Dear Doug,
I used selxavg3-sess in freesurfer 5.0 to run a big GLM analysis on a monkey data set containing 147 runs. But When I check the results, I found that in the voxels where the t-values should be very high were turned out to have a value of zero. I checked the brainmask, the ces and the cesvar values, they all look fine. It seems that only the ces values in these voxels are a bit high compared to the nearby normal voxels (e.g. 1.17 (ces)/0.00048(cesvar) compared to 0.74/0.00040 in a nearby normal voxel). And it seems that not only me have this problem (similar problem posted in the mailinglist before. http://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg13789.html). Do you know what the problem would be and how to solve it? Thanks very much for your help.
Attached please find an image of the problem.
Best,
Qi Zhu, PostDoc Laboratorium voor Neuro- en Psychofysiologie K.U.Leuven Medical School Herestraat 49, B-3000 Leuven (Belgium)
I think this is due to the high DOF and probably high t-value causing the significance in matlab to be computed as 0. This gets converted to Inf (infinity) when the -log10(p) calculation is made, then the Inf gets set to 0 when the volume is saved. The quick fix for you is to change fast_selxavg3.m (in $FREESURFER_HOME/fsfast/toolbox). There is a single line that reads:
fsigmat = -log10(pmat)
change this to
fsigmat = -log10(pmat + eps(0));
Let me know if this works and I'll update our source code.
doug
Qi Zhu wrote:
Dear Doug,
I used selxavg3-sess in freesurfer 5.0 to run a big GLM analysis on a monkey data set containing 147 runs. But When I check the results, I found that in the voxels where the t-values should be very high were turned out to have a value of zero. I checked the brainmask, the ces and the cesvar values, they all look fine. It seems that only the ces values in these voxels are a bit high compared to the nearby normal voxels (e.g. 1.17 (ces)/0.00048(cesvar) compared to 0.74/0.00040 in a nearby normal voxel). And it seems that not only me have this problem (similar problem posted in the mailinglist before. http://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg13789.html). Do you know what the problem would be and how to solve it? Thanks very much for your help.
Attached please find an image of the problem.
Best,
Qi Zhu, PostDoc Laboratorium voor Neuro- en Psychofysiologie K.U.Leuven Medical School Herestraat 49, B-3000 Leuven (Belgium)
btw, if you simply "touch" the contrast mat files in the analysis directory, it will only regenerate the contrasts. That should save you some time because it won't have to re-run all of selxavg. doug
Douglas N Greve wrote:
I think this is due to the high DOF and probably high t-value causing the significance in matlab to be computed as 0. This gets converted to Inf (infinity) when the -log10(p) calculation is made, then the Inf gets set to 0 when the volume is saved. The quick fix for you is to change fast_selxavg3.m (in $FREESURFER_HOME/fsfast/toolbox). There is a single line that reads:
fsigmat = -log10(pmat)
change this to
fsigmat = -log10(pmat + eps(0));
Let me know if this works and I'll update our source code.
doug
Qi Zhu wrote:
Dear Doug,
I used selxavg3-sess in freesurfer 5.0 to run a big GLM analysis on a monkey data set containing 147 runs. But When I check the results, I found that in the voxels where the t-values should be very high were turned out to have a value of zero. I checked the brainmask, the ces and the cesvar values, they all look fine. It seems that only the ces values in these voxels are a bit high compared to the nearby normal voxels (e.g. 1.17 (ces)/0.00048(cesvar) compared to 0.74/0.00040 in a nearby normal voxel). And it seems that not only me have this problem (similar problem posted in the mailinglist before. http://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg13789.html).
Do you know what the problem would be and how to solve it? Thanks very much for your help.
Attached please find an image of the problem.
