Dear FS experts
I am trying to run the commands for performing the longitudinal mixed effect statistics with the LME Matlab toolbox. I am following the instructions on the web page; however, I have two questions:
1) when I run the command lme_mass_fit_vw, the script is running for a very long time, after 3 hours the estimation was just at 10% (unfortunately, my laptop is equipped with Windows and is not particularly powerful). I had to stop the script at that point; before restarting it again I would like to be sure: is this behavior normal? Is it correct that lme_mass_fit_vw is running for such a long time?
2) when the script was running, there were several warnings on the Matlab screen, saying that the algorithm was not converging (at most of locations). Is it also normal? How can I check the correctness of my input image? When I open with freeview the "lh.thickness_sm10.mgh" volume, I just see a black screen with valid timeseries values in the minority of voxels (I imagine that these timeseries are concatenated thickness values of my study subjects). Is there another way to better visualize and check a concatenated 4d mgh file? Failure of convergence is referring to empty/zero voxels of this image?
Thank you for any advice on these two topics Kind regards Paola
Rispetta l'ambiente: non stampare questa mail se non ? necessario. Respect the environment: if it's not necessary, don't print this mail.
IL TUO 5XMILLE AL SAN RAFFAELE DI MILANO
Devolvi il tuo 5 per mille all'Ospedale San Raffaele: perch? al centro della Ricerca ci sei TU. CODICE FISCALE: 07636600962, nel riquadro RICERCA SANITARIA. Non c'? cura, senza ricerca. Non c'? ricerca, senza il tuo 5xmille. Scopri come su http://www.5xmille.org
Dear Martin
thank you so much for your answer. Actually, I tried to run the script with the parallel toolbox, but I got some errors and I had to run the version without parallel computing. Anyway, I will run the script till the end and let you know how results look like. The design matrix is really similar to that reported in the example Web page, since I have only one group, in which I want to estimate the effect of time and time x group interaction Best Paola
Da: freesurfer-bounces@nmr.mgh.harvard.edu [mailto:freesurfer-bounces@nmr.mgh.harvard.edu] Per conto di Martin Reuter Inviato: 14 July 2016 12:12 A: Freesurfer support list Oggetto: Re: [Freesurfer] problems by running longitudinal mass univariate analysis
Hi Paola,
1 . It really takes a long time, though I think you probably are missing the parallel toolbox, and running this sequentially will be really slow.
2. You need to wait till the end and look at the percent of places where it did not converge. It is normal that it does not converge in some places. Also make sure the design matrix is not (close to) singular, which can happen if you have nearly collinear columns.
Best Martin On Jul 14, 2016 10:43 AM, Valsasina Paola valsasina.paola@hsr.it wrote:
Dear FS experts
I am trying to run the commands for performing the longitudinal mixed effect statistics with the LME Matlab toolbox.
I am following the instructions on the web page; however, I have two questions:
1) when I run the command lme_mass_fit_vw, the script is running for a very long time, after 3 hours the estimation was just at 10% (unfortunately, my laptop is equipped with Windows and is not particularly powerful). I had to stop the script at that point; before restarting it again I would like to be sure: is this behavior normal? Is it correct that lme_mass_fit_vw is running for such a long time?
2) when the script was running, there were several warnings on the Matlab screen, saying that the algorithm was not converging (at most of locations). Is it also normal? How can I check the correctness of my input image? When I open with freeview the “lh.thickness_sm10.mgh” volume, I just see a black screen with valid timeseries values in the minority of voxels (I imagine that these timeseries are concatenated thickness values of my study subjects). Is there another way to better visualize and check a concatenated 4d mgh file? Failure of convergence is referring to empty/zero voxels of this image?
Thank you for any advice on these two topics
Kind regards
Paola
Rispetta l’ambiente: non stampare questa mail se non è necessario. Respect the environment: if it's not necessary, don't print this mail.
IL TUO 5XMILLE AL SAN RAFFAELE DI MILANO
Devolvi il tuo 5 per mille all’Ospedale San Raffaele: perché al centro della Ricerca ci sei TU. CODICE FISCALE: 07636600962, nel riquadro RICERCA SANITARIA. Non c’è cura, senza ricerca. Non c’è ricerca, senza il tuo 5xmille. Scopri come su http://www.5xmille.org [Immagine rimossa dal mittente.]
Rispetta l’ambiente: non stampare questa mail se non è necessario. Respect the environment: if it's not necessary, don't print this mail.
