[Insight-users] segmentation registration

[Luis Ibanez]

Hi Michael,


Mmmm,
is your supervisor suscribed to the users-list  ?   :-)


Just in case..., I think he/she is right !


Yes, you could use level sets (which are an implementation
of active contours) for segmenting the tonge. It depends
on how bright the tag lines are in the MRI...

----

I just sent you another email regarding deformable
registration.  It explains how to run Demons on the
saggital slices you have. Note that if you do this,
the deformation field can be used as a segmentation
mechanism since the tonge is the only moving object
in the image, the rest should be static, e.g. the
skull and the brain.

If you get deformation fields from Demons or FEM,
you can simply proceed to compute the magnitude
of the displacement vectors and create a scalar
image with them.  It should be relatively easy to
threshold this image and get a mask of the objects
that are moving from one slice to the next (along
time). (this moving object should be the tongue and
some of the soft tissue under the jaw.


Notice that this is *completely* equivalent to the
widespread technique of using optical-flow for
tracking movement in video sequences. This is commonly
used in video surveillance.  You may want to take
a look at computer vision literature, including
keywords such as compression of video, movement
detection, optical-flow, MPEG.



Regards,



    Luis



----------------------------------------------
a a wrote:

> Hi Luis and all,
>  
> however, my supervisor insisted on using some form of active contour or 
> deformable solid (B-spline etc.), also i have a volume of mri slices 
> without tagged lines for the first instance, so can i use the active 
> contour to segment the non-tagged volumetric image slices, then track 
> the tagged lines of the images of the same time frame, and combine them 
> together: if yes, how can i handle this problem?
>  
> Thank you and Regards
> Michael
> 
> */Luis Ibanez <luis.ibanez at kitware.com>/* wrote:
> 
> 
>     Hi Michael,
> 
>     In order to get a feeling of what kind of
>     volume rendering approach you want to use
>     you could test your data with the free
>     version of VolView.
> 
>     http://www.kitware.com/products/volview.html
> 
>     For segmenting the tissue, level sets may
>     not be appropriate since the tag lines are
>     probably more intense than the tissue
>     differences. You can easily figure out the
>     relative differences by computing the gradient
>     magnitude of the image. If in the gradient
>     magnitude, the edges of the tag lines are
>     more intense than the edges of the tonge,
>     then it is unlikely that level sets would
>     be ablo to segment the tonge itself.
> 
> 
>     You should give it a try to watersheds,
>     althought this will require some user
>     interaction/supervision.
> 
> 
>     You will find examples on watersheds on the
>     SoftwareGuide and an interactiv e watershed
>     segmenation tool in InsightApplications.
> 
> 
>     I would suggest you to do supervised segmentation
>     of only one initial tonge dataset. Then do deformable
>     registration of this dataset with the dataset of the
>     next time step, and then map the segmenation of the
>     first one through the deformation field. That is,
>     perform segmentation via registration with a segmented
>     model.
> 
> 
> 
>     Regards,
> 
> 
> 
>     Luis
> 
> 
> 
>     ---------------
>     a a wrote:
> 
>      > Hi to all,
>      >
>      > i have a set of tagged MRI slices of a tongue, which i have to
>     render
>      > the volume and measure the deformation based on the tagged lines.
>     For
>      > the image processing part, i've decided to use ITK, but i need some
>      > guidelines as i'm new to this area:
>      >
>      > Can the segmentation tools (i.e. GeodesicActiveContour) be used to
>      > segment the images in 3D (i.e. multiple slices loaded as a
>     volume), are
>      > there any e xamples for that?
>      > As i will need to visualize (VTK) and analyse the deformable
>     solid model
>      > of the tongue, are there any tools/examples for tracking the
>     deformation
>      > of the tongue and the tagged lines? Can anyone roughly tell me
>     the steps
>      > to take to achieve this, thanks.
>      >
>      >
>      > have a nice day
>      > Michael
>      >
>     \
> 
> 
>     _______________________________________________
>     Insight-users mailing list
>     Insight-users at itk.org
>     http://www.itk.org/mailman/listinfo/insight-users
> 
> *Y! Asia presents Lavalife* 
> <http://sg.rd.yahoo.com/mail/tagline/?http://asia.yahoo.com/lavalife>
> - Get clicking with thousands of local singles today!
>