MFSDA_Python is a Python based package for statistical shape analysis.

A multivariate varying coefficient model is introduced to build the association between the multivariate shape measurements and demographic information and other clinical variables.

Statistical inference, i.e., hypothesis testing, is also included in this package, which can be used in investigating whether some covariates of interest are significantly associated with the shape information.

The hypothesis testing results are further used in clustering based analysis, i.e., significant suregion detection.

The project loadable module allowing MFSDA to invoke the command line module MFSDA_run and MFSDA_createShapes. More details below.

Helper script for variable selection using pearson correlations with the principal components (PCA). Runs PCA in the dataset and selects the first 'n' first principal components. A correlation analysis is done with each variable in the dataset and the principal components. This analysis will give you an idea of how correlated each variable is to the first 'n' principal components, in other words, how much does each variable contribute to a principal components. Ideally, when running MFSDA, you should select variables that are not correlated between each other or select variables that contribute to different principal components.

Input table with header row (covariate names). One row per subject.

Output directory. It will generate a pdf file with the plots and a set of json and csv files with outputs of the analysis, i.e., the percentage of contribution for each variable to the principal component.

This is the main script of the application to run the Multivariate Functional Shape Data Analysis.

Run the script MFSDA_run.py with flag '-h' or '--help' for more information.

This is a text file that contains the vtk filenames to the shapes/3D models and the corresponding covariates for each shape. The fields must be separated by ','. The first column must be the filename to the shape.

This is the template that is used to align the shapes in a common space or spherical mapping. Please check SPHARM-PDM

Output directory

Once MFSDA is completed, you may add the resulting pvalues/efit coefficients to a shape from your population.

Run the script MFSDA_createShapes.py with flag '-h' or '--help' for more information.

Input vtk shape/3D model.

Resulting pvalues json file from MFSDA_run.

Resulting efit json file from MFSDA_run.

Covariate names, ex. '--covariates age gender group var1'. These names are used to add arrays to the .vtk file. You can later visualize them in '3DSlicer' or 'ShapePopulationViewer'

Output vtk file.

It is covered by the Apache License, Version 2.0:

http://www.apache.org/licenses/LICENSE-2.0

The license file was added at revision 3ccc5e9 on 2020-12-10, but you may consider that the license applies to all prior revisions as well.

This project was initially developed in early August 2017 by Chao Huang and Hongtu Zhu from the BIG-S2 lab at the University of North Carolina at Chapel Hill School of Medicine.

Then, between November 2017 and June 2018, Juan Carlos Prieto from the School of Dentistry at the University Of Michigan contributed the script for generating shapes with pvalue maps, and Loic Michoud, in the context of internship, contributed the MFSDA Slicer loadable module along with the initial infrastructure to support distribution as a Slicer extension (developed in an independent branch available here).

Then, in November 2018, in the context of an internship at NIRAL (Neuro Image Research and Analysis Laboratories) in the Department of Psychiatry at the University of North Carolina at Chapel Hill School of Medicine, Mateo Lopez further improved the Slicer loadable module by improving the user interface, re-organizing files and renaming the loadable module from ShapeDistanceAnalyzer to MFSDA.

Then, in 2019, Juan Carlos Prieto from NIRAL added the script MFSDA_selectVariablesPCA.py for performing the PCA of a set of variables and doing a correlation analysis with the N principal components.

Later, in 2019-2020, Beatriz Paniagua, Jared Vicory, Jean-Christophe Fillion-Robin and Sam Horvath from Kitware Inc. worked on improving the user interface and robustness, transitioning to Python 3.x and revamping the build-system to support distribution as a Slicer extension.