Workflow Entry: Local Shape Analysis (LSA) Workflow

This workflow takes raw un-skull-stripped volumes in 2 groups/populations and an ROI index to generate a color-coded shape object using the per-vertex local p-values of some local shape-measure (e.g., displacement, atrophy, Jacobian, curvature, etc). It also outputs the p-values color-map superimposed on the mean shape.

Problem addressed by this workflow

Identifying local group differences between 2 populations in one specific ROI. This workflow also generates mean-shape DX model and a CCBBM *.m file with vertex-coloring using p-values, which can be viewed as a scene in ShapeViewer. The p-values color-map is:

• Red: p<=0.05
• Green: 0.05<p<=0.1
• Blue: p>0.1.

Detailed Workflow Usage & Specifications

• Input: Any volumetric file format can be used for the raw unskull-stripped imput images (e.g., DICOM, Directories, Analyze, MNC, Nifti, etc.)
• Pre-processing
  • Image conversion to Analyze
  • Converting to 8-bit
  • Intensity Inhomogeneity correction
  • Skull-stripping
  • Affine Registration
• Auto Volume Parsing
  • Non-linear Registration to LPBA40 Atlas
  • 56 ROI Volume Parcellation
  • Mapping of ROI labels in native space
• Shape Modeling and Refinement
• Computing the local Mean and Gaussian curvatures per vertex on the shape of choice for all subjects
• Shape registration - mapping, shape-correspondences
• Mean (atlas) shape generation
• (Non-parametric) permutation test for statistical differences between the 2 groups locally computed on each vertex of the corresponding shapes
• Color-coding (mapping) the p-values on the mean-shape (shape-atlas)

URL: http://www.loni.ucla.edu/twiki/bin/view/CCB/PipelineWorkflows_LSA