Skip to content

Uses regression and factor model approach to correct for site effects in fMRI volumes

License

Notifications You must be signed in to change notification settings

neuroquant/fmri-site-effects

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

81 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

A toolbox to quantify and adjust for session and site effects in fMRI time-series

Gitter

This toolbox implements a number of approaches to correct fMRI time-series or volumes for session specific artifacts and site effects, aka "batch Effects".

Contents

This toolbox has 3 types of functions.

  • Primary functions to measure reliability, implement alternatives to the standard correlation matrix that adjust for nuisance variation or artifacts.

  • Demo functions that illustrate the function API and usage of primary functions. These functions start with demo_

  • Analysis functions that perform a detailed case study of a particular method or pipeline on a dataset. These functions start with analyze_

Primary functions

  • siteEffects: A class containing useful methods for diagnosis of site effects and other metrics of improvement.

    • siteEffects.within_group_error()
    • siteEffects.test_retest_mccc()
  • projpca, demo_projpca: Multivariate correction on time-series data tensor. Provides residual time-series after regressing out site effects can be used to infer correlation matrices. Disclaimer: This function is still being tested, unfinished.

Demo functions

  • demo_successive_normalization: Demonstrates examples of calling ggmClass for data cleaning and correlation estimation. Is a wrapper around local functions that combine different forms of standardization of data matrix with a variety of covariance estimators. Exported figures can be found in tmp/<date>/demo_successive_norm_*.png

  • demo_conditional_correlation.m: Demonstrates examples of calling ggmClass for data cleaning and correlation estimation. Is a wrapper around local functions that perform factor model based decomposition of the observed covariance by conditioning on covariates.

  • demo_robust_covariance.m: Demonstrates examples of calling ggmClass to call robust estimators of the covariance or correlation matrix.

Analysis functions

  • analyze_etkinlab_sessioneffects.m: To be documented

  • analyze_etkinlab_siteeffects.m: To be documented

  • analyze_abide_siteeffects.m: To be documented

Installation

Use the --recursive setting to make sure the ggmClass and connectivity-diagnostics submodule are added. Disclaimer: You will need to have keyless ssh login to github to access private repositories.

git clone --recursive git@github.com:TheEtkinLab/fmri-site-effects.git

Modify setup.m to specificy paths to matlab-library or your local equivalent. To add package dependencies, run setup using

run('setup.m')

About

Uses regression and factor model approach to correct for site effects in fMRI volumes

Topics

Resources

License

Stars

Watchers

Forks

Packages

No packages published