Howdy -
It seems that the skullstripping step is having some trouble removing bone marrow (see attached image).
Adjusting watershed values (I've tried as low as 5) does not make any significant improvement.
Does anyone else have this issue? Can anyone provide any hints as to how I might get a better automated skull-strip? I've been manually removing the marrow but it's getting real old.
Any thoughts would be much appreciated.
Bo
Hi Bo,
does it affect your pial surface? You may be fine with just leaving it in.
cheers, Bruce On Wed, 16 Aug 2006, Bo Shi wrote:
Howdy -
It seems that the skullstripping step is having some trouble removing bone marrow (see attached image).
Adjusting watershed values (I've tried as low as 5) does not make any significant improvement.
Does anyone else have this issue? Can anyone provide any hints as to how I might get a better automated skull-strip? I've been manually removing the marrow but it's getting real old.
Any thoughts would be much appreciated.
Bo
Hi Bruce,
The pial surface for the top regions of the brain do fairly well in ignoring the marrow, however, the unwanted regions still get classified as gray matter (image attached).
This will adversely affect the reported cortex volume no?
Thanks, Bo
On Wed, 2006-08-16 at 14:14 -0400, Bruce Fischl wrote:
Hi Bo,
does it affect your pial surface? You may be fine with just leaving it in.
cheers, Bruce On Wed, 16 Aug 2006, Bo Shi wrote:
Howdy -
It seems that the skullstripping step is having some trouble removing bone marrow (see attached image).
Adjusting watershed values (I've tried as low as 5) does not make any significant improvement.
Does anyone else have this issue? Can anyone provide any hints as to how I might get a better automated skull-strip? I've been manually removing the marrow but it's getting real old.
Any thoughts would be much appreciated.
Bo
no - we don't really use the cortex labels from the aseg for this reason (and others). Instead we use the surfaces (?h.white and ?h.pial) for computing cortical properties, so you should be all set.
cheers, Bruce
On Wed, 16 Aug 2006, Bo Shi wrote:
Hi Bruce,
The pial surface for the top regions of the brain do fairly well in ignoring the marrow, however, the unwanted regions still get classified as gray matter (image attached).
This will adversely affect the reported cortex volume no?
Thanks, Bo
On Wed, 2006-08-16 at 14:14 -0400, Bruce Fischl wrote:
Hi Bo,
does it affect your pial surface? You may be fine with just leaving it in.
cheers, Bruce On Wed, 16 Aug 2006, Bo Shi wrote:
Howdy -
It seems that the skullstripping step is having some trouble removing bone marrow (see attached image).
Adjusting watershed values (I've tried as low as 5) does not make any significant improvement.
Does anyone else have this issue? Can anyone provide any hints as to how I might get a better automated skull-strip? I've been manually removing the marrow but it's getting real old.
Any thoughts would be much appreciated.
Bo
Good to hear - but
On Wed, 2006-08-16 at 14:44 -0400, Bruce Fischl wrote:
no - we don't really use the cortex labels from the aseg for this reason (and others). Instead we use the surfaces (?h.white and ?h.pial) for computing cortical properties, so you should be all set.
cheers, Bruce
On Wed, 16 Aug 2006, Bo Shi wrote:
Hi Bruce,
The pial surface for the top regions of the brain do fairly well in ignoring the marrow, however, the unwanted regions still get classified as gray matter (image attached).
This will adversely affect the reported cortex volume no?
Thanks, Bo
On Wed, 2006-08-16 at 14:14 -0400, Bruce Fischl wrote:
Hi Bo,
does it affect your pial surface? You may be fine with just leaving it in.
cheers, Bruce On Wed, 16 Aug 2006, Bo Shi wrote:
Howdy -
It seems that the skullstripping step is having some trouble removing bone marrow (see attached image).
Adjusting watershed values (I've tried as low as 5) does not make any significant improvement.
Does anyone else have this issue? Can anyone provide any hints as to how I might get a better automated skull-strip? I've been manually removing the marrow but it's getting real old.
Any thoughts would be much appreciated.
Bo
Sorry, message cut off mid-write;
We're having some trouble with the pial surface for the temporal lobes. All four datasets we've processed so far have temporal lobe pial surfaces that are significantly smaller than they should be (attached)
I know there are control points for white matter - is there an equivalent for gray matter?
