BASC multiscale atlas#


See nilearn.datasets.fetch_atlas_basc_multiscale_2015.


This work is a derivative from the Cambridge sample found in the 1000 functional connectome project (Liu et al.[1]), originally released under Creative Commons – Attribution Non-Commercial. It includes group brain parcellations generated from resting-state functional magnetic resonance images for about 200 young healthy subjects. Multiple scales (number of networks) are available, and includes 7, 12, 20, 36, 64, 122, 197, 325, 444. The brain parcellations have been generated using a method called bootstrap analysis of stable clusters (BASC, Bellec et al.[2]) and the scales have been selected using a data-driven method called MSTEPS (Bellec[3]).

This release more specifically contains the following files:


a markdown (text) description of the release.

‘scale007’, ‘scale012’, ‘scale020’, ‘scale036’, ‘scale064’, ‘scale122’, ‘scale197’, ‘scale325’, ‘scale444’:

brain_parcellation_cambridge_basc_multiscale_(sym,asym)_scale(NNN).nii.gz: a 3D volume .nii format at 3 mm isotropic resolution, in the MNI non-linear 2009a space. Region number I is filled with Is (background is filled with 0s).

Note that two versions of the template are available, ending with either nii_sym or nii_asym. The asym flavor contains brain images that have been registered in the asymmetric version of the MNI brain template (reflecting that the brain is asymmetric), while with the sym flavor they have been registered in the symmetric version of the MNI template. The symmetric template has been forced to be symmetric anatomically, and is therefore ideally suited to study homotopic functional connections in fMRI: finding homotopic regions simply consists of flipping the x-axis of the template.


The datasets were analysed using the NeuroImaging Analysis Kit (NIAK) version 0.12.14, under CentOS version 6.3 with Octave version 3.8.1 and the Minc toolkit version 0.3.18. Each fMRI dataset was corrected for inter-slice difference in acquisition time and the parameters of a rigid-body motion were estimated for each time frame. Rigid-body motion was estimated within as well as between runs, using the median volume of the first run as a target. The median volume of one selected fMRI run for each subject was coregistered with a T1 individual scan using Minctracc (Collins and Evans[4]), which was itself non-linearly transformed to the Montreal Neurological Institute (MNI) template (Fonov et al.[5]) using the CIVET pipeline (Baldassarre et al.[6]). The MNI symmetric template was generated from the ICBM152 sample of 152 young adults, after 40 iterations of non-linear coregistration. The rigid-body transform, fMRI-to-T1 transform and T1-to-stereotaxic transform were all combined, and the functional volumes were resampled in the MNI space at a 3 mm isotropic resolution. The “scrubbing” method of (Power et al.[7]), was used to remove the volumes with excessive motion (frame displacement greater than 0.5 mm). A minimum number of 60 unscrubbed volumes per run, corresponding to ~180 s of acquisition, was then required for further analysis. The following nuisance parameters were regressed out from the time series at each voxel: slow time drifts (basis of discrete cosines with a 0.01 Hz high-pass cut-off), average signals in conservative masks of the white matter and the lateral ventricles as well as the first principal components (95% energy) of the six rigid-body motion parameters and their squares (Giove et al.[8]). The fMRI volumes were finally spatially smoothed with a 6 mm isotropic Gaussian blurring kernel.

Bootstrap Analysis of Stable Clusters#

Brain parcellations were derived using BASC (Bellec et al.[2]). A region growing algorithm was first applied to reduce the brain into regions of roughly equal size, set to 1000 mm3. The BASC used 100 replications of a hierarchical clustering with Ward’s criterion on resampled individual time series, using circular block bootstrap. A consensus clustering (hierarchical with Ward’s criterion) was generated across all the individual clustering replications pooled together, hence generating group clusters. The generation of group clusters was itself replicated by bootstrapping subjects 500 times, and a (final) consensus clustering (hierarchical Ward’s criterion) was generated on the replicated group clusters. The MSTEPS procedure (Bellec[3]) was implemented to select a data-driven subset of scales in the range 5-500, approximating the group stability matrices up to 5% residual energy, through linear interpolation over selected scales. Note that the number of scales itself was selected by the MSTEPS procedure in a data-driven fashion, and that the number of individual, group and final (consensus) number of clusters were not necessarily identical.