DBP2:UNC:Cortical Thickness Roadmap

Objective

We would like to create an end-to-end application within Slicer3 allowing individual and group analysis of regional and local cortical thickness. Such a workflow applied to the young brain (2-4 years old) is one goal of the UNC DBP. This page describes the technology roadmap for cortical thickness analysis in the NA-MIC Kit. The basic components necessary for this end-to-end application are:

• Tissue segmentation: Should be multi-modality, correcting for intensity inhomogeneity and work on non-skull-stripped data
• Cortical thickness measurement: Local cortical thickness needs measurements at every location of the white-gray matter boundary, as well as at the gray-csf boundary. Regional analysis does not need such a dense measurement.
• Cortical correspondence: Local analysis needs a full correspondence on both white-gray boundary and gray-csf boundary.
• Statistical analysis/Hypothesis testing: Measurements need to be compared and tested localy incorporating multiple-comparison correction, correlative analysis would be necessary too.

Roadmap

Starting with several MRI images (weighted-T1, weighted-T2...) we want to obtain cortical thickness maps for each subject, compute cortical correspondences between subjects, and analyze the cortical thickness at these corresponding locations. Ultimately, the NA-MIC Kit will provide a workflow for individual and group analysis of cortical thickness. It will be implemented as a set of Slicer3 modules that can be used interactively within the Slicer3 application as well as in batch on a computing cluster using BatchMake.

Next we discuss the main modules and details of current status and development work:

White matter/gray matter segmentation

• UNC has a open source segmentation tool called itkEMS (binary download) implemented in an ITK framework for segmenting white matter and gray matter in the young brain. This technique will be converted into a Slicer3 command line module
  • Since this segmentation technique exists in an ITK framework, the integration into Slicer3 is low risk and should be completed over the next couple of months (mid fall)
• UNC will also investigate adapting the Slicer3 EM Segment module to their young brain studies. Here, UNC will adapt the UNC atlas of the 2 year old brain to provide priors for the EM Segment module
  • This will be a good test case for applying the Slicer3 EM Segment module to a slightly different application. UNC should work through the training material on the Slicer3 EM Segment module and then refer to Brad Davis and Kilian Pohl as needed.
  • This also should be completed before the AHM.

Local cortical thickness measurement

• UNC has an algorithm to measure a local cortical thickness given a labeling of white matter and gray matter. This technique will be converted into a Slicer3 command line module
  • This technique is non-symmetric and sparse (only computing distances where they can be computed reliably).
  • This module is well suited for regional analysis, but additional work to interpolate dense measurements along the boundaries from the sparse ones would be necessary for a local analysis.
  • It is expected that this module should be available before the AHM (for the current, sparse measurements).
• Marc Niethammer developed a technique at a previous project week that would be symmetric. This could be used for the local cortical thickness analysis, if our comparison studies show appropriate results.

Local correspondence

• Regional as well as local subject comparisons are needed
• Regional analysis will require precise deformable registration to a young brain atlas
  • NA-MIC toolkit can be applied here, Slicer 3 has a b-spline based MI registration which needs to be tested
  • Parcellation atlas for regional analysis is the current atlas used by the DBP
• Local analysis requires techniques which are not currently in the NA-MIC Kit
  • Freesurfer can be used for the local analysis (but it is not in the NA-MIC Kit)
  • Ipek Oguz is developing local analysis tools and should be available in Fall 2008. The tool necessitates the following steps (also performed within Freesurfer)
    • 1) spherical topology of white matter: Marc Niethammer has Slicer module for this step, testing will be necessary to ensure appropriateness in the cortical case
    • 2) inflation of cortical surface: A Slicer module exists for this step, but likely not with the properties. Tests and possible additional implementation will need to be done.
    • 3) computation of correspondence on inflated surface using particle system: No module exists, first prototype programs are in development in collaboration with Utah (I Oguz, J Cates)

Performance characterization and validation

• Characterize response based on signal noise, patient motion, etc.
• Comparison to other tools (FreeSurfer)

To do

• Assign owners to tasks
• Define schedule

Schedule

• Tissue segmentation

• 15/10/2007 - White matter/gray matter segmentation of the young brain using itkEMS as a Slicer3 module (C Vachet) - at AHM
• 01/12/2007 - White matter/gray matter segmentation of the young brain using the Slicer3 EM Segment module (C Vachet) - at AHM
• 15/12/2007 - Comparison study itEMS vs EM Segment in Slicer 3 (C Vachet) - at AHM

• Cortical thickness

• 23/12/2007 - UNC Cortical thickness measurement as a Slicer3 module (C Vachet) - at AHM
• 23/12/2007 - Niethammer's Laplacian Cortical thickness measurement module code working at UNC as a Slicer3 module (C Vachet) - at AHM
• 01/03/2008 - Regional comparison between UNC and Laplacian cortical thickness (C Vachet)

• Cortical correspondence
• xx/xx/2007 - Deformable registration of young brain regional atlas
• xx/xx/2007 - Regional analysis of cortical thickness as a Slicer3 module
• xx/xx/2007 - BatchMake workflow
• xx/xx/2007 - Groupwise regional analysis of cortical thickness as a NA-MIC Workflow
• xx/xx/200x - Groupwise local analysis of cortical thickness as a NA-MIC Workflow

Team and Institute

• Co-PI: Heather Cody Hazlett, PhD, (heather_cody at med.unc.edu, Ph: 919-966-4099)
• Co-PI: Joseph Piven, MD
• NA-MIC Engineering Contact: Jim Miller, GE Research
• NA-MIC Algorithms Contact: Martin Styner, UNC