Hi Steve,
I have uploaded an example of a preprocessed scan (CP01_SIENAX_Gouw). the
upload ID is 462510.
Originally it was a sagittal MPRAGE(CP01_MPR), which was reoriented
(reorient_CP01_MPR), then downsampled (reorient_CP01_MPR_reduced), then the
neck was cut-off (xreorient_CP01_MPR_reduced), then BET with -f 0.45 -g -0.2
(xreorient_CP01_MPR_reduced_brain, eyes and some skull is left, however some
cortex was cut-off in the right parietal lobe), then sienax without BET.
It seems that the segmentation was not correct in especially the temporal
lobes (while the temporal lobes of the brain extracted image looks alright).
So we think this might be a bias field problem.
Do you have any suggestions?
Thanks a lot!
Best regards, Alida
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]]On
Behalf Of Stephen Smith
Sent: vrijdag 16 december 2005 16:24
To: [log in to unmask]
Subject: Re: [FSL] SIENAX 3D
Hi - it't not very clear from the JPGs that you sent what's going on;
mostly the image looks like the -f threshold should be lowered. In general
we would not recommend downsampling the data. If you would like to upload
an example original image we can have a quick look.
Please upload the files in a single compressed tarfile to
http://www.fmrib.ox.ac.uk/cgi-bin/upload.cgi
and then email me the upload ID.
Cheers, Steve.
On Fri, 16 Dec 2005, A Gouw wrote:
> Dear all, we are trying to run SIENAX on T1-3D images (1x1x1mm) in a
multi-center setting. However, we've run into several
> problems. We hope that you could help us out. First, BET does not give
satisfying results. Even after changing the options -f, -g
> and -c and cutting off the neck, manual editing is still necessary (parts
of the eyes and skull). We think that it could be caused
> by a wrong scanning angle. Therefore, while manual editing is not feasible
on 256 slices, we downsample the images to 4mm slices
> before running BET. Then we combine the manual and automatic masks and run
SIENAX without BET. For some centers, this works out
> well. For other centers however, BET and segmentation still give bad
results. For sagittal MPRAGES the temporal regions are not
> properly segmented (see fig 1), and for coronal MPRAGES this is the case
in the frontal and occipital regions. We believe that it
> is a bias field problem, and we are trying to solve it with optimal
N3-inhomogeneity correction. However, this has not been
> satisfactory yet. Do you have any suggestions, or have suggestions using
different FAST-options? Another problem is that for some
> centres the FOV is too small and some parts of the skull are not included
in the image. Therefore the skull image is distorted (see
> fig 2 and 3). Do you have any ideas of how to solve this? Thank you in
advance, best regards, Alida Gouw A.A. Gouw, MD VU Medical
> Centre Amsterdam The Netherlands
>
--
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
|