[pymvpa] Searchlight on a vstack-ed dataset: Which imghdr to use for niftiresults?

[Usman Sheikh]

Hi PyMVPA gurus,



This question more about nifiresults than it is about searchlight.




Specifically, it's about the image header in this command:




niftiresults = map2nifti(sl_map, imghdr=dataset.a.imghdr)




Let's say I create a big dataset from three of a subject's sessions, using vstack, while setting 'a' argument of the function to 'all'. This will return me a dataset with imghdr as a tuple of all three imghdrs (and imgtype and mapper and all others in the same format as well). Now if I run searchlight on this stacked dataset, which imghdr should I use to get the corresponding accuracy map in nifti format? And what problems I might face if I just select the first of all the tuples i.e. imgtype[0] and mapper[0] and so on?

Thanks in advance,
Usman Ayub Sheikh
Predoctoral Researcher
www.bcbl.eu

Legal disclaimer/Aviso legal/Lege-oharra: www.bcbl.eu/legal-disclaimer

[Yaroslav Halchenko]

Post by Usman Sheikh
Hi PyMVPA gurus,
This question more about nifiresults than it is about searchlight.
niftiresults = map2nifti(sl_map, imghdr=dataset.a.imghdr)
Let's say I create a big dataset from three of a subject's sessions, using
vstack, while setting 'a' argument of the function to 'all'. This will
return me a dataset with imghdr as a tuple of all three imghdrs (and
imgtype and mapper and all others in the same format as well). Now if I
run searchlight on this stacked dataset, which imghdr should I use to get
the corresponding accuracy map in nifti format? And what problems I might
face if I just select the first of all the tuples i.e. imgtype[0] and
mapper[0] and so on?

well -- it is your decision to make I guess ;)
What if they are all different etc. If you aligned them all into the
same space, then you would be the one to know that it is "kosher" to
take [0]th one

[Usman Sheikh]

Hi Mr. Yaroslav,

Thanks for your email. Sorry for an incomplete question. Yes, all the other session's functional images are already aligned to the first one. And this is exactly why I was biased towards using the first one's imghdr for the niftiresults.

Thanks,
Usman Ayub Sheikh
Predoctoral Researcher
www.bcbl.eu

Legal disclaimer/Aviso legal/Lege-oharra: www.bcbl.eu/legal-disclaimer

----- Original Message -----
From: "Yaroslav Halchenko" <***@onerussian.com>
To: "pkg-exppsy-pymvpa" <pkg-exppsy-***@lists.alioth.debian.org>
Sent: Friday, November 24, 2017 5:47:43 PM
Subject: Re: [pymvpa] Searchlight on a vstack-ed dataset: Which imghdr to use for niftiresults?

Post by Usman Sheikh
Hi PyMVPA gurus,
This question more about nifiresults than it is about searchlight.
niftiresults = map2nifti(sl_map, imghdr=dataset.a.imghdr)
Let's say I create a big dataset from three of a subject's sessions, using
vstack, while setting 'a' argument of the function to 'all'. This will
return me a dataset with imghdr as a tuple of all three imghdrs (and
imgtype and mapper and all others in the same format as well). Now if I
run searchlight on this stacked dataset, which imghdr should I use to get
the corresponding accuracy map in nifti format? And what problems I might
face if I just select the first of all the tuples i.e. imgtype[0] and
mapper[0] and so on?

well -- it is your decision to make I guess ;)
What if they are all different etc. If you aligned them all into the
same space, then you would be the one to know that it is "kosher" to
take [0]th one

[Yaroslav Halchenko]

Post by Usman Sheikh
Hi Mr. Yaroslav,
Thanks for your email. Sorry for an incomplete question. Yes, all the other session's functional images are already aligned to the first one. And this is exactly why I was biased towards using the first one's imghdr for the niftiresults.

if they were correctly aligned, then all the headers etc should be
identical... paranoid me would have first verified that (in the
script) and then chose the first (as good as any) one