Library and command-line tool to calculate the Vogt-Bailey index of a dataset
Vogt-Bailey index toolbox in Python
It is possible to simply copy the folder vb_toobox to your project folder and proceed from there. If this is the case, be sure you have the following packages installed
multiprocess nibabel numpy scipy
The preferred way to install is through pip. It is as easy as
pip install VBIndex
If your pip is properly configured, you can now use the program
your command line, and import any of the submodules in the
vb_toolbox in your python
If VBIndex was installed via
pip, the command line program
be available in your terminal. You can test if the program is correctly
installed by typing
in your terminal. If you see the following output, the program has been properly installed.
usage: app.py [-h] [-j N] [-n norm] [-fb] [-m file] [-c file] -s file -d file -o file Calculate the Vogt-Bailey index of a dataset. For more information, check https://github.com/VBIndex/py_vbindex. optional arguments: -h, --help show this help message and exit -j N, --jobs N Maximum number of jobs to be used. If abscent, one job per CPU will be spawned -n norm, --norm norm Laplacian normalization to be used. Defaults to unnorm -fb, --full-brain Calculate full brain spectral reordering. -m file, --mask file File containing the labels to identify the cortex, rather than the medial brain structures. This flag must be set for normal analyses and full brain analyses. -c file, --clusters file File containing the surface clusters. Cluster with index 0 are expected to denote the medial brain structures and will be ignored. required named arguments: -s file, --surface file File containing the surface mesh -d file, --data file File containing the data over the surface -o file, --output file Base name for the output files
If you copied the program source code, the executable is found in
You can test the program using
which should yield the results shown above.
There are three main uses for the
- Searchlight analyses
- Whole brain gradient maps
- Gradient maps in a specified set of regions of interest
The per vertex analyses can be carried with the following command
vb_tool --surface input_data/surface.surf.gii --data input_data/data.func.gii --mask input_data/cortical_mask.shape.gii --output search_light
The number of vertices in the surface mesh must match the number of entries in the data and in the mask.
The cortical mask must contain a logical array, with
True values in the
region on which the analyses will be carried out, and
False in the regions to
be left out. This is most commonly used to mask out midbrain structures which
would otherwise influence the analysis of the cortical regions.
Whole brain analyses
To perform full brain analyses, the flag
--full-brain must be set.
Otherwise, the flags are the same as in the searchlight analysis.
vb_tool --surface input_data/surface.surf.gii --data input_data/data.func.gii --mask input_data/cortical_mask.shape.gii --full-brain --output full_brain_gradient
Be warned, however, that this analysis can take long, use a large amount of RAM. In systems with 32k vertices, upwards of 30GB of RAM were used.
Regions of Interest analyses
Sometimes, one is interested only in a small set of ROIs. In this case, the way for calling the program changes slightly,
vb_tool --surface input_data/surface.surf.gii --data input_data/data.func.gii -c input_data/clusters.shape.gii --output clustered_analyses
The cluster file works similarly to the cortical mask in the previous modalities. However, its structure is slightly different. Instead of an array of logical values, the file must contain an array of integers, where each integer corresponds to a different cluster. The 0th cluster is special, and denotes an area which will not be analyzed. In these regards, it has a similar use to the cortical mask.
Notes on parallelism
vb_tool uses a high level of parallelism. How many threads are spawned by
vb_tool itself can be controlled using the
-j/--jobs flag. By default, it
will try to use all the CPUs in your computer at the same time to perform the
analyzes. Depending on the BLAS installation in your computer, this might not
be the best fastest approach, but rarely will be the slowest. If you are
unsure, leave the number of jobs at the default level.
Due to job structure of the
vb_tool, the level of parallelism it can achieve
on its own depends on the specific analyses being carried out.
- Searchlight analyses: High level of parallelism. Will spawn as many jobs are there are CPUs
- Whole brain analyses: Low lever of parallelism. Will only spawn one job
- Region of Interest analyses: Medium level of parallelism. Will spawn as many jobs as there are ROIs, or number of CPUS, whichever is the lowest.
Specially in the whole brain analyses, having a well optimized BLAS installation will grandly accelerate the process, and allow for a further paralelism. Both MKL and OpenBLAS have been shown to offer fast analyses. If you are using the Anaconda distribution, you will have a good BLAS pre-configured.
We use setuptool and wheel to build the distribution code. The process is described next. More information can be found here.
- Be sure that setuptools, twine, and wheel are up-to-dated
python3 -m pip install --user --upgrade setuptools wheel twine
- Run the build command
python3 setup.py sdist bdist_wheel
- Upload the package to pip
python3 -m twine upload dist/*
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
|Filename, size||File type||Python version||Upload date||Hashes|
|Filename, size VBIndex-0.0.4-py3-none-any.whl (24.2 kB)||File type Wheel||Python version py3||Upload date||Hashes View|
|Filename, size VBIndex-0.0.4.tar.gz (10.2 kB)||File type Source||Python version None||Upload date||Hashes View|