[pymvpa] surface-based analysis in PyMVPA.

Michael Hanke mih at debian.org
Tue Jul 5 15:14:52 UTC 2011


On Tue, Jul 05, 2011 at 07:59:10PM +0800, Zhen Zonglei wrote:
> Dear PyMVPAer:
> I have two questions about surface-based analysis in PyMVPA.
> 1.      Can the dataset mapper of PyMVPA handle surface-based data
> automatically?

As of today, there is no dedicated mapper for surface data. However, in
most cases (if just access to the data is required) that is not
necessary, because surface data is essentially already a 2D matrix that
can be wrapped in a dataset right away -- no need for mapping.

> 2.      Is there a convenient way to do surface-based searchlight analysis
> in PyMVPA?

Well, no -- or not entirely. The missing piece is a function that
enriches the dataset of surface data with geometry information for the
nodes. Like for the volumetric data, we need a coordinate per feature
(or matrix column) to feed to the distance computing/checking machinery.

I think the pieces are or ready to be stitched together. NiBabel has
support for GIFTI data, PyMVPA interfaces to NiBabel, PyMVPA as a
generic-enough way to generate searchlight ROIs on arbitrary
dimensions/spaces -- it just needs some glue.

I'd expect that sometime in the second half of this years, one of us
will implement that, but anyone is welcome to contribute that code!


Michael Hanke

More information about the Pkg-ExpPsy-PyMVPA mailing list