Re: [Freesurfer] Problem with longitudinal processing (MS lesions/samseg)

[Douglas N. Greve]

If you have run recon-all on the case and samseg on the conformed image 
(eg, orig.mgz or nu.mgz), then you can run mri_surf2volseg to insert the 
cortical parcellations into the volumetic seg (look in the recon-all.log 
to see the commandline used to create the aparc+aseg.mgz)



On 8/26/2022 7:53 AM, Wittayer, Matthias wrote:


External Email - Use Caution

Dear all,

thanks for the advice to use Samseg. That improved things a lot. Is 
there a way to assess more parcels than there are included in Samseg 
by default (i.e. something like aseg/astats- parcels)? I want to apply 
multivariate lonitudinal atrophy models on the data and need to 
separate i.e. frontal from parietal cortex.


Best Regards Matthias

*Von:*freesurfer-boun...@nmr.mgh.harvard.edu 
 *Im Auftrag von *Douglas N. Greve

*Gesendet:* Freitag, 24. Juni 2022 16:22
*An:* freesurfer@nmr.mgh.harvard.edu; Cerri, Stefano 

*Betreff:* Re: [Freesurfer] Problem with longitudinal processing (MS 
lesions/samseg)


On 6/20/2022 3:03 AM, Wittayer, Matthias wrote:

*External Email - Use Caution *

Hi, thanks for your answer.
Yes I did the longitudinal runs using the base as described in the
manual. So the pipeline is:
1.) Cross sectional Recon all
2.) recon-all -base Template.nii -tp timepoint1.nii -tp
timepoint2.nii -tp timepoint3.nii -all
3.) recon all -Long timepoint1.nii Template.nii -all
Then I read volumes of the segments from asegstats from the
folders of the longitudinal runs with a R script and subtract
values of two consecutive timepoints from one another.
Miraculously some grey matter areas seem to grow in MS patients
(Not only a few voxels but up to 46%, upon visual inspection I
cannot really tell the difference as these still are just some
voxels, but in the raw data areas are pretty much the same size).
Either my boss is heading for a highly remunerated price here or I
just did something wrong with the analysis. He hopes for the
former… I am guessing the latter. The problem is not
systematically distributed as there is not for example one
timepoint that is corrupted in several areas and the others are
fine and there is not one particular area that has a problem. So
garbage in garbage out is maybe not the main problem.
The only things that came to my mind (and did not improve the
problem) were:

1.Bias field correction (I used ANTS N4), though we only have 3T-
data.

2.Lesion filling with a binary lesion mask using FSL

Do I take the right tables (asegstats from the longitudinal runs
folders) to calculate my differences?

Yes, that is right

My understanding of the longitudinal pipeline was that all
timeponits available (and not only the timeponit I do the
longitudinal run with and the next one) are used to form a
template that is warped to atlas space to measure volumes by
combinig transformation matrices of the warping of the raw
timepoint- image of the longitudinal run to the common template
and the transformation matrix of the warping of the template to
atlas space. Is that roughly the way it works?

I did not entirely follow all of that, but it sounds right. How do the 
volumes look in the cross analysis? Do you see the same trend of 
increase? When you look at the segmentations, are you saying that only 
a few voxels change even though the volume is changing by 46%? One 
thing you can try is samseg longitudinal analysis with lesion 
segmentation. It was developed  using MS lesions in mind, so it would 
be good to check on a few cases at least. I'm cc'ing Stefano who 
developed the software.



Thanks a lot!

Matthias



*Von:* freesurfer-boun...@nmr.mgh.harvard.edu

<mailto:freesurfer-boun...@nmr.mgh.harvard.edu> im Auftrag von
Douglas N. Greve 
<mailto:dgr...@mgh.harvard.edu>
*Gesendet:* Sonntag, 19. Juni 2022 16:58:18
*An:* freesurfer@nmr.mgh.harvard.edu
*Betreff:* Re: [Freesurfer] Problem with longitudinal processing

When you say they are growing rather than shrinking, do you mean
in the longitudnial recon-all run? The reason I ask is that you
only mention the base and cross. When you do the longitudinal
analysis, you need to do cross, then base, then long.

