Re: [Freesurfer] Verifying ICV in FS5

[Bruce Fischl]

Hi Jeff

it looks pretty far off. Typically you want the surfaces to match the 
general shape of the brain


cheers
Bruce
On Tue, 7 Feb 2012, Jeff Sadino wrote:


Thank you Martin and Nick for the clarification about how to verify the 
talairach transformation in the newer FreeSurfers.  I think I am going to use 
the
-use-mritotal command, as this was giving more reasonable ICVs around 1.3L - 
1.7L.  I am still manually checking the outliers.  According to the ICV 
measurement,
this one seems way off, but I'm not sure how to judge the severity of the 
transform.  On a scale of 1 to 10, 1 being bad, how would this one rate?  How 
much would
this cause the ICV to be off?
Thank you,
Jeff

On Mon, Jan 30, 2012 at 8:47 AM, Nick Schmansky  
wrote:
  Jeff,

  to check the tal reg, use:

  tkregister2 --s subjid --fstal  --surfs

  i've corrected the eTIV page to reflect this.  --fstal loads mni305 as
  the target.  the target used in the tal reg step (711-2C...) is
  registered with the mni305.  note that --surfs is optional, as the
  surfaces may not exist yet.

  Nick


  On Fri, 2012-01-27 at 18:53 -0500, Martin Reuter wrote:
  > Hi Jeff,
  >
  > I am not sure about which atlas, but concerning the transforms:
  >
  > 4.2 and earlier: talairach_with_skull.lta
  > later versions: talairach.xfm
  >
  > I am pretty convinced that the way to check as described here
  > http://surfer.nmr.mgh.harvard.edu/fswiki/eTIV
  >
  > cd $SUBJECTS_DIR/subject/mri/transforms
  > tkregister2 --xfm talairach.xfm \
  >   --targ $FREESURFER_HOME/average/RB_all_withskull_2008-03-26.gca \
  >   --mov ../nu_noneck.mgz --reg junk
  >
  > is correct in newer FS versions, but maybe double check with Nick.
  >
  > Best, Martin
  >
  >
  > On Fri, 2012-01-27 at 13:39 -1000, Jeff Sadino wrote:
  > > bump :)
  > >
  > > any feedback on this would be very much appreciated.  Thank you!
  > >
  > > Jeff
  > >
  > > On Mon, Jan 23, 2012 at 6:56 PM, Jeff Sadino 
 wrote:
  > > > Hello,
  > > >
  > > > Thank you Nick, Martin, and Doug for your previous feedback about 
ICV.
  > > >  I have a couple more questions that I am hoping you could clarify 
for
  > > > me.
  > > >
  > > > To verify my ICV value, some pages on the mailing list and wiki say 
to
  > > > use this command:
  > > > tkregister2 --s subject --fstal --surf
  > > >
  > > > Other places say to use this:
  > > > tkregister2 --xfm talairach.xfm --targ
  > > > $FREESURFER_HOME/average/RB_all_withskull_2008-03-26.gca --mov
  > > > ../nu_noneck.mgz --reg junk
  > > >
  > > > However, on this wiki page
  > > > (http://surfer.nmr.mgh.harvard.edu/fswiki/eTIV), it says that post
  > > > fs4.2.0, the template is this file:
  > > > 711-2C_as_mni_average_305.4dfp.img.  In the first tkregister2 
command
  > > > above, doesn't the --fstal flag use the talairach template?  But
  > > > shouldn't it be using the 711 template?  And since I am using 
fs5.0.0,
  > > > does that mean that I should not be using the RB template?
  > > > Ultimately, what is the correct way to verify the ICV in fs 5.0.0?
  > > >
  > > > Thank you very much,
  > > > Jeff Sadino
  > > >
  > > > On Mon, Mar 1, 2010 at 7:57 AM, Douglas N Greve
  > > >  wrote:
  > > >> It's hard to tell from the coronal, but it looks ok. The problem 
is more
  > > >> visible in the sagittal.
  > > >>
  > > >> doug
  > > >>
  > > >> Jeff Sadino wrote:
  > > >>>
  > > >>> Hello Doug,
  > > >>>
  > > >>> I loaded up the tkregister2 program and I think I see some poor
  > > >>> transformations.  I've never really used it before though.  Am I 
just
  > > >>> looking at whether or not the green line lines up with the target 
image wm?
  > > >>>  Can you look at my pictures real quick?  040002_S04_test is from 
fs4 that
  > > >>> is having problems.  040002_S04 is from fs3 and has a good icv 
value.  Some
  > > >>> of the blank spots are from my graphics drivers and unrelated to 
fs.
  > > >>>  tkregister2 also said the determinant of 040002_S04_test was 
1.49544.
  > > >>>
  > > >>> Martin,
  > > >>>
  > > >>> Thank you for the idea.  I did this, and got a much better icv.  
On fs3,
  > > >>> where the icv was good, the number was 1,767,481.  On fs4, it was 
1,302,696.
  > > >>>  Using your idea on fs4, it was 1,819,485.
  > > >>>
  > > >>> Thank you so much to everyone for your help!
  > > >>> Jeff
  > > >>>
  > > >>> On Thu, Feb 25, 2010 at 8:53 PM, Martin Kavec 
 > >>> > wrote:
  > > >>>
  > > >>>    Hi Jeff,
  > > >>>
  > > >>>    I see this relatively often. Eventhough the talairach 
registration
  > > >>>    is fine, the
  > > >>

