[caret-users] Voxels size ??

Donna Hanlon donna at brainvis.wustl.edu
Mon Aug 22 07:56:21 CDT 2005

Dear Dr. Giffard,

I haven't tried this (on purpose, anyway), so I'm not sure, but here are 
some guesses:

* Minor deviations (e.g., cubic 0.9) are unlikely to cause trouble, but 
0.5mm voxels (e.g., colin27) almost certainly will.  My sense is that 
somewhere around .75-.8mm things would start to turn ugly.
* Parts of the algorithm that disconnect eye fat (skull), hindbrain, and 
contralateral hemisphere (corpus callossum cut) are most likely to be 
vulnerable to significant deviations from cubic 1mm voxel dimensions, 
because they use landmarks that are based on number of slices from the 
anterior commissure along each axis.  (Someday, we hope to improve the 
way these routines work.)
* Any resulting surface probably will need to be scaled (using Surface: 
Transform) accordingly, so that they reflect true mm coordinates.  
Unless John Harwell has made a change I don't know about (he's on 
vacation for another week), then such scaling is required.

Of course, you can always resample to cubic 1mm, but I assume you have a 
good reason for wanting to do otherwise.  I'm sure you realize there are 
ways to apply affine transforms to functional data (e.g., fMRI, CSM 
sites, or cells) you want to map onto the surface.

Hope this helps,


On 08/22/2005 04:44 AM, Cyrille Giffard wrote:

>Dear all,
>Please could you tell me if it is possible to use human anatomical MRI 
>with an isotropic voxel size lower than 1 mm3 ??
>Thanks in advance for your answer
>Cyrille Giffard
>Faculteit der Geneeskunde
>Laboratorium voor Neuro- en Psychofysiologie
>Campus Gasthuisberg
>Herestraat 49, B. 801, B-3000 LEUVEN (Belgium)
>Tel. +32 (0) 16 34 59 62
>caret-users mailing list
>caret-users at brainvis.wustl.edu

More information about the caret-users mailing list