Cvrmap

The major difference with this release it that all dependencies to FSL have been removed.
FSL was used for only two things: non-aggressive denoising and time filtering. Both of these steps are simple and are now implemented using standard python libraries.

Another difference is in execution time. Before, the pipeline was resampling the probe signal to the fMRI sampling frequency for each delay and for each voxel. But this does not have to be done for each voxel and can be done once and for all (for each delay) before looping on the whole space. With this simple change in the code structure, duration for the whole pipeline was decreased by a factor of (approx.) 10.

Finally:
- reports are now done using NiPrep Report class
- configuration parameters can be specified using a configuration file (see help for more info)

In this release, we introduce a new feature called vesselsignal. It estimates relative CVR using only fmri data (no physiological data are required).

This is the first stable release, which can be used out-of-the-box to reproduce the results from the shared open dataset at openneuro.org/datasets/ds004604.