How distorted are your EPI images?  If they are severely distorted, then the
rigid-body registration from the coregistration step will not work
particularly well as it doesn't include any additional parameters for
correcting distortions.  This means that the warps that match the EPI to its
template will not be wonderful for spatially normalising the structural.
Similarly, warps derived from a structural won't be all that good for
spatially normalising the functional data.

For multi-subject studies, one of the important factors is to match
homologous regions together in the functional data.  For this reason, we tend
to spatially normalise our EPI data by matching it with the EPI template,
during which time, we should have the brain masking disabled.

User the <Check Reg> option to display your coregistered functional and
structural data.  How well matched are they?  Unless the EPI can be
distortion corrected, then there is no really good solution to getting the
functional and structural data well registered.  See e.g.:

        P. Jezzard and S. Clare.
          "Source Distortion in Functional {MRI} Data"
          Human Brain Mapping 8(2):80-85 (1999).
        C. Studholme, R. T. Constable and J. S. Duncan.
          "Accurate alignment of functional EPI data to anatomical MRI
          using a physics-based distortion model"
          IEEE Transactions on Medical Imaging 19(11):1115-1127 (2000).

SPM does not include any functions for attempting to do distortion correction.

Depending on many factors, you may get best results by spatially normalising
EPI and structural data seperately or by estimating parameters from one and
applying them to both.  These factors include the severity of the distortions,
how similar the contrast is between your images and the templates, how normal
the subjects brains are, the field of view of your data, the default settings
you use for spatial normalisation, etc...

Best regards,
-John

> we have problems normalizing whole brain EPI images using SPM99.
> Relating to an earlier mail on the list, we chose the following steps:
> after realignment: coregistering of the T1 structural scan to the mean
> EPI from realignment
> determination of the normalisation parameters for the mean EPI image
> (EPI template, bounding box: template, as we are interested in the whole
> brain, sinc interpol. voxel size 2x2x2) and applying those parameters to
> all functionals and the structurals.
>
> both, functional and structural scans look quite distrored, the size of
> the images is greater than the glass brain in both x- and y- direction.
>
> Now, I have read on the mailing list, that the 'mask brain' option might
> be responsible for the problem when normalizing EPI images. When turning
> off this option in the normalisation defaults, the normalised EPI scans
> look much better, however the T1 scans show some additional distortions
> (resembling those of the EPI images). The size of those normalized
> functionals and structurals is o.k.
>
> Can anybody tell us, what is the best approach to normalize both
> functionals and structurals:
> - disabling brain masking during normalisation?
> - seperately normalizing the T1 scans to the T1 template?
> - determination of normalisation parameters for the T1 scans (with brain
> masking) and applying them to the functionals?
> - is there a problem with applying the template bounding box instead of
> the defaults bb?

--
Dr John Ashburner.
Wellcome Department of Cognitive Neurology.
12 Queen Square, London WC1N 3BG, UK.
tel: +44 (0)20 78337491 or +44 (0)20 78373611 x4381
fax: +44 (0)20 78131420
http://www.fil.ion.ucl.ac.uk/~john
mail: [log in to unmask]