I'll add a little bit to Marko's comments.

When you click the Normalise button, the procedure you get will
attempt to minimise the sum of squares difference between your
template and a warped version of the image.  When the intensity
non-uniformities are as great as those in your images, this procedure
does not work so well (although it does try to account for linear
gradients in image non-uniformity).  To obtain a good match with data
such as yours, the bias field needs to be dealt with in addition to
the deformations.

The segmentation of SPM may be able to deal with the non-uniformities,
although the initial affine registration part does not also model the
bias field.  This means that this part of the algorithm is more likely
to fail.  However, if you manually position your data so they are in
reasonably close alignment with MNI space, you can turn off the
initial affine registration part.  I would also suggest decreasing the
amount of regularisation for the bias correction bit of the
segmentation.  This essentially tells the algorithm that the data is
badly corrupted by low frequency noise.  If the segmentation part
works well for you, then it can be used for spatially normalising the
data.

Alternatively, if your fMRI was collected in the same scanning session
as an anatomical image, the subject did not move and if you used the
DICOM conversion routines in SPM then you may not need to do any
additional coregistration to bring the fMRI and anatomical data
together. Coregistration in SPM does not deal very well with bias
artifact.  The fix for this would be relatively trivial to implement,
but - as ever - it is a matter of getting around to actually writing
the code.

Best regards,
-John



On 28 March 2012 07:31, Marko Wilke <[log in to unmask]> wrote:
> Hi Tracy,
>
> my best guess would be that the segmentation routine gets hung up on your
> image inhomogeneities, of which there seems to be quite a bit (note how the
> anterior part of the brain is already visibly much brighter). If you wanted
> to go with segmenting/normalizing your EPIs directly, I have suggested using
> a two-pass procedure in the past (run segmentation once on the mean EPI and
> only write out a bias-corrected image in native space, on which you then run
> segmentation again to achieve normalization), and only very few people
> complained to me afterwards :) This may also need some playing with the bias
> field options.
>
> Cheers,
> Marko
>
>
> Luks, Tracy wrote:
>>
>>
>>
>> I'm having trouble with the normalization of some 3T epi images acquired
>> with a 32 channel head coil. The normalization of the EPI images to the EPI
>> template runs properly, but the resulting images are clearly not matching
>> the template, particularly in the posterior brain. I haven't had problems
>> like this with images acquired on different scanners. I have also tried
>> coregistering the EPI images to the T1, normalizing the T1, and applying
>> those parameters to the EPIs. That is more successful, but I'd still like to
>> know what's going wrong with the EPI to EPI normalization, and what I can do
>> to rectify the problems. The attached images show the "check reg" results
>> with the "normalized" EPI and the template EPI images.
>>
>> Thanks
>> Tracy Luks
>> UCSF
>> [cid:f7a170cf-4ba5-46df-b68a-13725679a7b8@ucsf.edu]
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>
> --
> ____________________________________________________
> PD Dr. med. Marko Wilke
>  Facharzt für Kinder- und Jugendmedizin
>  Leiter, Experimentelle Pädiatrische Neurobildgebung
>  Universitäts-Kinderklinik
>  Abt. III (Neuropädiatrie)
>
>
> Marko Wilke, MD, PhD
>  Pediatrician
>  Head, Experimental Pediatric Neuroimaging
>  University Children's Hospital
>  Dept. III (Pediatric Neurology)
>
>
> Hoppe-Seyler-Str. 1
>  D - 72076 Tübingen, Germany
>  Tel. +49 7071 29-83416
>  Fax  +49 7071 29-5473
>  [log in to unmask]
>
>  http://www.medizin.uni-tuebingen.de/kinder/epn
> ____________________________________________________

This email is intended only for the use of the individual or entity to which it is addressed and may contain information that is privileged and confidential. If the reader of this email message is not the intended recipient, you are hereby notified that any dissemination, distribution, or copying of this communication is prohibited. If you have received this email in error, please notify the sender and destroy/delete all copies of the transmittal. Thank you.


This email is intended only for the use of the individual or entity to which it is addressed and may contain information that is privileged and confidential. If the reader of this email message is not the intended recipient, you are hereby notified that any dissemination, distribution, or copying of this communication is prohibited. If you have received this email in error, please notify the sender and destroy/delete all copies of the transmittal. Thank you.