Re: [Freesurfer] Verifying ICV in FS5

[Nick Schmansky]

Jeff,

to check the tal reg, use:

tkregister2 --s subjid --fstal  --surfs

i've corrected the eTIV page to reflect this.  --fstal loads mni305 as
the target.  the target used in the tal reg step (711-2C...) is
registered with the mni305.  note that --surfs is optional, as the
surfaces may not exist yet.

Nick


On Fri, 2012-01-27 at 18:53 -0500, Martin Reuter wrote:
> Hi Jeff,
> 
> I am not sure about which atlas, but concerning the transforms:
> 
> 4.2 and earlier: talairach_with_skull.lta
> later versions: talairach.xfm
> 
> I am pretty convinced that the way to check as described here
> http://surfer.nmr.mgh.harvard.edu/fswiki/eTIV
> 
> cd $SUBJECTS_DIR/subject/mri/transforms
> tkregister2 --xfm talairach.xfm \
>   --targ $FREESURFER_HOME/average/RB_all_withskull_2008-03-26.gca \
>   --mov ../nu_noneck.mgz --reg junk
> 
> is correct in newer FS versions, but maybe double check with Nick.
> 
> Best, Martin
> 
> 
> On Fri, 2012-01-27 at 13:39 -1000, Jeff Sadino wrote:
> > bump :)
> > 
> > any feedback on this would be very much appreciated.  Thank you!
> > 
> > Jeff
> > 
> > On Mon, Jan 23, 2012 at 6:56 PM, Jeff Sadino  
> > wrote:
> > > Hello,
> > >
> > > Thank you Nick, Martin, and Doug for your previous feedback about ICV.
> > >  I have a couple more questions that I am hoping you could clarify for
> > > me.
> > >
> > > To verify my ICV value, some pages on the mailing list and wiki say to
> > > use this command:
> > > tkregister2 --s subject --fstal --surf
> > >
> > > Other places say to use this:
> > > tkregister2 --xfm talairach.xfm --targ
> > > $FREESURFER_HOME/average/RB_all_withskull_2008-03-26.gca --mov
> > > ../nu_noneck.mgz --reg junk
> > >
> > > However, on this wiki page
> > > (http://surfer.nmr.mgh.harvard.edu/fswiki/eTIV), it says that post
> > > fs4.2.0, the template is this file:
> > > 711-2C_as_mni_average_305.4dfp.img.  In the first tkregister2 command
> > > above, doesn't the --fstal flag use the talairach template?  But
> > > shouldn't it be using the 711 template?  And since I am using fs5.0.0,
> > > does that mean that I should not be using the RB template?
> > > Ultimately, what is the correct way to verify the ICV in fs 5.0.0?
> > >
> > > Thank you very much,
> > > Jeff Sadino
> > >
> > > On Mon, Mar 1, 2010 at 7:57 AM, Douglas N Greve
> > >  wrote:
> > >> It's hard to tell from the coronal, but it looks ok. The problem is more
> > >> visible in the sagittal.
> > >>
> > >> doug
> > >>
> > >> Jeff Sadino wrote:
> > >>>
> > >>> Hello Doug,
> > >>>
> > >>> I loaded up the tkregister2 program and I think I see some poor
> > >>> transformations.  I've never really used it before though.  Am I just
