Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/bids-apps/CPAC
BIDS Application for the Configurable Pipeline for the Analysis of Connectomes (C-PAC)
https://github.com/bids-apps/CPAC
bids bidsapp mri preprocessing
Last synced: about 1 month ago
JSON representation
BIDS Application for the Configurable Pipeline for the Analysis of Connectomes (C-PAC)
- Host: GitHub
- URL: https://github.com/bids-apps/CPAC
- Owner: bids-apps
- License: apache-2.0
- Created: 2016-08-02T16:20:58.000Z (over 8 years ago)
- Default Branch: master
- Last Pushed: 2024-04-30T09:46:26.000Z (8 months ago)
- Last Synced: 2024-08-03T13:14:20.336Z (4 months ago)
- Topics: bids, bidsapp, mri, preprocessing
- Language: Python
- Homepage:
- Size: 65.4 MB
- Stars: 14
- Watchers: 6
- Forks: 18
- Open Issues: 11
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
- awesome-bids - CPAC - PAC) (BIDS Apps / others)
README
# C-PAC BIDS Application
## Documentation
Extensive information can be found in the [C-PAC User Guide](http://fcp-indi.github.io/docs/user/index.html).
## Description
The Configurable Pipeline for the Analysis of Connectomes [C-PAC](http://fcp-indi.github.io) is a software for
performing high-throughput preprocessing and analysis of functional connectomes data using high-performance computers.
C-PAC is implemented in Python using the Nipype pipelining [[1](#nipype)] library to efficiently combine tools from AFNI
[[2](#afni)], ANTS [[3](#ants)], and FSL [[4](#fsl)] to achieve high quality and robust automated processing. This
docker container, when built, is an application for performing participant level analyses. Future releases will include
group-level analyses, when there is a BIDS standard for handling derivatives and group models.
### A note about versioning
C-PAC BIDS Apps version tags are composed of the C-PAC version followed by an underscore and then the version of the
container. The container version restarts for every new C-PAC version and is a single integer that reflects the
modification number of the build. For example ```v1.0.1a_5``` corresponds to the 5th build of the container for C-PAC
version v1.0.1a.
### Usage notes
1. You can either perform a custom processing using a YAML configuration file or use the default processing pipeline. A
GUI can be invoked to assist in pipeline custimization by specifying `GUI` command line arguement (this currently only
works for Singularity containers).
2. The default behavior is to read in data that is organized in the BIDS format. This includes data that is in Amazon
AWS S3 by using the format `s3:///` for the `bids_dir` command line argument. Outputs can be
written to S3 using the same format for the output_dir. Credentials for accessing these buckets can be specified on the
command line (using `--aws_input_creds` or `--aws_output_creds`).
3. Non-BIDS organized data can processed using a C-PAC data configuration yaml file. This file can be generated using
the C-PAC GUI (start the app with the `GUI` argument, also see instructions [below](#GUI)) or can be created using other means, please refer to [CPAC
documentation](http://fcp-indi.github.io/docs/user/subject_list_config.html) for more information.
4. When the app is run, a data configuration file is written to the working directory. This file can be passed into
subsequent runs, which avoids the overhead of re-parsing the BIDS input directory on each run (i.e. for cluster or
cloud runs). These files can be generated without executing the C-PAC pipeline using the `test_run` command line
argument.
5. The `participant_label` and `participant_ndx` arguments allow the user to specify which of the many datasets should
be processed, this are useful when parallelizing the run of multiple participants.
### Default configuration
The default processing pipeline performs fMRI processing using four strategies, with and without global signal
regression, with and without bandpass filtering.
Anatomical processing begins with conforming the data to RPI orientation and removing orientation header information
that will interfere with further processing. A non-linear transform between skull-on images and a 2mm MNI brain-only
template are calculated using ANTs [[3](#ants)]. Images are them skull-stripped using AFNI's `3dSkullStrip` [[5](#bet)]
and subsequently segmented into WM, GM, and CSF using FSL’s `fast` tool [[6](#fast)]. The resulting WM mask was
multiplied by a WM prior map that was transformed into individual space using the inverse of the linear transforms
previously calculated during the ANTs procedure. A CSF mask was multiplied by a ventricle map derived from the
Harvard-Oxford atlas distributed with FSL [[4](#fsl)]. Skull-stripped images and grey matter tissue maps are written
into MNI space at 2mm resolution.
