2014 Summer Project Week:Slicer Murin Shape Analysis

From NAMIC Wiki
Jump to: navigation, search
Home < 2014 Summer Project Week:Slicer Murin Shape Analysis

Key Investigators

  • Murat Maga
  • Ryan Young


Project Description

Procrustes based shape analyses are a very establish set of geometric morphometric analysis in the realm of developmental and evolutionary biology. Although traditionally conducted on 2D pictures, with the general availability of the 3D (either volumetric or surface) data, field is moving more towards 3D analyses. In case of 3D volumetric data, the typical workflow is to convert the scan dataset into a surface mesh by significantly reducing and smoothing, render the 3D surface on a platform capable of annotating the landmark, export the landmark coordinates into the analysis software (e.g. R), conduct the Procrustes alignment and geometric analyses, and then visualize the results using thin plate splines (TPS). We are interested in creating a geometric morphometric analysis module within Slicer to uniform this experience. Our goal for the project week is to conduct visualization, data collection, statistical analysis and visualizations of shape variation decomposition using Slicer. Our ultimate goal is to be able to repositories (e.g. using Xnat) with already annotated specimens along with all their metadata (species, sex, age, genotype, genomic data, etc.) that can be queried within Slicer (e.g. through XnatSlicer module) and analyze them accordingly. See A Brief Introduction to Statistical Shape Analysis for mathematical details.

Images

Label map.png Threshold murin skull.png File:Murin skull.jpg

Objective

  • Create a GPA/PCA shape analysis and visualization module for Slicer.

Approach, Plan

  • Impliment GPA/PCA shape analysis in python
  • Visualize the deformation of a reference volume along the principle components using thin plate splines
  • Ability to create semi-landmarks to increase spatial coverage. (Using ideas from Morpho package in R)
    • User will create a uniformly sampled point cloud by entering the number of semi-landmarks. Existing “hard” landmarks will be used for their distribution. This will serve as the template to be transferred to all remaining volumes (atlas)
    • The template will be transferred to a new surface. Existing “hard” landmarks will allow for correspondence. The transferred points will then be moved along the surface of the volume by optimizing the bending energy function.
    • The coordinates of the slid landmarks will be saved into a new fiducial list, from which the GPA analysis can be conducted.

Progress

  • Generalized Procrustes Alignment
  • Principal Component and Singular Value Decomposition of the Procrustes aligned coordinates
  • Thin Plate Spline visualization of the shape variables from PCA and/or SVD (by either morphing a reference volume along the shape variable, or visualizing the TPS grid using Transformation Visualizer module).