[Nipy-devel] fif IO in Python

Alexandre Gramfort alexandre.gramfort@inria...
Fri Jan 7 09:51:56 CST 2011

> At the risk of restating the obvious, let me just add that I would
> definitely advocate separating these two as much as possible. IMHO,
> all the IO part needs to give you is the raw data in a numpy array +
> all the information in the file you might need for an analysis
> (dimensions, scale, etc.). This should be right at home within
> nibabel.

read raw files in fairly easy to support in nibabel. However starting from
raw files there are many steps that need to be taken before doing
brain imaging with M/EEG. It means that only reading raw files is not
very helpful if you don't provide the MEG tools with it. Once Matti
says yes for BSD licensing I'll fork nibabel.

> The processing part could live better in an MEG-specific context, or
> can always go into something like nitime ;)

I plan to look in the near future at the time frequency tools that are
in nitime. Just being curious, how would you compare nitime with fieldtrip
from TF analysis?

> Are there MEG-specific challenges with separating these two? I mean,
> is processing entangled with IO in some particular way in MEG data, in
> contrast to fMRI data, for example?

Indeed. MNE does all the preprocessing of raw data (artefact rejection,
filtering, averaging etc.) in pure C in a very efficient way. It would probably
be a waste of time now to do this in python.


More information about the Nipy-devel mailing list