Hello everyone,
I have many scans taken just under 10 years ago on a 1.5T Siemens scanner. For some reason, the images from this time period have consistent problems with gray matter / white matter contrast; consistently, Freesurfer measures almost all cerebral brain tissue as being white. I have experimented with mri_ca_normalize (turning -p down to 0.25 or so helps), with mri_normalize (using -prune helped somewhat,) but all of these trials are experimental and I haven't gotten a solid script written that handles these special cases for me in an efficient way.
My question: has anyone else dealt with this problem regularly, and come up with a good way to either use recon-all to take care of it (I can't find any useful flags, or workflow tutorials for dealing with this on the wiki) or written a handy script to do recon-all with several of these "less wm" options built in? I have seen people ask about issues relating to mri_ca_normalize on the mailing list in the past, but the conversations usually end with "send us some of your data and we'll take a look at it."
Or, am I missing something here? Did someone already discuss this in detail? Sorry if this seems obvious. Thanks in advance for any help...
Victor Laluz
UCSF Memory and Aging Center
Hi Victor,
have you tried messing with the segmentation parameters to mri_segment and mris_make_surfaces? They are written to be quite general across a bunch of T1-weighted acquistion types, but if you know the exact sequence and hence the gm/wm intensity values, you can set them explicitly using e.g. -wlo, -whi, etc... (and min_gray_at_white_border etc... to mris_make_surfaces)
cheers, Bruce
On Mon, 20 Oct 2008, Laluz, Victor wrote:
Hello everyone,
I have many scans taken just under 10 years ago on a 1.5T Siemens scanner. For some reason, the images from this time period have consistent problems with gray matter / white matter contrast; consistently, Freesurfer measures almost all cerebral brain tissue as being white. I have experimented with mri_ca_normalize (turning -p down to 0.25 or so helps), with mri_normalize (using -prune helped somewhat,) but all of these trials are experimental and I haven't gotten a solid script written that handles these special cases for me in an efficient way.
My question: has anyone else dealt with this problem regularly, and come up with a good way to either use recon-all to take care of it (I can't find any useful flags, or workflow tutorials for dealing with this on the wiki) or written a handy script to do recon-all with several of these "less wm" options built in? I have seen people ask about issues relating to mri_ca_normalize on the mailing list in the past, but the conversations usually end with "send us some of your data and we'll take a look at it."
Or, am I missing something here? Did someone already discuss this in detail? Sorry if this seems obvious. Thanks in advance for any help...
Victor Laluz
UCSF Memory and Aging Center
Ah, I see those now, under recon-all expert preferences. No, I haven't tried those, I suppose I was following a bad lead with the mri_ca_normalize script. Thanks for the quick answer; I'll experiment with these for now.
-Victor
-----Original Message----- From: Bruce Fischl [mailto:fischl@nmr.mgh.harvard.edu] Sent: Monday, October 20, 2008 2:30 PM To: Laluz, Victor Cc: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] Recurring issue with poor gray/white distinction - easy fix?
Hi Victor,
have you tried messing with the segmentation parameters to mri_segment and mris_make_surfaces? They are written to be quite general across a bunch of T1-weighted acquistion types, but if you know the exact sequence and hence the gm/wm intensity values, you can set them explicitly using e.g.
-wlo, -whi, etc... (and min_gray_at_white_border etc... to mris_make_surfaces)
cheers, Bruce
On Mon, 20 Oct 2008, Laluz, Victor wrote:
Hello everyone,
I have many scans taken just under 10 years ago on a 1.5T Siemens scanner. For some reason, the images from this time period have consistent problems with gray matter / white matter contrast; consistently, Freesurfer measures almost all cerebral brain tissue as being white. I have experimented with mri_ca_normalize (turning -p
down
to 0.25 or so helps), with mri_normalize (using -prune helped
somewhat,)
but all of these trials are experimental and I haven't gotten a solid script written that handles these special cases for me in an efficient way.
My question: has anyone else dealt with this problem regularly, and
come
up with a good way to either use recon-all to take care of it (I can't find any useful flags, or workflow tutorials for dealing with this on the wiki) or written a handy script to do recon-all with several of these "less wm" options built in? I have seen people ask about issues relating to mri_ca_normalize on the mailing list in the past, but the conversations usually end with "send us some of your data and we'll
take
a look at it."
Or, am I missing something here? Did someone already discuss this in detail? Sorry if this seems obvious. Thanks in advance for any help...
Victor Laluz
UCSF Memory and Aging Center
freesurfer@nmr.mgh.harvard.edu
-
Bruce Fischl -
Laluz, Victor