[Insight-users] Re: FastMarching

[Luis Ibanez]

Hi Yasser,

Please look at the SoftwareGuide

    http://www . itk . org/ItkSoftwareGuide . pdf

for a detailed explanation on the behavior
of the SigmoidImageFilter (section 5.2.2,
pdf-page 100), and its use for preparing the
speed image feed in the FastMarching filter
(section 8.3.1, pdf-page 269).

The Sigmoid parameters (alpha,beta) should
be selected by analyzing the GradientMagnitude
(image) of the input image.  You should estimate
the highest value of the region you want to
segment [highest basin](the value in the gradient
magnitude image), and also estimate the value
of the lowest section of the border surrounding
the region you want to segment. This lower border
is the place where the FastMarching method will
leak the first [lowest border].  Using these two
values you compute the Sigmoid's alpha and beta
as:


beta  =  ( [ lowest border ] + [ highest basin ] ) / 2
alpha = -( [ lowest border ] - [ highest basin ] ) / 3


Note that alpha must always be negative. This is to
ensure that edges in the input image result in values
close to zero on the speed image, and homogeneous
regions in the input image result in values close
to 1.0 in the speed image.

Check your speed image and make sure that it looks
like the one presented in the software guide in
Figure 8.12, page 275.  FastMarching results will
be as good (or as bad) as your speed image because
this is the only image input provided to the
FastMarching method.


   Regards,


    Luis


-----------------------------
yasser salman wrote:
> hi luis.,
> when i'm using Fast Marching segmentationLevel set in
> application 
> my volume go so dark i can't see any sclie after
> putting values of aplha=-1
> beta=5; so what is the best values for Alpha and Beta.
> also the seed point at which step i can update it (it
> can be one seed in x-y-z?).
> yasser
> 
> __________________________________
> Do you Yahoo!?
> Yahoo! Calendar - Free online calendar with sync to Outlook(TM).
> http://calendar . yahoo . com
>