Best,
Qi Zhu, PostDoc Laboratorium voor Neuro- en Psychofysiologie K.U.Leuven Medical School Herestraat 49, B-3000 Leuven (Belgium)
Dear Doug,
Thanks very much for your prompt response. I tried what you suggested and the sig map looks fine now, that's probably the best that we can get for the sig map, because of the limitation of the MATLAB I guess. But for the t map, the values are still not correct for those voxels that highly activated, because they're just calculated based on the sig values, and they can only go up to 17.98 in my case because of the limitation. And I am also quite confused by the way that selxavg3 calculated the t values, which are calculated as the square root of Fsig values, but not the F values. Is that correct?
Best,
Qi Zhu, PostDoc Laboratorium voor Neuro- en Psychofysiologie K.U.Leuven Medical School Herestraat 49, B-3000 Leuven (Belgium) phone +32 16 33 02 09
On Tue, Dec 20, 2011 at 8:54 PM, Douglas N Greve greve@nmr.mgh.harvard.edu wrote:
I think this is due to the high DOF and probably high t-value causing the significance in matlab to be computed as 0. This gets converted to Inf (infinity) when the -log10(p) calculation is made, then the Inf gets set to 0 when the volume is saved. The quick fix for you is to change fast_selxavg3.m (in $FREESURFER_HOME/fsfast/toolbox). There is a single line that reads:
fsigmat = -log10(pmat)
change this to
fsigmat = -log10(pmat + eps(0));
Let me know if this works and I'll update our source code.
doug
Qi Zhu wrote:
Dear Doug,
I used selxavg3-sess in freesurfer 5.0 to run a big GLM analysis on a monkey data set containing 147 runs. But When I check the results, I found that in the voxels where the t-values should be very high were turned out to have a value of zero. I checked the brainmask, the ces and the cesvar values, they all look fine. It seems that only the ces values in these voxels are a bit high compared to the nearby normal voxels (e.g. 1.17 (ces)/0.00048(cesvar) compared to 0.74/0.00040 in a nearby normal voxel). And it seems that not only me have this problem (similar problem posted in the mailinglist before. http://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg13789.html). Do you know what the problem would be and how to solve it? Thanks very much for your help.
Attached please find an image of the problem.
Best,
Qi Zhu, PostDoc Laboratorium voor Neuro- en Psychofysiologie K.U.Leuven Medical School Herestraat 49, B-3000 Leuven (Belgium)
-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422
Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html
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.
The t is computed as the signed square root of the F (not Fsig) volume:
t.vol = sqrt(Fvol.vol) .* sign(ces.vol);
The Fvol (and do t) is not computed from the sig (the other way around). Are you sure that the t-value does not look right in those areas? doug
Qi Zhu wrote:
Dear Doug,
Thanks very much for your prompt response. I tried what you suggested and the sig map looks fine now, that's probably the best that we can get for the sig map, because of the limitation of the MATLAB I guess. But for the t map, the values are still not correct for those voxels that highly activated, because they're just calculated based on the sig values, and they can only go up to 17.98 in my case because of the limitation. And I am also quite confused by the way that selxavg3 calculated the t values, which are calculated as the square root of Fsig values, but not the F values. Is that correct?
Best,
Qi Zhu, PostDoc Laboratorium voor Neuro- en Psychofysiologie K.U.Leuven Medical School Herestraat 49, B-3000 Leuven (Belgium) phone +32 16 33 02 09
On Tue, Dec 20, 2011 at 8:54 PM, Douglas N Greve greve@nmr.mgh.harvard.edu wrote:
I think this is due to the high DOF and probably high t-value causing the significance in matlab to be computed as 0. This gets converted to Inf (infinity) when the -log10(p) calculation is made, then the Inf gets set to 0 when the volume is saved. The quick fix for you is to change fast_selxavg3.m (in $FREESURFER_HOME/fsfast/toolbox). There is a single line that reads:
fsigmat = -log10(pmat)
change this to
fsigmat = -log10(pmat + eps(0));
Let me know if this works and I'll update our source code.
doug
Qi Zhu wrote:
Dear Doug,
I used selxavg3-sess in freesurfer 5.0 to run a big GLM analysis on a monkey data set containing 147 runs. But When I check the results, I found that in the voxels where the t-values should be very high were turned out to have a value of zero. I checked the brainmask, the ces and the cesvar values, they all look fine. It seems that only the ces values in these voxels are a bit high compared to the nearby normal voxels (e.g. 1.17 (ces)/0.00048(cesvar) compared to 0.74/0.00040 in a nearby normal voxel). And it seems that not only me have this problem (similar problem posted in the mailinglist before. http://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg13789.html). Do you know what the problem would be and how to solve it? Thanks very much for your help.
