Concatenated

Atlas creation and registration

Images were registered according to the numerical method outlined by Holland and Dale (2011), extended to include multimodal inputs. This consisted of rigid body, affine and nonlinear transformations. In total eleven multimodal channels were used to align scans and create the MINT ABCD atlas, however the input channels varied according to the registration step. Three structural MRI (sMRI) channels were included: T1w images, white matter segmentation, and grey matter segmentation; and eight diffusion MRI (dMRI) channels: the zeroth and second order spherical harmonics (SH) coefficients of the restricted orientation distribution function (ODF) and the zeroth order SH coefficients from the hindered and free water FODs from the restriction spectrum imaging (RSI) model. sMRI and dMRI channels were first aligned within-subject using rigid body registration transforms. Then inter-subject anatomical alignment was achieved using rigid body, affine and nonlinear transforms.

An iterative procedure was used to generate a registration target and the final ABCD atlas. The initial registration target was chosen as the atlas image used by Hagler et al. (2019) for the AtlasTrack method. A weighted group average was calculated for all eleven channels and the registration target was updated with the new multimodal average. The registration was repeated and the registration target updated again. This procedure was repeated three times, producing the iteratively refined multimodal ABCD atlas.

All available scans from all observations included in Data Release 6.0 were registered to the final group average, the ABCD atlas. The warps were applied to numerous imaging measures across dMRI, sMRI and tfMRI. This produced 3D volumes which were anatomically aligned across observations and modalities.

Concatenation and data storage

To reduce the total data size and facilitate data sharing, a brain mask was applied to the 3D volumes for each imaging measure and only voxels within the mask were stored as a 2D vector. The masked vector data were concatenated across observations and according to each measure. For each imaging measure (FA, MD, RNI etc) .mat files of the concatenated data contain a variable volmat that is a matrix of size N x P, where N = number of observations and P=number of voxels within the mask. Please see the Concatenated Imaging Data Documentation for details. Imaging measures are located in modality specific directories along with a vol_info.mat file which contains participant and session IDs and high and low resolution 3D volumes of the ABCD atlas mask. The ABCD atlas masks can be used to convert the 2D vectors into 3D volumes. The following Matlab code provides an example of how to do this using the fullvol.m function provided by the Fast and Efficient Mixed-effects Algorithm (FEMA) available from https://github.com/cmig-research-group/cmig_tools.

Load the data for a given imaging measure, say FA (this may take a while)

load('/path/to/ABCD/data/6.0/imaging_concat/voxelwise/dti/fa.mat')

To transform a single observation from a 2D vector to a 3D volume, select the row number you want. If you want a specific participant and session you can select that data finding the corresponding row in the vol_info.mat.

load('/path/to/ABCD/data/6.0/imaging_concat/voxelwise/dti/vol_info.mat');                    
idx = find(**contains**(**participant_id**, 'sub-0001') & contains(**session_id**, 'ses-00A')) 
vec_subj = volmat(**idx**,:);

Transform the 2D vector to a 3D volume using the fullvol function and vol_mask_sub

vol_subj = fullvol(vec_subj, vol_mask_sub);
showVol(vol_subj)

or saved as a nifti using the niftiwrite_amd function also part of FEMA.

fname_nii = 'vol_subj.nii.gz';
niftiwrite_amd(vol_subj, fname_nii, M_atl);

The fullvol function can be applied to multiple row vectors, say if you wanted to look at all sessions for a given participant

idx = find(**contains**(**participant_id**, 'sub-0001'))

idx

idx =

     1     2     3     4

vec_subj = volmat(**idx**,:); 

vol_subj = fullvol(vec_subj, vol_mask_sub);

Where vol_subj would now be a 4D volume. It is not recommended that you convert the complete volmat variable into a 4D volume as that would be a 100x100x130x29173 matrix which would require an exceptionally large amount of computer memory.

Data types

voxelwise voxel-based derived measures in ABCD atlas space

smri: structural MRI derived measures; e.g., T1, T2, etc.

dti: Diffusion Tensor Imaging derived; e.g., FA, MD, etc.

rsi: Restriction Spectrum Imaging derived; e.g., RND, HNT, etc.

tfmri: task fMRI derived beta coefficients

vertexwise cortical surface-based derived measures on FreeSurfer spherical average

surf: cortical surface coordinates; i.e., white and pial surfaces

smri: sMRI-derived measures; e.g., thickness (thk), area, sulc, T1, T2

dti: Diffusion Tensor Imaging derived; e.g., FA, MD, etc.

rsi: Restriction Spectrum Imaging derived; e.g., RND, HNT, etc.

tfmri: task fMRI derived beta and SEM coefficients

corrrmat: ROI time series and correlation matrix data with Gordon, Desikan, Destrieux, and aseg ROIs

rsfmri: resting-state fMRI

tfmri: task fMRI

Vertex numbering and visualization for vertexwise concatenated imaging data

The number of vertices is 10242. They form a 5th order icosahedral surface, and correspond to the first 10242 vertices of the 7th order icosahedral surface used for the FreeSurfer fsaverage subject and spherical average surface atlas. Cortical surface based measures on this ico 5 surface can be visualized using the cmig_tools showSurf package.

Volume information files

Each concatenated imaging data type (e.g., voxelwise smri, vertexwise dti, corrmat tfmri, etc.) is accompanied by a volume information file (vol_info.mat) which contains participant and session IDs needed to link the rows of concatenated data matrices to the corresponding rows in the tabulated data.

Contents of vol_info.mat

Variable name Description
participant_id

cell array of participant IDs

with one entry per visit (participant-event)

  e.g., sub-XXXXXXXX
session_id

cell array of session IDs

with one entry per visit

e.g., `ses-00A`, `ses-02A`, `ses-04A`, `ses-06A`
visitidvec

cell array of visit IDs

with one entry per visit

e.g., S042_INVXXXXXXXX_baseline

   {siteID}\_{subjID}\_{eventID}
dirlist

cell array of input directory names; e.g.,

FSURF_{siteID}_{subjID}_{eventID}_{datetime}

DTIPROC_{siteID}_{subjID}_{eventID}_{datetime}

BOLDPROC_{siteID}_{subjID}_{eventID}_{datetime}

DTIREG_{siteID}_{subjID}_{eventID}_{datetime}
voxelwise data types
vol_mask_aseg high resolution 3D volume of ABCD atlas mask
vol_mask_sub lower resolution 3D volume of ABCD atlas mask
M_atl 4x4 voxel to RAS transform of ABCD atlas
vertexwise data types
vol_mask_lh

cortical mask vector for left hemisphere vertices

excludes non-cortical midbrain region
vol_mask_rh

cortical mask vector for right hemisphere vertices

excludes non-cortical midbrain region
corrmat data types
vol_mask empty mask vector – all ROIs are valid


Reading concatenated imaging data files

Concatenated imaging data are stored as Version 7.3 MAT-files, which use an HDF5-based format and can be read into memory using the MATLAB load function. HDF5 is an open source format that can also be read in R (e.g., using the hdf5r package) and python (e.g, using the numpy library and h5py extension). The memory (RAM) required to read each file depends on the contents of the file and memory overhead requirements while running matlab. The time it takes to read each file into memory depends on the contents of the file as well factors such as CPU speed, disk read speed, and network bandwidth. Approximate values below should be taken as rough estimates of what to expect when using the MATLAB load function.

