Hi Xiangchuan,
In recent versions of Freesurfer, the atlas $FREESURFER_HOME/average/711-2C_as_mni_average_305.4dfp.img and the registration to that atlas found in $SUBJECTS_DIR/subjid/mri/transforms/talairach.xfm (created by talairach_avi, a script which calls imgreg_4dfp) is used to estimate ICV (the '711-2C' atlas target image includes skull).
So only talairach.xfm needs to be accurate for ICV to be good. You need to check it with the corresponding atlas (maybe Nick can verify that this is correct). Here http://surfer.nmr.mgh.harvard.edu/fswiki/eTIV is some background.
Also: "ICV (sum of effects, but is estimated)". No, ICV should not show any effects (total intracranial volume should not change with atrophy). That is why it is usually used for normalizing. But Micheal is right, if you want to show 'extra' effect of specific regions, you're probably fine.
Cheers, Martin
On Sat, 2010-01-09 at 20:08 -0500, Xiangchuan Chen wrote:
Hi Martin, Bruce and Mike,
Thank you for all your suggestions.
For Martin's suggestion: I found that it is hard to really fix the talairach transform for a lot of my subjects. In most cases, I tried to match the outlines between target and moveable images when running "tkregister2 --mgz --fstal --s subjid" (step1). For some of these subjects, I also checked the talairach transform by running "tkregister2 --xfm talairach.xfm --targ $FREESURFER_HOME/average/RB_all_withskull_2008-03-26.gca --mov ../nu_noneck.mgz --reg junk" (step2), and found that the moveable and target images were not aligned very well (looked worse than in step1). In this case, do I need to further edit the talairach transform in step2?
The reason for adjusting with total brain volume: In our study, we found difference in ICV and a lot of cortical and sub-cortical regions between groups. Then, we want to know if some of these regions showed extra effect than other regions. For this purpose, using total brain volume (sum of effects on these regions) may be better than using ICV (sum of effects, but is estimated). Please correct me if I am wrong here.
For Bruce's suggestion: the cerebral white matter volume (calculated with mris_wm_volume) is already in aseg.stats. Is there a way to calculate the cerebral gray matter volume (surface based, more accurate)? I assume that I can use the sum of the volumes from aparc.stats.
Thanks, Xiangchuan
-----Original Message----- From: Martin Reuter [mailto:mreuter@nmr.mgh.harvard.edu] Sent: Saturday, January 09, 2010 13:34 To: Bruce Fischl Cc: Michael Harms; Xiangchuan Chen; freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] ICV/brain volume
Or you can fix the talairach transform and get the icv the standart way.
When dividing by total brain vol, you undo atrophy. That's why icv is used.
Martin
On Jan 9, 2010, at 13:01, Bruce Fischl fischl@nmr.mgh.harvard.edu wrote:
or you could use mris_wm_volume, which excludes the interior of the ventricles and gray matter structures, and mris_volume and the aseg
to
calculate it (adding the structures you are interested back in). Maybe we should start generating this # by default. Would enough people find
it
useful?
cheers, Bruce
On Sat, 9 Jan 2010, Michael Harms wrote:
You're correct that a small amount of hippocampus and amygdala falls outside the pial surface. However, it will be a TINY percentage of the whole brain volume, and MUCH smaller than variation of brain volume across individuals. So the little bit of hippocampus and amygdala that is "missed" will be inconsequential in my opinion.
cheers, Mike H.
Hi Mike,
Thanks for your suggestions. I have learned this idea from previous posts (including yours), but found that there is an issue for this approach when I checked the pial surface with tkmedit (tkmedit subjid T1.mgz
lh.pial
-aux-surface rh.pial): part of two important brain structures, amygdala and hippocampus, is not enclosed in the pial surface. Is there a way to resolve this issue?
Thanks, Xiangchuan
-----Original Message----- From: Michael Harms [mailto:mharms@conte.wustl.edu] Sent: Friday, January 08, 2010 15:55 To: Xiangchuan Chen Cc: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] ICV/brain volume
Here is an approach that will be easier, and involves less back and forth between the surface and volume-based streams (and thus should be more accurate): Just use 'mris_volume' to get the volume of everything enclosed by the pial surface, which if I recall correctly will include the lateral ventricles, choroid plexus, and usually most of the inferior lateral ventricles (but not the 3rd and 4th ventricles). Then, if you want to exclude ventricles from your definition of "brain volume", just subtract the Lateral-Ventricle, choroid-plexus, and Inf-Lat-Vent volumes (taken from the aseg.stats).
cheers, Mike H.
