DBP2:UNC:Local Cortical Thickness Pipeline

Back to UNC Cortical Thickness Roadmap

Cortical thickness on white matter cortical surface

Objective

We would like to create an end-to-end application within Slicer3 allowing group-wise automatic mesh-based analysis of cortical thickness as well as other surface measurements (surface area...)

This page describes the related pipeline with its basic components, as well as its validation.

Pipeline overview

A Slicer3 high-level module for group-wise cortical thickness analysis has been developed: GAMBIT (Group-wise Automatic Mesh Based analysis of cortIcal Thickness)

Input: CSV file containing RAW images (T1-weighted, T2-weighted, PD-weighted images)

• 1. Individual pipeline
  • 1.1. Tissue segmentation
    • Probabilistic atlas-based automatic tissue segmentation via an Expectation-Maximization scheme
    • Tool: itkEMS (UNC Slicer3 external module)
  • 1.2. Atlas-based ROI segmentation: subcortical structures, lateral ventricles, parcellation
    • 1.2.1. Skull stripping using previously computed tissue segmentation label image
      • Tool: SegPostProcess (UNC Slicer3 external module)
    • 1.2.2. T1-weighted atlas deformable registration
      • B-spline pipeline registration
      • Tool: RegisterImages (Slicer3 module)
    • 1.2.3. Applying transformations to the structures
      • Tool: ResampleVolume2 (Slicer3 module)
  • 1.3. White matter map creation
    • Brainstem and cerebellum extraction
    • Adding subcortical structures except amygdala and hippocampus
    • Tool: ImageMath (UNC Slicer3 external module)
  • 1.4. Cortical thickness computation
    • Asymmetric cortical thickness
    • Tool: UNCCortThick(UNC Slicer3 external module)
  • 1.5. White matter map post-processing
    • Largest component computation
    • Smoothing: Level set smoothing or weighted average filter
    • Connectivity enforcement (6-connectivity)
    • White matter filling
    • Tool: WMSegPostProcess (UNC Slicer3 external module)
  • 1.6. Genus zero white matter map image and surface creation
    • Tool: GenusZeroImageFilter (UNC Slicer3 external module)
  • 1.7. White matter surface inflation
    • Iterative smoothing using relaxation operator (considering average vertex) and L2 norm of the mean curvature as a stopping criterion
    • Iteration stopped if vertices that have too high curvature (some extremities)
    • Tool: MeshInflation (UNC Slicer3 external module)
  • 1.7 bis(Optional). White matter image fixing if necessary
    • Correction of the white matter map image (corresponding to vertices that have high curvature) with connectivity enforcement
    • Tool: FixImage (UNC Slicer3 external module)
    • Go back to step 1.6
  • 1.8. Sulcal depth
    • Sulcal depth computation using genus-zero surface and inflated one
    • Tool: MeshMath (UNC module)
  • 1.9. Surface area computation
    • Lobar surface area measurement on smoothed genus-zero surface
    • Tool: MeshMath (UNC module)
  • 1.10. Particles initialization for cortical correspondence
    • Initializing particles on inflated genus-zero surface using 98-lobe parcellation map and genus zero surface
    • Tool: ParticleInitializer (UNC Slicer3 external module)
• 2. Particle-based shape correspondence
  • Correspondence on inflated surfaces using particle system
  • 2.1. Preprocessing
    • Distance maps creation from inflated genus-zero surfaces with slight gaussian blurring
    • Tool: ParticleCorrespondencePreProcessing (UNC Slicer3 external module)
  • 2.2. Correspondence optimization
    • Particle-based shape correspondence optimization (using sulcal depth) with Procrustes alignement
    • Tool: ShapeWorksRun (Utah Slicer3 external module)
  • 2.3. Postprocessing
    • Re-meshing using template
    • Tool: ParticleCorrespondencePostProcessing (UNC Slicer3 external module)
  • 2.4. Cortical thickness interpolation
    • Cortical thickness interpolation on surface in correspondence
    • Tool: MeshMath (UNC module)
• 3. Group statistical analysis
  • Tool: QDEC Slicer module or StatNonParamPDM