IL TUO 5XMILLE AL SAN RAFFAELE DI MILANO
Devolvi il tuo 5 per mille all’Ospedale San Raffaele: perché al centro della Ricerca ci sei TU. CODICE FISCALE: 07636600962, nel riquadro RICERCA SANITARIA. Non c’è cura, senza ricerca. Non c’è ricerca, senza il tuo 5xmille. Scopri come su http://www.5xmille.org
Dear Martin
to follow-up mu previous question, I was finally able to run the lme_mass_fit_vw script till the end. Unfortunately, it came out that converge of the algorithm was failed in 99% of the locations. I have only one study group (with a variable number of follow-up , and I used as design matrix exactly the same proposed in the online guide: X=[ones(length(M),1) M M(:,1).*M(:,2)]
so, I don’t see any specific region of failure associated with the design matrix
Therefore, I suppose that this is due to the fact that input images are wrong (empty or with some problems?). How can I investigate further the correctness of my input data? The longitudinal pipeline was performed correctly, and mris_preproc command to generate the mass univariate file seems to work without errors. Can I eventually load somewhere my image file (i.e., the concatenated lh.thickness.mgh including all scans), so that someone chan check that input images for lme estimation are OK? Thank you so much in advance Kind regards Paola
Da: freesurfer-bounces@nmr.mgh.harvard.edu [mailto:freesurfer-bounces@nmr.mgh.harvard.edu] Per conto di Martin Reuter Inviato: 14 July 2016 12:12 A: Freesurfer support list Oggetto: Re: [Freesurfer] problems by running longitudinal mass univariate analysis
Hi Paola,
1 . It really takes a long time, though I think you probably are missing the parallel toolbox, and running this sequentially will be really slow.
2. You need to wait till the end and look at the percent of places where it did not converge. It is normal that it does not converge in some places. Also make sure the design matrix is not (close to) singular, which can happen if you have nearly collinear columns.
Best Martin On Jul 14, 2016 10:43 AM, Valsasina Paola valsasina.paola@hsr.it wrote:
Dear FS experts
I am trying to run the commands for performing the longitudinal mixed effect statistics with the LME Matlab toolbox.
I am following the instructions on the web page; however, I have two questions:
1) when I run the command lme_mass_fit_vw, the script is running for a very long time, after 3 hours the estimation was just at 10% (unfortunately, my laptop is equipped with Windows and is not particularly powerful). I had to stop the script at that point; before restarting it again I would like to be sure: is this behavior normal? Is it correct that lme_mass_fit_vw is running for such a long time?
2) when the script was running, there were several warnings on the Matlab screen, saying that the algorithm was not converging (at most of locations). Is it also normal? How can I check the correctness of my input image? When I open with freeview the “lh.thickness_sm10.mgh” volume, I just see a black screen with valid timeseries values in the minority of voxels (I imagine that these timeseries are concatenated thickness values of my study subjects). Is there another way to better visualize and check a concatenated 4d mgh file? Failure of convergence is referring to empty/zero voxels of this image?
Thank you for any advice on these two topics
Kind regards
Paola
Rispetta l’ambiente: non stampare questa mail se non è necessario. Respect the environment: if it's not necessary, don't print this mail.
IL TUO 5XMILLE AL SAN RAFFAELE DI MILANO
Devolvi il tuo 5 per mille all’Ospedale San Raffaele: perché al centro della Ricerca ci sei TU. CODICE FISCALE: 07636600962, nel riquadro RICERCA SANITARIA. Non c’è cura, senza ricerca. Non c’è ricerca, senza il tuo 5xmille. Scopri come su http://www.5xmille.org [Immagine rimossa dal mittente.]
Rispetta l’ambiente: non stampare questa mail se non è necessario. Respect the environment: if it's not necessary, don't print this mail.
IL TUO 5XMILLE AL SAN RAFFAELE DI MILANO
Devolvi il tuo 5 per mille all’Ospedale San Raffaele: perché al centro della Ricerca ci sei TU. CODICE FISCALE: 07636600962, nel riquadro RICERCA SANITARIA. Non c’è cura, senza ricerca. Non c’è ricerca, senza il tuo 5xmille. Scopri come su http://www.5xmille.org
Hi Paola,
there is probably a problem with the design matrix. What exactly are the columns of M ?
Also how many subjects (and time points) do you have?
I am not sure, but I think you can open a thickness stack on top of the corresponding surface (from fsaverag) in freeview. It will think it is a time course, but that is OK. Not sure how to switch between the thicknesses, but there is a way to click a vertex and see the time course (all the thickness values at that location). Also Qdec may do that.
Best, Martin
On 07/20/2016 10:35 AM, Valsasina Paola wrote:
Dear Martin
to follow-up mu previous question, I was finally able to run the lme_mass_fit_vw script till the end. Unfortunately, it came out that converge of the algorithm was failed in 99% of the locations.