On 6/15/2022 11:43 AM, Wittayer, Matthias wrote:

*External Email - Use Caution *

Dear community,

I tried to process MS- patient’s MRIs (mostly same scanner,
same settings) Longitudinally over a long period of time. I
first processed all timeponits crosssectionally and then
initialised  the base image by recon- all – base TP1 TP2 TP3
etc. Now I am trying to run label based morphometry and it
seems some areas are growing rather than shrinking. Which is
highly unlikely. I tried to exclu

Re: [Freesurfer] Problem with longitudinal processing (MS lesions/samseg)

[Wittayer, Matthias]

External Email - Use Caution

Dear all,

thanks for the advice to use Samseg. That improved things a lot. Is there a way 
to assess more parcels than there are included in Samseg by default (i.e. 
something like aseg/astats- parcels)? I want to apply multivariate lonitudinal 
atrophy models on the data and need to separate i.e. frontal from parietal 
cortex.

Best Regards Matthias


Von: freesurfer-boun...@nmr.mgh.harvard.edu 
 Im Auftrag von Douglas N. Greve
Gesendet: Freitag, 24. Juni 2022 16:22
An: freesurfer@nmr.mgh.harvard.edu; Cerri, Stefano 
Betreff: Re: [Freesurfer] Problem with longitudinal processing (MS 
lesions/samseg)


On 6/20/2022 3:03 AM, Wittayer, Matthias wrote:

External Email - Use Caution
Hi, thanks for your answer.
Yes I did the longitudinal runs using the base as described in the manual. So 
the pipeline is:
1.) Cross sectional Recon all
2.) recon-all -base Template.nii -tp timepoint1.nii -tp timepoint2.nii -tp 
timepoint3.nii -all
3.) recon all -Long timepoint1.nii Template.nii -all
Then I read volumes of the segments from asegstats from the folders of the 
longitudinal runs with a R script and subtract values of two consecutive 
timepoints from one another. Miraculously some grey matter areas seem to grow 
in MS patients (Not only a few voxels but up to 46%, upon visual inspection I 
cannot really tell the difference as these still are just some voxels, but in 
the raw data areas are pretty much the same size).
Either my boss is heading for a highly remunerated price here or I just did 
something wrong with the analysis. He hopes for the former… I am guessing the 
latter. The problem is not systematically distributed as there is not for 
example one timepoint that is corrupted in several areas and the others are 
fine and there is not one particular area that has a problem. So garbage in 
garbage out is maybe not the main problem.
The only things that came to my mind (and did not improve the problem) were:

1.   Bias field correction (I used ANTS N4), though we only have 3T- data.

2.   Lesion filling with a binary lesion mask using FSL
Do I take the right tables (asegstats from the longitudinal runs folders) to 
calculate my differences?
Yes, that is right

My understanding of the longitudinal pipeline was that all timeponits available 
(and not only the timeponit I do the longitudinal run with and the next one) 
are used to form a template that is warped to atlas space to measure volumes by 
combinig transformation matrices of the warping of the raw timepoint- image of 
the longitudinal run to the common template and the transformation matrix of 
the warping of the template to atlas space. Is that roughly the way it works?
I did not entirely follow all of that, but it sounds right. How do the volumes 
look in the cross analysis? Do you see the same trend of increase? When you 
look at the segmentations, are you saying that only a few voxels change even 
though the volume is changing by 46%? One thing you can try is samseg 
longitudinal analysis with lesion segmentation. It was developed  using MS 
lesions in mind, so it would be good to check on a few cases at least. I'm 
cc'ing Stefano who developed the software.