Attached please find an image of the problem.
Best,
Qi Zhu, PostDoc Laboratorium voor Neuro- en Psychofysiologie K.U.Leuven Medical School Herestraat 49, B-3000 Leuven (Belgium)
-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422
Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html
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.
Yes, seems weird. The image that I attached to you in the first mail is actually a t map, so the t value in those areas are also zero. And according to what I read from fast_selxavg3.m, the t value is actually calculated from the fsig, which is: t.vol = sqrt(fsig.vol).*sign(ces.vol), and fsig.vol is calculated from the p value that came from the Ftest. So the t value is calculated wrong in fs 5.0. I am now changing this line to:
Fvol = fast_mat2vol(Fmat,mri.volsize); t.vol = sqrt(Fvol).*sign(ces.vol);
Is this correct?
Best,
Qi Zhu, PostDoc Laboratorium voor Neuro- en Psychofysiologie K.U.Leuven Medical School Herestraat 49, B-3000 Leuven (Belgium)
On Wed, Dec 21, 2011 at 5:31 PM, Douglas N Greve greve@nmr.mgh.harvard.edu wrote:
The t is computed as the signed square root of the F (not Fsig) volume:
t.vol = sqrt(Fvol.vol) .* sign(ces.vol);
The Fvol (and do t) is not computed from the sig (the other way around). Are you sure that the t-value does not look right in those areas?
doug
Qi Zhu wrote:
Dear Doug,
Thanks very much for your prompt response. I tried what you suggested and the sig map looks fine now, that's probably the best that we can get for the sig map, because of the limitation of the MATLAB I guess. But for the t map, the values are still not correct for those voxels that highly activated, because they're just calculated based on the sig values, and they can only go up to 17.98 in my case because of the limitation. And I am also quite confused by the way that selxavg3 calculated the t values, which are calculated as the square root of Fsig values, but not the F values. Is that correct?
Best,
Qi Zhu, PostDoc Laboratorium voor Neuro- en Psychofysiologie K.U.Leuven Medical School Herestraat 49, B-3000 Leuven (Belgium) phone +32 16 33 02 09
On Tue, Dec 20, 2011 at 8:54 PM, Douglas N Greve greve@nmr.mgh.harvard.edu wrote:
I think this is due to the high DOF and probably high t-value causing the significance in matlab to be computed as 0. This gets converted to Inf (infinity) when the -log10(p) calculation is made, then the Inf gets set to 0 when the volume is saved. The quick fix for you is to change fast_selxavg3.m (in $FREESURFER_HOME/fsfast/toolbox). There is a single line that reads:
fsigmat = -log10(pmat)
change this to
fsigmat = -log10(pmat + eps(0));
Let me know if this works and I'll update our source code.
doug
Qi Zhu wrote:
Dear Doug,
I used selxavg3-sess in freesurfer 5.0 to run a big GLM analysis on a monkey data set containing 147 runs. But When I check the results, I found that in the voxels where the t-values should be very high were turned out to have a value of zero. I checked the brainmask, the ces and the cesvar values, they all look fine. It seems that only the ces values in these voxels are a bit high compared to the nearby normal voxels (e.g. 1.17 (ces)/0.00048(cesvar) compared to 0.74/0.00040 in a nearby normal voxel). And it seems that not only me have this problem (similar problem posted in the mailinglist before.
http://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg13789.html). Do you know what the problem would be and how to solve it? Thanks very much for your help.