> > >>> looking at whether or not the green line lines up with the target image 
> > >>> wm?
> > >>>  Can you look at my pictures real quick?  040002_S04_test is from fs4 
> > >>> that
> > >>> is having problems.  040002_S04 is from fs3 and has a good icv value.  
> > >>> Some
> > >>> of the blank spots are from my graphics drivers and unrelated to fs.
> > >>>  tkregister2 also said the determinant of 040002_S04_test was 1.49544.
> > >>>
> > >>> Martin,
> > >>>
> > >>> Thank you for the idea.  I did this, and got a much better icv.  On fs3,
> > >>> where the icv was good, the number was 1,767,481.  On fs4, it was 
> > >>> 1,302,696.
> > >>>  Using your idea on fs4, it was 1,819,485.
> > >>>
> > >>> Thank you so much to everyone for your help!
> > >>> Jeff
> > >>>
> > >>> On Thu, Feb 25, 2010 at 8:53 PM, Martin Kavec  > >>> > wrote:
> > >>>
> > >>>Hi Jeff,
> > >>>
> > >>>I see this relatively often. Eventhough the talairach registration
> > >>>is fine, the
> > >>>ICV is underestimated. In this case I take old ICV, which helps.
> > >>>
> > >>>mri_segstats --subject $MYSUBJECT --old-etiv-only
> > >>>
> > >>>This is in 99% of these cases a cure. This is not just a blind
> > >>>attempt. In
> > >>>these cases I then compare each ICV normalized aseg volume against
> > >>>an age
> > >>>matched group and only if the old-etiv-only ICV gives better
> > >>>estimates I use
> > >>>it.
> > >>>
> > >>>Cheers,
> > >>>
> > >>>Martin
> > >>>
> > >>>On Thursday 25 February 2010 23:05:10 Jeff Sadino wrote:
> > >>>> Thanks Doug.  That is very good news to hear!
> > >>>>
> > >>>> Maybe I spoke too soon about fixing the problem.  I looked at 4 
> > >>> more
> > >>>> subjects, and even though the icv value is unique, its magnitude
> > >>>seems
> > >>>> wrong.  Often times, it is smaller than the BrainSegVol, which
> > >>>doesn't make
> > >>>> much sense and it is also low overall.
> > >>>>
> > >>>>
> > >>>> # subjectname 090167_F01
> > >>>> # Measure BrainMask, BrainMaskNVox, Number of Brain Mask Voxels,
> > >>>1639959,
> > >>>> unitless
> > >>>> # Measure BrainMask, BrainMaskVol, Brain Mask Volume,
> > >>>1639959.00, mm^3
> > >>>> # Measure BrainSegNotVe

Re: [Freesurfer] Verifying ICV in FS5

[Martin Reuter]

Hi Jeff,

I am not sure about which atlas, but concerning the transforms:

4.2 and earlier: talairach_with_skull.lta
later versions: talairach.xfm

I am pretty convinced that the way to check as described here
http://surfer.nmr.mgh.harvard.edu/fswiki/eTIV

cd $SUBJECTS_DIR/subject/mri/transforms
tkregister2 --xfm talairach.xfm \
  --targ $FREESURFER_HOME/average/RB_all_withskull_2008-03-26.gca \
  --mov ../nu_noneck.mgz --reg junk

is correct in newer FS versions, but maybe double check with Nick.

