Hi Philipp, Hi FSL users
thanks a lot for your sophisticated advices.
please can you say me where (file and line?) I can change the -c option in
BET so
that I can run only BET to find adequate values for the
-f and -c option. I know where I can change this value in SIENA but
then I have to run always the whole SIENA procedure.
You have also recommended that the -f values should be about the same for
all images in a sequence. I have T1 MRI images from four different time
points.
I intend to use the same -f value for all images that came from one subject.
For different subjects I intend to use different -f values. Is there a limit
of the range (e.g. 0.5-0.7) of -f values for different subjects?
Thanks in advance
Best regards
Juergen
> -----Ursprüngliche Nachricht-----
> Von: FSL - FMRIB's Software Library
> [mailto:[log in to unmask]] Im Auftrag von Philipp Saemann
> Gesendet: Mittwoch, 31. August 2005 14:35
> An: [log in to unmask]
> Betreff: Re: [FSL] -f and -c option in BET
>
> Hello,
>
> the best f - value can be found by trial and error. Just try
> different f-values in steps of e. g. 0.05 (e. g. via a batch)
> and then try to finetune it. The f-value should be about the
> same for all images of one seuqnce type.
>
> The technical papers explain well, what the f-value is about.
>
> c stands for centre.
> By defining this value you start the extraction algorithm at
> a specific point within your brain. As this point is usually
> calculated automatically from signal intensities it may
> happen that it is set into neck tissue (or run intoi neck
> tissue soon) which should not be in the volume in the optimal
> case. This is why in cases with lots of neck on the image you
> can specify this value yourself.
>
> To see what happens use a very low radius value (lie 5) and
> some c-ccordinates you define and you can see where the
> algorithm starts and stops. By moving c, try to find a good
> starting position, the leave out the r-option again (the
> radius of the starting sphere) and manipulate f as long as the result
> (oberlay-image) looks acceptable.
>
> Combination of c and f:
> Well, try to find coordinates that lie within the brain (e.
> g. close to the thalamus). From there trie to additionally
> manipulate the f-value.
>
> f-values and SIENA and automatic analysis:
>
> To my opinion it is not a good idea to pick an f-value and
> run multiple SIENA analysis without carefully checking the
> result of BET.
>
> Your results of course get distorted as soon as non-brain
> tissue is analysed or in cases where brain tissue is missing.
> In large groups the whole analysis is still a robust method,
> BUT the BET step needs to be checked and optimized to some degree.
>
> Best regards,
> Philipp
>
|