On Fri, 2010-01-08 at 15:13 -0500, Xiangchuan Chen wrote:
Dear FreeSurfer experts,
After searching on FreeSurfer wiki and mailing list, I know that the ICV reported in aseg.stats is an estimated value based on talairach transformation. So, if the talairach transformation for a subject's head
is
not good, the ICV for this subject is not accurate, am I right?
If this is the case, it may not be suitable to use ICV to adjust the volumetric data in aseg.stats and ?h.aparc.stats for some special
subjects.
I am thinking of using "brain volume" for this purpose. Here are the steps to get it:
- Get the surface-based volume of cerebral white matter
(aseg.stats).
- Add up the volumes of all cortical regions to get the surface-
based volume of cerebral cortex (aparc.stats).
- Add up the above volumes (including both hemispheres) and the
volumes
of
all other regions listed in aseg.stats (except for cerebral white matter
and
cerebral cortex) to get the brain volume.
Is this a suitable way to obtain more accurate brain volume?
Thanks in advance and any suggestions are greatly appreciated,
Xiangchuan Chen
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
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Hi Martin,
I usually follow the "FreeSurfer Troubleshooting Reconstruction Work Flow" to do recon-all. So, I run "tkregister2 --mgz --s your_subject_name --fstal" to edit talairach transform. On ~/ fswiki/eTIV it is suggested to run " tkregister2 --xfm talairach.xfm --targ $FREESURFER_HOME/average/RB_all_withskull_2008-03-26.gca --mov ../nu_noneck.mgz --reg junk " to check the quality of talairach.xfm. I guess I should follow the method on the eTIV wiki to fix talairach transform, and the "Work Flow" wiki might need to be revised?
For ICV: I also think that for normally developed, and then impaired or degenerated brains, ICV is a good index for normalization purpose. For brains with development problems, total brain volume may be a better index than ICV. In either case, we need accurate value for these indices. I remember that several previous posts mentioned that ICV could be "measured" based on T2 images. Hope that this feature will be implemented in FreeSurfer.
Thanks, Xiangchuan
-----Original Message----- From: Martin Reuter [mailto:mreuter@nmr.mgh.harvard.edu] Sent: Saturday, January 09, 2010 22:05 To: Xiangchuan Chen; Nick Schmansky Cc: 'Bruce Fischl'; 'Michael Harms'; freesurfer@nmr.mgh.harvard.edu Subject: RE: [Freesurfer] ICV/brain volume
Hi Xiangchuan,
In recent versions of Freesurfer, the atlas $FREESURFER_HOME/average/711-2C_as_mni_average_305.4dfp.img and the registration to that atlas found in $SUBJECTS_DIR/subjid/mri/transforms/talairach.xfm (created by talairach_avi, a script which calls imgreg_4dfp) is used to estimate ICV (the '711-2C' atlas target image includes skull).
So only talairach.xfm needs to be accurate for ICV to be good. You need to check it with the corresponding atlas (maybe Nick can verify that this is correct). Here http://surfer.nmr.mgh.harvard.edu/fswiki/eTIV is some background.
Also: "ICV (sum of effects, but is estimated)". No, ICV should not show any effects (total intracranial volume should not change with atrophy). That is why it is usually used for normalizing. But Micheal is right, if you want to show 'extra' effect of specific regions, you're probably fine.
Cheers, Martin
On Sat, 2010-01-09 at 20:08 -0500, Xiangchuan Chen wrote:
Hi Martin, Bruce and Mike,
Thank you for all your suggestions.
For Martin's suggestion: I found that it is hard to really fix the
talairach
transform for a lot of my subjects. In most cases, I tried to match
the
outlines between target and moveable images when running "tkregister2
--mgz
--fstal --s subjid" (step1). For some of these subjects, I also
checked the
talairach transform by running "tkregister2 --xfm talairach.xfm --
targ
$FREESURFER_HOME/average/RB_all_withskull_2008-03-26.gca --mov ../nu_noneck.mgz --reg junk" (step2), and found that the moveable and
target
images were not aligned very well (looked worse than in step1). In
this
case, do I need to further edit the talairach transform in step2?
