 |
 |
It's purely to do with how much brain coverage you have in your images. In other routines in SPM (and other fMRI software), regions are set to zero where there is no actual data. Then, when the data are smoothed, this introduces variance into the data that will correlate with how close voxels are to the edge of the field of view. In Dartel, the algorithm at least tries to do some sort of extrapolation so that as little data are lost as possible. This approach should reduce these correlations. Careful use of masking when you fit your GLM can be used to exclude unwanted regions. Best regards, -John On 11 May 2016 at 03:44, Colin Hawco <[log in to unmask]> wrote: > I am working with DARTEL to normalize some fMRI data. In some resulting > images, there is distortions in the data, such as what is shown. The data > does not have whole brain coverage, and based on previous posts it looks > like this is not really a problem, as DARTEL is pulling data to where > there is none, but the data in the voxels in which there is actually EPI > data are not affected by this. Thus, as long as I make sure these voxels > are not included in the final statistical analysis, we should be OK. > > But I want to confirm that is true. In both ofo the attached images, the > 'pulling' distortion corresponds to regions without whole brain > coverage. > > It also happens to be the case that this seems mainly present in the > brains of older participants (we have young and old, and the older people > seem to have greater distortion in general. > > Lastly, I'll post my work flow, to make sure there are no issues here. > > 1. The EPI data is slice time corrected and unwarped using field maps. > > 2. The anatomical images are reoriented manually, as the origin in most > scans is in the splenium of the corpus callosum. When I select > reorient, I apply this reorientation to the ua*.nii EPI files. I do this > after the unwapring because I need to keep the field maps in > regeister with the EPI scans and prefer to avoid potential registration > issues. > > 4. coreg the EPI (ua*.nii) to the T1. Looks good. > > 3. Segment the T1 scans, DARTEL outputs for c1 and c2 (grey and white > matter). > > 4. Run spm_dartel_template with default parameters, calling the rc1 and > rc2 files. > > 5. Run normalize using Template 6 and the ua*.nii EPI files. > > I have been trying to improve the normalization for several days, and > nothing seems to substantially reduce the distortion I see in the attached > image. I'd appreciate any insight into what I should do (e.g. ignore it, > change parameters, use an alternate normalize option, etc). > > Colin Hawco, PhD > Neuranalysis Consulting > Neuroimaging analysis and consultation > www.neuranalysis.com > [log in to unmask] > > > > >