Hi David
I can take a look at a subject of yours as well if you upload it (both 5.3
and 6.0 processed versions if you have them) and point me at voxel coords.
The white surface is by design insensitive to local variations which is
what makes it robust, but also makes it more difficult to modify. I'll take
a look and think about how to make it easier
cheers
Bruce
On Thu, 20 Apr 2017, David Semanek
wrote:
Thank you for looking into the deeper mechanics of this issue. In my case I
am dealing with motion contaminated 1.5T data which require extensive and
somewhat intricate edits to the white matter bleed into the pial surface, as
well as into dura. In working with a past dataset of similar characteristics
I was consistently impressed with fs 5.3’s ability to properly segment white
matter in most cases and the robustness of the editing response. I was left
with the overall impression that freesurfer handled noisy data quite well
with a little bit of help. Our results from the analysis of corrected vs
uncorrected data were very strong in favor of the editing protocol.
My experience with 6.0, aside from any issues of information not propagating
through the longitudinal stream, is that there are many more cases in which
the white matter surface will not conform to the edits suggested in the
white matter volume. A number of tricks to “convince” the software in cases
of non-response used in 5.3, such as making sure to complete edits on
adjacent slices, are simply not effective. I am left with the overall
impression that I have very little control over the white matter surfaces
really, beyond making suggestions which may only slightly change the degree
of error, but does not fundamentally change the structure of the white
matter surface enough to remove the misclassification entirely in many
cases.
Best,
David P. Semanek, HCISPP
Research Technician, Posner Lab
Division of Child and Adolescent Psychiatry
Columbia University Medical Center
New York State Psychiatric Institute
1051 Riverside Drive, Pardes Bldg. Rm. 2424
New York, NY 10032
PH: (646) 774-5885
IMPORTANT NOTICE: This e-mail is meant only for the use of the intended
recipient. It may contain confidential information which is legally
privileged or otherwise protected by law. If you received this e-mail in
error or from someone who was not authorized to send it to you, you are
strictly prohibited from reviewing, using, disseminating, distributing or
copying the e-mail. PLEASE NOTIFY US IMMEDIATELY OF THE ERROR BY RETURN
E-MAIL AND DELETE THIS MESSAGE FROM YOUR SYSTEM. Thank you for your
cooperation.
From: Antonin Skoch <a...@ikem.cz>
Date: Thursday, April 20, 2017 at 10:25 AM
To: David Semanek <seman...@nyspi.columbia.edu>,
"freesurfer@nmr.mgh.harvard.edu" <freesurfer@nmr.mgh.harvard.edu>
Subject: Re: [Freesurfer] Worse determination of ?h.white with v6.0 in
comparison to v5.3 - worse GM/WM contrast
Dear David,
thank you for the feedback; I saw your posts concerning edits and responded
to them, see
http://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg52549.html
Just my case is not concerning poor response to the edits (which I believe
is not systematically different between 5.3 and 6.0), my concern is that the
data processed by v6.0 need much more wm.mgz edits than data processed by
v5.3.
I think that my issue lies in -normalization2 step of recon-all. One of the
difference between v5.3 and v6.0 is that by default the -mprage flag is
passed to mri_normalize. This affects several parameters inside
mri_normalize. I tried to reprocess my subjects using v6.0 with -no-mprage,
but unfortunately this did not help.
See the example screenshots processed by v5.3 and v6.0 with -no-mprage:
The brain.mgz is still more aggressively filtered in v6.0 and there is much
more prominent leak of ?h.white outside brain, which is probably caused by
extended filtration which affects GM/WM contrast.
Looking at the source code of mri_normalize.c I did not comprehend where the
basis of the issue lies, but in any case there are big differences in
mri_normalize.c code between versions.
Antonin
From: David Semanek <seman...@nyspi.columbia.edu>
To: Antonin Skoch <a...@ikem.cz>, "freesurfer@nmr.mgh.harvard.edu"
<freesurfer@nmr.mgh.harvard.edu>
Sent: 4/20/2017 3:41 PM
Subject: Re: [Freesurfer] Worse determination of ?h.white with v6.0 in
comparison to v5.3 - worse GM/WM contrast
Agreed. A validated protocol run on a very large group of
subjects in 5.3 was attempted with similar data in 6.0 and not
only was the longitudinal edit stream nearly non-functional for
white matter edits, cross edit performance was disappointing.
I am currently waiting on a response to these potential issues
before pursuing further work with 6.0.
Best,
David P. Semanek, HCISPP
Research Technician, Posner Lab
Division of Child and Adolescent Psychiatry
Columbia University Medical Center
New York State Psychiatric Institute
1051 Riverside Drive, Pardes Bldg. Rm. 2424
New York, NY 10032
PH: (646) 774-5885
IMPORTANT NOTICE: This e-mail is meant only for the use of the
intended recipient. It may contain confidential information which is
legally privileged or otherwise protected by law. If you received
this e-mail in error or from someone who was not authorized to send it
to you, you are strictly prohibited from reviewing, using,
disseminating, distributing or copying the e-mail. PLEASE NOTIFY US
IMMEDIATELY OF THE ERROR BY RETURN E-MAIL AND DELETE THIS MESSAGE FROM
YOUR SYSTEM. Thank you for your cooperation.
From: Antonin Skoch <a...@ikem.cz>
Date: Wednesday, April 19, 2017 at 5:23 PM
To: <freesurfer@nmr.mgh.harvard.edu>
Subject: [Freesurfer] Worse determination of ?h.white with v6.0 in
comparison to v5.3 - worse GM/WM contrast
Dear experts,
I am sending just one more example to illustrate issue with white
surface estimation in v6.0. See the attached screenshots: In v6.0
there seems to be insufficient contrast in brain.finalsurfs.mgz, so
the white surface is leaking at three spots dramatically outwards
towards pial surface. The white surface in v5.3 looks much more
anatomically relevant in the same spot.
Could you please comment on how to avoid such issues in v.6.0?
Regards,
Antonin Skoch
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
The information in this e-mail is intended only for the person to whom it is
addressed. If you believe this e-mail was sent to you in error and the e-mail
contains patient information, please contact the Partners Compliance HelpLine at
http://www.partners.org/complianceline . If the e-mail was sent to you in error
but does not contain patient information, please contact the sender and properly
dispose of the e-mail.
