Skip to content

An fMRI dataset from rich naturalistic audio/video stimulation (movie "Raiders of the lost Ark"). Origin: http://datasets.datalad.org/?dir=/labs/haxby/raiders

Notifications You must be signed in to change notification settings

HaxbyLab/raiders_data

Repository files navigation

Dataset description

We scanned healthy adult subjects in an fMRI scanner while they watched a full-length feature film "Raiders of the Lost Ark".

Subjects

Current release has 11 healthy right-handed subjects (4 females; mean age: 24.6+/-3.7 years) with no history of neurological or psychiatric illness.

Design

The ~2 hour long film was presented in eight parts (corresponding to 8 runs) in two sessions with 4 runs in each session. Each movie part was of approximately the same duration (~14.5 min), with about 20 seconds of movie overlap between consecutive parts (For example, last 20 seconds of the movie at the end of the first run were shown at the beginning of the 2nd run).
An anatomical image was collected after the first session (4 parts), and subjects were take out of the scanner for a short break before they participated in the second session. A second anatomical image was collected at the end of the experiment after the movie.

Details on length could be found in a [google spreadsheet] (https://docs.google.com/spreadsheets/d/1vs37Tdr_ynA1c9skVvy4JhFJM1SqAYhX6Z1QSDuCUB4/edit?usp=sharing).

Stimuli

The movie stimulus was projected with an LCD projector onto a rear projection screen that the subject could view through a mirror. The video image subtended a visual angle of approximately 22.7° horizontally and 17° vertically. The soundtrack for the movie was played through MRConf MR-compatible headphones. There was no specific task and subjects were just asked to pay attention to the movie and enjoy.

In a version (private, non-redistributable) of the repository stimulus/task002/generate/ contains movie parts regenerated from a DVD (use script stimulus/scripts/dvd-to-mks to re-generate your own copy), and per-frame images for the first part (stimulus/task002/generate/INDIANA_JONES_RAIDERS_LOST_ARK_part_1-182x121/). Script stimulus/scripts/video2jpeg-x121 could be used to generate a sequence of images at that resolution (182x121).

fMRI protocol

Subjects were scanned in a Philips Intera Achieva 3T scanner with an 8 channel head coil at the Dartmouth Brain Imaging Center.
T1-weighted anatomical scans were acquired at the end of each session (MPRAGE, TR=9.85 s, TE=4.53 s, flip angle=8°, 256 x 256 matrix, FOV=240 mm, 160 1 mm thick sagittal slices). The voxel resolution was 0.938 mm x 0.938 mm x 1.0 mm.
Functional scans were acquired with an echo planar imaging sequence (TR=2.5 s, TE=35 ms, flip angle=90°, 80 x 80 matrix, FOV=240 mm x 240 mm) every 2.5 s with whole brain coverage (41x3 mm thick interleaved axial slices). We acquired a total of 2718 functional scans with 1350 TRs in four runs during the first session and 1368 TRs in four runs during the second session.

Dataset content overview

Data layout

Data is organized largely following openfmri.org convention. Currently it contains data only for the "task002", as described in task_key.txt being "Raiders movie. Dartmouth sample #1 (8ch coil, 8 runs)".

Top directory contains following sub-directories, many of which contains other sub-directories grouping data according to the modality (e.g. BOLD, Blood Oxygen Level Dependent, for functional MRI data) and task (e.g. task002) or even task and run (e.g. task002_run001).

Top directories:

  • aligned/ Data aligned across all subjects, e.g. aligned/BOLD/task002 contains hyper-aligned BOLD data
  • masks/ Common masks
  • scripts/ Scripts used for various stages in pre-processing or dataset preparation
  • stimulus/ Data (non-redistributable) and scripts/ for stimuli delivery
  • sub*/ Data for each subject

Functional data

Raw BOLD functional MRI

Filename examples for subject 001 and run 001

  • sub001/BOLD/task002_run001/bold.{PAR,REC} Original BOLD data as acquired from the scanner
  • sub001/BOLD/task002_run001/bold.nii.gz Original BOLD data in NIfTI format, as converted from {PAR,REC} using dcm2nii tool

Pre-processed BOLD functional MRI

Currently we provide a sample of data as it was pre-processed by Swaroop Guntupalli, hence suffix _sw1 and used in the Hyperalignment paper, e.g.

  • sub001/BOLD/task002_run001/bold_sw1_stc_reg_despike_dt_bp.nii.gz Pre-processed functional data for subject 1, run 1, in original subject space (so "voxels" do not "align" across subjects). Abbreviations used:

    • _stc -- slice timing correction
    • _reg -- registered (motion corrected)
    • _despite -- de-spiked (high-freq artifacts removed)
    • _bp -- band-pass filtered

    Basic detail on the file:

    nib-ls sub001/BOLD/task002_run001/bold_sw1_stc_reg_despike_dt_bp.nii.gz ... float32 [ 80, 80, 41, 336] 3.00x3.00x3.00x2.50 #exts: 1

  • sub001/BOLD/task002/bold_sw1_stc_reg_despike_dt_bp_mni.nii.gz Pre-processed data for subject 1, all runs concatenated, in MNI space, after linear anatomical alignment

      > nib-ls sub001/BOLD/task002/bold_sw1_stc_reg_despike_dt_bp_mni.nii.gz
      ... float32 [ 61,  73,  61, 2718] 3.00x3.00x3.00x2.50  #exts: 1
    
  • aligned/BOLD/task002/bold_surfslhyper_sw1_radius20_lr32_hr128_full_featsel0.7.hdf5.gz Hyperalignment projections estimated per each subject. It is an HDF5 dataset, relatively large!, which you can load using mvpa2 functionality in Python:

from mvpa2.base.hdf5 import h5load
bold_surfslhyper = h5load('aligned/BOLD/task002/bold_surfslhyper_sw1_radius20_lr32_hr128_full_featsel0.7.hdf5.gz')
# TODO

git-annex 101

This repository is commonly distributed as a git-annex repository. To initially obtain it, or to fetch additional changes you would need to

git clone for-now-secret-host:/secret/location.git raiders
cd raiders

To obtain content for any "large" file or entire directories

git annex get WHATEVER-FILES-YOU-NEED

To fetch new changes later:

git pull # or git fetch; git merge origin/master
git annex get WHATEVER-FILES-YOU-NEED

Hint: use git annex get -J 4 to establish 4 parallel streams to fetch the data -- should speed up transfer considerably.

To add new content (possibly in a new branch, see git branch):

  • add only small files (text, scripts, etc) directly to git with

      git add FILES
    
  • add large files (images, data) to git-annex with

      git annex add FILES
    
  • commit added content into git (large files added to annex will not be added directly but placed under .git/annex/objects)

      git commit -m "Description of the changes"
    

To submit your changes to the original server

  • push git repository

      git push
    
  • send the new annex content to the server

      git annex copy --to=origin . # or specific files/directories
    

Recommended analysis environment

We would recommend to use some Debian/Ubuntu Linux environment (native, docker (neurodebian), or VM such as NeuroDebian VM) with NeuroDebian repository enabled.

Then you could install a large collection of relevant tools which would facilitate access and processing of the data

 apt-get install git-annex-standalone python-mvpa2 python-nibabel fslview afni

and to generate/manipulate stimuli

 apt-get install handbrake-cli libav-tools

About

An fMRI dataset from rich naturalistic audio/video stimulation (movie "Raiders of the lost Ark"). Origin: http://datasets.datalad.org/?dir=/labs/haxby/raiders

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published