Category Modality File name pattern Time (min) Required RAM (GB)
voxelwise dti {measure}.mat 10 57
voxelwise rsi {measure}.mat 9 57
voxelwise smri {measure}.mat 10 66
voxelwise tfmri {task}_{condition}_{stat}_{run}.mat 10 57
voxelwise roi {segmentation}.mat 1 48
vertexwise surf {surface}-{hemi}.mat 1 10
vertexwise smri {measure}_sm{smoothing}-{hemi}.mat 6 7
vertexwise dti {measure}_{layer}_sm{smoothing}_{hemi}.mat 7 7
vertexwise rsi {measure}_{layer}_sm{smoothing}_{hemi}.mat 7 7
vertexwise tfmri {task}_{condition}_{stat}_{run}_{hemi}.mat 8 9
corrmat rsfmri tseries.mat 77 121
corrmat rsfmri gp_tseries.mat 67 78
corrmat rsfmri corr.mat 4 28
corrmat rsfmri gp_corr.mat 4 15
corrmat rsfmri gpnet_corr.mat <1 6
corrmat rsfmri gpnet_aseg_corr.mat <1 6
corrmat tfmri {task}_tseries.mat 12 75
corrmat tfmri {task}_gp_tseries.mat 8 52
corrmat tfmri {task}_corr_{type}.mat 9 29
corrmat tfmri {task}_gp_corr_{type}.mat 4 15
corrmat tfmri {task}_gpnet_corr_{type}.mat <1 4
corrmat tfmri {task}_gpnet_aseg_corr_{type}.mat <1 4

Note: File name patterns used above (e.g., {surface}, {hemi}, {measure}, {smoothing}, {layer}, {task}, {condition}, {stat}, {run}, {type}) represent groups of files for a given data type with varying values for the label in curly braces. The patterns for the respective data types are explained in the subsequent sections of this document.

dMRI-DTI Voxelwise Data

Location of data files: release/concat/imaging/voxelwise/dti

List of files

File name Description of contents
vol_info.mat participant and session IDs used to link to tabulated data and ABCD atlas brain masks
fa.mat measure: DTI fractional anisotropy
md.mat measure: DTI mean diffusivity

Contents of {measure}.mat

Variable name Description
volmat matrix of size N x P where N = number of observations and P=number of voxels

dMRI-RSI Voxelwise Data

Location of data files: release/concat/imaging/voxelwise/rsi

List of files

File name Description of contents
vol_info.mat participant and session IDs used to link to tabulated data and ABCD atlas brain masks
ri.mat measure: RSI restricted isotropic
rif.mat measure: RSI restricted isotropic fraction of total restricted signal
rni.mat measure: RSI restricted normalized isotropic
rd.mat measure: RSI restricted directional
rdf.mat measure: RSI restricted directional fraction of total restricted signal
rnd.mat measure: RSI restricted normalized directional
rt.mat measure: RSI restricted total
rnt.mat measure: RSI restricted normalized total
hi.mat measure: RSI hindered isotropic
hif.mat Measure: RSI hindered isotropic fraction of total hindered signal
hni.mat measure: RSI hindered normalized isotropic
hd.mat measure: RSI hindered directional
hdf.mat measure: RSI hindered directional fraction of total hindered signal
hnd.mat measure: RSI hindered normalized directional
ht.mat measure: RSI hindered total
hnt.mat measure: RSI hindered normalized total
fi.mat measure: RSI free isotropic
fni.mat measure: RSI free normalized isotropic

Contents of {measure}.mat

Variable name Description
volmat matrix of size N x P where N= number of observations and P=number of voxels

sMRI Voxelwise Data

Location of data files: release/concat/imaging/voxelwise/smri

List of files

File name Description of contents
vol_info.mat participant and session IDs used to link to tabulated data and ABCD atlas brain masks
nu.mat measure: T1-weighted signal intensity
t2.mat measure: T2-weighted signal intensity
ja.mat measure: the determinant of the Jacobian matrix of the nonlinear warp to atlas space
gm_new.mat measure: grey matter segmentation
wm.mat measure: white matter segmentation
bm.mat measure: brain mask
b0.mat measure: b=0 signal intensity
di1vol1.mat measure: nonlinear deformation field for y-direction
di2vol1.mat measure: nonlinear deformation field for x-direction
di3vol1.mat measure: nonlinear deformation field for z-direction

Contents of {measure}.mat

Variable name Description
volmat matrix of size N x P where N= number of observations and P=number of voxels

Task-based fMRI Voxelwise Data

Location of data files: release/concat/imaging/voxelwise/tfmri

List of files

File name Description of contents
vol_info.mat participant and session IDs used to link to tabulated data and ABCD atlas brain masks

mid_{condition}_{stat}_{run}.mat

e.g., mid_sra_beta_r01.mat

mid_sra_beta_r02.mat

task: mid (Monetary Incentive Delay)
conditions: see below

stats: beta

runs: r01, r02

nback_{condition}_{stat}_{run}.mat

e.g., nback_2b_beta_r01.mat

nback_2b_beta_r02.mat

task: nback (Emotional n-Back)

conditions: see below

stats: beta

runs: r01, r02

sst_{condition}_{stat}_{run}.mat

e.g., sst_cg_beta_r01.mat

sst_cg_beta_r02.mat

task: sst (Stop Signal Task)

conditions: see below

stats: beta

runs: r01, r02

Notes:

  • File name patterns above with {condition}, {stat} and {run} represent a group of files for a given task (i.e.,”mid” = Monetary Incentive Delay, “nback” = Emotional N-Back, and “sst” = Stop Signal Task) with varying conditions (see tables below), GLM statistic beta = time series model coefficient, two separate runs (“r01” = run 1, and “r02” = run 2) represented.
  • Outlier beta values (i.e., outside the mean ± 3 x standard deviation) are present for some particular voxels, runs, conditions, and/or paticipant-events. See “Task fMRI Outliers” in Task-Based fMRI Data Documentation for more information and guidance.