Functional preprocessing begins with resampling the data to RPI orientation, and slice timing correction. Next, motion
correction is performed using a two-stage approach in which the images are first coregistered to the mean fMRI and then
a new mean is calculated and used as the target for a second coregistration (AFNI `3dvolreg` [[2](#afni)]). A 7 degree of
freedom linear transform between the mean fMRI and the structural image is calculated using FSL’s implementation of
boundary-based registration [[7](#bbr)]. Nuisance variable regression (NVR) is performed on motion corrected data using
a 2nd order polynomial, a 24-regressor model of motion [[8](#friston24)], 5 nuisance signals, identified via principal
components analysis of signals obtained from white matter (CompCor, [[9](#compcor)]), and mean CSF signal. WM and CSF
signals were extracted using the previously described masks after transforming the fMRI data to match them in 2mm space
using the inverse of the linear fMRI-sMRI transform. The NVR procedure is performed twice, with and without the
inclusion of the global signal as a nuisance regressor. The residuals of the NVR procedure are processed with and
without bandpass filtering (0.001Hz < f < 0.1Hz), written into MNI space at 3mm resolution and subsequently smoothed
using a 6mm FWHM kernel.
Several different individual level analysis are performed on the fMRI data including:
- **Amplitude of low frequency fluctuations (alff) [[10](#alff)]**: the variance of each voxel is calculated after
bandpass filtering in original space and subsequently written into MNI space at 2mm resolution and spatially smoothed
using a 6mm FWHM kernel.
- **Fractional amplitude of low frequency fluctuations (falff) [[11](#falff)]**: Similar to alff except that the
variance of the bandpassed signal is divided by the total variance (variance of non-bandpassed signal.
- **Regional homogeniety (ReHo) [[12](#reho)]**: a simultaneous Kendalls correlation is calculated between each
voxel's time course and the time courses of the 27 voxels that are face, edge, and corner touching the voxel. ReHo is
calculated in original space and subsequently written into MNI space at 2mm resolution and spatially smoothed using
a 6mm FWHM kernel.
- **Voxel mirrored homotopic connectivity (VMHC) [[13](#vmhc)]**: an non-linear transform is calculated between the
skull-on anatomical data and a symmetric brain template in 2mm space. Using this transform, processed fMRI data are
written in to symmetric MNI space at 2mm and the correlation between each voxel and its analog in the contralateral
hemisphere is calculated. The Fisher transform is applied to the resulting values, which are then spatially smoothed
using a 6mm FWHM kernel.
- **Weighted and binarized degree centrality (DC) [[14](#degree)]**: fMRI data is written into MNI space at 2mm
resolution and spatially smoothed using a 6mm FWHM kernel. The voxel x voxel similarity matrix is calculated by the
correlation between every pair of voxel time courses and then thresholded so that only the top 5% of correlations
remain. For each voxel, binarized DC is the number of connections that remain for the voxel after thresholding and
weighted DC is the average correlation coefficient across the remaining connections.
- **Eigenvector centrality (EC) [[15](#eigen)]**: fMRI data is written into MNI space at 2mm resolution and spatially
smoothed using a 6mm FWHM kernel. The voxel x voxel similarity matrix is calculated by the correlation between every
pair of voxel time courses and then thresholded so that only the top 5% of correlations remain. Weighted EC is
calculated from the eigenvector corresponding to the largest eigenvalue from an eigenvector decomposition of the
resulting similarity. Binarized EC, is the first eigenvector of the similarity matrix after setting the non-zero values
in the resulting matrix are set to 1.
- **Local functional connectivity density (lFCD) [[16](#lfcd)]**: fMRI data is written into MNI space at 2mm resolution
and spatially smoothed using a 6mm FWHM kernel. For each voxel, lFCD corresponds to the number of contiguous voxels that
are correlated with the voxel above 0.6 (r>0.6). This is similar to degree centrality, except only voxels that it only
includes the voxels that are directly connected to the seed voxel.
- **10 intrinsic connectivity networks (ICNs) from dual regression [[17](#dr)]**: a template including 10 ICNs from a
meta-analysis of resting state and task fMRI data [[18](#icn10)] is spatially regressed against the processed fMRI data
in MNI space. The resulting time courses are entered into a multiple regression with the voxel data in original space to
calculate individual representations of the 10 ICNs. The resulting networks are written into MNI space at 2mm and then
spatially smoothed using a 6mm FWHM kernel.
