Hi.
On Thu, 21 Feb 2002, Wei Wen wrote:
> Before using FAST to segment g/w matters, we need to use BET. The output of
> FAST is 3 segmented parts: gm, wm and csf. If we want to have an accurate
> estimation of csf (cortical csf), then output of BET must not introduce
> loss/gain of csf. However, I have the impression that BET is a
> pre-processing tool basically designed to remove non-brain tissues (cortical
> csf was not one of the major aims to be targeted as to be kept accurately?).
> This is fine with g/w segmentation (ventricle csf is fine too), but not fine
> if I wish to have an accurate csf estimation. Am I right on this about BET?
This is all correct, although the limitation is not so much with what the
software is designed to find, as with the problem of separating CSF from
other dark "tissues" such as skull, in T1-weighted imaging (which I assume
you're talking about). If you use T2-weighted images then BET will tend to
include cortical CSF in the "brain" and this will I guess be more useful
to you.
> The reason for me to consider getting a reliable csf volume is to have a
> measure of atrophy: the ratio of total brain tissue against total csf.
In this case, there is a better way of doing this anyway - use SIENAX
(part of siena in FSL) - this normalises using the external surface of the
skull and then finds the normalised (G&W) brain volume - so you don't need
to worry about CSF at all.
> One more thing: is there any way of extracting ventricular csf in order to
> get the measure of ventricle size and atrophy ratio please?
This is something we've been doing more recently - Jackie Chen has a
presenetation at ISMRM, partly on doing exactly that, using ventricular
masks.
Thanks, Steve.
Stephen M. Smith
Head of Image Analysis, FMRIB
Oxford University Centre for Functional MRI of the Brain
John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
|