Automatic Statistical Shape Analysis
Project Description
Summary
This project will create an automated tool to peform
a statistical analysis of the differences between populations
of hippocampi. This analysis differs from the medial analysis
protocol now in use in the following ways:
- it's based on surfaces (shapes).
- it's "true 3d" (does not depend on 2d planar approxmations of UCF contours).
- it does not depend on the assumption that you can reasonably fit a line to the (Hippocampus) shape.
- it will be cross-platform (e.g. Java, and C++ binaries for Sun,Mac,Windows and selected Linux variations).
- source code will be available.
This tool will be composed of several computer
programs, applied sequentially, whose final result will be a
an atlas created from the analyzed hippocampus surfaces.
Each vertex of this atlas will have an associated p-value indicating
the statistical significance of a selected metric used to characterize
the difference between an individual hippocampus and the group atlas.
In this proposal, we will use the Euclidean distance as the difference metric, and a Students T-Test
as the statistical method. Our modular approach to the computations (e.g. via a distinct
computer program) means that other metrics and statistics may easily substituted, as they become
available.
Expected Implementation
We expect to implement this as
- a set of Java programs to create the atlas and statistics (Craig's group provides this)
- a C++ program to perform the Direct Mapping of Hippocampus (DMH) (from Yongang).
- a Pipeline to tie it all together.
- web pages documenting all the parts, and how they fit together
- Various LONI software download package containing all the above.
Personnel and Resources
- Yongang Shi - DMH coding, DMH documentation
- Craig Schwartz - ShapeTools coding, project coordination, download packaging, documentation
- Ryan Cabeen - ShapeTools coding.
- BaeCheol Shin - ShapeTools coding.
- Ivo Dinov - coordinates with test researchers.
- Nancy Sicotte - test researcher
Schedule
- 15 June - Draft Project Definition
- 21 June - Final Project Definition
- 20 July - Project Approval
- 22 October - Beta release to friendly, external testers
- 19 November - Final release
Products
- Tools
- Shape Matching software - Direct Mapping of Hippocampus (DMH), in these forms:
- Windows Compiled binary, as LONI software download package
- Sun compiled binary , as LONI software download package
- Macintosh compiled binary , as LONI software download package
- Ubuntu Linux compiled binary , as LONI software download package
- Source code, as a LONI software download package
- Source code, in a CCB source code archive (SVN).
- Shape optimal rigid transform calculation software (java), as LONI software download package.
- Shape transform software (java), as LONI software download package.
- Shape difference metric calculation tool (java), as LONI software download package.
- Shape averaging software (java), as LONI software download package.
- Shape t-test calculation (java), as LONI software download package.
- A Loni Pipeline to perform the analysis, configured to use the LONI Pipeline Server to perform the calculations.
- Documentation
- Web pages for all tools
- A user manual ,in PDF form, including
- a description of how to use the plpeline.
- illustrations of the various stages of the analysis.
- Sample Analyses
- A sample data set, divided into two populations.
- The results for running the sample data set
- A pipeline configured to analyze the sample data set
- Instructions on unpacking this sample data set and running the pipeline analysis.
Analysis Procedure
Data: a group of shapes from controls (G_Ctrl) and a group of shapes from patients (G_Pat):
1) Determine the Atlas from the control group (G_Ctrl)
a) pick one shape from the control group.
b) rigidly align other shapes to this one
c) compute average of (b) and the ref. shape -> AVG0
d) align all shapes of the control group to AVG0 -> G1
e) compute average of G1 -> AVG1
f) measure change between || AVG1 - AVG0 ||
g) repeat (d),(e),(f) until (f) < tolerance
h) final AVG_N is the "AtlasG0"
2) at each point in each control shape (G_Ctrl):
a) find difference vector from the shape to the Atlas (dilation)
b) project this vector onto the atlas surface normal at that point
c) append the magnitude of this projection to a list for this vertex
3) do (2) except for disease shapes G_Pat)
4) calculate a statistical test (T-Test, initally) comparing the two lists for significance (p-value)
5) assign this P-value to a copy of the atlas (p<0.05 is probably significant).
Tasks
- modify pipeline to use direct hippocampal mapping program (via pipeline server)
- Get sample data from NS
- Define and write atlas creation application
- Debug RigidAtlas program
Accomplishments