- **Seed correlation analysis (SCA)**: preprocessed fMRI data is to match template that includes 160 regions of interest
defined from a meta-analysis of different task results [[19](#do160)]. A time series is calculated for each region from
the mean of all intra-ROI voxel time series. A seperate functional connectivity map is calculated per ROI by correlating
its time course with the time courses of every other voxel in the brain. Resulting values are Fisher transformed,
written into MNI space at 2mm resolution, and then spatiall smoothed using a 6mm FWHM kernel.
- **Time series extraction**: similar the procedure used for time series analysis, the preprocessed functional data is
written into MNI space at 2mm and then time series for the various atlases are extracted by averaging within region
voxel time courses. This procedure was used to generate summary time series for the automated anatomic labelling atlas
[[20](#aal)], Eickhoff-Zilles atlas [[21](#ez)], Harvard-Oxford atlas [[22](#ho)], Talaraich and Tournoux atlas
[[23](#tt)], 200 and 400 regions from the spatially constrained clustering voxel timeseries [[24](#cc)], and 160 ROIs
from a meta-analysis of task results [[19](#do160)]. Time series for 10 ICNs were extracted using spatial regression.
## Usage
This App has the following command line arguments:
usage: run.py [-h] [--pipeline_file PIPELINE_FILE]
[--data_config_file DATA_CONFIG_FILE]
[--aws_input_creds AWS_INPUT_CREDS]
[--aws_output_creds AWS_OUTPUT_CREDS] [--n_cpus N_CPUS]
[--mem_mb MEM_MB] [--mem_gb MEM_GB] [--save_working_dir]
[--participant_label PARTICIPANT_LABEL [PARTICIPANT_LABEL ...]]
[--participant_ndx PARTICIPANT_NDX]
bids_dir output_dir {participant,group,test_config,GUI}
C-PAC Pipeline Runner
positional arguments:
bids_dir The directory with the input dataset formatted
according to the BIDS standard. Use the format
s3://bucket/path/to/bidsdir to read data directly from
an S3 bucket. This may require AWS S3 credentials
specificied via the --aws_input_creds option.
output_dir The directory where the output files should be stored.
If you are running group level analysis this folder
should be prepopulated with the results of the
participant level analysis. Us the format
s3://bucket/path/to/bidsdir to write data directly to
an S3 bucket. This may require AWS S3 credentials
specificied via the --aws_output_creds option.
{participant,group,test_config,GUI}
Level of the analysis that will be performed. Multiple
participant level analyses can be run independently
(in parallel) using the same output_dir. GUI will open
the CPAC gui (currently only works with singularity)
and test_config will run through the entire
configuration process but will not execute the
pipeline.
optional arguments:
-h, --help show this help message and exit
--pipeline_file PIPELINE_FILE
Name for the pipeline configuration file to use
--data_config_file DATA_CONFIG_FILE
Yaml file containing the location of the data that is
to be processed. Can be generated from the CPAC gui.
This file is not necessary if the data in bids_dir is
organized according to the BIDS format. This enables
support for legacy data organization and cloud based
storage. A bids_dir must still be specified when using
this option, but its value will be ignored.
--aws_input_creds AWS_INPUT_CREDS
Credentials for reading from S3. If not provided and
s3 paths are specified in the data config we will try
to access the bucket anonymously
--aws_output_creds AWS_OUTPUT_CREDS
Credentials for writing to S3. If not provided and s3
paths are specified in the output directory we will
try to access the bucket anonymously
--n_cpus N_CPUS Number of execution resources available for the
pipeline
--mem_mb MEM_MB Amount of RAM available to the pipeline in megabytes.
Included for compatibility with BIDS-Apps standard,
but mem_gb is preferred
--mem_gb MEM_GB Amount of RAM available to the pipeline in gigabytes.
if this is specified along with mem_mb, this flag will
take precedence.
--save_working_dir Save the contents of the working directory.
--participant_label PARTICIPANT_LABEL [PARTICIPANT_LABEL ...]
The label of the participant that should be analyzed.
The label corresponds to sub- from
the BIDS spec (so it does not include "sub-"). If this
parameter is not provided all subjects should be
analyzed. Multiple participants can be specified with
a space separated list. To work correctly this should
come at the end of the command line
--participant_ndx PARTICIPANT_NDX
The index of the participant that should be analyzed.
This corresponds to the index of the participant in
the subject list file. This was added to make it
easier to accomodate SGE array jobs. Only a single
participant will be analyzed. Can be used with
participant label, in which case it is the index into
the list that follows the particpant_label flag.
