Hi - I suspect that you are right that something is changing in the
histogram that makes the segmentation less robust. The values will be
interpolated from their original values, increasing partial volume effect
and creating a new range of values around the brain edge between the
original values and zero.
In general we recommend segmenting the original images and then applying
registration transformations to the output of segmentation, if you need
this in another space.
Cheers, Steve.
On Sun, 16 Nov 2003, Mike Milham wrote:
> Thank you for your response - I appreciate your time. Following up on
> your response - I've noticed that when I segment before going to MNI space
> (via flirt or SPM), the behavior is much as you described. However, when
> segmenting in MNI space, the behavior becomes more erratic with respect to
> the labeling of tissue types (actually, if SPM is used for transformation,
> I have found an increase in misclassifications as well). Would you then
> say this is probably due to an unusual histogram produced by the
> normalization? Would you expect this?
>
> Thanks for your time and consideration.
> Mike
>
> -----Original Message-----
> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
> Of Stephen Smith
> Sent: Saturday, November 15, 2003 5:02 AM
> To: [log in to unmask]
> Subject: Re: [FSL] Assignment of class # in FAST
>
>
> Hi - in the latest version the numbers should be systematically assigned.
> If this is not happening then that probably means that the histogram is
> unusual for a given MR image modality, or that BET hasn't been used. The
> notation should be:
>
>
> -------------------------------------------------------------
> FAST outputs:
> -------------------------------------------------------------
>
>
> numbering for image outputs:
>
> CSF G W
>
> T1&T2 3-channel _seg_? 0 1 2
> pbmap & pve 0 1 2
>
> 2-channel _seg_? 0 1
> pbmap & pve 0 1
>
> -------------------------------------------------------------
>
> for the _seg.hdr image, the indexing is the same but with 1 added.
>
> -------------------------------------------------------------
>
> Cheers, Steve.
>
>
>
>
>
> On Wed, 12 Nov 2003, Mike Milham wrote:
>
> > Hi! While running the latest version of FAST I noticed that the class
> > numbers (0, 1 and 2) are not systematically assigned to tissue types
> > (e.g., 0 for white, 1 for CSF, 2 for gray) which obviously makes
> > scripting a bit more difficult. Is this an intended feature (I
> > believe earlier versions were systematic in their assignment of class
> > # to tissue type)? If not, is there any quick fix other than using an
> > older version.
> >
> > Thanks
> > Mike Milham
> >
>
> Stephen M. Smith DPhil
> Associate Director, FMRIB and Analysis Research Coordinator
>
> 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
>
Stephen M. Smith DPhil
Associate Director, FMRIB and Analysis Research Coordinator
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
|