Best, Martin


On Fri, 2012-01-27 at 13:39 -1000, Jeff Sadino wrote:
> bump :)
> 
> any feedback on this would be very much appreciated.  Thank you!
> 
> Jeff
> 
> On Mon, Jan 23, 2012 at 6:56 PM, Jeff Sadino  wrote:
> > Hello,
> >
> > Thank you Nick, Martin, and Doug for your previous feedback about ICV.
> >  I have a couple more questions that I am hoping you could clarify for
> > me.
> >
> > To verify my ICV value, some pages on the mailing list and wiki say to
> > use this command:
> > tkregister2 --s subject --fstal --surf
> >
> > Other places say to use this:
> > tkregister2 --xfm talairach.xfm --targ
> > $FREESURFER_HOME/average/RB_all_withskull_2008-03-26.gca --mov
> > ../nu_noneck.mgz --reg junk
> >
> > However, on this wiki page
> > (http://surfer.nmr.mgh.harvard.edu/fswiki/eTIV), it says that post
> > fs4.2.0, the template is this file:
> > 711-2C_as_mni_average_305.4dfp.img.  In the first tkregister2 command
> > above, doesn't the --fstal flag use the talairach template?  But
> > shouldn't it be using the 711 template?  And since I am using fs5.0.0,
> > does that mean that I should not be using the RB template?
> > Ultimately, what is the correct way to verify the ICV in fs 5.0.0?
> >
> > Thank you very much,
> > Jeff Sadino
> >
> > On Mon, Mar 1, 2010 at 7:57 AM, Douglas N Greve
> >  wrote:
> >> It's hard to tell from the coronal, but it looks ok. The problem is more
> >> visible in the sagittal.
> >>
> >> doug
> >>
> >> Jeff Sadino wrote:
> >>>
> >>> Hello Doug,
> >>>
> >>> I loaded up the tkregister2 program and I think I see some poor
> >>> transformations.  I've never really used it before though.  Am I just
> >>> looking at whether or not the green line lines up with the target image 
> >>> wm?
> >>>  Can you look at my pictures real quick?  040002_S04_test is from fs4 that
> >>> is having problems.  040002_S04 is from fs3 and has a good icv value.  
> >>> Some
> >>> of the blank spots are from my graphics drivers and unrelated to fs.
> >>>  tkregister2 also said the determinant of 040002_S04_test was 1.49544.
> >>>
> >>> Martin,
> >>>
> >>> Thank you for the idea.  I did this, and got a much better icv.  On fs3,
> >>> where the icv was good, the number was 1,767,481.  On fs4, it was 
> >>> 1,302,696.
> >>>  Using your idea on fs4, it was 1,819,485.
> >>>
> >>> Thank you so much to everyone for your help!
> >>> Jeff
> >>>
> >>> On Thu, Feb 25, 2010 at 8:53 PM, Martin Kavec  >>> > wrote:
> >>>
> >>>Hi Jeff,
> >>>
> >>>I see this relatively often. Eventhough the talairach registration
> >>>is fine, the
> >>>ICV is underestimated. In this case I take old ICV, which helps.
> >>>
> >>>mri_segstats --subject $MYSUBJECT --old-etiv-only
> >>>
> >>>This is in 99% of these cases a cure. This is not just a blind
> >>>attempt. In
> >>>these cases I then compare each ICV normalized aseg volume against
> >>>an age
> >>>matched group and only if the old-etiv-only ICV gives better
> >>>estimates I use
> >>>it.
> >>>
> >>>Cheers,
> >>>
> >>>Martin
> >>>
> >>>On Thursday 25 February 2010 23:05:10 Jeff Sadino wrote:
> >>>> Thanks Doug.  That is very good news to hear!
> >>>>
> >>>> Maybe I spoke too soon about fixing the problem.  I looked at 4 more
> >>>> subjects, and even though the icv value is unique, its magnitude
> >>>seems
> >>>> wrong.  Often times, it is smaller than the BrainSegVol, which
> >>>doesn't make
> >>>> much sense and it is also low overall.
> >>>>
> >>>>
> >>>> # subjectname 090167_F01
> >>>> # Measure BrainMask, BrainMaskNVox, Number of Brain Mask Voxels,
> >>>1639959,
> >>>> unitless
> >>>> # Measure BrainMask, BrainMaskVol, Brain Mask Volume,
> >>>1639959.00, mm^3
> >>>> # Measure BrainSegNotVent, BrainSegVolNotVent, Brain
> >>>Segmentation Volume
> >>>> Without Ventricles, 1289194.00, mm^3
> >>>> # Measure BrainSeg, BrainSegNVox, Number of Brain Segmentation
> >>>Voxels,
> >>>> 1299686, unitless
> >>>> # Measure BrainSeg, BrainSegVol, Brain Segmentation Volume,
> >>>1299686.00,
> >>>> mm^3
> >>>> # Measure IntraCranialVol, ICV, Intracranial Volume,
> >>>1105595.114652, mm^3
> >>>>
> >>>> # subjectname 090182_S01
> >>>> # Measure BrainMask, BrainMaskNVox, Number of Brain Mask Voxels,
> >>>1599981,
> >>>> unitless
> >>>> # Measure BrainMask, BrainMaskVol, Bra

