Implement a new analyzer: spatial ICA

Spatial ICA is a widely used technique to analyze MRI data (volumetric time series).
MRI data give information on brain activation patterns with a decent spatial resolution
and the activation maps obtained via sICA could help to identify parts of the brain that
are related to specific tasks.

-----------------------------------------------------------------------------------------------------------------------------
Spatial ICA should be general enough to work with:

• TimeSeriesRegion,
• TimeSeriesSurface and
• TimeSeriesVolume datatypes.

Input dimensions:
..........................

T: total number of time points
L: total number of voxels (spatial unit: could be regions or vertices)

• The spatial three-dimensional data structure is rearranged to a one-dimensional vector where the index will run from 1 to L, where L is the total number of voxels.
• The input will contain T vectors of length L.

Issues:
...........
The performance of ICA algorithms degrades when the number of sources increases.
In the case of sICA, we will have to compute at most T sources. Performance will
degrade as a function of time-points.
We can suggest to temporally subsample the signal.

NOTE: In BOLD signals, the components are ordered as a function of time. Thus sICA could be performed in sliding windows.

Objectives:
................

• extract non-stimulus related components or,
• decompose stimulus related components into various independent subcomponents.