I have only one study group (with a variable number of follow-up , and I used as design matrix exactly the same proposed in the online guide:
X=[ones(length(M),1) M M(:,1).*M(:,2)]
so, I don’t see any specific region of failure associated with the design matrix
Therefore, I suppose that this is due to the fact that input images are wrong (empty or with some problems?).
How can I investigate further the correctness of my input data? The longitudinal pipeline was performed correctly, and mris_preproc command to generate the mass univariate file seems to work without errors. Can I eventually load somewhere my image file (i.e., the concatenated lh.thickness.mgh including all scans), so that someone chan check that input images for lme estimation are OK?
Thank you so much in advance
Kind regards
Paola
*Da:*freesurfer-bounces@nmr.mgh.harvard.edu [mailto:freesurfer-bounces@nmr.mgh.harvard.edu] *Per conto di *Martin Reuter *Inviato:* 14 July 2016 12:12 *A:* Freesurfer support list *Oggetto:* Re: [Freesurfer] problems by running longitudinal mass univariate analysis
Hi Paola,
1 . It really takes a long time, though I think you probably are missing the parallel toolbox, and running this sequentially will be really slow.
- You need to wait till the end and look at the percent of places
where it did not converge. It is normal that it does not converge in some places. Also make sure the design matrix is not (close to) singular, which can happen if you have nearly collinear columns.
Best Martin
On Jul 14, 2016 10:43 AM, Valsasina Paola valsasina.paola@hsr.it wrote:
Dear FS experts
I am trying to run the commands for performing the longitudinal mixed effect statistics with the LME Matlab toolbox.
I am following the instructions on the web page; however, I have two questions:
- when I run the command lme_mass_fit_vw, the script is running for a
very long time, after 3 hours the estimation was just at 10% (unfortunately, my laptop is equipped with Windows and is not particularly powerful). I had to stop the script at that point; before restarting it again I would like to be sure: is this behavior normal? Is it correct that lme_mass_fit_vw is running for such a long time?
- when the script was running, there were several warnings on the
Matlab screen, saying that the algorithm was not converging (at most of locations). Is it also normal? How can I check the correctness of my input image? When I open with freeview the “lh.thickness_sm10.mgh” volume, I just see a black screen with valid timeseries values in the minority of voxels (I imagine that these timeseries are concatenated thickness values of my study subjects). Is there another way to better visualize and check a concatenated 4d mgh file? Failure of convergence is referring to empty/zero voxels of this image?
Thank you for any advice on these two topics
Kind regards
Paola
/Rispetta l’ambiente: non stampare questa mail se non è necessario. Respect the environment: if it's not necessary, don't print this mail./
IL TUO 5XMILLE AL SAN RAFFAELE DI MILANODevolvi il tuo 5 per mille all’Ospedale San Raffaele: perché al centro della Ricerca ci sei TU. CODICE FISCALE: 07636600962, nel riquadro RICERCA SANITARIA. Non c’è cura, senza ricerca. Non c’è ricerca, senza il tuo 5xmille. Scopri come su http://www.5xmille.org
Immagine rimossa dal mittente.
/Rispetta l’ambiente: non stampare questa mail se non è necessario. Respect the environment: if it's not necessary, don't print this mail./
IL TUO 5XMILLE AL SAN RAFFAELE DI MILANODevolvi il tuo 5 per mille all’Ospedale San Raffaele: perché al centro della Ricerca ci sei TU. CODICE FISCALE: 07636600962, nel riquadro RICERCA SANITARIA. Non c’è cura, senza ricerca. Non c’è ricerca, senza il tuo 5xmille. Scopri come su http://www.5xmille.org
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Dear Martin
thank you for your answer. Yes, it was a collinearity problem in the design matrix. I modified it and now everything works fine Best paola
Da: freesurfer-bounces@nmr.mgh.harvard.edu [mailto:freesurfer-bounces@nmr.mgh.harvard.edu] Per conto di Martin Reuter Inviato: 27 July 2016 17:16 A: Freesurfer support list Oggetto: Re: [Freesurfer] R: problems by running longitudinal mass univariate analysis
Hi Paola,
there is probably a problem with the design matrix. What exactly are the columns of M ?
Also how many subjects (and time points) do you have?
I am not sure, but I think you can open a thickness stack on top of the corresponding surface (from fsaverag) in freeview. It will think it is a time course, but that is OK. Not sure how to switch between the thicknesses, but there is a way to click a vertex and see the time course (all the thickness values at that location). Also Qdec may do that.
