Multi-domain task battery data set (King et al. 2019, Nat. Neurosci)

This code repository contains preprocessing and basic code to load and access preprocessed code from the MDTB data set

For usage on Yale's CRC Grace cluster (http://cluster.ycrc.yale.edu/grace/)

E-mail for questions: taku [dot] ito1 [at] gmail [dot] com

Last updated: 9/27/21

Basic data directory structure:

Note that files will be read-only access, so it's recommended to copy template code to one's own directory on Grace and read-in data from your local project repo

Base directory: /gpfs/loomis/pi/n3/Studies/MurrayLab/taku/mdtb_data/

Raw BIDS formatted data: /gpfs/loomis/pi/n3/Studies/MurrayLab/taku/mdtb_data/bids/

QuNex preprocessed data: /gpfs/loomis/pi/n3/Studies/MurrayLab/taku/mdtb_data/qunex_mdtb/sessions

Fully preprocessed data: Includes QuNex preprocessed + post-processed nuisance regression (including removal of white matter, ventricle and motion time series). This includes processed data for both task and rest data at parcellated and vertex-wise time series: /gpfs/loomis/pi/n3/Studies/MurrayLab/taku/mdtb_data/derivatives/postprocessing/

Local code repo (which is synced to this GitHub repo): /gpfs/loomis/pi/n3/Studies/MurrayLab/taku/mdtb_data/docs/

Demo code to load in processed resting-state and task-state fMRI

Note that processed task-state fMRI data has been processed using finite impulse response modeling (FIR) for task-state correlation/functional connectivity analyses, following procedures in Cole et al. (2019) NeuroImage http://www.sciencedirect.com/science/article/pii/S1053811918322043

Task conditions that are modeled in FIR are at the block-level, and so there are only 26 conditions for each task-specific time series.

Processed data is in parcellated time series (vertex-wise data to come) using the Glasser et al. (2016) parcellation scheme (http://www.nature.com/doifinder/10.1038/nature18933)

Demo code (Grace): /gpfs/loomis/pi/n3/Studies/MurrayLab/taku/mdtb_data/docs/scripts/rest_task_fMRI_demo.ipynb

Demo code (This repo): scripts/rest_task_fMRI_demo.ipynb (https://github.com/ito-takuya/mdtb_data/blob/main/scripts/rest_task_fMRI_demo.ipynb)

Useful python dependencies (python version 3.8)

Helpful to start with anaconda python environment

Code requires nibabel, nipy, h5py

For advanced users (e.g., fMRI users familiar with preprocessing)

For those who wish to preprocess and run GLMs themselves, the following code can be used as templates. A good knowledge of shell-scripting and python is required, along with best practices for preprocessing.

See QuNex documentation for preprocessing, and recommended post-processing/nuisance regression references:

Ciric, R., Wolf, D.H., Power, J.D., Roalf, D.R., Baum, G.L., Ruparel, K., Shinohara, R.T., Elliott, M.A., Eickhoff, S.B., Davatzikos, C., Gur, R.C., Gur, R.E., Bassett, D.S., Satterthwaite, T.D., 2017. Benchmarking of participant-level confound regression strategies for the control of motion artifact in studies of functional connectivity. NeuroImage 154, 174–187. https://doi.org/10.1016/j.neuroimage.2017.03.020

Cole, M.W., Ito, T., Schultz, D., Mill, R., Chen, R., Cocuzza, C., 2019. Task activations produce spurious but systematic inflation of task functional connectivity estimates. NeuroImage 189, 1–18. https://doi.org/10.1016/j.neuroimage.2018.12.054

Preprocessing from BIDS was implemented using QuNex.

Qunex shell script: scripts/preproc_qunex_turnkey_v2.sh

Postprocessing (i.e., nuisance regression for resting-state data and task GLMs) were implemented with custom python code.

Resting-state nuisance regression: scripts/glm_scripts/postproc_rest.py

Task-state FIR regression (for FC and timescale analyses): scripts/glm_scripts/postproc_taskFIR.py

Task-state GLM activation estimation (uses a beta series type model, see Rissman et al. (2004), NeuroImage 10.1016/j.neuroimage.2004.06.035: scripts/glm_scripts/postproc_taskbetaseries.py

Generic post-processing tools: scripts/glm_scripts/postproc_tools.py