Thanks a lot!
Matthias

Von: 
freesurfer-boun...@nmr.mgh.harvard.edu<mailto:freesurfer-boun...@nmr.mgh.harvard.edu>
 
<mailto:freesurfer-boun...@nmr.mgh.harvard.edu>
 im Auftrag von Douglas N. Greve 
<mailto:dgr...@mgh.harvard.edu>
Gesendet: Sonntag, 19. Juni 2022 16:58:18
An: freesurfer@nmr.mgh.harvard.edu<mailto:freesurfer@nmr.mgh.harvard.edu>
Betreff: Re: [Freesurfer] Problem with longitudinal processing

When you say they are growing rather than shrinking, do you mean in the 
longitudnial recon-all run? The reason I ask is that you only mention the base 
and cross. When you do the longitudinal analysis, you need to do cross, then 
base, then long.
On 6/15/2022 11:43 AM, Wittayer, Matthias wrote:

External Email - Use Caution
Dear community,

I tried to process MS- patient’s MRIs (mostly same scanner, same settings) 
Longitudinally over a long period of time. I first processed all timeponits 
crosssectionally and then initialised  the base image by recon- all – base TP1 
TP2 TP3 etc. Now I am trying to run label based morphometry and it seems some 
areas are growing rather than shrinking. Which is highly unlikely. I tried to 
exclude timepoint of a relapse to rule out perifocal edema interfering with 
measures but the problem remains.
Did anybody have the same problem?
Is it a potential bug or just a garbage in garbage out problem (though I don’t 
know what would be wrong with our scans)?
Does it make sense to make an intermediate template for two timeponits only? 
I.e. recon- all –base TP1 TP2; recon –all –base TP2 TP3 … and use them for 
longitudinal runs?

Thanks for your opinions. Best
Matthias




___

Freesurfer 

Re: [Freesurfer] Problem with longitudinal processing (MS lesions/samseg)

[Douglas N. Greve]

On 6/20/2022 3:03 AM, Wittayer, Matthias wrote:


External Email - Use Caution

Hi, thanks for your answer.
Yes I did the longitudinal runs using the base as described in the 
manual. So the pipeline is:

1.) Cross sectional Recon all
2.) recon-all -base Template.nii -tp timepoint1.nii -tp timepoint2.nii 
-tp timepoint3.nii -all

3.) recon all -Long timepoint1.nii Template.nii -all
Then I read volumes of the segments from asegstats from the folders of 
the longitudinal runs with a R script and subtract values of two 
consecutive timepoints from one another. Miraculously some grey matter 
areas seem to grow in MS patients (Not only a few voxels but up to 
46%, upon visual inspection I cannot really tell the difference as 
these still are just some voxels, but in the raw data areas are pretty 
much the same size).
Either my boss is heading for a highly remunerated price here or I 
just did something wrong with the analysis. He hopes for the former… I 
am guessing the latter. The problem is not systematically distributed 
as there is not for example one timepoint that is corrupted in several 
areas and the others are fine and there is not one particular area 
that has a problem. So garbage in garbage out is maybe not the main 
problem.
The only things that came to my mind (and did not improve the problem) 
were:


1.)Bias field correction (I used ANTS N4), though we only have 3T- data.

2.)Lesion filling with a binary lesion mask using FSL

Do I take the right tables (asegstats from the longitudinal runs 
folders) to calculate my differences?



Yes, that is right


My understanding of the longitudinal pipeline was that all timeponits 
available (and not only the timeponit I do the longitudinal run with 
and the next one) are used to form a template that is warped to atlas 
space to measure volumes by combinig transformation matrices of the 
warping of the raw timepoint- image of the longitudinal run to the 
common template and the transformation matrix of the warping of the 
template to atlas space. Is that roughly the way it works?


I did not entirely follow all of that, but it sounds right. How do the 
volumes look in the cross analysis? Do you see the same trend of 
increase? When you look at the segmentations, are you saying that only a 
few voxels change even though the volume is changing by 46%? One thing 
you can try is samseg longitudinal analysis with lesion segmentation. It 
was developed  using MS lesions in mind, so it would be good to check on 
a few cases at least. I'm cc'ing Stefano who developed the software.



Thanks a lot!

Matthias



*Von:* freesurfer-boun...@nmr.mgh.harvard.edu 
 im Auftrag von Douglas N. 
Greve 

*Gesendet:* Sonntag, 19. Juni 2022 16:58:18
*An:* freesurfer@nmr.mgh.harvard.edu
*Betreff:* Re: [Freesurfer] Problem with longitudinal processing

When you say they are growing rather than shrinking, do you mean in 
the longitudnial recon-all run? The reason I ask is that you only 
mention the base and cross. When you do the longitudinal analysis, you 
need to do cross, then base, then long.