Best, Martin On 07/20/2016 10:35 AM, Valsasina Paola wrote: Dear Martin
to follow-up mu previous question, I was finally able to run the lme_mass_fit_vw script till the end. Unfortunately, it came out that converge of the algorithm was failed in 99% of the locations. I have only one study group (with a variable number of follow-up , and I used as design matrix exactly the same proposed in the online guide: X=[ones(length(M),1) M M(:,1).*M(:,2)]
so, I don’t see any specific region of failure associated with the design matrix
Therefore, I suppose that this is due to the fact that input images are wrong (empty or with some problems?). How can I investigate further the correctness of my input data? The longitudinal pipeline was performed correctly, and mris_preproc command to generate the mass univariate file seems to work without errors. Can I eventually load somewhere my image file (i.e., the concatenated lh.thickness.mgh including all scans), so that someone chan check that input images for lme estimation are OK? Thank you so much in advance Kind regards Paola
Da: freesurfer-bounces@nmr.mgh.harvard.edumailto:freesurfer-bounces@nmr.mgh.harvard.edu [mailto:freesurfer-bounces@nmr.mgh.harvard.edu] Per conto di Martin Reuter Inviato: 14 July 2016 12:12 A: Freesurfer support list Oggetto: Re: [Freesurfer] problems by running longitudinal mass univariate analysis
Hi Paola,
1 . It really takes a long time, though I think you probably are missing the parallel toolbox, and running this sequentially will be really slow.
2. You need to wait till the end and look at the percent of places where it did not converge. It is normal that it does not converge in some places. Also make sure the design matrix is not (close to) singular, which can happen if you have nearly collinear columns.
Best Martin On Jul 14, 2016 10:43 AM, Valsasina Paola valsasina.paola@hsr.itmailto:valsasina.paola@hsr.it wrote:
Dear FS experts
I am trying to run the commands for performing the longitudinal mixed effect statistics with the LME Matlab toolbox.
I am following the instructions on the web page; however, I have two questions:
1) when I run the command lme_mass_fit_vw, the script is running for a very long time, after 3 hours the estimation was just at 10% (unfortunately, my laptop is equipped with Windows and is not particularly powerful). I had to stop the script at that point; before restarting it again I would like to be sure: is this behavior normal? Is it correct that lme_mass_fit_vw is running for such a long time?
2) when the script was running, there were several warnings on the Matlab screen, saying that the algorithm was not converging (at most of locations). Is it also normal? How can I check the correctness of my input image? When I open with freeview the “lh.thickness_sm10.mgh†volume, I just see a black screen with valid timeseries values in the minority of voxels (I imagine that these timeseries are concatenated thickness values of my study subjects). Is there another way to better visualize and check a concatenated 4d mgh file? Failure of convergence is referring to empty/zero voxels of this image?
Thank you for any advice on these two topics
Kind regards
Paola
Rispetta l’ambiente: non stampare questa mail se non è necessario. Respect the environment: if it's not necessary, don't print this mail.
IL TUO 5XMILLE AL SAN RAFFAELE DI MILANO
Devolvi il tuo 5 per mille all’Ospedale San Raffaele: perché al centro della Ricerca ci sei TU. CODICE FISCALE: 07636600962, nel riquadro RICERCA SANITARIA. Non c’è cura, senza ricerca. Non c’è ricerca, senza il tuo 5xmille. Scopri come su http://www.5xmille.org [Immagine rimossa dal mittente.]
Rispetta l’ambiente: non stampare questa mail se non è necessario. Respect the environment: if it's not necessary, don't print this mail.
IL TUO 5XMILLE AL SAN RAFFAELE DI MILANO
Devolvi il tuo 5 per mille all’Ospedale San Raffaele: perché al centro della Ricerca ci sei TU. CODICE FISCALE: 07636600962, nel riquadro RICERCA SANITARIA. Non c’è cura, senza ricerca. Non c’è ricerca, senza il tuo 5xmille. Scopri come su http://www.5xmille.org [Immagine rimossa dal mittente.]
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edumailto:Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
--
Martin Reuter, PhD
Assistant Professor of Radiology, Harvard Medical School
Assistant Professor of Neurology, Harvard Medical School
A.A.Martinos Center for Biomedical Imaging
Massachusetts General Hospital
Research Affiliate, CSAIL, MIT
Phone: +1-617-724-5652
Web : http://reuter.mit.edu
Rispetta l’ambiente: non stampare questa mail se non è necessario. Respect the environment: if it's not necessary, don't print this mail.
IL TUO 5XMILLE AL SAN RAFFAELE DI MILANO
Devolvi il tuo 5 per mille all’Ospedale San Raffaele: perché al centro della Ricerca ci sei TU. CODICE FISCALE: 07636600962, nel riquadro RICERCA SANITARIA. Non c’è cura, senza ricerca. Non c’è ricerca, senza il tuo 5xmille. Scopri come su http://www.5xmille.org
freesurfer@nmr.mgh.harvard.edu
-
Martin Reuter -
Valsasina Paola