glm-express

Automated linear models for functional neuroimaging data


License
MIT
Install
pip install glm-express==2.0.3

Documentation

GLM-Express

This is a package for modeling functional neuroimaging tasks. As the name implies, it's optimized to be simple and straightforward! The task_information.json file stores all of the regressors and modeling specifications for each task; modifying this file allows you to test a range of analytical outcomes.

Included

This package comes equipped with the following modeling objects:

  • Subject is a first-level modeler for subject-specific functional neuroimaging data
  • GroupLevel is a second-level modeler that is optimized to aggregate contrast maps derived by the Subject object
  • Aggregator applies first-level models to all subjects in your BIDS project (not efficient for larger datasets)
  • RestingState for analyses of subject-level resting state functional connectivity

Assumptions

We assume the following about your data:

  • Your data is in valid BIDS format
  • Your data has been preprocessed via fmriprep
  • Your preprocessed data are stored in a derivatives folder nested in your BIDS project
  • You have adequate events TSV files for all of your functional tasks
  • Any parametric modulators are stored within each event file
    • Otherwise, you can build custom design matrices and feed them into the modeling function

About task_information.json

Glossary of keys in the task_information file; manipulating these allow you to quickly and effectively customize your modeling parameters without editing any source code

  • tr: Repetition time (defined here, but can be overriden for any one subject)
  • excludes: Subjects in your project you need to exclude for a given task
  • condition_identifier: Column in your events file that denotes trial type; NOTE this will be changed to trial_type in the script
  • confound_regressors: Regressors to include from fmriprep output
  • modulators: Parametric modulators to weight trial type (these should be in your events file)
  • block_identifier: Column in your events file that denotes block type; defaults to null
  • design_contrasts: Your defined contrasts! Include as few or as many as you see fit