are deformable templates that have a special property that the skeleton (medial axis) of the template preserves its branching topology (number and configuration of its branches) during deformation. Furthermore, the skeleton of the deforming template is modeled explicitly as a continuous manifold and the exact geometric relationship between the skeleton and the template's boundary is captured. The skeleton-boundary relationship makes it possible to define a special shape-based coordinate system on the interior of the template. One of the coordinate axes connects points on the skeleton to the corresponding (nearest) points on the boundary, while the other two axes parameterize the skeleton as well as the boundary. As argued in [5], this coordinate system is a natural way to parameterize the interiors of anatomical structures to which cm-rep templates have been fitted. It allows us to analyze the distribution of appearance features (intensity, fMRI, DTI) within the structures more or less independently of the differences in the shape of the structures. In this paper, we illustrate the application of such analysis to functional MRI data associated with the hippocampus. fMRI Study. We designed an fMRI study in which hippocampal activation was highly anticipated. Healthy volunteers (n=18) performed a complex scene encoding task in the scanner. The study followed the block design paradigm. In addition to EPI functional imaging, T1 anatomical scans were acquired. Standard techniques (SPM, [2]) were used to align functional images to the anatomy, to correct for motion artifacts, to smooth the data temporally and spatially and to compute subject-level statistical maps using the general linear model. Structure-Specific Group Analysis of fMRI. Whole-brain fMRI group analysis methods such as SPM [2] rely on deformable image registration to normalize data between subjects. Registration does not always align the hippocampus accurately [1] and such misalignment will, in theory, reduce the statistical significance of intersubject statistical maps. To improve intersubject alignment in the hippocampus, we fit cm-rep templates to manual segmentations of the hippocampus (Fig. 1) and use the cm-rep coordinate system to project subject-level fMRI statistics to a common reference space3. We then perform random effects analysis to compute a hippocampus-specific groupwise statistical map.
Document type :
Conference papers
Complete list of metadata

Cited literature [4 references]  Display  Hide  Download
Contributor : Service Ist Inria Sophia Antipolis-Méditerranée / I3s Connect in order to contact the contributor
Submitted on : Wednesday, October 26, 2011 - 4:11:13 PM
Last modification on : Wednesday, October 9, 2019 - 1:58:04 PM
Long-term archiving on: : Sunday, December 4, 2016 - 4:58:04 PM


Files produced by the author(s)


  • HAL Id : inria-00636061, version 1



Paul Yushkevich, John Detre, Kathy Tang, Angela Hoang, Dawn Mechanic-Hamilton, et al.. Hippocampus-Specific fMRI Group Activation Analysis with Continuous M-Reps. 1st MICCAI Workshop on Mathematical Foundations of Computational Anatomy: Geometrical, Statistical and Registration Methods for Modeling Biological Shape Variability, Oct 2006, Copenhagen, Denmark. pp.112-114. ⟨inria-00636061⟩



Record views


Files downloads