From matthijs at phenogenomics.ca  Tue Dec 21 13:50:55 2010
From: matthijs at phenogenomics.ca (Matthijs van Eede)
Date: Tue, 21 Dec 2010 13:50:55 -0500
Subject: [MINC-development] mincresample issue/bug at smaller voxel sizes
	using concatenated transforms
Message-ID: <4D10F70F.1030504@phenogenomics.ca>

Hi all,

We recently ran into an interesting resampling issue using 60 micron 
data. We have a concatenated transform that uses a few non linear 
transformation grids, and created a equivalent transformation with a 
single transformation grid from it (using minc_displacement written by 
Jason). However, resampling using the concatenated transform gave a 
visually noticeable different result from resampling with the transform 
with a single non linear grid. In many places there was a discrepancy of 
about a voxel. Main note about all this is that (understandably) the 
more transforms are being concatenated, the larger the error in the 
resampling.

Looking at the function grid_inverse_transform_point in 
grid_transforms.c, we found that there is a tolerance variable (ftol at 
line 254) set to 0.05 (defined in mm). Assuming that the default voxel 
size is 1mm, we adjusted the value of ftol to 0.06*0.05=0.003. This 
mostly eliminated the differences between resampling with 
concatenated/single transformation grids. Setting the value smaller 
still, removed any remaining differences.

A fix for this problem would be to have the ftol variable be dependent 
on the voxel size of the input file, i.e., ftol = voxelsize*0.05 (or an 
even smaller value), but the function does not receive any information 
about the volume, so to be able to do that the api would need to be 
changed. Not sure whether that is a desired solution, or whether there 
is an alternative solution.

Any input would be appreciated,
Matthijs



From claude at bic.mni.mcgill.ca  Tue Dec 21 15:23:28 2010
From: claude at bic.mni.mcgill.ca (Claude LEPAGE)
Date: Tue, 21 Dec 2010 15:23:28 -0500
Subject: [MINC-development] mincresample issue/bug at smaller voxel
	sizes using concatenated transforms
In-Reply-To: <4D10F70F.1030504@phenogenomics.ca>
Message-ID: <201012212023.oBLKNSa3032658@cassio.bic.mni.mcgill.ca>

Hi,

Have a look at what I did in minc-2.1.00 - which I use with histology
data at 10 microns. I based the ftol as a function of the grid size
used in the non-linear transform. It's better the previous hard-coded
value of 0.05 (fine at 1mm). Ideally, as you say, ftol should be based
on voxel size. However, good luck at obtaining this information in
this piece of code.

Claude

From matthijs at phenogenomics.ca  Thu Dec 23 11:48:29 2010
From: matthijs at phenogenomics.ca (Matthijs van Eede)
Date: Thu, 23 Dec 2010 11:48:29 -0500
Subject: [MINC-development] mincresample issue/bug at smaller voxel
	sizes using concatenated transforms
In-Reply-To: <201012212023.oBLKNSa3032658@cassio.bic.mni.mcgill.ca>
References: <201012212023.oBLKNSa3032658@cassio.bic.mni.mcgill.ca>
Message-ID: <4D137D5D.5050803@phenogenomics.ca>

Hi Claude,

Initially I thought the current solution in minc 2.1 would work for use, 
but we quite frequently use transformation grids with resolutions that 
are a lot coarser than the volumes we transform with them, so we would 
still run into the same problem. Is it possible to have an environment 
variable that can override the value that is set for ftol?

Cheers,
Matthijs

Claude LEPAGE wrote:
> Hi,
>
> Have a look at what I did in minc-2.1.00 - which I use with histology
> data at 10 microns. I based the ftol as a function of the grid size
> used in the non-linear transform. It's better the previous hard-coded
> value of 0.05 (fine at 1mm). Ideally, as you say, ftol should be based
> on voxel size. However, good luck at obtaining this information in
> this piece of code.
>
> Claude
> _______________________________________________
> MINC-development mailing list
> MINC-development at bic.mni.mcgill.ca
> http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-development
>