Re: [Freesurfer] Verifying ICV in FS5

[Jeff Sadino]

bump :)

any feedback on this would be very much appreciated.  Thank you!

Jeff

On Mon, Jan 23, 2012 at 6:56 PM, Jeff Sadino  wrote:
> Hello,
>
> Thank you Nick, Martin, and Doug for your previous feedback about ICV.
>  I have a couple more questions that I am hoping you could clarify for
> me.
>
> To verify my ICV value, some pages on the mailing list and wiki say to
> use this command:
> tkregister2 --s subject --fstal --surf
>
> Other places say to use this:
> tkregister2 --xfm talairach.xfm --targ
> $FREESURFER_HOME/average/RB_all_withskull_2008-03-26.gca --mov
> ../nu_noneck.mgz --reg junk
>
> However, on this wiki page
> (http://surfer.nmr.mgh.harvard.edu/fswiki/eTIV), it says that post
> fs4.2.0, the template is this file:
> 711-2C_as_mni_average_305.4dfp.img.  In the first tkregister2 command
> above, doesn't the --fstal flag use the talairach template?  But
> shouldn't it be using the 711 template?  And since I am using fs5.0.0,
> does that mean that I should not be using the RB template?
> Ultimately, what is the correct way to verify the ICV in fs 5.0.0?
>
> Thank you very much,
> Jeff Sadino
>
> On Mon, Mar 1, 2010 at 7:57 AM, Douglas N Greve
>  wrote:
>> It's hard to tell from the coronal, but it looks ok. The problem is more
>> visible in the sagittal.
>>
>> doug
>>
>> Jeff Sadino wrote:
>>>
>>> Hello Doug,
>>>
>>> I loaded up the tkregister2 program and I think I see some poor
>>> transformations.  I've never really used it before though.  Am I just
>>> looking at whether or not the green line lines up with the target image wm?
>>>  Can you look at my pictures real quick?  040002_S04_test is from fs4 that
>>> is having problems.  040002_S04 is from fs3 and has a good icv value.  Some
>>> of the blank spots are from my graphics drivers and unrelated to fs.
>>>  tkregister2 also said the determinant of 040002_S04_test was 1.49544.
>>>
>>> Martin,
>>>
>>> Thank you for the idea.  I did this, and got a much better icv.  On fs3,
>>> where the icv was good, the number was 1,767,481.  On fs4, it was 1,302,696.
>>>  Using your idea on fs4, it was 1,819,485.
>>>
>>> Thank you so much to everyone for your help!
>>> Jeff
>>>
>>> On Thu, Feb 25, 2010 at 8:53 PM, Martin Kavec >> > wrote:
>>>
>>>    Hi Jeff,
>>>
>>>    I see this relatively often. Eventhough the talairach registration
>>>    is fine, the
>>>    ICV is underestimated. In this case I take old ICV, which helps.
>>>