Conditions and contrasts

Concatenated data files containing beta estimates derived from AFNI’s 3dDeconvolve were created for each of the conditions and contrasts defined in supplementary tables for each task. Contrasts were implemented as general linear tests with AFNI’s 3dDeconvolve, defined as a subtraction between two groups of conditions that can each include zero, one, or more conditions, depending on the contrast definition. See Supplementary Tables for more info.

Contents of {task}\_{condition}\_{stat}\_{run}.mat

Variable name Description
volmat matrix of size N x P where N= number of observations and P=number of voxels

ROI Voxelwise Data

Location of data files: release/concat/imaging/voxelwise/smri

List of files

File name Description of contents
aparcaseg.mat probabilistic ROIs of the Freesurfer automatic cortical parcellation (aparc) and the automatic subcortical segmentation (aseg)
aseg.mat probabilistic ROIs of the Freesurfer automatic subcortical segmentation (ages)
fiber.mat probabilistic ROIs of white matter tracts labelled using AtlasTrack
pauli_subcortnuclei_abcd3_cor10.mat subcortical ROIs from Pauli, Nili, and Tyszka (2018) |
thalamus_abcd3_cor10.mat thalamic nuclei ROIs from Najdenovska et al. (2018) |

Contents of {segmentation}.mat

Variable name Description
volmat_prob matrix of size X x Y x Z x R where X,Y,Z voxels dimension and R is number of ROIs
volmat_prob_rgb 5D matrix of size X x Y x Z x R x 3 where X,Y,Z voxels dimension and R is number of ROIs
volmat_prob_rgbsum 4D matrix of size X x Y x Z x 3 where X,Y,Z voxels dimensions.
roicodes array of ROI numeric identifiers
roinames ROI names
roirgb ROI RGB color codes
subjlist list of subjects included in cohort average

Contents of pauli_subcortnuclei_abcd3_cor10.mat

Variable name Description
D_fwd combined 3D nonlinear warp from subcortical nuclei atlas to ABCD atlas
M_atl M transform of ABCD atlas
M_reg_af reslice matrix from ABCD atlas to subcortical nuclei atlas
di1vol1 nonlinear deformation field for y-direction
di2vol1 nonlinear deformation field for x-direction
di3vol1 nonlinear deformation field for z-direction
muvols_T1_ctx ABCD atlas T1 volume
pauli_lut look up table for ROI code, name and RGB color code
vol_T1 subcortical nuclei atlas T1 volume in original space
vol_T1_af subcortical nuclei atlas T1 volume affine transformed to ABCD atlas space
vol_T1_rb subcortical nuclei atlas T1 volume rigid body transformed to ABCD atlas space
vol_T1_reg subcortical nuclei atlas T1 volume nonlinearly transformed to ABCD atlas space
vol_T2 subcortical nuclei atlas T2 volume in original space
vol_T2_af subcortical nuclei atlas T2 volume affine transformed to ABCD atlas space
vol_T2_rb subcortical nuclei atlas T2 volume rigid body transformed to ABCD atlas space
vol_T2_reg subcortical nuclei atlas T2 volume nonlinearly transformed to ABCD atlas space
vol_labels_T1_reg_rgb color coded segmentation volume nonlinearly transformed to ABCD atlas space
vol_labels_af segmentation volume affine transformed to ABCD atlas space
vol_labels_reg segmentation volume rigid body transformed to ABCD atlas space
vol_labels_reg_rgb segmentation volume nonlinearly transformed to ABCD atlas space
vol_labels_res_T1 segmentation volume in original space, T1 resolution
volresfun_reg function handle to transform subcortical nuclei atlas to ABCD atlas space

Contents of thalamus_abcd3_cor10.mat

Variable name Description
D_fwd combined 3D nonlinear warp from thalamic nuclei atlas to ABCD atlas
M_atl M transform of ABCD atlas
M_reg_af reslice matrix from ABCD atlas to thalamic nuclei atlas
di1vol1 nonlinear deformation field for y-direction
di2vol1 nonlinear deformation field for x-direction
di3vol1 nonlinear deformation field for z-direction
muvols_T1_ctx ABCD atlas T1 volume
thalamus_lut look up table for ROI code, name and RGB color code
vol_T1 thalamic nuclei atlas T1 volume in original space
vol_T1_af thalamic nuclei atlas T1 volume affine transformed to ABCD atlas space
vol_T1_rb thalamic nuclei atlas T1 volume rigid body transformed to ABCD atlas space
vol_T1_reg thalamic nuclei atlas T1 volume nonlinearly transformed to ABCD atlas space
vol_labels_T1_reg_rgb color coded T1 volume nonlinearly transformed to ABCD atlas space
vol_labels_af segmentation volume affine transformed to ABCD atlas space
vol_labels_reg segmentation volume nonlinearly transformed to ABCD atlas space
vol_labels_reg_rgb color coded segmentation volume nonlinearly transformed to ABCD atlas space
vol_labels_res_T1 segmentation volume in original space, T1 resolution
volresfun_reg function handle to transform thalamic atlas to ABCD atlas space

Surfaces Vertexwise Data

Location of data files: release/concat/imaging/vertexwise/surfs

List of files

File name Description of contents
vol_info.mat

participant IDs and event names

used to link to tabulated data

pial-{hemi}.mat
i.e., pial-lh.mat

   `pial-rh.mat`
 measure: cortical pial surface
vertex coordinates
hemispheres: lh, rh

Note: File names above with {hemi} represent a group of files for a given measure (i.e., white or pial surfaces) with both hemispheres (lh and rh) represented

Contents of vol_info

Variable name Description
subjidvec

cell array of subject IDs

with one entry per visit (participant-event)

  e.g., “XXXXXXXX”

compare with old style*:

  “NDAR_INVXXXXXXXX”
eventvec

cell array of event names

with one entry per visit (participant-event)

    e.g., 00A, 02A, 04A, 06A

compare with old style:

    baseline_year_1_arm_1

    2_year_follow_up_y_arm_1

    4_year_follow_up_y_arm_1

    6_year_follow_up_y_arm_1
visitidvec

cell array of visit IDs

with one entry per visit (participant-event)

e.g., S042_INVXXXXXXXX_baseline
vol_mask_lh

cortical mask vector for left hemisphere

e.g., excludes non-cortical midbrain region
vol_mask_rh

cortical mask vector for right hemisphere

e.g., excludes non-cortical midbrain region

Note: examples shown for subject IDs and event names reflect new convention for Release 6.0, which removes the “NDAR_INV” from the subject ID and changes the event names to be “00A”, “02A”, “04A”, and “06A”.