On Wed, 2006-08-16 at 14:44 -0400, Bruce Fischl wrote:
no - we don't really use the cortex labels from the aseg for this reason (and others). Instead we use the surfaces (?h.white and ?h.pial) for computing cortical properties, so you should be all set.
cheers, Bruce
On Wed, 16 Aug 2006, Bo Shi wrote:
Hi Bruce,
The pial surface for the top regions of the brain do fairly well in ignoring the marrow, however, the unwanted regions still get classified as gray matter (image attached).
This will adversely affect the reported cortex volume no?
Thanks, Bo
On Wed, 2006-08-16 at 14:14 -0400, Bruce Fischl wrote:
Hi Bo,
does it affect your pial surface? You may be fine with just leaving it in.
cheers, Bruce On Wed, 16 Aug 2006, Bo Shi wrote:
Howdy -
It seems that the skullstripping step is having some trouble removing bone marrow (see attached image).
Adjusting watershed values (I've tried as low as 5) does not make any significant improvement.
Does anyone else have this issue? Can anyone provide any hints as to how I might get a better automated skull-strip? I've been manually removing the marrow but it's getting real old.
Any thoughts would be much appreciated.
Bo
can you send an image without the aseg overlay? Is the wm darker there than elsewhere (i.e. <110)? On Wed, 16 Aug 2006, Bo Shi wrote:
Sorry, message cut off mid-write;
We're having some trouble with the pial surface for the temporal lobes. All four datasets we've processed so far have temporal lobe pial surfaces that are significantly smaller than they should be (attached)
I know there are control points for white matter - is there an equivalent for gray matter?
On Wed, 2006-08-16 at 14:44 -0400, Bruce Fischl wrote:
no - we don't really use the cortex labels from the aseg for this reason (and others). Instead we use the surfaces (?h.white and ?h.pial) for computing cortical properties, so you should be all set.
cheers, Bruce
On Wed, 16 Aug 2006, Bo Shi wrote:
Hi Bruce,
The pial surface for the top regions of the brain do fairly well in ignoring the marrow, however, the unwanted regions still get classified as gray matter (image attached).
This will adversely affect the reported cortex volume no?
Thanks, Bo
On Wed, 2006-08-16 at 14:14 -0400, Bruce Fischl wrote:
Hi Bo,
does it affect your pial surface? You may be fine with just leaving it in.
cheers, Bruce On Wed, 16 Aug 2006, Bo Shi wrote:
Howdy -
It seems that the skullstripping step is having some trouble removing bone marrow (see attached image).
Adjusting watershed values (I've tried as low as 5) does not make any significant improvement.
Does anyone else have this issue? Can anyone provide any hints as to how I might get a better automated skull-strip? I've been manually removing the marrow but it's getting real old.
Any thoughts would be much appreciated.
Bo
The WM in the temporal lobe has intensities ranging [89-111] - hard to tell the average, but my estimate would be around ~95-100.
gray matter inside the pial surface ranges from 73-85, and outside the pial surface is approximately the same.
WM in parietal lobe is generally 110.
On Wed, 2006-08-16 at 15:50 -0400, Bruce Fischl wrote:
can you send an image without the aseg overlay? Is the wm darker there than elsewhere (i.e. <110)? On Wed, 16 Aug 2006, Bo Shi wrote:
Sorry, message cut off mid-write;
We're having some trouble with the pial surface for the temporal lobes. All four datasets we've processed so far have temporal lobe pial surfaces that are significantly smaller than they should be (attached)
I know there are control points for white matter - is there an equivalent for gray matter?
On Wed, 2006-08-16 at 14:44 -0400, Bruce Fischl wrote:
no - we don't really use the cortex labels from the aseg for this reason (and others). Instead we use the surfaces (?h.white and ?h.pial) for computing cortical properties, so you should be all set.
cheers, Bruce
On Wed, 16 Aug 2006, Bo Shi wrote:
Hi Bruce,
The pial surface for the top regions of the brain do fairly well in ignoring the marrow, however, the unwanted regions still get classified as gray matter (image attached).
This will adversely affect the reported cortex volume no?
Thanks, Bo
On Wed, 2006-08-16 at 14:14 -0400, Bruce Fischl wrote:
Hi Bo,
does it affect your pial surface? You may be fine with just leaving it in.
cheers, Bruce On Wed, 16 Aug 2006, Bo Shi wrote:
Howdy -
It seems that the skullstripping step is having some trouble removing bone marrow (see attached image).
