Delphine Ribes

1

• Delphine Ribes
• (Internship UNC 2005/2006)
• Guido Gerig

2
Project
Measure Cortical Thickness by using the
Danielsson Distance Transform
White matter Boundary
Grey Matter exterior boundary
?
Cortical Thickness
?
3
Just another cortical thickness method

• Given imperfect, somewhat noisy tissue images
• Perfect, topologically correct surface cannot be
  obtained
• Not necessary to inflate cortex
• Display on wm and gm surfaces
• Separate measure of thickness inside folds and on
  outer cortex
• Choice Engineering method based on Euclidean
  distance transform
• Good enough to collect statistics across cortical
  areas, not precise measurements at each voxel

4
Pipeline
5

• Danielssons DT propagates a vectorial
  information, the location of the nearest pixel of
  the set S

6
Grey matter
Distance map
White matter

• Danielsson distance transform of white matter

• At each grey matter voxel, we know the
  distance to the white matter

gt At each grey matter voxel, store vector to the
closest point
7
Grey Matter exterior boundary
White matter Boundary
Vector Map

• Problem
• Each grey matter voxel points to white matter
  surface
• Gray matter surface in folds is not known

• Solution
• At each wm surface point, choose the grey matter
  voxel distance with maximum distance

Method Maximum Operator
gt For each white Matter point, find the furthest
grey Matter point
8
White matter
Mean 5,02 mm
Grey matter
Danielsson Distance Map
White Distance Map
Grey Distance Map
9
(No Transcript)
10
Results differ Reason is that thickness within
folds is spatially limited!!!
11
Only thickness of outer cortex but not inside
folds Values match validation sphere!
12
(No Transcript)
13
Traveling Phantom Validation Duke QC study
(Cecil Charles)
14
(No Transcript)
15
Full Brain Cortical Thickness as fct of Age N22
Full brain cortical thickness as function of age
shows peak at around 12 years of age.
16
Summary Procedure
Distance gray matter surface to white matter
boundary
Tissue Segmentation
Cortical thickness for full brain and for
individual lobes
Lobe Parcellation
17
Web download of tool and manual at
www.ia.unc.edu/dev
18
Status

• Tool fully written in ITK
• Toolkit available on UNC website and compiled on
  UNC dashboard
• Visualization with NA-MIC MeshVisu (Oguz)
• Comprehensive manual, very easy to use
• Might go very well with Slicer3 EMS (needs tissue
  segmentation and parcellation)