|
Description
|
The GSP1000 Processed Connectome is derived from data acquired by the Brain Genomics Superstruct Project (GSP), which contained 1570 subjects in total (ages 18-36). From this dataset, 1000 subjects (1:1 M/F) were chosen and processed using publicly available tools to generate a normative functional connectivity dataset. This release contains one T1w anatomical image warped to the MNI152 2mm isovolumetric space distributed with FSL and either one or two preprocessed resting state fMRI BOLD runs. This dataset was created to provide a new version of the "Yeo1000" connectome that was created 10 years ago from software that is no longer available (Yeo et al., J Neurophys 2011), but has been used by a number of laboratories and research groups. Of note, the CBIG pipeline has been slightly modified, as have the default pipeline settings, to approximate those used for the original "yeo1000" dataset. Both our modified pipeline and the configuration file as well as code to apply this pipeline to BIDS-formatted data are linked below. (2020-11-12)
We have updated the GSP1000 cohort to consist of a fully 1:1 M:F dataset. The original version consisted of 346:654 M:F participants, while the earlier unreleased "yeo1000" dataset consisted of 426:574 M:F participants. All data in both versions are usable, the associated text files delineate the entire GSP1570 cohort, and sample pairwise correlations and correlation maps are not substantially different at the group average (n=1000) level, with a very high degree of spatial correlation between "same-seed" functional connectivity maps. (2021-03-22)
File GSP1000/GSP1000_v2_16.tar was made incorrectly (only contained 1 participant's data). This has been corrected. (2021-04-17)
|
|
Notes
| ## Data Acquisition:
The original GSP data was acquired on matched Siemens 3T MAGNETOM Tim Trio MRI systems (Erlangen, Germany) using the vendor-supplied 12-channel phased-array head coil. Sequences, parameters, and instructions were unchanged throughout the collection process. However, not all subjects were acquired on the same scanner, as five different scanners were used. In addition, during the scanning period, the scanner console changed from B13 to B15 to B17. The scanner (Scanner_Bin) and console version (Console) for each imaging session are available within the CSV files included in the original data release (GSP_list_140630.csv and GSP_retest_140630.csv). The test-retest data include individuals scanned twice on the different scanners and across different console versions. The data may be useful in assessing any subtle differences. Finally, as a precaution, the original GSP authors highly recommend regressing the scanner and console from analyses.
## Data Processing:
1. GSP subjects included in this connectome were chosen based on a combined 3-way score of normalized DVARS, normalized Entropy Focused Criterion (EFC), and normalized Framewise Displacement (FD) values generated from mriqc (https://mriqc.readthedocs.io/en/stable/) consistent with the goal of minimizing the effects of motion without "scrubbing" (Power et al. 2012), as this was not standard practice for the original "yeo1000" dataset (Yeo et al., J Neurophys 2011). The 1000 GSP subjects with the best combined scores were selected for connectome inclusion. Additionally, if a subject contains 2 resting-state BOLD runs, then the 2 motion quality values are averaged to produce a single value.
2. Anatomical surfaces were created with FreeSurfer v4.5.
3. Functional preprocessing was then performed with a 'slightly' modified version of Thomas Yeo’s Computational Brain Imaging Group (CBIG) fMRI preprocessing pipeline to generate the BOLD runs used in the connectome (https://github.com/bchcohenlab/CBIG/blob/master/README.md).
-- Note: Our "config file" for the CBIG pipeline is also included in the current upload to facilitate methodological replication and the CBIG preprocessing scripts themselves can be freely accessed here: https://github.com/bchcohenlab/CBIG/tree/master/stable_projects/preprocessing/CBIG_fMRI_Preproc2016.
## License:
This data upload abides by the original GSP Open Access Data Use Terms agreement released with the initial GSP project; all data contained in this upload are deidentified, defaced, and no code (or other analysis techniques) used to process the data endangers the security of subjects' Protected Health Information or any additional confidential information that would otherwise violate Terms of the agreement.
Given the size of this dataset, you may find better performance downloading it from Harvard Dataverse using the available command line tools:
# The API token is obtainable after you log-in to Harvard dataverse from your profile page
export API_TOKEN=xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx
export SERVER_URL=https://dataverse.harvard.edu
export PERSISTENT_ID=doi:10.7910/DVN/ILXIKS
export VERSION=2.1
curl -O -J -H "X-Dataverse-key:$API_TOKEN" $SERVER_URL/api/access/dataset/:persistentId/versions/$VERSION?persistentId=$PERSISTENT_ID
|
