[Nipy-devel] Fwd: 1st/2nd level analysis modules in python

Matthew Brett matthew.brett@gmail....
Fri Nov 6 18:07:38 CST 2009


Hi,

Ah - that's much better - specific questions!

> - will load_image on a 4D file return a shape of the form (x,y,z,t) in
> the near term ( referred to as ITNT below) ? or do i always have to
> check the coord_names?

Yes, and no.

Yes, for our standard image, you can depend on the 4th dimension being time.

No, you can't depend on the first 3 dimensions of the array being MNI
xyz, they can be in any order.  This is same as for NIFTI.  You can
work out which of the array dimensions is closest to x, y, z in MNI
space, but of course they may not be very close in an oblique
acquisition.

> - is there some specific notion of what i,j,k means for complex data?
> or is that a bad use of the api? let's say i store k-space data. now
> this doesn't really map directly to xyz. only after a complex
> transform does it map to xyz. this question is more about my ignorance
> and potential for storing different data (e.g.,  parallel acquisition
> k-space data - kx,ky,kslice,ch,t). it seems that the class is flexible
> enough to handle arbitrary names and order. is that correct? and if so
> will this flexibility remain in the base class.

I guess this is an example where a coordinate in the array does not
map to - say - MNI world, but Fourier world.

It's an interesting use case, I'd like to hear from Jonathan and Gael
as to how they'd handle that case.

Can you give an example where you care about this in images, instead
of arrays?   I mean, I guess you want to be able to load these data
from something like the scanner, as arrays I suppose, then save them
as NIFTI or similar, then load them again, and maintain the knowledge
of spatial relationships?

> - is the affine ITNT going to be a mathematically true affine that is
> dependent on the domain and range? or will it change to a 4x4 affine
> that is commonly used in other software?

I didn't really understand the question, but yes you will be able to
get a 4x4 affine with a single method call.

> usage question:
>
> i tried to load a 4D PACE corrected data set, where there is a
> different vox2world mapping matrix associated with each
> volume/timepoint. how do i get hold of each of these? i got the first
> one, but couldn't get the rest.

Where are the vox2world mappings stored in this dataset?

Best,

Matthew



More information about the Nipy-devel mailing list