On 6/15/2022 11:43 AM, Wittayer, Matthias wrote:

*External Email - Use Caution *

Dear community,

I tried to process MS- patient’s MRIs (mostly same scanner, same
settings) Longitudinally over a long period of time. I first
processed all timeponits crosssectionally and then initialised 
the base image by recon- all – base TP1 TP2 TP3 etc. Now I am
trying to run label based morphometry and it seems some areas are
growing rather than shrinking. Which is highly unlikely. I tried
to exclude timepoint of a relapse to rule out perifocal edema
interfering with measures but the problem remains.

Did anybody have the same problem?

Is it a potential bug or just a garbage in garbage out problem
(though I don’t know what would be wrong with our scans)?

Does it make sense to make an intermediate template for two
timeponits only? I.e. recon- all –base TP1 TP2; recon –all –base
TP2 TP3 … and use them for longitudinal runs?

Thanks for your opinions. Best

Matthias



___

Freesurfer mailing list

Freesurfer@nmr.mgh.harvard.edu

BLOCKEDmail[.]nmr[.]mgh[.]harvard[.]edu/mailman/listinfo/freesurferBLOCKED


___
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
___
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 Mass General Brigham 
Compliance HelpLine at https

Re: [Freesurfer] Problem with longitudinal processing

[Wittayer, Matthias]

External Email - Use Caution

Hi, thanks for your answer.
Yes I did the longitudinal runs using the base as described in the manual. So 
the pipeline is:
1.) Cross sectional Recon all
2.) recon-all -base Template.nii -tp timepoint1.nii -tp timepoint2.nii -tp 
timepoint3.nii -all
3.) recon all -Long timepoint1.nii Template.nii -all
Then I read volumes of the segments from asegstats from the folders of the 
longitudinal runs with a R script and subtract values of two consecutive 
timepoints from one another. Miraculously some grey matter areas seem to grow 
in MS patients (Not only a few voxels but up to 46%, upon visual inspection I 
cannot really tell the difference as these still are just some voxels, but in 
the raw data areas are pretty much the same size).
Either my boss is heading for a highly remunerated price here or I just did 
something wrong with the analysis. He hopes for the former... I am guessing the 
latter. The problem is not systematically distributed as there is not for 
example one timepoint that is corrupted in several areas and the others are 
fine and there is not one particular area that has a problem. So garbage in 
garbage out is maybe not the main problem.
The only things that came to my mind (and did not improve the problem) were:

1.)Bias field correction (I used ANTS N4), though we only have 3T- data.

2.)Lesion filling with a binary lesion mask using FSL
Do I take the right tables (asegstats from the longitudinal runs folders) to 
calculate my differences?
My understanding of the longitudinal pipeline was that all timeponits available 
(and not only the timeponit I do the longitudinal run with and the next one) 
are used to form a template that is warped to atlas space to measure volumes by 
combinig transformation matrices of the warping of the raw timepoint- image of 
the longitudinal run to the common template and the transformation matrix of 
the warping of the template to atlas space. Is that roughly the way it works?

Thanks a lot!
Matthias

Von: freesurfer-boun...@nmr.mgh.harvard.edu 
 im Auftrag von Douglas N. Greve 

Gesendet: Sonntag, 19. Juni 2022 16:58:18
An: freesurfer@nmr.mgh.harvard.edu
Betreff: Re: [Freesurfer] Problem with longitudinal processing

When you say they are growing rather than shrinking, do you mean in the 
longitudnial recon-all run? The reason I ask is that you only mention the base 
and cross. When you do the longitudinal analysis, you need to do cross, then 
base, then long.
On 6/15/2022 11:43 AM, Wittayer, Matthias wrote:

External Email - Use Caution
Dear community,

I tried to process MS- patient's MRIs (mostly same scanner, same settings) 
Longitudinally over a long period of time. I first processed all timeponits 
crosssectionally and then initialised  the base image by recon- all - base TP1 
TP2 TP3 etc. Now I am trying to run label based morphometry and it seems some 
areas are growing rather than shrinking. Which is highly unlikely. I tried to 
exclude timepoint of a relapse to rule out perifocal edema interfering with 
measures but the problem remains.
Did anybody have the same problem?
Is it a potential bug or just a garbage in garbage out problem (though I don't 
know what would be wrong with our scans)?
Does it make sense to make an intermediate template for two timeponits only? 
I.e. recon- all -base TP1 TP2; recon -all -base TP2 TP3 ... and use them for 
longitudinal runs?

