Re[1] [Insight-users] MRI segmentation

[Elsa Angelini]

Lydia,

Our group  at UPenn and Columbia has implemented various segmentation
algorithms:=20
(1) Vectorial fuzzy connectedness
(2) 3D Deformable model=20
(3) Voronoi Diagram classification

You can find test examples for the three methods in the Insight folder you
checked out. All examples use raw data. The group also had a paper at the
MICCAI01 conference where you can find more details on how these
segmentation methods were used for both Human Visible data and brain MRIs.=
=20
For each method you need to input information on the structure you want to
segment (such as statistics, seed points,...)

I hope this helps.

Elsa



On Tue, 8 Jan 2002, lydia ycl wrote:

> Hi,
> I have an MR serie of pelvis data. I need to get the bone
> from it. The format of the images is dicom format. I
> converted it to raw data but if I have to convert again to
> metaimage format, I'm not sure it will work cause until now
> I could not create this kind of metaimage data.
> I just want to know if someone already tested the
> segmentation algorithms of itk on MRI data whatever could
> be the anatomical structure.
> Thanks
>=20
> > -------Message d'origine-------
> > De : Luis Ibanez <ibanez@choroid.cs.unc.edu>
> > Date : 08/01/2002 14:24:08
> >
> >
> > Hi Lydia,
> >
> > Reading the MRI data is just a problem of file formats.
> > That should not be very hard if you know the format in
> > which the data is available.
> >
> > About segmenting, this is different. Segmentation is an
> > ill posed problem and you can only talk about segmenting
> > particular anatomical structures from particular kinds
> > of image modalities.
> >
> > 1) What kind of anatomical structures do you
> > want to segment from the MRI data ?
>  > 2) What MRI modality do you have ?
> >
> >
> > Thanks
> >
> >
> > Luis
> >
> > -------------------------------------------
> > On Tue, 8 Jan 2002, lydia ycl wrote:
> >
> > > Hi folks,
> > > Could any one read and segment MRI data using ITK?
> > >
> >
> >
> >
> ______________________________________________________
> Bo=EEte aux lettres - Caramail - http://www.caramail.com
>=20
>=20