Exploring the Impact of Analysis Software on fMRI Results

Supporting code to perform the analyses and create the figures of the manuscript with the same title available at https://doi.org/10.1101/285585.

Table of contents

• How to cite
• How to reproduce the figures
  • Figs. 1a & 1b
  • Fig. 2
  • Fig. 3a
  • Fig. 3b
  • Fig. 4
  • Figs. 5a & 5b
  • Fig. 6
  • Fig. 7
  • Fig. 8
• How to rerun the full analysis
• Contents overview

How to cite

To cite this repository, please cite the corresponding manuscript:

"Exploring the Impact of Analysis Software on fMRI Results" Alexander Bowring, Camille Maumet*, Thomas Nichols*. bioRxiv 285585; doi: 10.1101/285585

How to reproduce the figures

Dependencies

To reproduce the figures, you will need to install the dependencies listed in requirements.txt, this can be done using pip with:

pip install -r requirements.txt

You will also need to have Jupyter notebook installed, we recommend using Anaconda to perform the install.

Figs. 1a & 1b

Left column

1. From a Terminal, run:

   jupyter notebook ./figures/ds001_notebook.ipynb
   

2. In the notebook, run all the cells up to the cell starting with from lib import plot_excursion_sets. In this cell, make sure that x_coords=[4, 32]. Running this cell will then reproduce the first column of Figure 1a and first row of Figure 1b.

Middle column

Same as Figs. 1a & 1b, but using ./figures/ds109_notebook.ipynb; again, make sure that x_coords=[0, 32].

Right column

Same as Figs. 1a & 1b, but using ./figures/ds120_notebook.ipynb again, make sure that x_coords=[0, 32].

Fig. 2

Left column

1. From a Terminal, run:

   jupyter notebook ./figures/ds001_notebook.ipynb
   

2. In the notebook, run all the cells up to the cell starting with from lib import plot_stat_images. In this cell, make sure that the coordinates being inputted to the plot_stat_images function are [-17, 1, 15]. Running this cell will then repoduce the first column of figure 2.

Middle column

Same as Fig. 2, but using ./figures/ds109_notebook.ipynb.

Right column

Same as Fig. 2, but using ./figures/ds120_notebook.ipynb.

Fig. 3a

Left column

1. From a Terminal, run:

   jupyter notebook ./figures/ds001_notebook.ipynb
   

2. In the notebook, run all the cells up to the cell starting with from lib import bland_altman which will reproduce the first column of figure 3a.

Right column

Same as Fig. 3a left but using ./figures/ds109_notebook.ipynb.

Fig. 3b

Same as 3a but using ./figures/ds120_notebook.ipynb.

Fig. 4

Left sub-plot

1. From a Terminal, run:

   jupyter notebook ./figures/ds001_notebook.ipynb
   

2. In the notebook, run all the cells up to the cell starting with from lib import dice.

Middle sub-plot

Same as Fig. 4 left but using ./figures/ds109_notebook.ipynb.

Right sub-plot

Same as Fig. 4 left but using ./figures/ds120_notebook.ipynb.

Figs. 5a & 5b

Left column of both figures

1. From a Terminal, run:

   jupyter notebook ./figures/ds001_notebook.ipynb
   

2. In the notebook, run all the cells up to the cell starting with from lib import euler_characteristics.

Right column of both figures

Same as Figs. 5a & 5b left, but using ./figures/ds109_notebook.ipynb.

Fig. 6

1. From a Terminal, run:

   jupyter notebook ./figures/ds001_notebook.ipynb
   

2. In the notebook, run all the cells up to the cell starting with from lib import plot_excursion_sets.

Fig. 7

Left column

1. From a Terminal, run:

   jupyter notebook ./figures/ds001_notebook.ipynb
   

2. In the notebook, run all the cells up to the cell starting with bland_altman.bland_altman('Bland-Altman Plots: Permutation Tests').

Right column

Same as Fig. 7 left but using ./figures/ds109_notebook.ipynb.

Fig. 8

Left column

1. From a Terminal, run:

   jupyter notebook ./figures/ds001_notebook.ipynb
   

2. In the notebook, run all the cells up to the cell starting with reload(bland_altman) bland_altman.bland_altman_intra('Bland-Altman Plots: Parametric vs Permutation').

Right column

Same as Fig. 8 left but using ./figures/ds109_notebook.ipynb.

How to rerun the full analysis

Raw data

To run the experiments included in the manuscript, raw data must be downloaded from OpenfMRI.org and copied on your local computer:

• ds000001 revision 2.0.4: https://openfmri.org/dataset/ds000001/
• ds000109 revision 2.0.2: https://openfmri.org/dataset/ds000109/
• ds000120 revision 1.0.0: https://openfmri.org/dataset/ds000120/

Analysis of ds000001

Data processing using SPM, FSL and AFNI

Given:

• <PATH_TO_RAW_DATA>: the path to the raw data for ds000001 and
• <PATH_TO_OUTPUT>: the path to the output folder where the results should be stored (must end with a ds001 sub-folder).

1. In scripts/process_ds001_SPM.m replace the values of study_dir and results_dir by <PATH_TO_RAW_DATA> and <PATH_TO_OUTPUT> respectively.

2. In scripts/process_ds001_FSL.py and scripts/process_ds001_AFNI.py replace the values of raw_dir and results_dir by <PATH_TO_RAW_DATA> and <PATH_TO_OUTPUT> respectively.

3. For the SPM analysis, inside Matlab run:

   addpath('scripts')
   addpath(fullfile('scripts', 'lib'))
   process_ds001_SPM
   

   This will create onsets, preprocess the data, and run first and group level analyses.

4. For the FSL analysis, from a terminal run:

   python scripts/process_ds001_FSL.py
   

5. For the AFNI analysis, from a terminal run:

   python scripts/process_ds001_AFNI.py
   

Derived data

Derived images

The derived data available on NeuroVault at https://neurovault.org/collections/4110/ can be reproduced as follows:

1. NIDM-Results packs for SPM and FSL are available in <PATH_TO_OUTPUT>/SPM/LEVEL2 and <PATH_TO_OUTPUT>/FSL/LEVEL2 respectively.

2. For the resliced images, inside Matlab run

   addpath('scripts')
   addpath(fullfile('scripts', 'lib'))
   ds001_reslice_images
   

Other derived data

The csv files containing the Euler characteristics can be recomputed in Matlab, using: addpath('scripts') addpath(fullfile('scripts', 'lib')) ds001_euler_chars

Analysis of ds000109

Same as for ds000001, replacing all occurences of 001 by 109 and https://neurovault.org/collections/4110/ by https://neurovault.org/collections/4099/.

Analysis of ds000120

Same as for ds000001, replacing all occurences of 001 by 120 and https://neurovault.org/collections/4110/ by https://neurovault.org/collections/4100/.

Contents overview

 	ds001: Part of ds001 output data (excluding images)
	ds109: Part of ds109 output data (excluding images)
	ds120: Part of ds120 output data (excluding images)
	figures: Scripts and notebooks to reproduce the figures
	scripts: Scripts to rerun the analysis 
	.gitignore: git configuration file
	README.md: current file