Dear David,

I did not see your data, but maybe couple of relevant remarks could be useful:

1. By default the wm.mgz edits are not propagated from cross to the base, but from the cross to long. And surfaces in long are initialized from the base. Therefore the wm.mgz has to be OK both in the cross and in the base. My view and current experience is, that this means that in many cases it is probably necessary to edit wm.mgz both in cross and base. This would explain your observation that the wm.mgz edits in cross are not propagated to the base and the edits in base (but not in cross) are not seen in the long.

I implemented a custom script which transfers INTERSECTION of the wm.mgz edits from all cross points to the base. I think that to transfer the intersection of edits is safe approach to prevent to introduce errors in wm.mgz in base due to the false propagation of wm.mgz edits from cross to base in case of significant changes in the anatomy between cross and base. My preliminary testing shows that this approach works well and saves time in wm.mgz editing of large errors which are consistently present in all timepoints. But in case of significant errors, the wm.mgz in the base still needs some additional editing. Therefore I think that there is not any single intervention point in wm.mgz edits.

I was discussing the approach of transferring wm.mgz edits with Martin Reuter in the following thread:

http://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg51092.html

2. When there are spots in the cross, where wm.mgz edits do not take an effect, I would suggest to inspect the brain.finalsurfs.mgz to see if there is sufficient contrast of the white/gray matter. In many such cases where I observe my wm.mgz edits do not take an effect, there is local issue with insufficient white/gray matter contrast. The white surface placement is primarily driven by the local gradient of intensity in brain.finalsurfs.mgz, the boundaries/edits of wm.mgz play only supplementary role (as an starting point for gradient search, via ?h.orig, at least in default invocation of mris_make_surfaces, AFAIK).

If I have enough clues where the white surface should lie (i.e. from FLAIR) and there is insufficient white/gray matter contrast (typically there are spots in gray matter with spuriously high signal intensity, similar to white matter), I am currently experimenting with workaround, by directly editing gray matter voxels of 001.mgz in problematic spots by setting their values to typical gray matter values to enforce better white surface placement. I decided to place my edits just in the beginning to the recon-all stream (001.mgz) to ensure the edits are not overwritten in case of new recon-all rerun.

Maybe other experts could comment on my workaround approach, it would also help me out.

Regards,

Antonin Skoch

I believe what I am seeing in the long stream is the edits not properly taking effect.I uploaded a subject with two timepoints, and only baseline edits to the wm as dsemanek2.zip about a week and a half ago. There doesn’t seem to be any significant changes to the surfaces in the base or either of the long runs. I will redo this subject on my end with some more extreme edits for testing to see if I can “force” something.

Thanks again for looking,

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: Martin Reuter mreu...@nmr.mgh.harvard.edu Date: Wednesday, April 5, 2017 at 2:19 AM To: David Semanek seman...@nyspi.columbia.edu Cc: "freesurfer@nmr.mgh.harvard.edu" freesurfer@nmr.mgh.harvard.edu, "Hoopes, Andrew" ahoo...@mgh.harvard.edu Subject: Re: [Freesurfer] wm.mgz Edits Ignored With Current Dataset in FS 5.3/6 Cross and Long Streams

Hi David,

so edits are affecting the surfaces in cross, just not as much as you would like. We need to check if there is a problem there.

For the longitudinal stream, edits in the base should affect surfaces in the base, which in turn will affect surfaces in the long. In long, surfaces would start from a better starting position. Furthermore, wm edits from cross should be propagated to long, to make sure that surfaces don’t move back to the wrong position. So far the theory. If you don’t see any effect of WM edits (in the base) on surfaces (in the base) then there is a problem. This step should be basically identical to how we treat edits in cross.

Best, Martin

On 04 Apr 2017, at 23:13, David Semanek <seman...@nyspi.columbia.edumailto:seman...@nyspi.columbia.edu> wrote:

Thanks for looking into this. I have uploaded another two subject folders as dsemanek3.zip, this time from a single time point for the same subject having been run through the cross stream of recon-all, the one with “test” appended at the end has had white matter edits on sagittal slices 126-142.

So, I have been able to get the surfaces to change in a single cross run as demonstrated in this upload. For this test, I did some edits that are typical on this dataset. Although the surfaces did change to reflect the edits, in a couple of places the surfaces did not change enough, and still encompassed space in the wm volume which had the white matter voxels removed.

For the first pass through these data before I realized there was a potential problem, I had initially edited the cross folders for the subjects and then ran the base, expecting the edits to be transferred over, but they did not seem to be. So I am still left with two issues: 1) incomplete incorporation of cross edits on the cross folders, 2) so far being unable to find an intervention point which will change the surfaces in the long stream, I have edited the base and then ran the long folders and these were not carried over as demonstrated in my previous uploads.