### To run it in participant level mode (for one participant):
docker run -i --rm \
-v /tmp:/scratch \
-v /Users/filo/data/ds005:/bids_dataset \
-v /Users/filo/outputs:/outputs \
bids/cpac \
/bids_dataset /outputs participant --participant_label 01
### Running the GUI (requires mapping your X socket)
#### Running docker container on Linux
1. Start the docker container, mapping the X socket (change /Users/filo to a local directory on your computer)
```bash
docker run -i --rm \
--privileged \
-e DISPLAY=$DISPLAY \
-v /tmp/.X11-unix:/tmp/.X11-unix \
-v /tmp:/scratch \
-v /Users/filo/data/ds005:/bids_dataset \
-v /Users/filo/outputs:/outputs \
bids/cpac \
/bids_dataset /outputs GUI
```
#### Running docker container on Mac OSX
1. [Install XQuartz](https://www.xquartz.org/)
2. Start XQuartz (from terminal)
```bash
open -a XQuartz
```
3. Enable XQuartz connections from network clients
XQuartz -> preferences -> security -> "Allow connections from network clients"
4. Get your ip address (e.g., might have to change eth0 to match the name of your network interface.)
```bash
ip=$(ifconfig en0 | grep inet | awk '$1=="inet" {print $2}')
```
5. Tell xhost to accept connections from the localhost
```bash
xhost + ${ip}
```
6. Start the docker container, mapping the X socket (change /Users/filo to a local directory on your computer)
```bash
docker run -i --rm \
--privileged \
-e DISPLAY=$ip:0 \
-v /tmp/.X11-unix:/tmp/.X11-unix \
-v /tmp:/scratch \
-v /Users/filo/data/ds005:/bids_dataset \
-v /Users/filo/outputs:/outputs \
bids/cpac \
/bids_dataset /outputs GUI
```
#### Running singularity container on Linux
1. Start the docker container (it just works!, provided you change /Users/filo to a local directory on your computer)
```bash
singularity run \
-B /home/ubuntu:/mnt \
-B /mnt:/scratch \
-B /Users/filo/data/ds005:/bids_dataset \
-B /Users/filo/outputs:/outputs \
/home/ubuntu/workspace/container_build/singularity_images/cpac_latest.img \
/bids_dataset \
/outputs\
GUI
```
### To convert the Docker container to a [Singularity](http://singularity.lbl.gov/) container :
docker run --privileged -ti --rm \
-v /var/run/docker.sock:/var/run/docker.sock \
-v /home/srycajal/singularity_images:/output \
filo/docker2singularity \
bids/cpac
### Example submit script for running as a Singularity container on sun grid engine:
#! /bin/bash
## SGE batch file - bgsp
#$ -S /bin/bash
## bgsp is the jobname and can be changed
#$ -N bgsp
## execute the job using the mpi_smp parallel enviroment and 8 cores per job
#$ -pe mpi_smp 8
## create an array of 1112 jobs
#$ -t 1-1112
#$ -V
## change the following working directory to a persistent directory that is
## available on all nodes, this is were messages printed by the app (stdout
## and stderr) will be stored
#$ -wd /home/ubuntu/workspace/cluster_files
sudo chmod 777 /mnt
mkdir -p /mnt/log/reports
sge_ndx=$(( SGE_TASK_ID - 1 ))
# random sleep so that jobs dont start at _exactly_ the same time
sleep $(( $SGE_TASK_ID % 10 ))
singularity run -B /home/ubuntu:/mnt -B /mnt:/scratch \
/home/ubuntu/workspace/container_build/singularity_images/cpac_latest.img \
--n_cpus 8 --mem 12 \
--aws_input_creds /mnt/workspace/cluster_files/s3-keys.csv \
--aws_output_creds /mnt/workspace/cluster_files/s3-keys.csv \
--data_config_file /mnt/workspace/cluster_files/bgsp_data_config.yml \
s3://fcp-indi/data/Projects/BrainGenomicsSuperstructProject/orig_bids/ \
s3://fcp-indi/data/Projects/BrainGenomicsSuperstructProject/cpac_out/ \
participant --participant_ndx ${sge_ndx}
#### Notes:
1. With the exception of your home directory, which is mounted from the local filesystem, the filesystem in Singularity
containers is read-only. Files can be easily transferred in and out of the container by mapping local directories to
directories inside the container using the `-B from:to` command line argument, where the `from` dir is mapped to `to`.
When using mapped directories, remember that the paths specified on the command line are in relation to the directory
inside the container (e.g. the `to` directory).