Attached please find an image of the problem.
Best,
Qi Zhu, PostDoc Laboratorium voor Neuro- en Psychofysiologie K.U.Leuven Medical School Herestraat 49, B-3000 Leuven (Belgium)
-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422
Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html
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.
-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422
Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html
Yes that is correct. That was fixed with version 5.1 so current distributions do the right thing. doug
Qi Zhu wrote:
Yes, seems weird. The image that I attached to you in the first mail is actually a t map, so the t value in those areas are also zero. And according to what I read from fast_selxavg3.m, the t value is actually calculated from the fsig, which is: t.vol = sqrt(fsig.vol).*sign(ces.vol), and fsig.vol is calculated from the p value that came from the Ftest. So the t value is calculated wrong in fs 5.0. I am now changing this line to:
Fvol = fast_mat2vol(Fmat,mri.volsize); t.vol = sqrt(Fvol).*sign(ces.vol);
Is this correct?
Best,
Qi Zhu, PostDoc Laboratorium voor Neuro- en Psychofysiologie K.U.Leuven Medical School Herestraat 49, B-3000 Leuven (Belgium)
On Wed, Dec 21, 2011 at 5:31 PM, Douglas N Greve greve@nmr.mgh.harvard.edu wrote:
The t is computed as the signed square root of the F (not Fsig) volume:
t.vol = sqrt(Fvol.vol) .* sign(ces.vol);
The Fvol (and do t) is not computed from the sig (the other way around). Are you sure that the t-value does not look right in those areas?
doug
Qi Zhu wrote:
Dear Doug,
Thanks very much for your prompt response. I tried what you suggested and the sig map looks fine now, that's probably the best that we can get for the sig map, because of the limitation of the MATLAB I guess. But for the t map, the values are still not correct for those voxels that highly activated, because they're just calculated based on the sig values, and they can only go up to 17.98 in my case because of the limitation. And I am also quite confused by the way that selxavg3 calculated the t values, which are calculated as the square root of Fsig values, but not the F values. Is that correct?
Best,
Qi Zhu, PostDoc Laboratorium voor Neuro- en Psychofysiologie K.U.Leuven Medical School Herestraat 49, B-3000 Leuven (Belgium) phone +32 16 33 02 09
On Tue, Dec 20, 2011 at 8:54 PM, Douglas N Greve greve@nmr.mgh.harvard.edu wrote:
I think this is due to the high DOF and probably high t-value causing the significance in matlab to be computed as 0. This gets converted to Inf (infinity) when the -log10(p) calculation is made, then the Inf gets set to 0 when the volume is saved. The quick fix for you is to change fast_selxavg3.m (in $FREESURFER_HOME/fsfast/toolbox). There is a single line that reads:
fsigmat = -log10(pmat)
change this to
fsigmat = -log10(pmat + eps(0));
Let me know if this works and I'll update our source code.
doug
Qi Zhu wrote:
Dear Doug,
I used selxavg3-sess in freesurfer 5.0 to run a big GLM analysis on a monkey data set containing 147 runs. But When I check the results, I found that in the voxels where the t-values should be very high were turned out to have a value of zero. I checked the brainmask, the ces and the cesvar values, they all look fine. It seems that only the ces values in these voxels are a bit high compared to the nearby normal voxels (e.g. 1.17 (ces)/0.00048(cesvar) compared to 0.74/0.00040 in a nearby normal voxel). And it seems that not only me have this problem (similar problem posted in the mailinglist before.
http://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg13789.html). Do you know what the problem would be and how to solve it? Thanks very much for your help.
Attached please find an image of the problem.
Best,
Qi Zhu, PostDoc Laboratorium voor Neuro- en Psychofysiologie K.U.Leuven Medical School Herestraat 49, B-3000 Leuven (Belgium)
-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422
Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html
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.
-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422
Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html
Ok, that's great. Thanks very much for your help.
Best wishes and Merry Christmas!
