This is a known effect of DARTEL. The data within the brain is fine and you can use a mask to eliminate the parts outside the brain. This is based on the following comments.

(1) In the case of unsmoothed data: "In the new way, the algorithm scans through the original image and
pushes the voxel values into the appropriate place in the warped image
(adding it to the existing values at that location).  This means that
some voxels in the warped image may not receive any data, whereas other
voxels receive data from several voxels.  This is the cause of the
aliasing." -John Ashburner.

(2) In the case of smoothed data, the above process is performed and then the image is smoothed. Here is the explanation of the artifact: "The reason for the effect is the following.  When we wish to compute an
average of the signal intensity within a region, the idea is to sum up
the values in the region and divide by the number of voxels.  This is
essentially what is done when images are warped using the normalise to
MNI space, except that weighted averaging is used.  For spatially
normalised images, the count is the number of voxels in the original
image that end up in the region, and the sum of the values is a sum over
the intensities of the original voxels that wind up in the region.

The "region" referred to is defined by the smoothing kernel centred at
each voxel, so the contribution to the average is greater for voxels in
the middle and decreases towards zero as you move away. In theory, this
contribution never quite reaches zero (although in practice it does,
because we only use the parts of the Gaussian that are less than a few
standard deviations away from the centre).  If signal is undefined in
some region of the image, the smoothed values in that region will be a
weighted average over those parts of the image where values are defined.
Values may be undefined because they have been set to NaN." -John Ashburner

Hope this helps.

Best Regards, Donald McLaren
=================
D.G. McLaren, Ph.D.
Postdoctoral Research Fellow, GRECC, Bedford VA
Research Fellow, Department of Neurology, Massachusetts General Hospital and Harvard Medical School
Office: (773) 406-2464
=====================
This e-mail contains CONFIDENTIAL INFORMATION which may contain PROTECTED HEALTHCARE INFORMATION and may also be LEGALLY PRIVILEGED and which is intended only for the use of the individual or entity named above. If the reader of the e-mail is not the intended recipient or the employee or agent responsible for delivering it to the intended recipient, you are hereby notified that you are in possession of confidential and privileged information. Any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited and may be unlawful. If you have received this e-mail unintentionally, please immediately notify the sender via telephone at (773) 406-2464 or email.


On Tue, Dec 21, 2010 at 10:27 PM, Thackery Brown <[log in to unmask]> wrote:
I have collected a series of fMRI data for analysis in SPM8. BOLD scans were acquired 3x3x4mm with a slice orientation aligned with the anterior-posterior commissure. All subjects had a little clipping (3 or 4mm) off of the very top of the parietal cortex due the the bounding box size/positioning during acquisition.

Data were preprocessed up to normalization in SPM8 using a standard preprocessing stream. When I look at the raw BOLD images up to the motion corrected images, the brains look perfectly fine - save for signal dropout there are no strange distortions in the images. When I run MNI normalization and smoothing (6mm FWHM) using SPM8's built in functionality, nothing bad comes of it and the swua images look fine as well.

However, when I run normalization/smoothing using DARTEL instead of SPM's built in functionality, I get a strange vertical expansion artifact coming out of the top of the functional brain - seemingly around the point of the scanner bounding box cut-off, although perhaps this is just a coincidence.

I've attempted to attach images of 1) a "healthy" raw image, 2) a "healthy" normalized and smoothed image using SPM8's functionality, and 3) a distorted normalized and smoothed image generated by DARTEL for one subject

If anyone has any experience with this I'd really appreciate the help!