3  Data management

In neuroimaging

General

(Juavinett 2022)

State in 2022

Image credit: (Juavinett 2022)

BIDS

1. Psychopy

BIDS plugin for PsychoPy

Should be installed and enabled to save events (conditions and button presses) in a comprehensive and standardized manner.

This will also make our data analysis easier

https://psychopy-bids.readthedocs.io/en/stable/

Installation

Open psychopy and press “Get more” on the right side. Alternatively you can open “plugin/pachages manager” via “Tools” menue. Psychopy-plugin should be listed among the available plugins.

On the university computers, bids plugin is available through software center.

Experiment adjustment1

Thanks to

Lukas currently took over the BIDS plugin development.

Properties

  1. Experiment name will correspond to the task-[taskname] properety in the bids dataset. It is therefore useful to have a short name without capital letters. This did not work on my mac, but maybe this is an OS issue?
  2. Participant: switched from random code generator to 001. Since we keep the naming consistent with the MRI dataset, and have multiple runs, it is not useful to generate a new code on every run repetition
  3. Session: changed from 001 to 1. This is just our naming convention. Nobody has hundreds of sessions, so we do not expect a number with more than two digits.

Renaming?

File

Project folder

For the task label to work properly in the BIDS plugin, I also tried renaming the experiment file and the folder, but this did not help. Gambling is still spelled out with a capital “G” no mater what I do. Let’s stay tuned for the psychopy-bids updates.

Research question

What information do we actually need for analysis?

We should keep the research question in mind.

H1 is formulated as a one-way ANOVA with three pairwise post-hoc tests

H2 is formulated as 2x2 two-way ANOVA (at least we will simplify it to that)

Unique conditions

  • small win

  • small loss

  • large win

  • large loss

  • pass

trials.xlsx

For every trial, we should be able to identify whether it is win, loss or pass.

For every win or loss trial, we need to know if it is a win/loss with high stakes or with low stakes.

Note: right now we are not taking the win probability into account, but we know that it is always low for large stakes, and large for low stakes, so it is redundant.

Side note:

Ideal paradigm

  • Each trial type appears the same number of times

  • Carryover counterbalancing - each condition is equally likely to be preceded by each other condition (Brooks 2012)

Events of interest in a trial

gamble_trial or gamble_outcome?

Do we need key presses?

Adding bids events

Scanner trigger

I am not sure if it is needed after all, but I added a bids component to a routine that waits 10 seconds after the scanner trigger and logged its onset.

Linking to a routine # 1

In the BIDS Event Type field you can choose between TaskEvent and BehEvent. For now the difference does not matter much, let’s choose task event.

Linking to a routine #2

Right now time 0 in our experiment is the start of the welcome screen. We want time 0 to be the scanner trigger. Will take care of it later.

Adding bids events of interest

New variable outcome_trial

We need one variable to take 3 possible outcomes: win, loss or pass. Therefore, I adjusted the code, adding a variable “outcome_trial” for this purpose.

points_text

We are creating a bids event and linking it to points_text within the trial (because this routine appears on every loop interation).

Time stamp

Time 0 - scanning onset

Now we want to adjust the onset time to the scanner trigger arrival. This is done by ticking “Manually set values” and putting the difference between the onset of the current routine and the start of the blank period.

It would have been better and more accurate to subtract the actually key press time, but I ran out of time.

Other events

Here I added two other bids events that may or may not be useful, just in case: Trial onset and key press, including reaction times.

BIDS export

It is important to add the bidsExport routine at the end of the experiment.

  1. For fMRI, the data type should be “func”
  2. Choose. your favourite license
  3. Tick the box “Add Run Nubmers” - this should attach run numbers to the file names in bids format for every new run. I did not get it to work properly on mac (it was always run 1, overwriting the previous file). But supposedly it works on windows and linux.

Complete BIDS dataset

Example

This is how a complete dataset for fmri analysis should look like. You should have functional _bold.nii.gz files and _events.tsv files in one func/ folder.

