Spring Meeting 2006

Tensor AtlasMorphometry BIRN

Analysis, Visualization, and Interpretation

Goals from Miami Provide tools to support testing of hypotheses related to in white matter

morphometry • Are depression-related white matter lesions more likely to occur in the area

of certain tracts? (BELL Project)• How does white matter vary among twin pairs compared to the general

population? (VETSA Project) Proposed Process

• Draw on JHU Tract Atlas experience Use existing database Have mBIRN researchers learn tools to expand atlas

• Draw on MGH Topographic White Matter Parcellation• Use BWH Tensor Registration code to map atlases into subject space• Integrate probabalistic white matter atlases into FreeSurfer to subdivide

white matter (using both Structural and Diffusion Data

Topographic White Matter Parcellation

Radiate Sagittal Bridging

Individual fasciculardivisions

Anterior-Posterior orderedParcellation Units

Parcellated accordingto proximity to cortical

Parcellation Units

•Geometric constructs for:-Sagittal Strata (superior, inferior, temporal, cingulum)-Corpus Callosum-Internal Capsule, Fimbrial-Fornix, Amygdalofugal

Peripheral Deep

DTI - Interoperability

VETSA Datasets• 35-Gradient DWI

Acquisitions• FreeSurfer

Analysis of Each Subject

• ~300 Twin Pairs

Conversion of DWI data to NRRD has been completed

DTI Analysis - Atlas

JHU (Mori) Multi-Subject White Matter Labeled Tensor Atlas Available in Slicer (NRRD Format)

30 Subjects with full tensor data sets and tract label maps

Status and Outstanding Issues

JHU Tract Label Maps Not in Same Space as Average Tensor• Revise strategy to register each atlas subject individually

to target. Create probabilistic atlas in each target space.

FA-based Registration is Working and Should be Tested (more in later slides)• Full Tensor Registration Code Not Robust, Further

Algorithmic Development Needed• Collaboration with NAC and NA-MIC. Others?

Integration with Topographic White Matter Parcellation TBD

Integration with FreeSurfer TBD

Batch registration in 3D-Slicer

Using the AG-module

Matthan Caan

Batch registration in 3D-Slicer

Register a series of scans to an average brain

Automatic procedure, no interaction needed.

Can be done in the background or overnight.

Coregistration of other scans is possible.

Using 3D-Slicer from the command-line

Register a series of scans without any interaction:

1) Set up a virtual display Xvfb :2 -screen 0 800x600x16

2) Run Slicerslicer2-linux-x86 -y --no-tkcon --exec AGCommandLine avgFA.hdr FA_1.hdr tr_1.hdr ParamFile.tcl , exit -display :2

3) Repeat 2) for each scan to register using a simple script

average scan

individual scan

scan to coregister

parameter-file

Output

Registered image will be saved.

Registration parameters are stored for future use.

1.091 -0.0209 -0.0802 3.8730.029 1.0827 0.1108 -19.7320.063 -0.0933 0.8805 -2.2770 0 0 1

affine transformationmatrix

non-rigid transformationfile (x,y,z,3)

Results (using lo-res data)

before after affine after non-rigid

atlasindividual

average of 60 registered subjects

Individual

resolution:128*128*1

9(2x2x6mm)