Collaboration:UNC-PNL

Goal

• Reviewing shape analysis with current NAMIC data
• Establishing a concrete plan and schedule for collaborations. Target projects are shape analysis and statistical analysis of anisotropy measure along fibers.

Agenda

• 10am: UNC presentations for Psychiatrists/Psychologists:
  • Shape Analysis of Caudate and Corpus Callosum Data (30 min)File:2005 NAMIC ShapeSchizoBWH.ppt
  • ITK SNAP: A level Set Semi Automatic Segmentation tool and its use at UNC for the segmentation of brain structures File:2005 NAMIC ShapeSchizoBWH.ppt
  • UNC Tools for the Statistical Analysis of Anisotropy of Diffusion Tensor Measures along White Matter Fibers. (30min) File:Isabelle100705-1.ppt

• 11:15am: Open Discussion with PNL on immediate and future collaborations plans.
  • Establish what UNC needs from PNL and PNL from UNC.
  • Resolve technical issues such authorship, clinical data, etc...

• 12pm: Lunch
• 1pm: Wrap up of PNL/UNC collaboration plans
• 2pm: UNC presentations for Computer Scientists.
  • Shape Analysis: UNC Shape Analysis Pipeline in ITK/NAMIC File:2005 10 ShapePipeline.ppt
  • Shape Analysis: Statistical Non-Parametric Permutation tests for shape hypothesis testing, File:2005 10 NonParamShapeStats.ppt
  • DTI Tensor Analysis File:Isabelle100705-2.ppt

Participants

• Martin Styner (UNC)
• Isabelle Corouge (UNC)
• Martha Shenton (PNL)
• James Levitt (PNL)
• Marek Kubicki (PNL)
• C-F Westin (BWH)
• Marc Niethammer (PNL)
• Sylvain Bouix (PNL)

Discussion overview

Statistical shape analysis:

Caudate (SPD):

• Spherical harmonics for shape description
• local/global/parcellated significance and anlysis
• Local and global differences on R, but only local on L
• main shape difference in caudate head, small effect in caudate tail
• volume reduction in SPD

Corpus callosum:

• model based 2D segmentation of corpus callosum
• DTI fiber tracking based corpus callosum subdivision method (coloring based on connection to white matter parcellation -- given as a template) results in probabilistic subdivision of the corpus callosum

DTI:

• quantitative analysis of DTI data (along fibers; statistics within fiber bundles)

PNL provides:

• an interpretation of the shape differences seen in SPD vs control group caudates
• a larger dataset to improve the statistical significance (chronic SZ?)

UNC provides:

• statistical shape analysis software
• fiber viewer and fiber tracking software (in Namic sandbox see also software here)
• tools for quantitiative analysis of DTI data (statistics of diffusion metrics and tensors along fibers and across fibers within a fiber bundle)

Issues:

• very small sample sizes (-> PNL will provide more and different data)
• local shape difference significance measures are very conservative or not conservative enough (is there a middle groud?)
• not clear how to compare fiber tracts between subjects (how to align?)

Possible future work:

• Do statistical caudate analysis for chronic SZ (expect larger changes than for SPD)?
• Hippocampus study using the UNC software?