[MINC-users] Conversion to MINC and visual inspection of the MINC-file

[Jason Lerch]

On 2012-09-06, at 8:17 AM, Andrew Janke <a.janke at gmail.com> wrote:

> Hi Jörg
> 
> Unfortunately attachments aren't allowed on this list but I'm going to
> take a stab at this anyhow.
> 
> On 6 September 2012 20:12, Jörg Pfannmöller
> <pfannmoelj at uni-greifswald.de> wrote:
>> Subsequently mydata.mnc is converted to .raw format using:
>> minctoraw -short -nonormalize mydata.mnc > mydata.raw
>> 
>> This is displayed in imageJ and seem to be correct (example image attached as number 02)
>> 
>> If mydata.mnc is depicted using mincpik:
>> 
>> mincpik -verbose -depth 16 -scale 40 out.mnc MIFF:- | display -
>> 
>> a rather strange image occurs (example image attached as number 03)
> 
> My guess is that you don't want a scale of 40 here, you probably want
> to use a -width option instead.  The problem is that a lot of the minc
> tools have a few constants hard-coded in there for "human head" sized
> data so working with little voxels will occasionally cause grief.

That should be considered a bug to be identified and fixed - do you know where that's still the case? We run everything at native resolution (i.e. voxel sizes between 5 and 200 micron isotropic) and haven't been bitten by that in a while with the exception of some ITK algorithms.

> 
>> Sample computations using nu_correct for the reduction of intensity in homogeneities obviously are not correct if displayed with imageJ and converted using
> 
> For a start you are going to have to specify a -distance parameter to
> nu_correct to suit your data. This will depend on field strength. One
> of the common approaches when working with mouse data is to simple
> scale it all to "human" size. Typically this is about a scale of 30.

Eek - we really shouldn't have to do that. Step sizes should accurately reflect the acquired data. For what it's worth, we use these nu_correct parameters on mouse brain MRIs (we keep it constant regardless of whether we use ex or in-vivo data, so voxel sizes ranging from 30 to 125 micron isotropic):

-distance 8
-iterations 100
-stop 0.001
-fwhm 0.15
-shrink 4
-lambda 5.0e-02

Jason