[Insight-developers] PhysicalObjects for defining anatomical context

[Celina Imielinska]

 Peter,

 if you want to support queries like:

 "within less then 20 mm from the kidney"
 "if on the left, it must be spleen"...

  the knowledge organization (ontology) for anatomical entities
(structures) must capture complex 3D relationships between the
entities. Plus you must have a spatial database, linked to the ontology,
which will support all kind of geometric/spatial queries. We have done
some work on ontology and spatial database/query language with Nina
Wacholder and Ken Ross. These are not trivial problems, and the ontology
organized around 3D anatomy is not yet resolved... But this is something
important and exciting to look at. Do you want to talk about it?

 -Celina

  

On Wed, 28 Nov 2001, Peter Ratiu wrote:

> Does anyone's code use such parameters? I have been working for a while now
> on defining such generic relationships between subjects. I am using the VHM
> and VHF to consistently map homologous regions and landmarks. I am doing
> this for anatomical reasons, but I was thinking about the possibility to
> define broad boundaries for various structures and derive incremental rules
> such as: 
> A) if this is within less then 20 mm from the kidney, it cannot be the
> bladder. This seems coarse, but next, we the next step would be
> B) if on the left, it must be spleen, if on the right must be liver.
> C) if liver, and the superior boundary is not straight (I.e. Incomplete data
> set, we must find diaphragm, and next left lung, etc.
> 
> One of the validation methodologies I proposed also uses this principle, at
> a more refined level. If anyone is interested in this approach, please let
> me know.
> Peter
> 
> 
> > 
> > At 09:34 AM 11/27/2001 -0500, Luis Ibanez wrote:
> >> 
> >> PhysicalObjects are oriented to support processes like:
> >> 
> >> - defining anatomical context for image analysis:
> >>  e.g. when looking for a Liver in a CT we can start by
> >>  identifying the spine, place a 3D model there, then move
> >>  on to identifying the ribs and place rib 3D models there,
> >>  .... and so on. It is well known that radiolgist identify
> >>  structures by their context.  e.g something it may look like
> >>  a liver... but it if not in the right place it should not
> >>  be a liver.
> >>
> 
> _______________________________________________
> Insight-developers mailing list
> Insight-developers@public.kitware.com
> http://public.kitware.com/mailman/listinfo/insight-developers
>