[Insight-users] Any advantage of BSpline ?

[Luis Ibanez]

Hi Eduard,

None of the three methods:

     A- FEM
     B- BSpline transform
     C- Demons

is particularly customized for representing
the deformations resulting from lung motion.


Here you may have to differentiate between
these two *different* goals:

   a) Registering two images
   b) Representing the real lung motion


---

Demons simply deforms an image using forces
generated by optical flow, and constrain the
deformations by using Gaussian smoothing on
the displacement field. This is *theoretically*
equivalent to modeling elasticity. The basic
implementation of Demons will not consider at
all the physical characteristics of the lung
tissue.  It *may* however be good enough for
doing (a), that is, just registering the local
features of the images.


BSplines do not have *any* physical analogy.
They are simply providing a concise representation
of the displacement field and implicitly provide
a smooth displacement field. If you carefully
*manually* customize the placement of the
BSpline nodes, you could create a BSpline grid
that is "lung friendly", meaning, having nodes
in the regions of the lung that will displace
the most.  This however is a fully manual task,
it will require a lot of coffee and you probably
don't want to do it twice.
Again, just as with Demons, BSplines may be effective
enough for just registering the two images, without
making any claims regarding the physical correctness
of the resulting deformation field.

FEM is probably the method that will give you more
expression power if what you want is to go after the
*real* physical deformation of the Lung. NOTE that
the FEM models that we use for deformable registration
are *totally artificial*. They do not represent the
reality of the biological tissue at all. Their are
only used as an analogy in order to recreate a desired
behavior of the deformation field. Values such as
*density*, *elasticity* and so on, are usually manipulated
just as tunning parameters in order to generate a plausible
deformation field.  HOWEVER, if you feel motivated and you
really care about deriving the real physical deformation
of the Tissue, you could customize a FEM grid in order to
represent the real physical characteristics of the Tissues.
Notice that you will have to deal in that case with the
fact that air is flowing in and out of the lungs and there
fore their volume is changing.  Probably the critical aspect
of such a modeling attempt will be to create the initial
Mesh that will  fit the lung, along with the surrounding
tissue. You will have to then to classify all the FEM nodes
according to tissue classes and assign physical properties
to them.




Please let us know if you have further questions.



   Regards,


     Luis



----------------------------
Eduard Schreibmann wrote:

>     Hello everybody,
> 
>      
> 
>     Was wondering what method is theoretically more suited to model lung
>     motion. There are some papers using demons, there is an example in
>     the ITK theory book using the FEM model.
> 
>      
> 
>     Are these methods better suited than let’s say the BSpline model, is
>     there an a-priory reason to select one of these methods to register
>     lung images, at inspire and expire ?
> 
>      
> 
>     Also, is there any additional information on the results presented
>     in the teory book, page 330, such as the code or
>     elasticity/viscosity parameters used, or any additional information
>     ? The book does not give many details.
> 
>      
> 
>     Thank you
> 
>     Edi
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Insight-users mailing list
> Insight-users at itk.org
> http://www.itk.org/mailman/listinfo/insight-users