[MINC-users] mincresample along a dimension with *-width variables
Andrew Janke
a.janke at gmail.com
Thu Feb 6 18:23:12 EST 2014
Hi all,
As others have commented, the width parameters are typically for data
with a time dimension, but there are use cases for irregular spacing
in an image dimension. I have used it for microscopy data in which we
dynamically change a vibratome slice thickness from 20um -> 80um
depending if we are in an "interesting" part of the tissue or not.
Peter, Bert and I had a long discussion about the use case that Alex
mentioned of capturing slice thickness and gap. This used to be common
in MRI acquisition to avoid slice cross talk but has (thankfully)
faded from use with interleaved sequences. I argued strongly that we
shouldn't support this type of acquisition as it would introduce too
many special cases, what for example do you do when you resample? you
can no longer capture the gaps. Peter as I recall agreed so Bert
didin't get to vote!
History out of the way, I am surprised that your volume has both slice
widths and has a regular__ spacing type. These from my understanding
should be mutually exclusive of each other and I'm guessing is what is
causing the problem.
Why also the use of mincresample to add a few slices? I'd generally do
this (no interpolation needed).
mincreshape -dimrange zspace=-1,49 in.mnc out.mnc
Do your dimension widths survive the above? I'd suspect not given the
regular__ type. Where did this file originate from?
a
On 7 February 2014 06:00, Alex Zijdenbos <zijdenbos at gmail.com> wrote:
> In my conceptual understanding, the -width attributes allow you to separate
> the "thickness" from the sample step size. Indeed most commonly used along
> a time dimension; but I could imagine an image volume being a stack of 3mm
> thick slices with a slice gap of 1mm; making the step size 4mm (a lot of
> MRI data used to be acquired like that). But I doubt that in reality this
> is used much, and I have no idea whether something like mincresample
> actually uses this information during resampling (I suspect not) - but it
> seems to me that it shouldn't break, either.
>
> -- A
>
>
> On Thu, Feb 6, 2014 at 1:17 PM, Andrew Wood <andrew at biospective.com> wrote:
>
>> Are they ever used for spatial dimensions? It was actually a mistake that
>> it ended up in my affected volumes.
>>
>>
>> On Thu, Feb 6, 2014 at 12:31 PM, Vladimir S. FONOV <
>> vladimir.fonov at gmail.com
>> > wrote:
>>
>> > I've seen them used in dynamic acquisition where they are used with
>> > time dimension, for example in PET data
>> >
>> > On Thu, Feb 6, 2014 at 12:25 PM, Andrew Wood <andrew at biospective.com>
>> > wrote:
>> > > HI All,
>> > >
>> > > I ran into a issue when resampling a volume along a dimension which has
>> > > *-width attribute values. It seems that such a dimension can't be
>> > resampled
>> > > to have a different number of elements.
>> > >
>> > > An example input volume:
>> > > $ mincinfo in.mnc
>> > > file: in.mnc
>> > > image: signed__ short -32768 to 32767
>> > > image dimensions: zspace yspace xspace
>> > > dimension name length step start
>> > > -------------- ------ ---- -----
>> > > zspace 47 3.375 -1027.16
>> > > yspace 128 -2.65336 363.575
>> > > xspace 128 -2.65336 171.012
>> > >
>> > > $ mincheader in.mnc | grep zspace-width
>> > > double zspace-width(zspace) ;
>> > > zspace-width:dimorder = "zspace" ;
>> > > zspace-width:varid = "MINC standard variable" ;
>> > > zspace-width:vartype = "dim-width____" ;
>> > > zspace-width:version = "MINC Version 1.0" ;
>> > > zspace-width:filtertype = "square____" ;
>> > > zspace-width:spacing = "regular__" ;
>> > > zspace-width:width = -3.375 ;
>> > > zspace-width = -3.375, -3.375, -3.375, -3.375, -3.375, -3.375, -3.375,
>> > >
>> > > Now, add a voxel to each dimension:
>> > > $ mincresample -clobber in.mnc out.mnc -nelements 129 129 48
>> > > (from miattputstr): Function 'dimrename' not implemented
>> > >
>> > > $ mincinfo out.mnc
>> > > file: out.mnc
>> > > image: signed__ short -32768 to 32767
>> > > image dimensions: zspace yspace xspace
>> > > dimension name length step start
>> > > -------------- ------ ---- -----
>> > > zspace 47 3.375 -1027.16
>> > > yspace 129 -2.65336 363.575
>> > > xspace 129 -2.65336 171.012
>> > >
>> > >
>> > > I dug around the mincresample source code, and the the first thing done
>> > > when creating the output volume is a copy of all variable definitions
>> > from
>> > > the input, excluding {x,y,z}space, image, and image-{min,max}. It
>> appears
>> > > that copying the zspace-width actually forces zspace to be copied,
>> > > resulting in a naming conflict between the original zspace and our new
>> > > zspace. To work around this, it tries to rename the copied zspace to
>> > > zspace0, which is where we see the "Function 'dimrename' not
>> implemented"
>> > > error message. The result is being stuck with the size of the original
>> > > zspace.
>> > >
>> > > The problem goes away if {x,y,z}space-width is added to the exclusion
>> > list
>> > > for the original variable copy operation. I'm having trouble imagining
>> > when
>> > > those *-width variables are used though, so I'm not sure if this'd be a
>> > > sane fix.
>> > >
>> > > Does anyone have any thoughts that might clear up this situation?
>> > >
>> > > Thanks,
>> > > Andrew
>> > > _______________________________________________
>> > > MINC-users at bic.mni.mcgill.ca
>> > > http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users
>> >
>> >
>> >
>> > --
>> > Best regards,
>> >
>> > Vladimir S. Fonov ~ vladimir <dot> fonov <at> gmail <dot> com
>> > _______________________________________________
>> > MINC-users at bic.mni.mcgill.ca
>> > http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users
>> >
>> _______________________________________________
>> MINC-users at bic.mni.mcgill.ca
>> http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users
>>
>>
> _______________________________________________
> MINC-users at bic.mni.mcgill.ca
> http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users
More information about the MINC-users
mailing list