Qi Zhu, PostDoc Laboratorium voor Neuro- en Psychofysiologie K.U.Leuven Medical School Herestraat 49, B-3000 Leuven (Belgium)
On Wed, Dec 21, 2011 at 7:40 PM, Douglas N Greve greve@nmr.mgh.harvard.edu wrote:
Yes that is correct. That was fixed with version 5.1 so current distributions do the right thing. doug
Qi Zhu wrote:
Yes, seems weird. The image that I attached to you in the first mail is actually a t map, so the t value in those areas are also zero. And according to what I read from fast_selxavg3.m, the t value is actually calculated from the fsig, which is: t.vol = sqrt(fsig.vol).*sign(ces.vol), and fsig.vol is calculated from the p value that came from the Ftest. So the t value is calculated wrong in fs 5.0. I am now changing this line to:
Fvol = fast_mat2vol(Fmat,mri.volsize); t.vol = sqrt(Fvol).*sign(ces.vol);
Is this correct?
Best,
Qi Zhu, PostDoc Laboratorium voor Neuro- en Psychofysiologie K.U.Leuven Medical School Herestraat 49, B-3000 Leuven (Belgium)
On Wed, Dec 21, 2011 at 5:31 PM, Douglas N Greve greve@nmr.mgh.harvard.edu wrote:
The t is computed as the signed square root of the F (not Fsig) volume:
t.vol = sqrt(Fvol.vol) .* sign(ces.vol);
The Fvol (and do t) is not computed from the sig (the other way around). Are you sure that the t-value does not look right in those areas?
doug
Qi Zhu wrote:
Dear Doug,
Thanks very much for your prompt response. I tried what you suggested and the sig map looks fine now, that's probably the best that we can get for the sig map, because of the limitation of the MATLAB I guess. But for the t map, the values are still not correct for those voxels that highly activated, because they're just calculated based on the sig values, and they can only go up to 17.98 in my case because of the limitation. And I am also quite confused by the way that selxavg3 calculated the t values, which are calculated as the square root of Fsig values, but not the F values. Is that correct?
Best,
Qi Zhu, PostDoc Laboratorium voor Neuro- en Psychofysiologie K.U.Leuven Medical School Herestraat 49, B-3000 Leuven (Belgium) phone +32 16 33 02 09
On Tue, Dec 20, 2011 at 8:54 PM, Douglas N Greve greve@nmr.mgh.harvard.edu wrote:
I think this is due to the high DOF and probably high t-value causing the significance in matlab to be computed as 0. This gets converted to Inf (infinity) when the -log10(p) calculation is made, then the Inf gets set to 0 when the volume is saved. The quick fix for you is to change fast_selxavg3.m (in $FREESURFER_HOME/fsfast/toolbox). There is a single line that reads:
fsigmat = -log10(pmat)
change this to
fsigmat = -log10(pmat + eps(0));
Let me know if this works and I'll update our source code.
doug
Qi Zhu wrote:
Dear Doug,
I used selxavg3-sess in freesurfer 5.0 to run a big GLM analysis on a monkey data set containing 147 runs. But When I check the results, I found that in the voxels where the t-values should be very high were turned out to have a value of zero. I checked the brainmask, the ces and the cesvar values, they all look fine. It seems that only the ces values in these voxels are a bit high compared to the nearby normal voxels (e.g. 1.17 (ces)/0.00048(cesvar) compared to 0.74/0.00040 in a nearby normal voxel). And it seems that not only me have this problem (similar problem posted in the mailinglist before.
http://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg13789.html). Do you know what the problem would be and how to solve it? Thanks very much for your help.
Attached please find an image of the problem.
Best,
Qi Zhu, PostDoc Laboratorium voor Neuro- en Psychofysiologie K.U.Leuven Medical School Herestraat 49, B-3000 Leuven (Belgium)
-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422
Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html
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.
-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422
Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html
-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422
Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html
freesurfer@nmr.mgh.harvard.edu
-
Douglas N Greve -
Qi Zhu