That's a bit strange. The segmentation seems OK, and in order to be OK, it
needs to have registered the images well with the tissue probability maps.
Is there any chance that the template headers have become corrupted? If you
do a check reg with the images in the template, canonical and tpm folders, do
they appear to be in proper alignment with each other? That's all that I can
think of (and I know that it is something that can happen).
Best regards,
-John
On Monday 01 December 2008 17:16, Julien Dubois wrote:
> Thank you for your response. I looked at the results of segmentation,
> and the only thing that looks weird to me is, there might be a little
> bit of CSF left on the GM image. All my images (GM,WM,Bias-Field
> Corrected, original T1) are in good registration. I normalized the GM
> and WM images and attached them here. If you see anything that looks
> wrong please let me know. Also, you can download the actual images
> from http://www.klab.caltech.edu/~jdubois/spmlist/ if you want to be
> able to look at them more closely. Thank you for your time.
> - Julien
>
> On Mon, Dec 1, 2008 at 3:46 AM, John Ashburner <[log in to unmask]>
wrote:
> > It looks very much as if it is going wrong, but there is not enough
> > information here to know why. There are lots of possibilities, but the
> > more likely ones are:
> >
> > 1) Segmentation didn't work well, possibly because of poor initial
> > registration. Take a look at a spatially normalised GM image produced by
> > the segmentation. This will tell you how well this step has worked.
> >
> > 2) User error. If the anatomical scans that were segmenented got moved
> > afterwards. If you segment the anatomical, and then coregister it to
> > some other image (so that it is treated as the "source" image or one of
> > the "other" images), then things will be messed up.
> >
> > Best regards,
> > -John
> >
> > On Saturday 29 November 2008 00:41, Julien Dubois wrote:
> >> Dear list,
> >> I want to do some statistics at the group level, hence I have to work
> >> in the MNI space. However, on inspection of the normalization results,
> >> I see that the subjects' normalized anatomicals don't match well, and
> >> actually that they don't match the MNI template well (which is why
> >> they don't match well...). I believe I'm using the recommended steps:
> >> segment, make sure origin is close enough to AC, then use segmentation
> >> parameters to estimate non-linear warping parameters. I attach here
> >> results for 2 subjects: T1.png is the bias-corrected anatomical for
> >> the first subject; normalizedT1_vs_templateT1.png is the result of
> >> normalization compared to the MNI template. Same for second subject,
> >> with _2 at the end of the filename.
> >> Is this kind of result expected, or is there clearly something going
> >> wrong (that's my hunch)? does any have an idea of what I should do to
> >> make it better?
> >> - Julien
|