Contents of {surface}-{hemi}.mat

Variable name Description
measmat

cortical surface coordinates

matrix with size = [ndirs,nverts,3]

with ndirs = number of directories (visits)

with nverts = number of vertices = 10242

and 3 coordinate axes (x, y, z)
mask

binary mask

(e.g., exclude non-cortical midbrain region)

vector with size = [nverts,1]

nverts: number of vertices
vctx

vertex numbers of fsaverage vertices in mask

vector with size = [nmask,1]

nmask: number of vertices with mask = 1
dirlist

cell array of directory names with length ndirs

with ndirs = number of directories (visits)

e.g., FSURF_ S042_INVXXXXXXXX_baseline_20181001.095833.791000_1

Note: Loading the file {measure}\_sm{smoothing}-{hemi}.mat file takes ~1 minute (depending on CPU speed) and requires ~10 GB

*Note: The number of vertices is 10242. They form a 5th order icosahedral surface, and correspond to the first 10242 vertices of the 7th order icosahedral surface used for the FreeSurfer fsaverage subject and spherical average surface atlas. Cortical surface based measures on this ico 5 surface can be visualized using the cmig_tools showSurf package.

sMRI Vertexwise Data

Location of data files: release/concat/imaging/vertexwise/smri

List of files

File name Description of contents
vol_info.mat

participant and session IDs

used to link to tabulated data

thk_sm{smoothing}_{hemi}.mat

e.g., thk_sm0_lh.mat

    `thk_sm0_rh.mat`

    `thk_sm16_lh.mat`

    `thk_sm16_rh.mat`

    `thk_sm256_lh.mat`

    `thk_sm256_rh.mat`

    `thk_sm1000_lh.mat`

    `thk_sm1000_rh.mat`

measure: cortical thickness
smoothing: 0, 16, 256, 1000 iterations

               0, 5,   20,   40 mm FWHM

hemispheres: lh, rh
area_sm{smoothing}_{hemi}.mat
e.g., area_sm0_lh.mat

measure: cortical surface area

smoothing: 0, 16, 256, 1000 iterations

               0, 5,   20,   40 mm FWHM

hemispheres: lh, rh

sulc_sm{smoothing}_{hemi}.mat

e.g., sulc_sm0_lh.mat

measure: cortical sulcal depth

smoothing: 0, 16, 256, 1000 iterations

               0, 5,   20,   40 mm FWHM

hemispheres: lh, rh

t1_gwc_sm{smoothing}_{hemi}.mat

e.g., t1_gwc_sm0_lh.mat

measure: cortical T1w contrast

smoothing: 0, 16, 256, 1000 iterations

               0, 5,   20,   40 mm FWHM

hemispheres: lh, rh

t2_gwc_sm{smoothing}_{hemi}.mat

e.g., t2_gwc_sm0_lh.mat

measure: cortical T2w contrast

smoothing: 0, 16, 256, 1000 iterations

               0, 5,   20,   40 mm FWHM

hemispheres: lh, rh

Note: File names above with {smoothing} and {hemi} represent a group of files for a given measure (i.e., thk = thickness, area = surface area, sulc = sulcal depth, t1_gwc = T1w gray-white contrast, t2_gwc = T2w gray-white contrast) with varying smoothing levels (0, 16, 256, 1000 iterations) and two hemispheres (lh and rh) represented.

Contents of {measure}\_sm{smoothing}\_{hemi}.mat

Variable name Description
measmat

cortical surface measure

matrix with size = [ndirs,nverts]

with ndirs = number of directories (visits)

with nverts = number of vertices = 10242

dMRI-DTI Vertexwise Data

Location of data files: release/concat/imaging/vertexwise/dti

List of files

File name Description of contents
vol_info.mat

participant and session IDs

used to link to tabulated data

fa_{layer}_sm{smoothing}_{hemi}.mat

e.g., fa_gm_sm0_lh.mat

    `fa_gm_sm0_rh.mat`

    `fa_wm_sm0_lh.mat`

    `fa_wm_sm0_rh.mat`

    `fa_gm_sm1000_lh.mat`

    `fa_gm_sm1000_rh.mat`

    `fa_wm_sm1000_lh.mat`

    `fa_wm_sm1000_rh.mat`

measure: DTI fractional anisotropy

layers: gm, wm
smoothing: 0, 16, 256, 1000 iterations

               0, 5,   20,   40 mm FWHM

hemispheres: lh, rh

md_{layer}_sm{smoothing}_{hemi}.mat

e.g., md_gm_sm0_lh.mat

    `md_gm_sm0_rh.mat`

    `md_wm_sm0_lh.mat`

    `md_wm_sm0_rh.mat`

measure: DTI mean diffusivity

layers: gm, wm
smoothing: 0, 16, 256, 1000 iterations

               0, 5,   20,   40 mm FWHM

hemispheres: lh, rh

ld_{layer}_sm{smoothing}_{hemi}.mat

e.g., ld_gm_sm0_lh.mat

    `ld_gm_sm0_rh.mat`

    `ld_wm_sm0_lh.mat`

    `ld_wm_sm0_rh.mat`

measure: DTI longitudinal (axial) diffusivity

layers: gm, wm
smoothing: 0, 16, 256, 1000 iterations

               0, 5,   20,   40 mm FWHM

hemispheres: lh, rh

td_{layer}_sm{smoothing}_{hemi}.mat

e.g., td_gm_sm0_lh.mat

    `td_gm_sm0_rh.mat`

measure: DTI transverse (radial) diffusivity

layers: gm, wm
smoothing: 0, 16, 256, 1000 iterations

               0, 5,   20,   40 mm FWHM

hemispheres: lh, rh

Note: File names above with {layer}, {smoothing}, and {hemi} represent a group of files for a given measure (i.e., DTI measures: FA, MD, LD, TD) with varying layers (gm = gray matter, wm = white matter), smoothing levels (0, 16, 256, 1000 iterations), and hemispheres (lh = left hemisphere and rh = right hemisphere) represented