T1-weighted image

T1 corrected image

Label image

White matter mesh

T1-weigthed atlas with subcortical structures

ROI segmentation on T1-weigthed stripped image

Genus-zero cortical surface

Inflated cortical surface

Cortical thickness on genus-zero cortical surface

Cortical thickness on inflated genus-zero cortical surface

Sulcal depth on genus-zero cortical surface

Sulcal depth on inflated genus-zero cortical surface

Particles on inflated genus-zero cortical surface

GAMBIT Download

Source code, executables and tutorial are available on NITRC

Brain atlases

Four brain atlases are available on MIDAS and on NITRC:

• Pediatric atlas
• Adult atlas
• Elderly atlas
• Primate atlas

Pediatric MRI Brain data

Data of 2 autistic children and 2 normal controls (male, female) scanned at 2 years with follow up at 4 years from a 1.5T Siemens scanner. Files include structural data, tissue segmentation label map and subcortical structures segmentation.

Pipeline validation

Analysis on a small pediatric dataset

Tests will be computed on a small pediatric dataset which includes 2 year-old and 4 year-old cases.

• 16 autistic cases
• 1 developmental delay
• 3 normal control

Comparison to state of the art

We would like to compare our pipeline with FreeSurfer. We will thus perform a regional statistical analysis using Pearson's correlation coefficient on an adult dataset (FreeSurfer's publicly available tutorial dataset) including 40 cases.

Planning

Done

• Workflow for group analysis (Slicer3 external module using BatchMake):
  • Development of UNC Slicer3 modules
  • Modules applied on small pediatric dataset from the Autism study
• Pediatric and adult brain atlases available to the community via MIDAS
  • T1-weighted atlas
  • Tissue segmentation probability maps
  • Subcortical structures probability maps
  • LBinary mask images
• GAMBIT available to the community via NITRC: executables (UNC external modules for Slicer3) and tutorial dataset
• Tutorial with application example on a small dataset
• GAMBIT source code (CVS) available to the community

In progress

• Step 1.7: Parameter exploration on autism dataset to improve inflation-fixing steps
• Step 2: Particle correspondence testing with pediatric surfaces (Meeting with Josh Cates at UNC - February 2010)
• New version of GAMBIT including quality control through MRML scene, and WM, GM models generation
• GAMBIT executable can be downloaded directly within Slicer3 using the extension wizard

References

• C. Vachet, H.C. Hazlett, M. Niethammer, I. Oguz, J.Cates, R. Whitaker, J. Piven, M. Styner, Mesh-based Local Cortical Thickness Framework, UNC Radiology Research Day 2010 abstract
• I. Oguz, M. Niethammer, J. Cates, R. Whitaker, T. Fletcher, C. Vachet, and M. Styner, Cortical Correspondence with Probabilistic Fiber Connectivity, Information Processing in Medical Imaging, IPMI 2009, LNCS, 21:651-63
• C. Vachet, H.C. Hazlett, M. Niethammer, I. Oguz, J.Cates, R. Whitaker, J. Piven, M. Styner, Mesh-based Local Cortical Thickness Framework, UNC Radiology Research Day 2009 abstract
• Oguz, I., Cates, J., Fletcher, T., Whitaker, R., Cool, D., Aylward, S., Styner, M., Cortical correspondence using entropy-based particle systems and local features, IEEE Symposium on Biomedical Imaging ISBI 2008. 1637– 1640
• J. Cates, P. Fletcher, M. Styner, H. Hazlett, R. Whitaker, Particle-based shape analysis of multi-object complexes, MICCAI 2008, 477-85
• Cates, J., Fletcher, P., Whitaker, R.: Entropy-based particle systems for shape correspondence. MFCA Workshop, MICCAI 2006, 90–99