[MINC-users] inormalize-2.0.18: Volume and model dimensions do not match!

[Alex Zijdenbos]

Hi Burt,

Well, I am not sure why in your case the behaviour (or at least the
error message) has changed; but I am pretty sure that inormalize never
supported volumes and models with different dimensions, as it works
entirely in the voxel domain. I think the problem (perhaps now in the
past) has been that it silently "worked" anyways, but I think it is
unlikely that you actually got the results you were expecting.

Internally, it will pull out the voxels into a linear array (or
arrays) and then operate on those - so the only way it would actually
work correctly is if *all* input volumes are sampled on the same
lattice - at least in cases where it does some kind of voxel-by-voxel
operation such as medianOfRatios (for ratioOfMedians for instance it
probably won't matter, as it would not need to perform a
voxel-by-voxel operation).

If you can still reproduce this (looks like you are), I would try and
resample the volumes to the same lattice (mincresample -nearest -like)
and re-run the "old" inormalize you have; my guess is that the result
would be different.

Rule of thumb would be to always make sure your volumes are sampled
the same way (and inormalize *should* warn you if they are not, like
it now appears to do).

-- A

On Tue, Apr 13, 2010 at 11:06 AM, Burt Crépeault
<burt.crepeault at crulrg.ulaval.ca> wrote:
> Anyone?
>
> Burt.
>
>
>
> On Tue, Apr 6, 2010 at 08:59, Burt Crépeault <
> burt.crepeault at crulrg.ulaval.ca> wrote:
>
>> Hi all,
>>
>> I've recently built the latest MINC suite for OS X and am currently testing
>> everything. It appears that inormalize's behaviour has changed when using a
>> model file.
>>
>> With version 2.0.09:
>>
>> neon:install:# /opt/MEDICS/EX/etc/minc2/bin/inormalize -clobber -model
>> colin27.mnc 70.mnc 71.mnc
>> Reading volume
>> /Users/burt/Desktop/build/install/70.mnc...............................................................................................................................................................................................................................................................Done
>> Reading volume /Users/burt/Desktop/build/install/colin27.mncDone
>> Considering 7536640 voxels (100% of total)
>> Selected normalization: medianOfRatios
>> Converting
>> volume................................................................................................................................................................................................................................................................Done
>> Converting
>> volume.....................................................................................................................................................................................Done
>> Normalizing (0.365348, 0)
>> CPU time elapsed in inormalize: 1.184
>> SYS time elapsed in inormalize: 0.117
>> USR time elapsed in inormalize: 1.459
>>
>> With version 2.0.18:
>>
>> neon:install:# inormalize -clobber -model colin27.mnc 70.mnc 71.mnc
>> Reading volume
>> /Users/burt/Desktop/build/install/70.mnc...............................................................................................................................................................................................................................................................Done
>> Reading volume /Users/burt/Desktop/build/install/colin27.mncDone
>> Considering 7536640 voxels (100% of total)
>> Selected normalization: medianOfRatios
>> Volume and model dimensions do not match!
>>
>> What has changed since version 2.0.09? The changelog doesn't seem to
>> contain the answer. Is there a reason why we can no longer normalize images
>> whose dimensions differ from the model? Here are the dimensions for both:
>>
>>
>> neon:install:# mincinfo 70.mnc colin27.mnc
>> file: 70.mnc
>> image: unsigned short 0 to 65535
>> image dimensions: zspace yspace xspace
>>     dimension name         length         step        start
>>     --------------         ------         ----        -----
>>     zspace                    256      1.05469     -139.275
>>     yspace                    256      1.05469     -132.873
>>     xspace                    115          1.3     -79.0931
>>
>>
>> file: colin27.mnc
>> image: unsigned short 0 to 65535
>> image dimensions: zspace yspace xspace
>>     dimension name         length         step        start
>>     --------------         ------         ----        -----
>>     zspace                    181            1          -72
>>     yspace                    217            1         -126
>>     xspace                    181            1          -90
>>
>> Thanks,
>>
>> Burt.
>>
>>
>> -------------
>> Burt Crépeault
>> Centre de recherche de l'université Laval - Robert-Giffard
>> 2601 de la Canardière, suite F-4400
>> Québec QC G1J 2G3
>> 418-663-5741, ext 6844
>>
> _______________________________________________
> MINC-users at bic.mni.mcgill.ca
> http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users
>
>