Contents of {measure}\_{layer}\_sm{smoothing}\_{hemi}.mat

Variable name Description
measmat

cortical surface measure

matrix with size = [ndirs,nverts]

with ndirs = number of directories (visits)

with nverts = number of vertices = 10242

dMRI-RSI Vertexwise Data

Location of data files: release/concat/imaging/vertexwise/rsi

List of files

File name Description of contents
vol_info.mat

participant and session IDs

used to link to tabulated data

rni_{layer}_sm{smoothing}_{hemi}.mat

e.g., rni_gm_sm0_lh.mat

    `rni_gm_sm0_rh.mat`

measure: RSI restricted normalized isotropic

layers: gm, wm
smoothing: 0, 16, 256, 1000 iterations

               0, 5,   20,   40 mm FWHM

hemispheres: lh, rh

rnd_{layer}_sm{smoothing}_{hemi}.mat

e.g., rnd_gm_sm0_lh.mat

    `rnd_gm_sm0_rh.mat`

measure: RSI restricted normalized directional

layers: gm, wm
smoothing: 0, 16, 256, 1000 iterations

               0, 5,   20,   40 mm FWHM

hemispheres: lh, rh

rnt_{layer}_sm{smoothing}_{hemi}.mat

e.g., rnt_gm_sm0_lh.mat

    `rnt_gm_sm0_rh.mat`

measure: RSI restricted normalized total

layers: gm, wm
smoothing: 0, 16, 256, 1000 iterations

               0, 5,   20,   40 mm FWHM

hemispheres: lh, rh

hnt_{layer}_sm{smoothing}_{hemi}.mat

e.g., hnt_gm_sm0_lh.mat

    `hnt_gm_sm0_rh.mat`

measure: RSI hindered normalized total

layers: gm, wm
smoothing: 0, 16, 256, 1000 iterations

               0, 5,   20,   40 mm FWHM

hemispheres: lh, rh

fni_{layer}_sm{smoothing}_{hemi}.mat

e.g., fni_gm_sm0_lh.mat

    `fni_gm_sm0_rh.mat`

measure: RSI free normalized isotropic

layers: gm, wm
smoothing: 0, 16, 256, 1000 iterations

               0, 5,   20,   40 mm FWHM

hemispheres: lh, rh

Note: File names above with {layer}, {smoothing}, and {hemi} represent a group of files for a given measure (i.e., RSI measures: RNI, RND, RNT, HNT, FNI) with varying layers (gm = gray matter, wm = white matter), smoothing levels (0, 16, 256, 1000 iterations), and hemispheres (lh = left hemisphere and rh = right hemisphere) represented.

Contents of {measure}\_{layer}\_sm{smoothing}\_{hemi}.mat

Variable name Description
measmat

cortical surface measure

matrix with size = [ndirs,nverts]

with ndirs = number of directories (visits)

with nverts = number of vertices = 10242

Task-based fMRI Vertexwise Data

Location of data files: release/concat/imaging/vertexwise/tfmri

List of files

File name Description of contents
vol_info.mat

participant and session IDs

used to link to tabulated data

mid_{condition}_{stat}_{run}_{hemi}.mat

e.g., mid_sra_beta_r01_lh.mat

    `mid_sra_sem_r01_lh.mat`

    `mid_sra_beta_r02_lh.mat`

    `mid_sra_sem_r02_lh.mat`

task: mid (Monetary Incentive Delay)
conditions: see below

stats: beta, sem

runs: r01, r02

hemispheres: lh, rh

nback_{condition}_{stat}_{run}_{hemi}.mat

e.g., nback_2b_beta_r01_lh.mat

    `nback_2b_sem_r01_lh.mat`

task: nback (Emotional n-Back)

conditions: see below

stats: beta, sem

runs: r01, r02

hemispheres: lh, rh

sst_{condition}_{stat}_{run}_{hemi}.mat

e.g., sst_cg_beta_r01_lh.mat

    `sst_cg_sem_r01_lh.mat`

task: sst (Stop Signal Task)

conditions: see below

stats: beta, sem

runs: r01, r02

hemispheres: lh, rh

Notes:

  • File name patterns above with {condition}, {stat}, {run}, and {hemi} represent a group of files for a given task (i.e.,”mid” = Monetary Incentive Delay, “nback” = Emotional N-Back, and “sst” = Stop Signal Task) with varying conditions (see tables below), two GLM statistics (beta = time series model coefficient, sem = standard error of the mean), two separate runs (“r01” = run 1, and “r02” = run 2), and two hemispheres (lh = left and rh = right) represented.
  • Outlier beta values (i.e., outside the mean ± 3 x standard deviation) are present for some particular vertices, runs, conditions, and/or paticipant-events. See “Task fMRI Outliers” in Task-Based fMRI Data Documentation for more information and guidance.

Conditions and contrasts

Concatenated data files containing beta estimates and standard errors derived from AFNI’s 3dDeconvolve were created for each of the conditions and contrasts defined in supplementary tables for each task. Contrasts were implemented as general linear tests with AFNI’s 3dDeconvolve, defined as a subtraction between two groups of conditions that can each include zero, one, or more conditions, depending on the contrast definition. See Supplementary Tables for more info.

Contents of {task}\_{condition}\_{stat}\_{run}\_hemi}.mat

Variable name Description
measmat

beta or sem on cortical surface

matrix with size = [ndirs,nverts]

with ndirs = number of directories (visits)

with nverts = number of vertices = 10242

Conditions and contrasts

Concatenated data files containing beta estimates and standard errors derived from AFNI’s 3dDeconvolve were created for each of the conditions and contrasts defined in supplementary tables for each task. Contrasts were implemented as general linear tests with AFNI’s 3dDeconvolve, defined as a subtraction between two groups of conditions that can each include zero, one, or more conditions, depending on the contrast definition.

