Hi Jef, cc list,
although what you suggest is a common strategy, I do not completely
agree with the claim that such a non linear 'normalization' or
deformation is not required because both scans have the same size and
shape (they don't). EPI scans are considerably distorted spatially in a
nonlinear fashion (due to field inhomogeneities at susceptibility
transitions), in a very different way as compared to T1 weighted
recordings. Most notably at higher field strengths and longer echo
trains (eg in high resolution fMRI). This difference cannot be accounted
for by a rigid body or even affine transformation, and is often ignored,
leading to faulty co registration.
Of course 0th order unwarping by using field maps can somewhat correct
for EPI distortion, but that depends greatly on the quality of the
fieldmaps. To my knowledge this is only implemented in spm5 for axial
single shot EPI with phase encoding AP or PA, which is a very limited
application with respect to the many alternative acquisition schemes
that are used (coronal/saggital EPI, EVI, PRESTO, etc).
It might therefore be interesting to do non-linear warping between Echo
Planar Images and T1 weighted structural scans to improve native space
registration before template normalizatoin. I suspect the algorithms
used in normalization are optimized for warping to average atlases or
tissue probability maps, but has this option been seriously investigated
somewhere?
Cheers,
Bas
Jonathan Peelle schreef:
> Hi Jef
>
> Because your subject's brain is the same size and shape in the
> structural and functional images, there is no "normalization"
> required, only moving one image around to match the other (i.e. rigid
> body registaration). This is done using the "coregister" option.
> Before doing so, you should realign all functional images using
> "realign", and written out the mean functional image. Then you can
> coregister the mean functional to that subject's structural image.
> Specify the mean functional as the reference image (you want that to
> remain stationary), and the structural image as the source image.
> There is no need to reslice, as the image headers will contain the
> updated location. When you're done use checkReg to select both
> volumes and make sure they are indeed lined up in a reasonable way.
>
> Hope this helps,
>
> Jonathan
>
> On Wed, Feb 4, 2009 at 4:50 PM, Jef West <[log in to unmask]> wrote:
>
>> hello
>>
>> i am attempting to conduct a single-subject analysis and i need to
>> coregister and normalize an individual's mean fmri volume to the
>> individual's smri volumes. therefore, i do not need to normalize to the
>> templates in spm5, but to the individual's smri volume.
>>
>> i would greatly appreciate any advice on how to normalize to the
>> individual's smri and not the standard templates, since only one person is
>> being analyzed. i want to ultimately superimpose the mean fmri onto their
>> smri, to determine the location to apply tms. any suggestions on these
>> matters would be greatly appreciated.
>>
>> j
>>
>>
--
--------------------------------------------------
Dr. S.F.W. Neggers
Division of Brain Research
Rudolf Magnus Institute for Neuroscience
Utrecht University Medical Center
Visiting : Heidelberglaan 100, 3584 CX Utrecht
Room B.01.1.03
Mail : Huispost B.01.206, P.O. Box 85500
3508 GA Utrecht, the Netherlands
Tel : +31 (0)88 7559609
Fax : +31 (0)88 7555443
E-mail : [log in to unmask]
--------------------------------------------------
|