Structure-Specific fMRI Analysis

description

We are using shape-based normalization (with the cm-rep method) to study activation patterns in the hippocampus and other structures of interest in functional MRI studies. In contrast to whole-brain techniques, our analysis restricts normalization and statistical analysis to structures of interest in an effort to tease out as much detail as possible about local activation patterns. This type of detailed local fMRI analysis will become especially relevant with the new generation of MRI scanners and head coils that will allow functional imaging at nearly 1mm resolution.

aims

• Automatic hippocampus segmentation

  We are working to develop a technique for automatic segmentation of the hippocampus using cm-rep deformable models. Cm-rep models are an effective way of modeling variability in the shape and appearance of the hippocampus, and thus allow the construction of shape and intensity priors that can be incorporated into an automatic segmentation algorithm.

• Asymmetry maps

  We plan to use the shape-based coordinate systems to derive maps of asymmetry in hippocampal function across hemispheres. These maps may prove useful for predicting the outcome of hippocampus resection surgery in temporal lobe epilepsy

• Thickness-function correlation

  We are developing techniques for co-analyzing differences in the thickness of the hippocampus with differences in hippocampal function. Gray matter thickness is physiologically related to the BOLD fMRI response and statistical analyses that correct for differences in thickness would be possible with our methods.

• Application to dementia data

  We aim to apply the hippocampus-specific analysis technique to functional imaging data from dementia studies. We are interested in localizing differences in hippocampal recruitment in Alzheimer's disease, semantic dementia and healthy aging

images

Multi-subject analysis of brain activation restricted to the hippocampus. The color map represents the statistical significance of activation. Our approach allows us to examine activation in the context of a specific structure, and the way in which data from different subjects is combined in the analysis is designed to maximize specificity.

A model of the hippocampus with explicit description of thickness and the medial axis. Such models offer a way to parameterize the interior of the hippocampus using a shape-based coordinate system, and are the technology underlying hippocampus SPM.

bibliography