Contents of {task}\_{condition}\_{stat}\_{run}\_hemi}.mat

Variable name Description
measmat

beta or sem on cortical surface

matrix with size = [ndirs,nverts]

with ndirs = number of directories (visits)

with nverts = number of vertices = 10242

Resting-state fMRI ROI Correlation and Time Series Data

Location of data files: release/concat/imaging/corrmat/rsfmri

List of files

File name Description of contents
vol_info.mat

participant and session IDs

used to link to tabulated data
corr.mat ROI correlation matrices for all ROIs (includes Desikan, Destrieux, and Gordon parcels plus subcortical volumetric ROIs)
dsk_aseg_corr.mat ROI correlation matrices for Desikan parcels plus subcortical volumetric ROIs
dst_aseg_corr.mat ROI correlation matrices for Destrieux parcels plus subcortical volumetric ROIs
gp_aseg_corr.mat ROI correlation matrices for Gordon parcels plus subcortical volumetric ROIs
gpnet_corr.mat Within and between network correlation matrices derived from ROI correlations for Gordon parcels
gpnet_aseg_corr.mat Network to subcortical ROI correlation matrices derived from ROI correlations for Gordon parcels and subcortical volumetric ROIs
tseries.mat ROI time series data (includes Desikan, Destrieux, and Gordon parcels plus subcortical volumetric ROIs)
dsk_aseg_tseries.mat ROI time series data for Desikan parcels plus subcortical volumetric ROIs
dst_aseg_tseries.mat ROI time series data for Destrieux parcels plus subcortical volumetric ROIs
gp_aseg_tseries.mat ROI time series data for Gordon parcels plus subcortical volumetric ROIs

Notes:

  • File names above with dsk, dst, or gp, and aseg indicate ROI data for one specific cortical surface atlas (i.e., dsk = Desikan, dst = Destrieux, and gp = Gordon) plus a selection of subcortical volumetric ROIs (aseg = FreeSurfer automated segmentation).
  • File names with gpnet indicate correlations within or between networks defined by the Gordon parcel “communities”.
  • File names with trim5 or trim10 indicate correlations derived from time series data “trimmed” to either 5 or 10 minutes of low-motion (FD < 0.2 mm) time points. Time points were randomly selected from available time points. Excess time points were discarded. Correlations for visits with fewer available time points were set to NaN (see Responsible Use Warning: Head motion in Resting-State fMRI Data Documentation).

Contents of corr.mat and {atlas}\_aseg_corr.mat

Variable name Description
corrmat

ROI-ROI correlations

matrix with size = [ndirs,nnodes]

notes:

correlation values range from -1 to 1

no R to Z transform applied

for reduced memoy usage, corrmat contains

only the “upper triangle” of the matrrix

not the full matrix
varmat

ROI temporal variance

matrix with size = [ndirs,nroi]
meanfdvec

mean framewise displacement

vector with size = [ndirs,1]
ntpointvec

number of time points

after censoring frames with excessive motion

vector with size = [ndirs,1]
ndirs

number of input directories

matches lengths of participant_id and session_id in vol_info
nroi number of ROIs
roinames

cell array of ROI names with length nroi

e.g., ctx-lh-bankssts

    ctx-lh-caudalanteriorcingulate

    ctx-lh-caudalmiddlefrontal

    …

    ctx_lh_G_and_S_frontomargin'

    ctx_lh_G_and_S_occipital_inf'

    ctx_lh_G_and_S_paracentral'

   …

    ctx-lh-gp1

    ctx-lh-gp2

    ctx-lh-gp3

   …

    Left-Cerebral-White-Matter

    Left-Cerebral-Cortex

    Left-Lateral-Ventricle

   …
nnodes

number of nodes (ROI-ROI pairs)

equal to (nroi x (nroi + 1))/2

includes upper triangle of ROI-ROI matrix
roi1vec specifies ROI number (index to roinames) for first ROI of each node of correlation matrix
roi2vec specifies ROI number (index to roinames) for second ROI of each node of correlation matrix

Example: extract correlation matrix for one visit

x = find(strcmp(visitidvec,’ S042_INVXXXXXXXX_baseline’))

rvec = reshape(corrmat(x,:),\[nroi,nroi\])

idx_triu = find(triu(ones(nroi,nroi)));

idx_tril = find(tril(ones(nroi,nroi)));

R_upper = nan(nroi,nroi);

R_upper(idx_triu) = rvec;

R_lower = R_upper';

R = R_upper;

R(idx_tril) = R_lower(idx_tril);

Contents of gpnet_corr.mat

Variable name Description
corrmat

ROI-ROI correlations

matrix with size = [ndirs,nnodes]

notes:

correlation values range from -1 to 1

no R to Z transform applied

for reduced memoy usage, corrmat contains

only the “upper triangle” of the matrrix

not the full matrix
meanfdvec

mean framewise displacement

vector with size = [ndirs,1]
ntpointvec

number of time points

after censoring frames with excessive motion

vector with size = [ndirs,1]
ndirs

number of input directories

matches lengths of participant_id and session_id in vol_info
nnet number of networks
network_names

cell array of network names with length nnet

i.e., Auditory

   CinguloOperc

   CinguloParietal

   Default

   DorsalAttn

   FrontoParietal

   None

   RetrosplenialTemporal

   SMhand

   SMmouth

   Salience

   VentralAttn

   Visual
nnodes

number of nodes (network-network pairs)

equal to (nnet x (nnet + 1))/2

includes upper triangle of ROI-ROI matrix
net1vec specifies network number (index to network_names) for first network of each node of correlation matrix
net2vec specifies network number (index to roinames) for second network of each node of correlation matrix

Contents of gpnet_aseg_corr.mat

Variable name Description
corrmat

network-ROI correlations

matrix with size = [ndirs,nnodes]

notes:

correlation values range from -1 to 1

no R to Z transform applied
meanfdvec

mean framewise displacement

vector with size = [ndirs,1]
ntpointvec

number of time points

after censoring frames with excessive motion

vector with size = [ndirs,1]
ndirs

number of input directories

matches lengths of participant_id and session_id in vol_info
nnet number of networks
network_names

cell array of network names with length nnet

i.e., Auditory

   CinguloOperc

   CinguloParietal

   Default

   DorsalAttn

   FrontoParietal

   None

   RetrosplenialTemporal

   SMhand

   SMmouth

   Salience

   VentralAttn

   Visual
nroi number of subcortical ROIs
roinames

cell array of ROI names with length nroi

i.e., Left-Cerebellum-Cortex

   Left-Thalamus-Proper

   Left-Caudate

   Left-Putamen

   Left-Pallidum

   Brain-Stem

   Left-Hippocampus

   Left-Amygdala

   Left-Accumbens-area

   Left-VentralDC

   Right-Cerebellum-Cortex

   Right-Thalamus-Proper

   Right-Caudate

   Right-Putamen

   Right-Pallidum

   Right-Hippocampus

   Right-Amygdala

   Right-Accumbens-area

   Right-VentralDC
nnodes

number of nodes (network-ROI pairs)

equal to nnet x nroi
net1vec specifies network number (index to network_names) for first network of each node of correlation matrix
roi2vec specifies ROI number (index to roinames) for second ROI of each node of correlation matrix

