the trouble is I can't get it right to my satisfaction. I tried a lot!

my problem is finding the right parameters that you show below.

My experience (since you have not tired yet) is that you will have to use these parameters, absolutely. you have to mess with them till the surfaces seem right.

your results may look pretty nice. Mine have, in 250um monkey. but they don't do justice to the resolution. the GM/WM boundary is a single voxel in this data pretty much, but the surfaces don't reflect that.

I really want it super, insanely perfect for this bit of the project (150um marmoset). GM/WM and GM/CSF.

so part of what I'm asking is how to "fake CSF". another is to ensure that gaps between gyri look like CSF, not like low GM. and for e.g. extreme capule to look like WM, not high GM.

it seems to do with nu-correct in the end.

you can't just use the WM segmentation and set all WM to 110 unfortunately. it seems that there needs to be variation even though the nean should be 110

also that there needs to be a gradient between WM and GM and GM and CSF.

It's basically to complex for me to understand. I don't quite get the results I need, and I just bash away with the parameters trial and error.

it seems to me that I need to adjust the relative intensities differently in different parts of the volume. I don't know how. by faking coil locations and distances maybe? but where? and how?. At one point a year ago I tried with about 16,000 control points that I placed manually.

getting something that is good is entiely doable bear inn mind.

getting something that exactly follows a GM/WM boundary (and GM/CSF) that you know is in the data is really super tough.

And I need to unfortunately. I really do. I need it exact.

if you are mostly guessing ... but not really understanding (that means me, not you), it is hard.

On 5 August 2012 17:44, Ritobrato Datta <ridatta@mail.med.upenn.edu> wrote:

Hi Colin,

This is Rito. I am stuck at a similar step. I have acquired high-res ex-vivo T1 images at 7T from canine brains.

I got the following (P.S. below) from Nick Schmansky. Havent tried it though.

Please let me know how it goes for the images u r working with.

Best

Rito

P.S. The flags to mris_make_surfaces are:

-wlo <val> minimum for wm at border with gm
-ghi <val> maximum for gm at border with wm
-max_gray_at_csf_border <val> maximum for gm at border with csf
-min_gray_at_csf_border <val> minimum for gm at border with csf

so its that latter two that you would try (the first two are what people
normally use to mess with the gm/wm junction).

use tkmedit's brightness and contrast adjustments to find where you think
gm actually ends. the danger is that you select a min-gray-at-csf-border
value that is too low and it starts putting the surface out into open
space.

n.

Ritobrato Datta, Ph.D.
Post Doctoral Researcher
Department of Neurology
University of Pennsylvania School of Medicine
3rd Floor, Room 312
3710 Hamilton Walk (Goddard Laboratories)
Philadelphia, PA 19104-6241

----- Original Message -----
From: Colin Reveley <cmr25@sussex.ac.uk>
To: Freesurfer Mailinglist <freesurfer@nmr.mgh.harvard.edu>
Sent: Sun, 05 Aug 2012 12:10:31 -0400 (EDT)
Subject: [Freesurfer] no CSF

when running mris_make_surfaces it's possible to manually set CSF,GM,WM and
related thresholds.

now, what of the case where one simply doesn't have any CSF?

i.e. an ex-vivo scan.

I find it really hard to get a good pial surface despite much
experimentation. It wants to contract the GM and make the outer edges CSF,
no matter what I do.

I should say also that the T1 is really an MTR scan (nearly, but not
exactly the same intensity profile) and the brains are macaque and marmoset.

I also find it hard to get the intensities such that frontal WM and e.g.
the extreme capsule have WM range intensity. I'm happy to do manually, but
I canot get the gradient from WM to GM right manually in these areas to
satisfy mris_make_surfaces.

the data is really high resolution. it would be really great to capture the
whole GM ribbon.

putting in noise at a sensible average intensity for CSF doesn't seem to
help.

many thanks,

Colin