Dear Valentin,
I have just managed to get around to re-running feat-fugue on your data.
It actually appears to me that in fact 552us is giving the best results.
Certainly the brain stem ends up much more symmetrical and not shifted to
one side when using 552 versus the smaller values.
I noticed that you only sent me the following files:
EF_UD_example_func (at various dwell times)
EF_D_example_func
fieldmap_*
FM_D_fmap_mag_brain (at various dwell times)
However, when comparing the outputs you want to look at the
unwarped, and registered, example_func versus the fieldmap
*in the same space*. Note that the start of the filename represents
the space that the image is in. Hence EF_UD_fmap_mag_brain
is the sensible image to compare to EF_UD_example_func as both
are in example func (EF), undistorted (UD) space. These are the
images that are included on the unwarping report page generated
by FEAT.
So maybe it is just that you have not been comparing the right images.
Have a look at EF_UD_fmap_mag_brain vs EF_UD_example_func
and see if you agree that the correct dwell time (552) gives the best
result.
All the best,
Mark
Valentin Piech wrote:
>Dear Mark,
>
>Thank you very much for your reply.
>
>I've talked to our physicist, and he says the value for the dwell time in
>the DICOM field should be the correct one. I've also found a program at
>UCSD's fmri center to unwarp EPI volumes called "ppge2", and this programs
>searches for the same DICOM header field to get the dwell time. Moreover,
>according to their website, they are also using a GE Excite 3.0T scanner,
>and thus presumably the same software.
>
>I am judging that the unwarping is to strong by looking, for example, at the
>curvature of midline of the brain in an axial view in the volumes that have
>the phase direction from left to right because the field is changing
>relatively slowly in the center of the brain. Another place where it becomes
>obvious is by looking at the position of the cerebellum in the saggital view
>of the volumes where the phase directions is anterior-posterior, but there
>are many places it can be seen, particularly in the moving GIFs from the
>html report. I don't pay too much attention to the regions with drop-out, as
>they can't be really restored anyhow.
>
>The actual dwell time is 552 microseconds, and I've tried 552, 460, 368, 276
>and 184 microseconds, both with and without using BET, and for volumes with
>the phase directions being AP or LR. I've uploaded the EPI distorted and
>undistorted EPI volumes, the fieldmap and its magnitude, the fieldmap
>magnitude maps being forward warped according to the field, and a
>high-resolution MPRAGE of the subject. The six-digit reference number for
>this upload session is 177941.
>
>By the way, the field map used for unwarping the EPI volumes with phase
>direction AP was aquired with the phase direction LR, but I have a different
>data set where the phase direction for the fieldmap was also AP, and the
>effect is the same (and the field maps are very similar).
>
>I think that the best match can be found with half of the dwell time (276
>microseconds), or perhaps 2/3 (368 us) of the actual dwell time.
>
>I don't have any further details on the ramp sampling, but if you agree that
>the unwarping effect is to strong, and there isn't a mistake on my part in
>the way I am running FUGUE, the fMRI center will contact GE to double check
>on the field values, and we'll run a test session with and without ramp
>sampling, see if the DICOM field changes, and whether the geometric
>distortion depends on the ramp sampling.
>
>Thank you very much for your help.
>
>Have a great weekend,
>
>Valentin
>
>
>
|