Hi all. I thought I should mention another batch processing tool that
might be of use. As part of a separate project (VoxBo,
www.voxbo.org), we have an engine for distributed processing of fMRI
data. We've been using it for cluster-based distributed processing of
fMRI data for the past six years, but until recently no one had used
it with SPM. (It's a fairly generic piece of software, but in the
past we've only written config files for our own software as yet.)
Over the past week or so, I've started writing config files for
various SPM2 preprocessing steps, so that our users can integrate them
into their preprocessing. Only one (spatial normalization) is done so
far, but I'm hoping to have a reasonably rich set fairly soon, and it
would certainly achieve most of the goals of a good batch mode --
automation, ease of tweaking, accounting, etc. Plus parallel
processing, if you have more than one CPU to devote to the task. It
could potentially be combined with some of the other batch tools folks
have been developing, or used independently.
The system has some rough edges -- it's mildly tricky to install and
configure, and the manual is severely outdated, which in practice
means it's easiest to just have us help. But it also has some
distinct advantages, most notably the facility for scheduling
processing jobs in parallel across a cluster, which can make
processing and re-processing large datasets a lot more bearable.
Also, it can be readily used to integrate disparate packages, whether
or not they're MATLAB-based.
Obviously this will be of limited use to SPM users until I have a
complete set of config files for SPM2 (help with this gratefully
accepted, of course). They're not hard to write -- the first one is
the hardest, and I've done that -- and even easier to customize if
needed. So at least for preprocessing I think we should have a solid
set soon, and I imagine it would be fairly helpful if you have a few
CPUs to throw at your data. I'll make an announcement when I think
it's likely to be more broadly useful, but if anyone would like to be
on the bleeding edge with this, please drop me a line. Note that aAt
the moment, it's Linux-only (most of the code is close for OSX,
Solaris, and Cygwin).
dan
|