[caret-users] Using SPM5 bias corrected image improved segmentation results

[Donna Dierker]

Markus gave me permission to pass on this message.  Paragraph 1 seemed 
worth passing along.  There are, of course, many bias correction 
algorithms and we don't officially endorse one over another.  We only 
know that for some data, bias correction can make a huge difference.

I'm working with Markus on his volume-surface alignment issues.

-------- Original Message --------
Subject: 	Solved problem and question on mapping metric files
Date: 	Wed, 04 Feb 2009 15:50:47 +0100
From: 	Markus Goldhacker <Markus.Goldhacker at biologie.uni-regensburg.de>
To: 	Donna Dierker <donna at brainvis.wustl.edu>



Hello Donna,

some time ago I wrote you because of a problem with an inhomogenous
individual, structural image, which I wanted to segment. The problem
was that in the Segmentation window  just one big peak appeared instead
of a grey- and white-matter peak. We solved this problem with using the
Bias Corrected image from SPM5. With this image we get an almost
perfect grey/white-matter distinction.
Perhaps this helps other people with the same problem.



Unfortunately I have another problem. As it is no problem anymore to get
the Fiducial image of every subject, I tried mapping SPM and ICA
contrasts on the volume. When I display the metric files on the
surface, no metric file fits the underlying Volume. And the displacement
 differs form subject to subject.

The origin of every contrast and its structural image was set at the AC
so that in CARET I didn't have to change the crosshair anymore. When
loading the picture I just had to take the first crosshair position as
AC location. So in my opinion, if the AC is the origin of every image
(contrasts, structural) CARET should map it right, with no
displacement.

Maybe the underlying problem is the voxelsize of the structural image
(-0.977 x 0.977 x 1.25) and the resampling (to 1x1x1) I do before going
on with the segmentation process?

I attached a screenshot of such a displacement. I hope you can help me.

Thanks in advance!

Markus





-------------- next part --------------
A non-text attachment was scrubbed...
Name: caret_problem.jpg
Type: image/jpeg
Size: 58391 bytes
Desc: not available
Url : http://brainvis.wustl.edu/pipermail/caret-users/attachments/20090204/b6f26581/attachment-0001.jpg