>>>    mri_segstats --subject $MYSUBJECT --old-etiv-only
>>>
>>>    This is in 99% of these cases a cure. This is not just a blind
>>>    attempt. In
>>>    these cases I then compare each ICV normalized aseg volume against
>>>    an age
>>>    matched group and only if the old-etiv-only ICV gives better
>>>    estimates I use
>>>    it.
>>>
>>>    Cheers,
>>>
>>>    Martin
>>>
>>>    On Thursday 25 February 2010 23:05:10 Jeff Sadino wrote:
>>>    > Thanks Doug.  That is very good news to hear!
>>>    >
>>>    > Maybe I spoke too soon about fixing the problem.  I looked at 4 more
>>>    > subjects, and even though the icv value is unique, its magnitude
>>>    seems
>>>    > wrong.  Often times, it is smaller than the BrainSegVol, which
>>>    doesn't make
>>>    > much sense and it is also low overall.
>>>    >
>>>    >
>>>    > # subjectname 090167_F01
>>>    > # Measure BrainMask, BrainMaskNVox, Number of Brain Mask Voxels,
>>>    1639959,
>>>    > unitless
>>>    > # Measure BrainMask, BrainMaskVol, Brain Mask Volume,
>>>    1639959.00, mm^3
>>>    > # Measure BrainSegNotVent, BrainSegVolNotVent, Brain
>>>    Segmentation Volume
>>>    > Without Ventricles, 1289194.00, mm^3
>>>    > # Measure BrainSeg, BrainSegNVox, Number of Brain Segmentation
>>>    Voxels,
>>>    > 1299686, unitless
>>>    > # Measure BrainSeg, BrainSegVol, Brain Segmentation Volume,
>>>    1299686.00,
>>>    > mm^3
>>>    > # Measure IntraCranialVol, ICV, Intracranial Volume,
>>>    1105595.114652, mm^3
>>>    >
>>>    > # subjectname 090182_S01
>>>    > # Measure BrainMask, BrainMaskNVox, Number of Brain Mask Voxels,
>>>    1599981,
>>>    > unitless
>>>    > # Measure BrainMask, BrainMaskVol, Brain Mask Volume,
>>>    1599981.00, mm^3
>>>    > # Measure BrainSegNotVent, BrainSegVolNotVent, Brain
>>>    Segmentation Volume
>>>    > Without Ventricles, 1199694.00, mm^3
>>>    > # Measure BrainSeg, BrainSegNVox, Number of Brain Segmentation
>>>    Voxels,
>>>    > 1205382, unitless
>>>    > # Measure BrainSeg, BrainSegVol, Brain Segmentation Volume,
>>>    1205382.00,
>>>    > mm^3
>>>    > # Measure IntraCranialVol, ICV, Intracranial Volume,
>>>    1114008.197334, mm^3
>>>    >
>>>    > # subjectname 090187_F01
>>>    > # Measure BrainMask, BrainMaskNVox, Number of Brain Mask Voxels,
>>>    1826694,
>>>    > unitless
>>>    > # Measure BrainMask, BrainMaskVol, Brain Mask Volume,
>>>    1826694.00, mm^3
>>>    > # Measure BrainSegNotVent, BrainSegVolNotVent, Brain
>>>    Segmentation Volume
>>>    >

