[VoxBo] normalizing coronal images

[Ling Wong]

Hi Dan,

I tried normalizing the data and then converting to coronal, but I
suspect that might be missing somehow (the brains look a little
flat-topped).  I'd like to try reorienting the template and
normalizing to that, but vborient does not recognize the .mnc
extension.  Any ideas?

Thanks!
Ling

>
> On 4/24/07, Daniel Y Kimberg <kimberg at mail.med.upenn.edu> wrote:
> > Ling Wong wrote:
> >
> > > I'm trying to convert normalized data that was axially acquired to the
> > > coronal orientation, but since the MNI template is in the axial orientation,
> > > I'm not sure if my method is mathematically correct.  What I did was to
> > > convert the nAnatomical, nFunctional, and nDisplay all to coronal.  This
> > > makes an ncrDisplay whose X-dimension matches that of the ncrAnatomical
> > > (159), whose Y-dimension matches that of the ncrFunctional (46), but whose
> > > Z-dimension matches that of the ncrAnatomical (189).  I then resampled
> > > ("-d") the cr_nDisplay so that its Z-dimension matches that of the
> > > ncrFunctional (189 --> 63).  Is this an appropriate method?  Obviously, the
> > > image is very squished with such a short Y-axis.  Is there a coronal MNI
> > > template that would be more appropriate to use for normalization, or a more
> > > suitable approach to this?
> >
> > Hi Ling, sorry about the delay in replying to this.  You probably want
> > a slightly different approach for converting your images to coronal.
> > The reason is that usually you want your display volume to have high
> > resolution in x and y (so that the slices you're looking at have good
> > anatomic detail), but low resolution in z (so that as you move from
> > slice to slice, you get a fresh functional image for each slice).
> >
> > You could certainly re-orient the MNI template, as well as your data,
> > and then just normalize as before.  As long as this doesn't disturb
> > the alignment of your functional and structural volumes, you should be
> > okay.  If it does, you may need to play with the origin a little.
> >
> > Alternatively, you could just resample your ncrAnatomical volume,
> > which should be high-resolution in all three axes.  The bounding box
> > you used for normalization should tell you how the edges of the
> > volumes line up, so all you need to do is reslice in z to the lower
> > resolution.
> >
> > It's a little tricky sometimes figuring out the offset of volumes from
> > different resolutions, but once things are normalized well, it's
> > easier.  If you want to post the bounding boxes you're using and what
> > normalization software, I can make some educated guesses.
> >
> > dan
> >
> >
>
>
> --
> Ling M. Wong
> Research Assistant
> Jha Lab
> Center for Cognitive Neuroscience
> University of Pennsylvania
> lingmw at psych.upenn.edu
>


-- 
Ling M. Wong
Research Assistant
Jha Lab
Center for Cognitive Neuroscience
University of Pennsylvania
lingmw at psych.upenn.edu