The reason for adjusting with total brain volume: In our study, we
found
difference in ICV and a lot of cortical and sub-cortical regions
between
groups. Then, we want to know if some of these regions showed extra
effect
than other regions. For this purpose, using total brain volume (sum
of
effects on these regions) may be better than using ICV (sum of
effects, but
is estimated). Please correct me if I am wrong here.
For Bruce's suggestion: the cerebral white matter volume (calculated
with
mris_wm_volume) is already in aseg.stats. Is there a way to calculate
the
cerebral gray matter volume (surface based, more accurate)? I assume
that I
can use the sum of the volumes from aparc.stats.
Thanks, Xiangchuan
-----Original Message----- From: Martin Reuter [mailto:mreuter@nmr.mgh.harvard.edu] Sent: Saturday, January 09, 2010 13:34 To: Bruce Fischl Cc: Michael Harms; Xiangchuan Chen; freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] ICV/brain volume
Or you can fix the talairach transform and get the icv the standart
way.
When dividing by total brain vol, you undo atrophy. That's why icv
is
used.
Martin
On Jan 9, 2010, at 13:01, Bruce Fischl fischl@nmr.mgh.harvard.edu wrote:
or you could use mris_wm_volume, which excludes the interior of
the
ventricles and gray matter structures, and mris_volume and the
aseg
to
calculate it (adding the structures you are interested back in). Maybe we should start generating this # by default. Would enough people
find
it
useful?
cheers, Bruce
On Sat, 9 Jan 2010, Michael Harms wrote:
You're correct that a small amount of hippocampus and amygdala
falls
outside the pial surface. However, it will be a TINY percentage
of
the whole brain volume, and MUCH smaller than variation of brain
volume
across individuals. So the little bit of hippocampus and amygdala that
is
"missed" will be inconsequential in my opinion.
cheers, Mike H.
Hi Mike,
Thanks for your suggestions. I have learned this idea from previous posts (including yours), but found that there is an issue for this approach when I checked the pial surface with tkmedit (tkmedit subjid T1.mgz
lh.pial
-aux-surface rh.pial): part of two important brain structures, amygdala and hippocampus, is not enclosed in the pial surface. Is there a
way to
resolve this issue?
Thanks, Xiangchuan
-----Original Message----- From: Michael Harms [mailto:mharms@conte.wustl.edu] Sent: Friday, January 08, 2010 15:55 To: Xiangchuan Chen Cc: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] ICV/brain volume
Here is an approach that will be easier, and involves less back
and
forth between the surface and volume-based streams (and thus should be more accurate): Just use 'mris_volume' to get the volume of everything enclosed by the pial surface, which if I recall correctly will include the lateral ventricles, choroid plexus, and usually most of the inferior lateral ventricles (but not the 3rd and 4th ventricles). Then,
if
you want to exclude ventricles from your definition of "brain
volume",
just subtract the Lateral-Ventricle, choroid-plexus, and Inf-Lat-
Vent
volumes (taken from the aseg.stats).
cheers, Mike H.
On Fri, 2010-01-08 at 15:13 -0500, Xiangchuan Chen wrote: > Dear FreeSurfer experts, > > After searching on FreeSurfer wiki and mailing list, I know
that
> the ICV > reported in aseg.stats is an estimated value based on
talairach
> transformation. So, if the talairach transformation for a > subject's head is > not good, the ICV for this subject is not accurate, am I right? > > If this is the case, it may not be suitable to use ICV to
adjust
> the > volumetric data in aseg.stats and ?h.aparc.stats for some
special
subjects. > I am thinking of using "brain volume" for this purpose. Here
are
> the > steps > to get it: > > 1. Get the surface-based volume of cerebral white matter > (aseg.stats). > > 2. Add up the volumes of all cortical regions to get the
surface-
> based > volume of cerebral cortex (aparc.stats). > > 3. Add up the above volumes (including both hemispheres) and
the
> volumes of > all other regions listed in aseg.stats (except for cerebral
white
> matter and > cerebral cortex) to get the brain volume. > > Is this a suitable way to obtain more accurate brain volume? > > Thanks in advance and any suggestions are greatly appreciated, > > Xiangchuan Chen > > > _______________________________________________ > 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
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
freesurfer@nmr.mgh.harvard.edu
-
Martin Reuter -
Xiangchuan Chen