Whole brain atlases generated by parcellation.
Available below are atlases that have been constructed from various datasets and preprocessing strategies.
- Craddock 2011 parcellations [Download]
- ADHD 200 parcellations [Download]
Craddock 2011 parcellations
These are the parcellations from the Craddock et. al 2011 publication “A whole brain fMRI atlas generated via spatially constrained spectral clustering”. These atlases were generated using a variety of parcellation strategies applied to a dataset of 41 healthy controls.
Data acquisition
Forty-one healthy control participants (age: 18–55; mean 31.2; std. dev. 7.8; 19 females) were recruited in accordance with Emory University Institutional Review Board policy. To qualify for inclusion, participants were required to be between the ages of 18–65, have no contraindications for MRI, to be medication free, and have no current or past neurological or psychiatric conditions. Subjects were scanned on a 3.0 T Siemens Magnetom TIM Trio scanner (Siemens Medical Solutions; Malvern PA) using a 12-channel head matrix coil. Anatomic images were acquired at 1 x 1 x 1 mm3 resolution with an MPRAGE sequence using: field of view (FOV) 224 x 256 x 176 mm3, repetition time (TR) 2,600 ms, echo time (TE) 3.02 ms, inversion time (TI) 900 ms, flip angle (FA) 8º, and GRAPPA factor 2. Resting state fMRI data were acquired with the Z-SAGA sequence to minimize susceptibility artifacts. One hundred and fifty functional volumes were acquired in 30 4-mm axial slices using the parameters: TR 2920 ms, TE1/TE2 30/66 ms, FA 90º,64x64 matrix, in-plane resolution 3.44 x 3.44 mm2.
Preprocessing
All preprocessing of MRI data was performed using SPM5 running in MATLAB 2008a. Anatomic and fMRI data were evaluated for imaging artifacts such as excessive ghosting, banding, and other imaging errors; no images were removed. Anatomic scans were simultaneously segmented into white matter (WM), gray matter (GM), and cerebral- spinal fluid (CSF) and normalized to the MNI152 normal- ized brain atlas using SPM5’s unified segmentation procedure. fMRI volumes were slice timing corrected, motion corrected, written into MNI152 space at 4 x 4 x 4mm3 resolution using the trans- formation calculated on the corresponding anatomic images, and spatially smoothed using a 6-mm FWHM Gaussian kernel. No images were removed due to excessive head motion (motion 0.2–1.82 mm; mean 0.84 mm; std. dev. 0.41 mm). fMRI data were restricted to GM and denoised by regressing out motion parameters, WM timecourse, as well as CSF timecourse. Each voxel timecourse was band-pass filtered (0.009 Hz < f < 0.08 Hz) to remove frequencies not implicated in resting state FC.
Clustering
Clustering was accomplished using the spatially-constrained normalized-cut spectral clustering algorithm with various different similarity metrics and group-level clustering schemes. The similarity metrics used were temporal correlation between voxel time-courses (tcorr), spatial correlation between whole-brain functional connectivity maps generated from voxel time-courses (scorr), and based exclusively on whether or not the voxels are touching (ones). Group level clustering was accomplished by either a two-level scheme in which the data of each participant are clustered seperately and the results combined for a group level clustering (2level) and a scheme were the participant level similarity matrices were averaged and then submitted to clustering (groupmean). Clustering was performed for K between 10 and 300 in steps of 10, and from 350 to 1000 in steps of 50.
The files
The download includes five ROI atlases, one Each nifti file contains 44 different volumes (sub-briks in AFNI lingo) which correspond to the level of clustering. Each volume contains the cluster maps, in which each voxel contains the integer value of the cluster to which it belongs. Do to the clustering method, it is very likely that the results of a clustering will result in fewer clusters then was requested, i.e. a 1000-level clustering might result in 937 clusters.
The correspondance between clustering number (K) and volume number (#) is listed below:
| # | K | # | K | # | K | # | K | # | K | # | K | |||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | 10 | 9 | 90 | 17 | 170 | 25 | 250 | 33 | 450 | 41 | 850 | |||||
| 2 | 20 | 10 | 100 | 18 | 180 | 26 | 260 | 34 | 500 | 42 | 900 | |||||
| 3 | 30 | 11 | 110 | 19 | 190 | 27 | 270 | 35 | 550 | 43 | 950 | |||||
| 4 | 40 | 12 | 120 | 20 | 200 | 28 | 280 | 36 | 600 | 44 | 1000 | |||||
| 5 | 50 | 13 | 130 | 21 | 210 | 29 | 290 | 37 | 650 | |||||||
| 6 | 60 | 14 | 140 | 22 | 220 | 30 | 300 | 38 | 700 | |||||||
| 7 | 70 | 15 | 150 | 23 | 230 | 31 | 350 | 39 | 750 | |||||||
| 8 | 80 | 16 | 160 | 24 | 240 | 32 | 400 | 40 | 800 |
ADHD 200 parcellations
These are the parcellations from the Athena Pipeline of the ADHD 200 preprocessing initiative. 200 and 400 ROI atlases were generated using 2-level parcellation of 650 individuals from the ADHD 200 Sample.
Data acquisition
The ADHD 200 Sample is an open science initiative that is sharing resting state fMRI and structural MRI data from 383 individuals suffering from ADHD and 598 typically developing controls (Ages 7-21 years old). The data was aggregated from eight sites each of which had permission from their local ethics boards to acquire and share the data. All data were anonymized prior to sharing, which included defacing structural images and randomizing the patient identifiers. The Neuro Bureau’s ADHD 200 Preprocessing Initiative systematically preprocessed the ADHD 200 Sample using three different preprocessing pipeline. ROI atlases containing 200 and 400 regions were calculated by the Athena Pipeline from the data of 650 individuals with the cleanest data as determined by registration quality and applying a max translation < 3mm and max rotation < 3 degree threshold to head motion estimates.
Preprocessing
Preprocessing was performed using the Athena Pipeline from the Neuro Bureau’s ADHD 200 Preprocessing Initiative. The first 4 EPI volumes were removed from each resting state fMRI scan to ensure that the imaging signal had reached T1 equilibrium. The data were then corrected for slice timing and realigned to correct for head motion. The mean EPI volume was registered to the corresponding structural image using a six degree of freedom affine transformation. A non-linear registration was calculated from the anatomical to the MNI152 template, which was subsequently combined with the fMRI-Anatomical transform to create a transformation from EPI to MNI space. WM and CSF signals were calculated from the fMRI data using masks derived by segmenting the corresponding anatomical image. fMRI data were denoised by regressing WM and CSF signals, a low order polynomial model of signal drift, and a six regressor model of head motion from each voxel’s time series. De-noised data was written into MNI space at 4x4x4 mm resolution and spatially smoothed using a 6-mm FHWM Gaussian filter.
Clustering
Functional parcellation was accomplished using the two-stage spatially-constrained functional procedure. A grey matter mask was constructed for the subjects by averaging grey matter masks derived from freesurfer automated segmentation. Subject specific connectivity graphs were constructed by treating each withen-gm-mask voxel as a node and edges corresponding to super-threshold temporal correlations to the voxels’ 3D (27 voxel) neighborhood. Each subjects’ graph was partitioned into 200 or 400 regions using normalized cut spectral clustering. Association matrices were constructed from the clustering results by setting the connectivity between voxels to 1 if they are in the same ROI and 0 otherwise. A group-level correspondance matrix was constructed by averaging the individual level association matrices and subsequently partitioned into 200 or 400 regions using normalized cut clustering.