Thanks for your opinions. Best
Matthias



___

Freesurfer mailing list

Freesurfer@nmr.mgh.harvard.edu<mailto:Freesurfer@nmr.mgh.harvard.edu>

BLOCKEDmail[.]nmr[.]mgh[.]harvard[.]edu/mailman/listinfo/freesurferBLOCKED

___
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 Mass General Brigham 
Compliance HelpLine at https://www.massgeneralbrigham.org/complianceline 
<https://www.massgeneralbrigham.org/complianceline> .
Please note that this e-mail is not secure (encrypted).  If you do not wish to 
continue communication over unencrypted e-mail, please notify the sender of 
this message immediately.  Continuing to send or respond to e-mail after 
receiving this message means you understand and accept this risk and wish to 
continue to communicate over unencrypted e-mail. 

Re: [Freesurfer] Problem with longitudinal processing

[Douglas N. Greve]

When you say they are growing rather than shrinking, do you mean in the 
longitudnial recon-all run? The reason I ask is that you only mention 
the base and cross. When you do the longitudinal analysis, you need to 
do cross, then base, then long.


On 6/15/2022 11:43 AM, Wittayer, Matthias wrote:


External Email - Use Caution

Dear community,

I tried to process MS- patient’s MRIs (mostly same scanner, same 
settings) Longitudinally over a long period of time. I first processed 
all timeponits crosssectionally and then initialised  the base image 
by recon- all – base TP1 TP2 TP3 etc. Now I am trying to run label 
based morphometry and it seems some areas are growing rather than 
shrinking. Which is highly unlikely. I tried to exclude timepoint of a 
relapse to rule out perifocal edema interfering with measures but the 
problem remains.


Did anybody have the same problem?

Is it a potential bug or just a garbage in garbage out problem (though 
I don’t know what would be wrong with our scans)?


Does it make sense to make an intermediate template for two timeponits 
only? I.e. recon- all –base TP1 TP2; recon –all –base TP2 TP3 … and 
use them for longitudinal runs?


Thanks for your opinions. Best

Matthias


___
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
___
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 Mass General Brigham 
Compliance HelpLine at https://www.massgeneralbrigham.org/complianceline 
 .
Please note that this e-mail is not secure (encrypted).  If you do not wish to 
continue communication over unencrypted e-mail, please notify the sender of 
this message immediately.  Continuing to send or respond to e-mail after 
receiving this message means you understand and accept this risk and wish to 
continue to communicate over unencrypted e-mail. 

[Freesurfer] Problem with longitudinal processing

[Wittayer, Matthias]

External Email - Use Caution

Dear community,

I tried to process MS- patient's MRIs (mostly same scanner, same settings) 
Longitudinally over a long period of time. I first processed all timeponits 
crosssectionally and then initialised  the base image by recon- all - base TP1 
TP2 TP3 etc. Now I am trying to run label based morphometry and it seems some 
areas are growing rather than shrinking. Which is highly unlikely. I tried to 
exclude timepoint of a relapse to rule out perifocal edema interfering with 
measures but the problem remains.
Did anybody have the same problem?
Is it a potential bug or just a garbage in garbage out problem (though I don't 
know what would be wrong with our scans)?
Does it make sense to make an intermediate template for two timeponits only? 
I.e. recon- all -base TP1 TP2; recon -all -base TP2 TP3 ... and use them for 
longitudinal runs?

Thanks for your opinions. Best
Matthias
___
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 Mass General Brigham 
Compliance HelpLine at https://www.massgeneralbrigham.org/complianceline 
 .
Please note that this e-mail is not secure (encrypted).  If you do not wish to 
continue communication over unencrypted e-mail, please notify the sender of 
this message immediately.  Continuing to send or respond to e-mail after 
receiving this message means you understand and accept this risk and wish to 
continue to communicate over unencrypted e-mail.