Contents of tseries.mat and {atlas}\_aseg_tseries.mat

Variable name Description
datamat_tsdata

ROI time series data

with size = [ndirs,nruns,ntpoints,nroi]
datamat_motion

motion time series data

with size = [ndirs,nruns,ntpoints,6]

6 motion parameters: dx, dy, dz, rx, ry, rz

notch-filtered to remove respiratory signals
censvec

censor vector

with size = [ndirs,nruns,ntpoints]

value of 1 indicates time point was censored

due to excessive motion
ndirs

number of input directories

matches lengths of participant_id and session_id in vol_info
nroi number of ROIs
roinames

cell array of ROI names with length nroi

e.g., ctx-lh-bankssts

    ctx-lh-caudalanteriorcingulate

    ctx-lh-caudalmiddlefrontal

    …

    ctx_lh_G_and_S_frontomargin'

    ctx_lh_G_and_S_occipital_inf'

    ctx_lh_G_and_S_paracentral'

   …

    ctx-lh-gp1

    ctx-lh-gp2

    ctx-lh-gp3

   …

    Left-Cerebral-White-Matter

    Left-Cerebral-Cortex

    Left-Lateral-Ventricle

   …
nruns number of imaging runs (scans)
ntpoints number of time points per run

Example: extract time series data for one run of one visit

x = find(strcmp(visitidvec,’ S042_INVXXXXXXXX_baseline’))

r = 2

tmp_censvec = censvec(x,r,:);

tmp_tsdata =  datamat_tsdata(x,r,:,:);

Task-based fMRI ROI Correlation and Time Series Data

Location of data files: release/concat/imaging/corrmat/tfmri

List of files and file name patterns

File name Description of contents
vol_info.mat

participant and session IDs

used to link to tabulated data

{task}_corr_{type}.mat
e.g., mid_corr_data.mat

    `nback_corr_data.mat`

    `sst_corr_data.mat`

    `mid_corr_task.mat`

    `mid_corr_residual.mat`

ROI correlation matrices for all ROIs (includes Desikan, Destrieux, and Gordon parcels plus subcortical volumetric ROIs)

task: mid, nback, sst

types: data, task, residual

{task}_{atlas}_aseg_corr_{type}.mat

e.g., mid_dsk_aseg_corr_data.mat

    `mid_dst_aseg_corr_data.mat`

    `mid_gp_aseg_corr_data.mat`

ROI correlation matrices for {atlas} parcels plus subcortical volumetric ROIs


atlases: dsk, dst, gp

task: mid, nback, sst

types: data, task, residual
{task}_gpnet_corr_{type}.mat

Within and between network correlation matrices derived from ROI correlations for Gordon parcels


atlas: gpnet

task: mid, nback, sst

types: data, task, residual
{task}_gpnet_aseg_corr_{type}.mat

Network to subcortical ROI correlation matrices derived from ROI correlations for Gordon parcels and subcortical volumetric ROIs


atlas: gpnet

task: mid, nback, sst

types: data, task, residual

{task}_tseries.mat

e.g., mid_tseries.mat

ROI time series data (includes Desikan, Destrieux, and Gordon parcels plus subcortical volumetric ROIs)


task: mid, nback, sst

{task}_{atlas}_aseg_tseries.mat

e.g., mid_dsk_aseg_tseries.mat

    `mid_dst_aseg_tseries.mat`

    `mid_gp_aseg_tseries.mat`

ROI time series data for {atlas} parcels plus subcortical volumetric ROIs


atlases: dsk, dst, gp

task: mid, nback, sst

Notes:

  • File names above with {task} represent a group of files for each task (i.e.,”mid” = Monetary Incentive Delay, “nback” = Emotional N-Back, and “sst” = Stop Signal Task).
  • File names above with {type} represent a group of files derived from time series data (i.e., data = preprocessed time series, residual = task residualized time series, task = estimated task-related time series).
  • File names above with {atlas} represent a group of files that include ROI data for one specific cortical surface atlas (i.e., dsk = Desikan, dst = Destrieux, and gp = Gordon) plus a selection of subcortical volumetric ROIs (aseg = FreeSurfer automated segmentation).
  • File names with “gpnet” indicate correlations within or between networks defined by the Gordon parcel “communities”.
  • File names with “trim5” indicate correlations derived from time series data “trimmed” to 5 minutes of low-motion (FD < 0.2 mm) time points. Time points were randomly selected from available time points. Excess time points were discarded. Correlations for visits with fewer available time points were set to NaN (see Responsible Use Warning: Head Motion in Resting-State fMRI Data Documentation).

Contents of {task}\_corr\_{type}.mat and {task}\_{atlas}\_aseg_corr\_{type}.mat

Variable name Description
corrmat

ROI-ROI correlations

matrix with size = [ndirs,nnodes]

notes:

correlation values range from -1 to 1

no R to Z transform applied

for reduced memoy usage, corrmat contains

only the “upper triangle” of the matrrix

not the full matrix
varmat

ROI temporal variance

matrix with size = [ndirs,nroi]
meanfdvec

mean framewise displacement

vector with size = [ndirs,1]
ntpointvec

number of time points

after censoring frames with excessive motion

vector with size = [ndirs,1]
ndirs

number of input directories

matches lengths of participant_id and session_id in vol_info
nroi number of ROIs
roinames

cell array of ROI names with length nroi

e.g., ctx-lh-bankssts

    ctx-lh-caudalanteriorcingulate

    ctx-lh-caudalmiddlefrontal

    …

    ctx_lh_G_and_S_frontomargin'

    ctx_lh_G_and_S_occipital_inf'

    ctx_lh_G_and_S_paracentral'

   …

    ctx-lh-gp1

    ctx-lh-gp2

    ctx-lh-gp3

   …

    Left-Cerebral-White-Matter

    Left-Cerebral-Cortex

    Left-Lateral-Ventricle

   …
nnodes

number of nodes (ROI-ROI pairs)

equal to (nroi x (nroi + 1))/2

includes upper triangle of ROI-ROI matrix
roi1vec specifies ROI number (index to roinames) for first ROI of each node of correlation matrix
roi2vec specifies ROI number (index to roinames) for second ROI of each node of correlation matrix

Notes:

  • Concatenated ROI-ROI correlation data is derived from the concatenated ROI time series data, which is provided in a relatively raw state, without pre-residualization or filtering.
  • Additional processing steps applied before calculation of ROI-ROI correlations match those used in the standard analysis pipeline (see Resting-State fMRI Data Documentation).
  • Additional processing steps:
    • Normalization and demean time series for each ROI
    • Residualization of motion, squared motion, derivatives, and squared derivatives
    • Residualization of global signals (white matter, brain, CSF)
    • Band pass filtering (between 0.009 and 0.08 Hz)
    • Motion censoring of bad frames
      • FD > 0.2 mm
      • < 5 contiguous uncensored frames
      • Outliers in standard deviation across ROIs