2. Unless the `--save_working_dir` flag is set, the C-PAC app will use the `/scratch` directory for intermediary files.
Since this directory is write protected, a directory from the local filesystem must be mapped to `/scratch` for the
pipeline to run successfully. This directory should be large enough to hold all of the intermediary files for the
datasets that are processed in parallel, as a rule of thumb we suggest 3 GB per dataset. Unless the `--save_working_dir`
flag is set, the working directory will be deleted when the pipeline has completed.
3. Use the `--save_working_dir` flag to retain all intermediary files, which can be useful for debugging. In this case,
the intermediary files will be saved in the `working_dir` subdirectory of the user specified `output` directory. This
will require about 3GB per dataset, but may require more for multiple or very long fMRI scans.
## Reporting errors and getting help
Please report errors on the [C-PAC github page issue tracker](https://github.com/FCP-INDI/C-PAC/issues). Please use the
[C-PAC google group](https://groups.google.com/forum/#!forum/cpax_forum) for help using C-PAC and this application.
## Acknowledgements
We currently have a publication in preparation, in the meantime please cite our poster from INCF:
Craddock C, Sikka S, Cheung B, Khanuja R, Ghosh SS, Yan C, Li Q, Lurie D, Vogelstein J, Burns R, Colcombe S,
Mennes M, Kelly C, Di Martino A, Castellanos FX and Milham M (2013). Towards Automated Analysis of Connectomes:
The Configurable Pipeline for the Analysis of Connectomes (C-PAC). Front. Neuroinform. Conference Abstract:
Neuroinformatics 2013. doi:10.3389/conf.fninf.2013.09.00042
@ARTICLE{cpac2013,
AUTHOR={Craddock, Cameron and Sikka, Sharad and Cheung, Brian and Khanuja, Ranjeet and Ghosh, Satrajit S
and Yan, Chaogan and Li, Qingyang and Lurie, Daniel and Vogelstein, Joshua and Burns, Randal and
Colcombe, Stanley and Mennes, Maarten and Kelly, Clare and Di Martino, Adriana and Castellanos,
Francisco Xavier and Milham, Michael},
TITLE={Towards Automated Analysis of Connectomes: The Configurable Pipeline for the Analysis of Connectomes (C-PAC)},
JOURNAL={Frontiers in Neuroinformatics},
YEAR={2013},
NUMBER={42},
URL={http://www.frontiersin.org/neuroinformatics/10.3389/conf.fninf.2013.09.00042/full},
DOI={10.3389/conf.fninf.2013.09.00042},
ISSN={1662-5196}
}
## References
1. Gorgolewski, K., Burns, C.D., Madison, C., Clark, D., Halchenko, Y.O., Waskom, M.L., Ghosh, S.S.: Nipype: A flexible, lightweight and extensible neuroimaging data processing framework in python. Front. Neuroinform. 5
(2011). doi:10.3389/fninf.2011.00013
2. Cox, R.W., Jesmanowicz, A.: Real-time 3d image registration for functional mri. Magn Reson Med 42(6),
1014–8 (1999)
3. Avants, B., Epstein, C., Grossman, M., Gee, J.: Symmetric diffeomorphic image registration with
cross-correlation: Evaluating automated labeling of elderly and neurodegenerative brain. Medical Image Analysis
12(1), 26–41 (2008). doi:10.1016/j.media.2007.06.004
4. Smith, S.M., Jenkinson, M., Woolrich, M.W., Beckmann, C.F., Behrens, T.E.J., Johansen-Berg, H., Bannister,
P.R., Luca, M.D., Drobnjak, I., Flitney, D.E., Niazy, R.K., Saunders, J., Vickers, J., Zhang, Y., Stefano, N.D.,
Brady, J.M., Matthews, P.M.: Advances in functional and structural mr image analysis and implementation as
fsl. NeuroImage 23, 208–219 (2004). doi:10.1016/j.neuroimage.2004.07.051
5. Smith, S.M.: Fast robust automated brain extraction. Human Brain Mapping 17(3), 143–155 (2002). doi:10.1002/hbm.10062
6. Zhang, Y., Brady, M., Smith, S.: Segmentation of brain mr images through a hidden markov random field
model and the expectation-maximization algorithm. IEEE Transactions on Medical Imaging 20(1), 45–57
(2001). doi:10.1109/42.906424
7. Greve, D.N., Fischl, B.: Accurate and robust brain image alignment using boundary-based registration.
NeuroImage 48(1), 63–72 (2009). doi:10.1016/j.neuroimage.2009.06.060
8. Friston, K.J., Williams, S., Howard, R., Frackowiak, R.S., Turner, R.: Movement-related effects in fmri time-series. Magn Reson Med 35(3), 346–55 (1996)
9. Behzadi, Y., Restom, K., Liau, J., Liu, T.T.: A component based noise correction method (compcor) for bold and perfusion based fmri. NeuroImage 37(1), 90–101 (2007). doi:10.1016/j.neuroimage.2007.04.042
10. Zang, Y.-F., He, Y., Zhu, C.-Z., Cao, Q.-J., Sui, M.-Q., Liang, M., Tian, L.-X., et al. (2007). Altered baseline brain activity in children with ADHD revealed by resting-state functional MRI. Brain & development, 29(2), 83–91.