2. DICOM to BIDS

Neurodesk

MRI data formats

  • DICOM - medical image format

    • .dcm .ima or nothing at all

  • NIFTI - neuroimaging format we work with

    • .nii or .nii.gz

bidscoin

Shell commands

#!/bin/bash

# you can copypaste the following commands into the terminal window that is opened by BIDSCOINER

# define directories
sourceFolder=/home/jovyan/completion2/dcm
sortedFolder=/home/jovyan/completion2/dcm_sorted

# check the output of the source folder
ls -l $sourceFolder

# make the target directory with subdirectories
mkdir -p $sortedFolder

# change to the script folder
cd /home/jovyan/dicomsort/

# run the script
./DICOMsort.sh $sourceFolder $sortedFolder

# check the output of the target folder
ls -l $sortedFolder

# run bidsmapper
bidsmapper $sortedFolder /home/jovyan/completion2/bids
# note: in the intendedFor of the fieldmap, add func/sub-001_ses-1_xxx_bold.*

# once you are happy with the mapping, run bidscoiner
bidscoiner $sortedFolder /home/jovyan/completion2/bids

bismapper

Currently, to discover the data we need to delete the content of the “subprefix” field and then save the bids map in the default location. After this, run the same bidsmapper command again.

Subject and session

Now the bidsmapper discovered everything, but the files are not named correctly.

Edit subject and session first.

Now we need to rename the files by clicking “Edit” next to every file.

Editing file names: anatomical

anatmprage -> mprage in the “acq” filed

Editing file names: functional

Edit the task name and acq field

Do this for every functional file

In the intendedFor of the fieldmap, add “func/sub-999_ses-1_task-completion_acq–noise_bold.*” with a wildcard. This will select all respective functional runs

Once you are happy, run bidscoiner to do the actual conversion

View files

MRICron

Now you can view the nifti files with any suitable software

3. Quality assurance

MRIQC

MRIQC (Esteban et al. 2017)is a tool used for quality assurance of bids-valid datasets. It is a bids app, meaning that it works on bids-conform data.

Running MRIQC

Should take around 10 minutes

Command line

# define relevant paths
bids_folder=/home/jovyan/gambling/bids/
mriqc_folder=/home/jovyan/gambling/bids/derivatives/mriqc/

# create mriqc directory
mkdir -p $mriqc_folder

# run mriqc
mriqc $bids_folder $mriqc_folder participant --participant-label 001

Output

You will see that there are html outputs for anatomical and functional scans. They can be viewed in a browser, without actually loading the images.

Anatomical

Noise

Actual image

This is useful to quickly identify problems with the data. A rating widget is also available, which is useful for large datasets.

Functional

For functional data, in addition to spatial noise (middle) and average image over time (right), there is a section showing temporal noise over time. Note how the values are highest in the eyes (due to eye movements) and in the center of the brain, due to blood pulsation.

Carpet plots for functional data

Carpet plot shows signal fluctuations over time (x-axis) for each voxel (y-axis). Note how there are some patterns initially and at the end. There should be no clear pattern during the experiment, this speaks for good signal quality.

References

Brooks, Joseph L. 2012. “Counterbalancing for Serial Order Carryover Effects in Experimental Condition Orders.” Psychological Methods 17 (4): 600–614. https://doi.org/10.1037/a0029310.
Esteban, Oscar, Daniel Birman, Marie Schaer, Oluwasanmi O. Koyejo, Russell A. Poldrack, and Krzysztof J. Gorgolewski. 2017. “MRIQC: Advancing the Automatic Prediction of Image Quality in MRI from Unseen Sites.” Edited by Boris C Bernhardt. PLOS ONE 12 (9): e0184661. https://doi.org/10.1371/journal.pone.0184661.
Juavinett, Ashley L. 2022. “The Next Generation of Neuroscientists Needs to Learn How to Code, and We Need New Ways to Teach Them.” Neuron 110 (4): 576–78. https://doi.org/10.1016/j.neuron.2021.12.001.

  1. valid for PsychoPy v2024.2.4↩︎