[Freesurfer] Verifying ICV in FS5

[Jeff Sadino]

Hello,

Thank you Nick, Martin, and Doug for your previous feedback about ICV.
 I have a couple more questions that I am hoping you could clarify for
me.

To verify my ICV value, some pages on the mailing list and wiki say to
use this command:
tkregister2 --s subject --fstal --surf

Other places say to use this:
tkregister2 --xfm talairach.xfm --targ
$FREESURFER_HOME/average/RB_all_withskull_2008-03-26.gca --mov
../nu_noneck.mgz --reg junk

However, on this wiki page
(http://surfer.nmr.mgh.harvard.edu/fswiki/eTIV), it says that post
fs4.2.0, the template is this file:
711-2C_as_mni_average_305.4dfp.img.  In the first tkregister2 command
above, doesn't the --fstal flag use the talairach template?  But
shouldn't it be using the 711 template?  And since I am using fs5.0.0,
does that mean that I should not be using the RB template?
Ultimately, what is the correct way to verify the ICV in fs 5.0.0?

Thank you very much,
Jeff Sadino

On Mon, Mar 1, 2010 at 7:57 AM, Douglas N Greve
 wrote:
> It's hard to tell from the coronal, but it looks ok. The problem is more
> visible in the sagittal.
>
> doug
>
> Jeff Sadino wrote:
>>
>> Hello Doug,
>>
>> I loaded up the tkregister2 program and I think I see some poor
>> transformations.  I've never really used it before though.  Am I just
>> looking at whether or not the green line lines up with the target image wm?
>>  Can you look at my pictures real quick?  040002_S04_test is from fs4 that
>> is having problems.  040002_S04 is from fs3 and has a good icv value.  Some
>> of the blank spots are from my graphics drivers and unrelated to fs.
>>  tkregister2 also said the determinant of 040002_S04_test was 1.49544.
>>
>> Martin,
>>
>> Thank you for the idea.  I did this, and got a much better icv.  On fs3,
>> where the icv was good, the number was 1,767,481.  On fs4, it was 1,302,696.
>>  Using your idea on fs4, it was 1,819,485.
>>
>> Thank you so much to everyone for your help!
>> Jeff
>>
>> On Thu, Feb 25, 2010 at 8:53 PM, Martin Kavec > > wrote:
>>
>>    Hi Jeff,
>>
>>    I see this relatively often. Eventhough the talairach registration
>>    is fine, the
>>    ICV is underestimated. In this case I take old ICV, which helps.
>>
>>    mri_segstats --subject $MYSUBJECT --old-etiv-only
>>
>>    This is in 99% of these cases a cure. This is not just a blind
>>    attempt. In
>>    these cases I then compare each ICV normalized aseg volume against
>>    an age
>>    matched group and only if the old-etiv-only ICV gives better
>>    estimates I use
>>    it.
>>
>>    Cheers,
>>
>>    Martin
>>
>>    On Thursday 25 February 2010 23:05:10 Jeff Sadino wrote:
>>    > Thanks Doug.  That is very good news to hear!
>>    >
>>    > Maybe I spoke too soon about fixing the problem.  I looked at 4 more
>>    > subjects, and even though the icv value is unique, its magnitude
>>    seems
>>    > wrong.  Often times, it is smaller than the BrainSegVol, which
>>    doesn't make
>>    > much sense and it is also low overall.
>>    >
>>    >
>>    > # subjectname 090167_F01
>>    > # Measure BrainMask, BrainMaskNVox, Number of Brain Mask Voxels,
>>    1639959,
>>    > unitless
>>    > # Measure BrainMask, BrainMaskVol, Brain Mask Volume,
>>    1639959.00, mm^3
>>    > # Measure BrainSegNotVent, BrainSegVolNotVent, Brain
>>    Segmentation Volume
>>    > Without Ventricles, 1289194.00, mm^3
>>    > # Measure BrainSeg, BrainSegNVox, Number of Brain Segmentation
>>    Voxels,
>>    > 1299686, unitless
>>    > # Measure BrainSeg, BrainSegVol, Brain Segmentation Volume,
>>    1299686.00,
>>    > mm^3
>>    > # Measure IntraCranialVol, ICV, Intracranial Volume,
>>    1105595.114652, mm^3
>>    >
>>    > # subjectname 090182_S01
>>    > # Measure BrainMask, BrainMaskNVox, Number of Brain Mask Voxels,
>>    1599981,
>>    > unitless
>>    > # Measure BrainMask, BrainMaskVol, Brain Mask Volume,
>>    1599981.00, mm^3
>>    > # Measure BrainSegNotVent, BrainSegVolNotVent, Brain
>>    Segmentation Volume
>>    > Without Ventricles, 1199694.00, mm^3
>>    > # Measure BrainSeg, BrainSegNVox, Number of Brain Segmentation
>>    Voxels,
>>    > 1205382, unitless
>>    > # Measure BrainSeg, BrainSegVol, Brain Segmentation Volume,
>>    1205382.00,
>>    > mm^3
>>    > # Measure IntraCranialVol, ICV, Intracranial Volume,
>>    1114008.197334, mm^3
>>    >
>>    > # subjectname 090187_F01
>>    > # Measure BrainMask, BrainMaskNVox, Number of Brain Mask Voxels,
>>    1826694,
>>    > unitless
>>    > # Measure BrainMask, BrainMaskVol, Brain Mask Volume,
>>    1826694.00, mm^3
>>    > # Measure BrainSegNotVent, BrainSegVolNotVent, Brain
>>    Segmentation Volume
>>    > Without Ventricles, 1381871.00, mm^3
>>    > # Measure BrainSeg, BrainSegNVox, Number of Brain Segmentation
>>    Voxels,
>>    > 1388801, unitless
>>    > # Measure BrainSeg, BrainSegVol, Brain Segmentation Volume,
>>    1388801.00,
>>    > mm^3
>>    > # Measure IntraCranialVol,