11. Zou, Q.-H., Zhu, C.-Z., Yang, Y., Zuo, X.-N., Long, X.-Y., Cao, Q.-J., Wang, Y.-F., et al. (2008). An improved approach to detection of amplitude of low-frequency fluctuation (ALFF) for resting-state fMRI: Fractional ALFF. Journal of neuroscience methods, 172(1), 137–141.
12. Zang, Y., Jiang, T., Lu, Y., He, Y., Tian, L., 2004. Regional homogeneity approach to fMRI data analysis. Neuroimage 22, 394-400.
13. Stark, D. E., Margulies, D. S., Shehzad, Z. E., Reiss, P., Kelly, A. M. C., Uddin, L. Q., Gee, D. G., et al. (2008). Regional variation in interhemispheric coordination of intrinsic hemodynamic fluctuations. The Journal of Neuroscience, 28(51), 13754–13764.
14. Buckner RL, Sepulcre J, Talukdar T, Krienen FM, Liu H, Hedden T, Andrews-Hanna JR, Sperling RA, Johnson KA. 2009. Cortical hubs revealed by intrinsic functional connectivity: mapping, assessment of stability, and relation to Alzheimer’s disease. J Neurosci. 29:1860–1873.
15. Lohmann G, Margulies DS, Horstmann A, Pleger B, Lepsien J, Goldhahn D, Schloegl H, Stumvoll M, Villringer A, Turner R. 2010. Eigenvector centrality mapping for analyzing connectivity patterns in fMRI data of the human brain. PLoS One. 5:e10232
16. Tomasi D, Volkow ND. 2010. Functional connectivity density mapping. PNAS. 107(21):9885-9890.
17. C.F. Beckmann, C.E. Mackay, N. Filippini, and S.M. Smith. Group comparison of resting-state FMRI data using multi-subject ICA and dual regression. OHBM, 2009.
18. Smith, S. M., Fox, P. T., Miller, K. L., Glahn, D. C., Fox, P. M., Mackay, C. E., et al. (2009). Correspondence of the brain’s functional architecture during activation and rest. Proceedings of the National Academy of Sciences of the United States of America, 106(31), 13040–13045. doi:10.1073/pnas.0905267106
19. Dosenbach, N. U. F., Nardos, B., Cohen, A. L., Fair, D. a, Power, J. D., Church, J. a, … Schlaggar, B. L. (2010). Prediction of individual brain maturity using fMRI. Science (New York, N.Y.), 329(5997), 1358–61. http://doi.org/10.1126/science.1194144
20. Tzourio-Mazoyer, N., Landeau, B., Papathanassiou, D., Crivello, F., Etard, O., Delcroix, N., … Joliot, M. (2002). Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. NeuroImage, 15(1), 273–89. http://doi.org/10.1006/nimg.2001.0978
21. Eickhoff, S. B., Stephan, K. E., Mohlberg, H., Grefkes, C., Fink, G. R., Amunts, K., & Zilles, K. (2005). A new SPM toolbox for combining probabilistic cytoarchitectonic maps and functional imaging data. NeuroImage, 25(4), 1325–35. http://doi.org/10.1016/j.neuroimage.2004.12.034
22. Harvard-Oxford cortical and subcortical structural atlases, http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/Atlases
23. Lancaster, J. L., Woldorff, M. G., Parsons, L. M., Liotti, M., Freitas, C. S., Rainey, L., … Fox, P. T. (2000). Automated Talairach atlas labels for functional brain mapping. Human Brain Mapping, 10(3), 120–31. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/10912591
24. Craddock, R. C., James, G. A., Holtzheimer, P. E., Hu, X. P., & Mayberg, H. S. (2011). A whole brain fMRI atlas generated via spatially constrained spectral clustering. Human Brain Mapping, 0(July 2010). http://doi.org/10.1002/hbm.21333