[Insight-users] geodesic active contours: problems with fast marching

[Lydia Ng]

> I did indeed play with the parameters, and the advection scaling makes
> almost no difference. How do I turn off the other terms? Setting them
to
> zero or one?

Setting the weight/scaling to zero will turn a term off.
I think with the GAC filter you will need to have some propagation
otherwise it is slow to expand.

- Lydia


> -----Original Message-----
> From: Sven Prevrhal [mailto:sven . prevrhal at oarg . ucsf . edu]
> Sent: Monday, July 14, 2003 2:57 PM
> To: Lydia Ng
> Subject: RE: [Insight-users] geodesic active contours: problems with
fast
> marching
>=20
> Hi Lydia,
>=20
> No I use the fast marching just to generate little spheres. I don't
know
> why
> the application example does this, I guess to save computation time. I
can
> and do use smaller spheres. I tried the fast marching on the speed
image,
> but with disappointing results - a lot of leaking. That's why I tried
the
> geodesic contouring so I can have an advection term.
>=20
> I did indeed play with the parameters, and the advection scaling makes
> almost no difference. How do I turn off the other terms? Setting them
to
> zero or one?
>=20
> Yes I could of course use more seed points but then I would get seeds
in
> the
> outside region for the slices where the bone has branched. (It is a
femur,
> and for slices above the neck you have the major trochanter and the
head
> disjoint.) So I guess the active contouring needs to do that for me.
>=20
> :-) sven
> --
> Sven Prevrhal, Ph.D.
> Assistant Adjunct Professor
> University of California, San Francisco
> Dept. of Radiology
> Osteoporosis and Arthritis Research Group
> 533 Parnassus Ave U368J
> San Francisco, CA 94143-1250
> Phone (415) 476-2855
> Fax (415) 502-2663
>=20
> email: sven . prevrhal at oarg . ucsf . edu
> http://www . oarg . ucsf . edu/~sprevrha
>=20
>=20
> -----Original Message-----
> From: Lydia Ng [mailto:lng at insightful . com]
> Sent: Monday, July 14, 2003 2:28 PM
> To: Sven Prevrhal
> Subject: RE: [Insight-users] geodesic active contours: problems with
fast
> marching
>=20
>=20
> > 1. Seed points and fast marching: The bone splits in two pieces in
the
> > "top"
> > slices. How can I avoid having the results of the fast marching leak
> into
> > the gap between the bone in these slices?
>=20
> In what context are you using FastMarching here?
> Are you just using it to generate little spheres around your seed
point
> or are you using it to segmentation (ie. by setting the output of the
> sigmoid as the speed image for fastmarching)?
>=20
> If the former, then you have to make the radius of the spheres smaller
> so that they are within your shape you want to segment.
>=20
> If the latter, then you have to reduce the threshold in which you are
> using to threshold the output time map.
>=20
> > 3. The iterations leak into the surrounding tissue almost
> independently of
> > the slice, ie the result looks the same on all slices.
>=20
> This might be a parameter tuning issue?
> Have you played around with the parameters?
> Have you look at the effect of say turning off all the terms except
> propagation and then run it for a small number of iterations etc ...
>=20
> Also if you are looking for something long and narrow, it might be a
> good idea to set several seed point along the length of the object.
> Since the speed is never zero - the boundary might start to unlock
> before the front gets to the extremes of the objects.
>=20
>=20
> - Lydia
>=20
>=20
> > -----Original Message-----
> > From: Sven Prevrhal [mailto:sven . prevrhal at oarg . ucsf . edu]
> > Sent: Monday, July 14, 2003 1:24 PM
> > To: insight-users at public . kitware . com
> > Subject: [Insight-users] geodesic active contours: problems with
fast
> > marching
> >
> > Hi,
> >
> > I use version 1.2.0 and want to use the geodesic active contours
> > application
> > example for segmenting some bone in a stack of CT images. I use
> positive
> > alpha for the sigmoid filter, so the speed image is > 0 inside the
> bone,
> > and
> > small or zero outside the bone. I place 2 seed points inside the
bone
> on
> > the
> > center slice. I have three difficulties:
> >
> > 1. Seed points and fast marching: The bone splits in two pieces in
the
> > "top"
> > slices. How can I avoid having the results of the fast marching leak
> into
> > the gap between the bone in these slices?
> >
> > 2. The advection term doesn't seem to do anything. I get the same
> results
> > even for terms > 1.
> >
> > 3. The iterations leak into the surrounding tissue almost
> independently of
> > the slice, ie the result looks the same on all slices.
> >
> > Help totally appreciated! (Thanks Luis for the last one, the
> > UpdateLargestPossibleRegion for chaining resampling output to
another
> > filter
> > worked!)
> >
> >
> > :-) sven
> > --
> >
> >
> > _______________________________________________
> > Insight-users mailing list
> > Insight-users at itk . org
> > http://www . itk . org/mailman/listinfo/insight-users