2005 AHM Planning: Integration of Morphometry & Functional Neuroimaging Data

Jump to: navigation, search
Home < 2005 AHM Planning: Integration of Morphometry & Functional Neuroimaging Data

'Integration of Morphometry & Functional Neuroimaging Data'

Meeting time

  • Tue Oct 18, 2pm-3:15pm

Meeting goals

  • Increase awareness and demand for multi-modal image integration

Potential Participants

  • Morphometry BIRN: Jorge Jovicich, Silvester Czanner, Anders Dale, Susumu Mori, Allen Song, James Fallon, Martina, Vid, Randy Gollub
  • Function BIRN: Gary Glover, Doug Greve, Vincent Magnotta, Bryon Mueller, Jeremy Bockholt, Steve Pieper, Sandy Wells, Cindy Wible, Vince Calhoun
  • Mouse BIRN: Al Johnson, Russ Jacobs, David Shattuck, Jyl Boline

Potential Agenda

  • [/Wiki/images/c/c8/Integration_morph_func_AHM2005.ppt Presentation (intro missing)]
  • Review of multi-modal integration (D. Greve)
    • Multi-Modal Analysis (MMA) refers to the combining of data from different modalities to draw conclusions that could not be drawn from each modality separately. Most people already use MMA in a trivial manner. Examples include displaying functional MRI results on structural MRI volumes and/or surfaces, registering fMRI volumes using structurals as intermediaries, constraining fMRI analysis to a structural ROI, and constraining EEG/MEG solutions to be on the cortical surface.
  • Proposed discussion (D. Greve)
    • We are proposing another level of integration called Multi-Modal General Linear Model(MMGLM) in the context of group GLM analysis. In a non-multi-modal group analysis, one usually constructs a design matrix based on the demographics of members of the group. This one design matrix is then applied to all voxels. In MMGLM, there is a different design matrix for each voxel. Some of the regressors may be global (eg,demographics), but some of the regressors may be dependent upon the voxel and can include data from other modalities. For example, one can use the cortical thickness at a voxel as a regressor for the functional activation. This requires a different design matrix at each voxel as the thickness will change from voxel to voxel. Such regressors can be constructed from any data source (as long as it can be registered with the functional), including DTI, ASL, EEG, and MEG.
  • Discuss further developments (All)
  • Wrap-up (All)
    • Suggest next steps and follow up plans.
    • Agree on who will prepare & present a 1-slide summary to the BIRN-wide group on Wed afternoon.

Other topics that would need volunteers to drive them

  • Integration of DTI data with structural and/or functional MRI
  • Image co-registration issues
  • Add your suggestions!!

A few suggestions (VDC)

1) From the same voxel or different voxels? The latter is useful for examining, for example, whether functional activation may be related to anatomic changes in different (or even distant) regions of the brain. One could also think of a multiscale approach where functional and/or anatomic clusters are defined within modality and related to the same in other modalities.

2) Capitalizing on the strength of each modality while minimizing the impact of the weakness (e.g. using fMRI spatially localized signal to help with 'where' and EEG temporal information to help with 'when').

3) How to select what to use? Sometimes adding modalities can hurt you. What is the criteria used, etc.