The Mindboggle software package automates shape analysis of anatomical labels and features extracted from human brain magnetic resonance image data. Mindboggle can be run as a single command, and can be installed as a cross-platform virtual machine for convenience and reproducibility of results. Behind the scenes, open source Python 3 and C++ code run within a modular Nipype pipeline framework on Linux (tested with Python 3.5.1 on Ubuntu 14.04).

Date:August 11, 2016


Installing Mindboggle

Mindboggle comes as a single installation script,, that may be directly called to install Mindboggle on a Linux machine ($ source However, for reasons of convenience and reproducibility of results, we recommend using a different script, install_mindboggle_vm, to perform the same installation on Linux, MacOSX, or Windows, but in a virtual machine (VM). Download this script to wherever you want the VM startup directory to be, and do the following (type commands in a terminal for steps 2 and 3).

  1. Install VM dependencies:

    Vagrant manages virtual machines.

    Vagrant provides reproducible and portable work environments that isolate dependencies and their configuration within a single disposable, consistent environment that can run on Linux, MacOSX, or Windows.

    Virtualbox provides virtual machines used by Vagrant.

2. Download and configure the virtual machine to access your local brain image (usually FreeSurfer output) data by typing the following in the same directory as the VM script. This generates a configuration file called “Vagrantfile”:

python install_mindboggle_vm

For help with more options, such as how to mount your local ANTs data directory, set the number of processors, etc., add “-h” to the above:

python install_mindboggle_vm -h

3. Henceforth, whenever running Mindboggle, first type the following in the same directory as the Vagrantfile:

vagrant up
vagrant ssh

Running Mindboggle

To run Mindboggle, you must first preprocess brain MR image data (see Preprocessing below). To get up and running with the following examples, download and unzip the directory (450 MB), which contains some example preprocessed data. More example input and output data can be found on Mindboggle’s examples site.

If using the Mindboggle virtual machine, type the commands in step 3 of Installing Mindboggle above to enter the VM before continuing.

For help after installing, type the following in a terminal window:

mindboggle -h

Example 1: The following bare-bones command runs Mindboggle on data processed by FreeSurfer but not ANTs:

mindboggle $HOME/mindboggle_input_example/freesurfer/subjects/arno

Example 2: The same command, but takes advantage of ANTs output (backslash denotes line return):

mindboggle $HOME/mindboggle_input_example/freesurfer/subjects/arno \
--ants $HOME/mindboggle_input_example/ants/subjects/arno/antsBrainSegmentation.nii.gz

Example 3: To generate only volume (and not surface) labels and shape measures from FreeSurfer data, using 8 processors:

mindboggle $HOME/mindboggle_input_example/freesurfer/subjects/arno --no_surfaces -p 8


As you may have inferred from the “Running Mindboggle” examples and example data above, Mindboggle currently takes output from FreeSurfer (v6 or higher recommended) and optionally from ANTs (v2.1.0rc3 or higher recommended). Example input data can be found on Mindboggle’s examples site on

FreeSurfer generates labeled cortical surfaces, and labeled cortical and noncortical volumes. Run recon-all on a T1-weighted $IMAGE file (e.g., subject1.nii.gz) and set the output $SUBJECT name (e.g., subject1):

recon-all -all -i $IMAGE -s $SUBJECT

ANTs provides brain volume extraction, segmentation, and registration-based labeling. To generate the ANTs transforms and segmentation files used by Mindboggle, run the script on the same $IMAGE file, set an output $PREFIX, and provide paths to the OASIS-30 Atropos template files (backslash denotes a line return): -d 3 -a $IMAGE -o $PREFIX \
  -e OASIS-30_Atropos_template/T_template0.nii.gz \
  -t OASIS-30_Atropos_template/T_template0_BrainCerebellum.nii.gz \
  -m OASIS-30_Atropos_template/T_template0_BrainCerebellumProbabilityMask.nii.gz \
  -f OASIS-30_Atropos_template/T_template0_BrainCerebellumExtractionMask.nii.gz \
  -p OASIS-30_Atropos_template/Priors2/priors%d.nii.gz

Processing steps

The following steps are performed by Mindboggle (with links to code on GitHub):

  1. Create hybrid gray/white segmentation from FreeSurfer and ANTs output (combine_2labels_in_2volumes).

  2. Fill hybrid segmentation with FreeSurfer- or ANTs-registered labels.

  3. Compute volume shape measures for each labeled region:

  4. Compute surface shape measures for every cortical mesh vertex:

  5. Extract cortical surface features:

  6. For each cortical surface label/sulcus, compute:

  7. Compute statistics (stats_per_label in for each shape measure in #4 for each label/feature:

    • median
    • median absolute deviation
    • mean
    • standard deviation
    • skew
    • kurtosis
    • lower quartile
    • upper quartile


Example output data can be found on Mindboggle’s examples site on By default, output files are saved in $HOME/mindboggled/SUBJECT, where $HOME is the home directory and SUBJECT is a name representing the person’s brain that has been scanned. Volume files are in NIfTI format, surface meshes in VTK format, and tables are comma-delimited. Each file contains integers that correspond to anatomical labels or features (0-24 for sulci). All output data are in the original subject’s space. The following include outputs from most, but not all, optional arguments.

Folder Contents Format
labels/ number-labeled surfaces and volumes .vtk, .nii.gz
features/ surfaces with features: sulci, fundi .vtk
shapes/ surfaces with shape measures (per vertex) .vtk
tables/ tables of shape measures (per label/feature/vertex) .csv

mindboggled / $SUBJECT /

labels /

freesurfer_wmparc_labels_in_hybrid_graywhite.nii.gz: hybrid segmentation filled with FS labels

ants_labels_in_hybrid_graywhite.nii.gz: hybrid segmentation filled with ANTs + FS cerebellar labels

[left,right]_cortical_surface / freesurfer_cortex_labels.vtk: DKT cortical surface labels

features / [left,right]_cortical_surface /

folds.vtk: (unidentified) depth-based folds

sulci.vtk: sulci defined by DKT label pairs in depth-based folds

fundus_per_sulcus.vtk: fundus curve per sulcus – UNDER EVALUATION –

cortex_in_MNI152_space.vtk: cortical surfaces aligned to an MNI152 template

shapes / [left,right]_cortical_surface /

area.vtk: per-vertex surface area

mean_curvature.vtk: per-vertex mean curvature

geodesic_depth.vtk: per-vertex geodesic depth

travel_depth.vtk: per-vertex travel depth

freesurfer_curvature.vtk: FS curvature files converted to VTK

freesurfer_sulc.vtk: FS sulc (convexity) files converted to VTK

freesurfer_thickness.vtk: FS thickness files converted to VTK

tables /

volume_per_freesurfer_label.csv: volume per FS label

volumes_per_ants_label.csv: volume per ANTs label

thickinthehead_per_freesurfer_cortex_label.csv: FS cortex label thickness

thickinthehead_per_ants_cortex_label.csv: ANTs cortex label thickness

[left,right]_cortical_surface /

label_shapes.csv: per-label surface shape statistics

sulcus_shapes.csv: per-sulcus surface shape statistics

fundus_shapes.csv: per-fundus surface shape statistics – UNDER EVALUATION –

vertices.csv: per-vertex surface shape statistics