fMRI Independent Component Analysis (ICA)

The CanICA interface neuro_pypes.interfaces.CanICAInterface can be used to perform Independent-Component Analysis (ICA) on fMRI images. It uses the CanICA and the DictLearning implementation in NiLearn. You can choose which implementation to use through the canica.algorithm setting.

The helper functions to attach this interface to a workflow are in [neuro_pypes.] There is one version for one functional image, in attach_canica and another for a group ICA (GICA) in attach_concat_canica. However, probably the GICA approach should be further tested on real data.

It depends on the RS-fMRI pipeline. This is implemented in neuro_pypes.postproc.decompose.

# INDEPENDENT COMPONENTS ANALYSIS
## True to perform CanICA
rest_preproc.canica: False

# CanICA settings
canica.algorithm: 'canica' # choices: 'canica', 'dictlearning'
canica.do_cca: True
canica.standardize: True
canica.n_components: 20
canica.threshold: 2.0
canica.smoothing_fwhm: 8
#canica.random_state: 0

canica_extra.plot: True
canica_extra.plot_thr: 2.0 # used if threshold is not set in the ICA

See also