Example: extract correlation matrix for one visit

x = find(strcmp(visitidvec,’ S042_INVXXXXXXXX_baseline’))

rvec = reshape(corrmat(x,:),\[nroi,nroi\])

idx_triu = find(triu(ones(nroi,nroi)));

idx_tril = find(tril(ones(nroi,nroi)));

R_upper = nan(nroi,nroi);

R_upper(idx_triu) = rvec;

R_lower = R_upper';

R = R_upper;

R(idx_tril) = R_lower(idx_tril);

Contents of {task}\_gpnet_corr\_{type}.mat

Variable name Description
corrmat

ROI-ROI correlations

matrix with size = [ndirs,nnodes]

notes:

correlation values range from -1 to 1

no R to Z transform applied

for reduced memoy usage, corrmat contains

only the “upper triangle” of the matrrix

not the full matrix
meanfdvec

mean framewise displacement

vector with size = [ndirs,1]
ntpointvec

number of time points

after censoring frames with excessive motion

vector with size = [ndirs,1]
ndirs

number of input directories

matches lengths of participant_id and session_id in vol_info
nnet number of networks
network_names cell array of network names with length nnet
nnodes

number of nodes (network-network pairs)

equal to (nnet x (nnet + 1))/2

includes upper triangle of ROI-ROI matrix
net1vec specifies network number (index to network_names) for first network of each node of correlation matrix
net2vec specifies network number (index to roinames) for second network of each node of correlation matrix

Notes:

  • Concatenated network-network correlation data is derived from the concatenated ROI-ROI correlation data.
  • Differences between the concatenated and tabulated network correlation data are attributable to differences in the order of pre-processing steps.
  • For the concatenated data, average ROI time series are first calculated from relatively unprocessed time series data, and residualization, filtering, and censoring are performed on the average ROI times series data.
  • For the tabulated data, residualization and filtering is performed on the voxelwise time series data, and then average ROI time series are calculated from the pre-processed voxelwise data.

Contents of {task}\_gpnet_aseg_corr\_{type}.mat

Variable name Description
corrmat

network-ROI correlations

matrix with size = [ndirs,nnodes]

notes:

correlation values range from -1 to 1

no R to Z transform applied

for reduced memoy usage, corrmat contains

only the “upper triangle” of the matrrix

not the full matrix
meanfdvec

mean framewise displacement

vector with size = [ndirs,1]
ntpointvec

number of time points

after censoring frames with excessive motion

vector with size = [ndirs,1]
ndirs

number of input directories

matches lengths of participant_id and session_id in vol_info
nnet number of networks
network_names

cell array of network names with length nnet

i.e., Auditory

   CinguloOperc

   CinguloParietal

   Default

   DorsalAttn

   FrontoParietal

   None

   RetrosplenialTemporal

   SMhand

   SMmouth

   Salience

   VentralAttn

   Visual
nroi number of subcortical ROIs
roinames

cell array of ROI names with length nroi

e.g., Left-Cerebral-White-Matter

    Left-Cerebral-Cortex

    Left-Lateral-Ventricle

   …
nnodes

number of nodes (network-ROI pairs)

equal to nnet x nroi
net1vec specifies network number (index to network_names) for first network of each node of correlation matrix
roi2vec specifies ROI number (index to roinames) for second ROI of each node of correlation matrix

Contents of {task}\_tseries and {task}\_{atlas}\_aseg_tseries.mat

Variable name Description
datamat_tsdata

ROI time series data

with size = [ndirs,nruns,ntpoints,nroi]
datamat_motion

motion time series data

with size = [ndirs,nruns,ntpoints,6]

6 motion parameters: dx, dy, dz, rx, ry, rz

notch-filtered to remove respiratory signals
censvec

censor vector

with size = [ndirs,nruns,ntpoints]

value of 1 indicates time point was censored

due to excessive motion
ndirs

number of input directories

matches lengths of participant_id and session_id in vol_info
nroi number of ROIs
roinames

cell array of ROI names with length nroi

e.g., ctx-lh-bankssts

    ctx-lh-caudalanteriorcingulate

    ctx-lh-caudalmiddlefrontal

    …

    ctx_lh_G_and_S_frontomargin

    ctx_lh_G_and_S_occipital_inf

    ctx_lh_G_and_S_paracentral

   …

    ctx-lh-gp1

    ctx-lh-gp2

    ctx-lh-gp3

   …

    Left-Cerebral-White-Matter

    Left-Cerebral-Cortex

    Left-Lateral-Ventricle

   …
nruns number of imaging runs (scans)
ntpoints number of time points per run

Notes:

  • Concatenated ROI time series data is provided in a relatively raw state, without pre-residualization or filtering.
  • ROI time series data is derived from the minimally processed data with these steps:
    • Removal of initial “dummy” volumes
    • Sampling of time series data to cortical surface
    • Calculation of average ROI times series for cortical and subcortical ROIs

Example: extract time series data for one run of one visit

x = find(strcmp(visitidvec,’ S042_INVXXXXXXXX_baseline’))

r = 2

tmp_censvec = censvec(x,r,:);

tmp_tsdata =  datamat_tsdata(x,r,:,:);

References

Hagler, Donald J., SeanN. Hatton, M. Daniela Cornejo, Carolina Makowski, Damien A. Fair, Anthony Steven Dick, Matthew T. Sutherland, et al. 2019. NeuroImage 202 (November): 116091. doi:10.1016/j.neuroimage.2019.116091.
Holland, H, and A Dale. 2011. Medical Image Analysis 15 (4): 489–97. doi:10.1016/j.media.2011.02.005.
Najdenovska, Elena, Yasser Alemán-Gómez, Giovanni Battistella, Maxime Descoteaux, Patric Hagmann, Sebastien Jacquemont, Philippe Maeder, Jean-Philippe Thiran, Eleonora Fornari, and Meritxell Bach Cuadra. 2018. Scientific Data 5 (1). Nature Publishing Group: 180270. doi:10.1038/sdata.2018.270.
Pauli, Wolfgang M., Amanda N. Nili, and J. Michael Tyszka. 2018. Scientific Data 5 (1). Nature Publishing Group: 180063. doi:10.1038/sdata.2018.63.