Thanks again,

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: Martin Reuter <mreu...@nmr.mgh.harvard.edumailto:mreu...@nmr.mgh.harvard.edu> Date: Tuesday, April 4, 2017 at 11:20 AM To: "freesurfer@nmr.mgh.harvard.edumailto:freesurfer@nmr.mgh.harvard.edu" <freesurfer@nmr.mgh.harvard.edumailto:freesurfer@nmr.mgh.harvard.edu> Subject: Re: [Freesurfer] wm.mgz Edits Ignored With Current Dataset in FS 5.3/6 Cross and Long Streams

So the way to debug this is to test if we can replicate the problem of WM edits not taking effect in a regular cross sectional stream. If true this means that there is a bug to the way WM edits are considered in FS6.0. It is unlikely that we missed that, but can happen. Probably something else is going on here.

For that it would be good to have a single time point (cross sectional step) with WM edits send over, so that Andrew can take a look at that.

Best, Martin

On 04/04/2017 05:03 PM, David Semanek wrote: I used –uselongbasewmedits only as a test to see if I could get some condition under which the wm edits would be taken into account when the surfaces were generated.

I did not use that flag for the data I uploaded. The wm edits to base did not influence the base wm surfaces or the cross wm surfaces. The cross wm edits (not uploaded) did not influence cross, base, or long wm surfaces.

I’ve found no condition under which wm edits have influenced the white matter surface in any of my data.

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: Martin Reuter mreu...@nmr.mgh.harvard.edumailto:mreu...@nmr.mgh.harvard.edu Date: Tuesday, March 28, 2017 at 1:27 PM To: freesurfer@nmr.mgh.harvard.edumailto:freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] wm.mgz Edits Ignored With Current Dataset in FS 5.3/6 Cross and Long Streams

This is how it should work: WM edits in the base should affect surfaces in the base, these surfaces should be used as initialization for the long processing and fix most of the things there. If not , you need to edit the cross, because WM edits are copied into long from cross.

Using the -uselongbasewmedits flag is usually not recommended. It copies WM edits from base to the long and that is only meaningful if there is very little longitudinal change.

Best, Martin

On 03/23/2017 04:12 PM, David Semanek wrote: Thanks for taking another look at our data.

I have uploaded another subject in the archive dsemanek2.zip . For this subject I have:

1) run the cross processing stream on both timepoints 2) created the base 3) edited the wm.mgz on the base 4) rerun –base –autorecon2-wm –autorecon3 5) run the long stream to create the long folders for the two time points

It doesn’t look like the white matter edits were factored into any of the surfaces at any point in the process. NB: some edits have been exaggerated over protocol for the purpose of testing the software.

These folders were processed on OSX 10.11. I am running the same subject on Linux and I will let you know if I get a qualitatively different outcome.

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: "Hoopes, Andrew" ahoo...@mgh.harvard.edumailto:ahoo...@mgh.harvard.edu Date: Monday, March 20, 2017 at 4:54 PM To: Freesurfer support list freesurfer@nmr.mgh.harvard.edumailto:freesurfer@nmr.mgh.harvard.edu, David Semanek seman...@nyspi.columbia.edumailto:seman...@nyspi.columbia.edu Subject: Re: [Freesurfer] wm.mgz Edits Ignored With Current Dataset in FS 5.3/6 Cross and Long Streams

Hi David,

That is interesting. Sorry to ask you to upload data again, but if you could upload this subject (with only base wm.mgz edits), I could try to replicate this and see why no wm changes are showing up in the longs.

best, Andrew

________________________________ From: freesurfer-boun...@nmr.mgh.harvard.edumailto:freesurfer-boun...@nmr.mgh.harvard.edu

freesurfer-boun...@nmr.mgh.harvard.edumailto:freesurfer-boun...@nmr.mgh.harvard.edu on behalf of David Semanekseman...@nyspi.columbia.edumailto:seman...@nyspi.columbia.edu Sent: Monday, March 20, 2017 1:59 PM To: Hoopes, Andrew; Freesurfer support list Subject: Re: [Freesurfer] wm.mgz Edits Ignored With Current Dataset in FS 5.3/6 Cross and Long Streams

Andrew, thanks for your response. I am still not seeing the white matter edit performance that I am expecting, or that I have seen from using the cross stream on 5.3 in the past with a different dataset.

I started with a new subject with two timepoints. I ran recon-all on both for the cross stream with no edits, and then ran the base. I edited the wm.mgz for the base, then ran “recon-all –autorecon2-wm –autorecon3 –base xx_base –tp xx_t1 –tp xx_t2”. I noticed the surfaces didn’t really change in the base, but I went ahead and ran the two long runs using “recon-all –all –long xx_tx xx_base” and although there are minor differences in the base and time point surfaces, the white matter edits I did on the base were largely ignored, and none of them were included in the time point long run wm.mgz files.