Adjusting watershed values (I've tried as low as 5) does not make any significant improvement.
Does anyone else have this issue? Can anyone provide any hints as to how I might get a better automated skull-strip? I've been manually removing the marrow but it's getting real old.
Any thoughts would be much appreciated.
Bo
it's probably fixable by adding some control points in the temporal lobe wm that is darker, but *not* partial volumed. Make sure they are in the bulk of the wm and not near the gray/white boundary. The control points can affect the whole region if it's too dark.
cheers, Bruce
On Wed, 16 Aug 2006, Bo Shi wrote:
The WM in the temporal lobe has intensities ranging [89-111] - hard to tell the average, but my estimate would be around ~95-100.
gray matter inside the pial surface ranges from 73-85, and outside the pial surface is approximately the same.
WM in parietal lobe is generally 110.
On Wed, 2006-08-16 at 15:50 -0400, Bruce Fischl wrote:
can you send an image without the aseg overlay? Is the wm darker there than elsewhere (i.e. <110)? On Wed, 16 Aug 2006, Bo Shi wrote:
Sorry, message cut off mid-write;
We're having some trouble with the pial surface for the temporal lobes. All four datasets we've processed so far have temporal lobe pial surfaces that are significantly smaller than they should be (attached)
I know there are control points for white matter - is there an equivalent for gray matter?
On Wed, 2006-08-16 at 14:44 -0400, Bruce Fischl wrote:
no - we don't really use the cortex labels from the aseg for this reason (and others). Instead we use the surfaces (?h.white and ?h.pial) for computing cortical properties, so you should be all set.
cheers, Bruce
On Wed, 16 Aug 2006, Bo Shi wrote:
Hi Bruce,
The pial surface for the top regions of the brain do fairly well in ignoring the marrow, however, the unwanted regions still get classified as gray matter (image attached).
This will adversely affect the reported cortex volume no?
Thanks, Bo
On Wed, 2006-08-16 at 14:14 -0400, Bruce Fischl wrote:
Hi Bo,
does it affect your pial surface? You may be fine with just leaving it in.
cheers, Bruce On Wed, 16 Aug 2006, Bo Shi wrote:
> Howdy - > > It seems that the skullstripping step is having some trouble removing > bone marrow (see attached image). > > Adjusting watershed values (I've tried as low as 5) does not make any > significant improvement. > > Does anyone else have this issue? Can anyone provide any hints as to > how I might get a better automated skull-strip? I've been manually > removing the marrow but it's getting real old. > > Any thoughts would be much appreciated. > > Bo > >
Bo,
It looks like control points will do the trick - in the wm. Your white surfaces are missing some regions too. Control points will fix this and subsequently fix your pial surfaces as well. Give them a try, but be sure to get the control points in wm voxels, not too close to the wm/gm boundary.
Jenni
On Wed, 16 Aug 2006, Bo Shi wrote:
Sorry, message cut off mid-write;
We're having some trouble with the pial surface for the temporal lobes. All four datasets we've processed so far have temporal lobe pial surfaces that are significantly smaller than they should be (attached)
I know there are control points for white matter - is there an equivalent for gray matter?
On Wed, 2006-08-16 at 14:44 -0400, Bruce Fischl wrote:
no - we don't really use the cortex labels from the aseg for this reason (and others). Instead we use the surfaces (?h.white and ?h.pial) for computing cortical properties, so you should be all set.
cheers, Bruce
On Wed, 16 Aug 2006, Bo Shi wrote:
Hi Bruce,
The pial surface for the top regions of the brain do fairly well in ignoring the marrow, however, the unwanted regions still get classified as gray matter (image attached).
This will adversely affect the reported cortex volume no?
Thanks, Bo
On Wed, 2006-08-16 at 14:14 -0400, Bruce Fischl wrote:
Hi Bo,
does it affect your pial surface? You may be fine with just leaving it in.
cheers, Bruce On Wed, 16 Aug 2006, Bo Shi wrote:
Howdy -
It seems that the skullstripping step is having some trouble removing bone marrow (see attached image).
Adjusting watershed values (I've tried as low as 5) does not make any significant improvement.
Does anyone else have this issue? Can anyone provide any hints as to how I might get a better automated skull-strip? I've been manually removing the marrow but it's getting real old.
Any thoughts would be much appreciated.
Bo
freesurfer@nmr.mgh.harvard.edu
-
Bo Shi -
Bruce Fischl -
Jenni Pacheco