apaper_simba

This repository maintains the code and scripts for "SIMBA: Scalable Image Modeling using a Bayesian Approach, A Consistent Framework for Including Spatial Dependencies in fMRI Studies".

The data applied to this code is stored at the OSF project.

Notes on running the Jupyter notebooks for the SIMBA code.

Directory setup

To run the examples here, the user's directory should be set up as follows:

apaper_simba/ : copy of GitHub repository of SIMBA code and notebooks data/ : directory with data for examples |-- NARPS : dir with input data pickles and NIFTI data for Ex. 2 `-- simulation : dir with input data pickles for Ex. 1

The simulation demo (Ex. 1) uses the mask dataset in the NARPS directory when generating baseline inputs.

Note that Ex. 3 uses ABCD which is not publicly shareable in the same way as the NARPS and simulation data that are openly included here. Users can download that separately to run the Ex. 3 notebook.

Environment/dependencies

The Python module and other dependencies needed to run these data projects are listed in the import .. lines of the repository notebooks. Additionally, the code repository's environment_apaper_simba.yml text file contains a list of all dependencies across the notebook examples.

To use conda to install the dependencies, one can install the program (using, say, Miniconda) and then execute the following to build it:

conda env create -f environment_apaper_simba.yml

Then, users can activate the environment with:

conda activate apaper_simba

Running the example notebooks

The following Jupyter notebooks are included in this code repository:

abcd_analysis.ipynb : notebook for Ex. 3 (ABCD surface data) narps_analysis.ipynb : notebook for Ex. 2 (NARPS volumetric analysis) simulation_analysis.ipynb : notebook for Ex. 1 (simulated data)

The directories of data for narps*.ipynb and sim*.ipynb can be downloaded directly from OSF. As noted above, data for abcd*.ipynb must be obtained separately.

To run any simulation, first organize the code and data set shown above. Then, start a Jupyter interface (with the appropriate dependencies available; see previous section) from the code directory, such as by running:

cd apaper_simba jupyter-notebook

Finally, select the appropriate notebook file, and execute the cells within it.