Please cite the following article:

Han, H. (2021). A method to adjust a prior distribution in Bayesian second-level fMRI analysis. PeerJ, 9, e10861. https://doi.org/10.7717/peerj.10861

Prior-Adjustment-BayesFactorFMRI

Adjusting prior distributions for Bayesian second-level fMRI analysis

This repository includes files for the prior adjustment for Bayesian second-level fMRI analysis. There are two case examples available: the simulation data test (/Simulation) and the working memory fMRI data test (/Working_memory_fMRI). Directions for each example are available in the following sections.
In order to run the tests, both Python (tested with 3.7.3; >= 3.7.3 required, but 3.8 is not recommended) and R (tested with 4.0.2; >= 3.5 required) are required. Here are the list of dependencies for each language:

1. Python: tkinter (developed with 8.6), shutil, pandas (developed with 0.24.2), nibabel (developed with 2.4.1), rpy2 (developed with 3.2.2), numpy (developed with 1.16.2), nilearn (developed with 0.6.2), subprocess. Specified directions about how to install required dependencies are available in https://github.com/hyemin-han/BayesFactorFMRI/blob/master/README.md
2. R: BayesFactor (developed with 0.9.12-4.2), metaBMA (developed with 0.6.1), oro.nifti (developed with 0.9.1), imager, tools

For further details about required software and dependencies, refer to and cite:

• Han, H. (in press). BayesFactorFMRI: Implementing Bayesian second-level fMRI analysis with multiple comparison correction and Bayesian meta-analysis of fMRI images with multiprocessing. Journal of Open Research Software. (currently available at bioRxiv https://bit.ly/34t7555)
  

In addition, please cite these papers as well:

• Bayesian second-level fMRI analysis with multiple comparison correction: Han, H. (2020). Implementation of Bayesian multiple comparison correction in the second-level analysis of fMRI data: With pilot analyses of simulation and real fMRI datasets based on voxelwise inference. Cognitive neuroscience, 11(3), 157-169.
• Bayesian meta-analysis of fMRI data: Han, H., & Park, J. (2019). Bayesian meta-analysis of fMRI image data. Cognitive neuroscience, 10(2), 66-76.

Simulation data test

Source code and data files for the simulated data test are available in /Simulation. The test can be performed with "run_cycle.py". Required parameters can be entered on command line. Here are guidelines about how to enter the parameters:
> run_cycle.py Proportion(R) Ns Percentiles(P) Iterations Cores

• Proportion (R): Only one R can be provided. R should be one of the nii filenames in "originals" subfolder. For instance, if we want to test .10%, then it should be "0.10"
• Ns: The number(s) of subjects (or image files) to be tested. Multiple values can be provided with ",". For instance, if we want to test Ns = 8, 12, 16, 20 as done in the paper, then "8,12,16,20".
• Percentiles (P): P value(s) to be tested. Multiple values can be provided with ",". If P = "0," then the default Cauchy prior (sigma = .707) after multiple comparison correction is used. For instance, if we want to test 80%, 85%, 90%, 95%, and the default Cauchy prior, then ".80,.85,.90,.95,0".
• Iterations: How many times each condition shall be tested?
• Cores: How many cores shall be utilized?

For instance, if we want to test the simulated image with .10% of true positives (R = .10%), and Ns = 8, 12, 16, 20, P = 80%, 85%, 90%, 95%, Default Cauchy prior with 10 iterations for each condition with 4 cores, then,
> run_cycle.py 0.10 8,12,16,20 .80,.85,.90,.95,0 10 4
The current original files were created with C (contrast) = 1 and N (noise strength in terms of standard deviation) = .50.

Working memory fMRI data test

If you want to test the working memory fMRI data as described in the paper, please use the files in /Working_memory_fMRI. There are three subfolders in this folder:

1. Meta_analysis: NIfTI files that report results from meta-analyses are contained in this subfolder. Run "meta_test.R" to estimate parameters required for the prior adjustment, C, N, and R.
2. BayesFactorFMRI: Files for Bayesian second-level fMRI analysis with prior adjustment are available. Run "run_meta_test.py" to perform Bayesian analysis (five conditions: no adjustment, P = 80%, 85%, 90%, 95%). Resultant nii files will be created in the same folder.
3. Performance_evaluation: Files for performance comparison are available. In this folder, resultant nii files from BayesFactorFMRI and SPM 12 are provided. In addition, meta-analysis results for evaluation created by BayesMeta, BrainMap, NeuroSynth, and NeuroQuery are also available. Run "test_overlap_working.R" to perform performance evaluation.

Working memory fMRI data test (2)

The same test can be done with another working memory fMRI dataset in /Working_memory_fMRI_2. Among three steps in the prior section, only 2 and 3 can be performed with the second working memory fMRI dataset.

Want to test the method with your own nii files?

Please refer to https://github.com/hyemin-han/Prior-Adjustment-BayesFactorFMRI/blob/master/Apply.md