neuropower-validation

This repository contains all code to produce the methods and results in our paper 'Power and sample size calculations for fMRI studies based on the prevalence of active peaks'. Everything was run on a SLURM system. The structure of this repository is as follows:

Folders:

• Example/ contains all code used for the example in the paper
• PythonScripts/: contains a few pythonscripts for the HCP or simulated validations
• SimulationFiles/: contains files used by the simulations (activation maps,...)
• HcpFiles/: contains files used by the HCP validations (userID's,...)
• Functions/: python functions called by the python scripts
• Figures/: contains scripts producing the figures in the paper

Scripts:

• SIM_tacc.sbatch: the batch script that executes all simulations
• HCP-prospective.sbatch: the batch script that executes all HCP validations
• collect.sh: a batch script (meant for interactive use) that combines all data
• config_tacc.sh: some parameters about local machine structure, see below

Workflow

Reproduce

To reproduce the results on a local machine, please customize config_tacc.sh with a personal folder structure. To reproduce the results in this paper, follow these pipelines:

Example

All files are inside folder Example/ Running Example.sh calls Example.py, while using Mask.nii.gz and Zstat1.nii.gz, which produces the figures.

Validations

1. SIM_prospective.sbatch performs the simulations and exports for each simulation (1) a file with the model estimation, (2) a file with the predicted power and (3) a file with the true power. Calls PythonScripts/SIM.py while using files in the folder SimulationFiles/. Should be run 20 times with environment variable i ranging from 0 to 20.
2. HCP_prospective.sbatch performs the resampling scheme using the Human Connectome Project data. Calls PythonScripts/HCP.py while using files in the folder HcpFiles/. Should be run 25 times with environment variable i ranging from 0 to 20.
3. collect.sh: depending on whether the parameter MODALITY is set to HCP or SIM, this script combines and reshapes all the data in 4 files:

• true_$MODALITY_predictive_2.3_RFT.csv: true values for power
• prediction_$MODALITY_predictive_2.3_RFT.csv: predicted values for power
• estimation_$MODALITY_predictive_2.3_RFT.csv: model estimation parameters
• conditional_$MODALITY_predictive_2.3_RFT.csv: predicted sample size versus true sample size for 0.8 power This script calls PythonScripts/aggregate_estimation.py

Computing time

These analyses should be performed on a high-performance computing environment. We performed the analysis on Texas Advanced Computing Center: 120 general compute nodes with dual socket Intel(R) Xeon(R) CPU E5-2650 v2 @ 2.60GHz (8 core/socket); 64 GB 1866 MHz DDR3.

• HCP analyses: computing information for 1 resample, 1 contrast:

User time (seconds): 557.53
System time (seconds): 52.56
Percent of CPU this job got: 84%
Elapsed (wall clock) time (h:mm:ss or m:ss): 12:04.25
Maximum resident set size (kbytes): 1821596
Average resident set size (kbytes): 0
Major (requiring I/O) page faults: 150
Minor (reclaiming a frame) page faults: 21774042
Voluntary context switches: 2223540
Involuntary context switches: 1269
File system inputs: 8738958
File system outputs: 7781542
Page size (bytes): 4096
Exit status: 0

• SIM analyses: computing information for 1 simulation, 1 condition:

User time (seconds): 199.64
System time (seconds): 7.72
Percent of CPU this job got: 84%
Elapsed (wall clock) time (h:mm:ss or m:ss): 4:04.51
Maximum resident set size (kbytes): 408764
Average resident set size (kbytes): 0
Major (requiring I/O) page faults: 43
Minor (reclaiming a frame) page faults: 1583111
Voluntary context switches: 75497
Involuntary context switches: 543
File system inputs: 3132563
File system outputs: 4279842
Page size (bytes): 4096
Exit status: 0