Thanks for these practical tips, Jonathan! As you have done both before,
could I just ask:
Jonathan Peelle wrote:
> Hello,
>
>> 1. How to make a batch file for coregistration + segmentation of all the
>> input T1 images?
>> What I did was: make a batch file for coregistration of 1 T1 image to the
>> MNI-space T1 template, and use a dependency to segment the output of the
>> coregistration. The other input image names were put into the same batch
>> file using a script
>> Somewhere along the way, 'new segment' always crashes. With 'segment' it is
>> (at least, was) possible to finish this step. Is it equivalent to use
>> 'segment' instead and then do the DARTEL import step separately (instead of
>> having the import step built-in with 'new segment')?
>
> I'll have to leave the batching part to someone else. However, just
> to make sure it's clear, using 'segment' + DARTEL import will be
> different than 'new segment' because the tissue probability maps are
> different (and probably more accurate) in 'new segment'. I.e. they
> are equivalent in that you will end up with rc* images, but because
> the segmentation routine isn't identical, you're going to end up with
> different results.
That is very clear - from the viewpoint of accuracy/correctness, by all
means use 'new segment' rather than 'segment'.
What bothers me is the crashing of matlab at random points (proper
crash, exiting matlab incl. stack trace and everything). Does that mean
that coregistraton and 'new segment' in one batch file for all images is
too much for the average computer? My study is 105 subjects x 2 time
points, so not huge as VBM studies go.
>> But is it a good idea to normalise to MNI space? Or is it better to warp the
>> images to the DARTEL-created template instead?
>
> When I've compared results in "DARTEL space" with those same results
> put into MNI space they are extremely similar (as you'd expect).
> Because having your results in MNI space greatly facilitates the
> sharing of data and reporting of results, I think it's definitely a
> good idea. (Note that normalizing to MNI uses the flow fields
> generated during template creation, along with the parameters required
> to get the template to MNI space, to bring each subject into MNI
> space. So you're still gaining a lot of benefit from doing this
> through DARTEL.)
Ah, so the procedure described in the VBM tutorial implicitly still goes
via the DARTEL-space?
Best wishes
Alle Meije
|