I am tempted to try these same analyses using Linux (I am running this on OSX 10.11 currently), as I experienced a completely different response from the surface generation modules to my edits in the past when using Linux. I’m thinking this is a real long shot, but I cannot otherwise figure out why the software would be behaving so differently from my past experiences.

Any thoughts? Thanks!

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: "Hoopes, Andrew" ahoo...@mgh.harvard.edumailto:ahoo...@mgh.harvard.edu Date: Wednesday, March 15, 2017 at 12:47 PM To: Freesurfer support list freesurfer@nmr.mgh.harvard.edumailto:freesurfer@nmr.mgh.harvard.edu, David Semanek seman...@nyspi.columbia.edumailto:seman...@nyspi.columbia.edu Cc: Bruce Fischl fis...@nmr.mgh.harvard.edumailto:fis...@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] wm.mgz Edits Ignored With Current Dataset in FS 5.3/6 Cross and Long Streams

Hi David

Try editing the base wm.mgz first instead of editing the long and cross wm files. Rerun autorecon2-wm and autorecon3 for the base dir, then completely rerun the longitudinals. The long surfaces are initialized from the base surfaces, so this could be why your wm fixes seem to have no effect. You can find more info here: https://surfer.nmr.mgh.harvard.edu/fswiki/LongitudinalEdits#CheatSheethttps://surfer.nmr.mgh.harvard.edu/fswiki/LongitudinalEdits If editing the base doesn't solve the problem, you can send me the commands you ran in order and I can look into this further.

best, Andrew

________________________________ From: freesurfer-boun...@nmr.mgh.harvard.edumailto:freesurfer-boun...@nmr.mgh.harvard.edu

freesurfer-boun...@nmr.mgh.harvard.edumailto:freesurfer-boun...@nmr.mgh.harvard.edu on behalf of David Semanekseman...@nyspi.columbia.edumailto:seman...@nyspi.columbia.edu Sent: Monday, March 13, 2017 11:55 AM To: Bruce Fischl; Freesurfer support list Subject: Re: [Freesurfer] wm.mgz Edits Ignored With Current Dataset in FS 5.3/6 Cross and Long Streams

Thanks, I have uploaded the cross and long stream processing from one subject which requires numerous white matter edits to correct defects in the white matter surfaces; the file is on the ftp server as dsemanek.zip.

Both of the cross subject folders, s02_t1 and s02_t2 have had edits done to both the brainmask as well as the wm files, and autorecon2-wm and autorecon-3 have been run on them, as well as the long folder for the first time point, s02_t1.long.s02_base.

It was in working with the rerun results of s02_t1.long.s02_base that I noticed the white matter surfaces after being regenerated with the edited wm.mgz did not reflect any of the edits. The easiest way to see this is to load the wm.mgz with the white matter surfaces and scroll through the slices, there are numerous areas where the contours of the white matter surfaces do not follow the voxels of the wm.mgz volume, mostly near what should be identified as hyperintense gray matter. I’m fairly certain the white matter surfaces didn’t change at all after running autorecon2-wm with the wm.mgz edits.

Thanks for taking a look at our data.

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.

On 3/12/17, 4:13 PM, "Bruce Fischl" fis...@nmr.mgh.harvard.edumailto:fis...@nmr.mgh.harvard.edu wrote:

Hi David

if you upload a subject to our ftp site and give us enough detail to replicate what you tried we will take a look

cheers Bruce On Fri, 10 Mar 2017, David Semanek wrote:

> > Hello, I have worked quite a bit in the past with fs 5.3 on datasets which > required a fair number of manual edits to the white matter volume in order > to correct defects in the white matter surface. Typically, these edits take > the form of removing voxels in the wm.mgz volume that have been incorrectly > identified as white matter, usually near the pial surface caused by > intensity artifacts resulting from motion. My experience in the past is that > generating the white matter surface after edits to the wm.mgz volume will > reliably change the geometry of the resulting surfaces. > > > > However, on my current dataset, 1.5T adolescent brains with pervasive motion > artifacts that do not meet the threshold for unusable data, absolutely no > intervention I have done on the wm.mgz volume has any impact at all on the > generation of the white matter surfaces. I am really very puzzled by this. > All of the files that result from wm.mgz reflect the edits, however the aseg > does not. > > > > The resulting white matter surfaces always follow the aseg white matter > definitions and never the wm.mgz edits. I feel as if there might be > something I am missing but this protocol has reliably been used to do white > matter edits in the past. I thought it may be an issue with fs 6 or the long > stream, but I have tried the same edits in 5.3, 6, long and cross streams > and nothing at all has worked. > > > > Does anyone have any suggestions, or perhaps a hint that I am overlooking > something common? > > > > Thanks, > > > > 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