[Insight-users] Re: problem(text) : Brain Shift

[Luis Ibanez]

Hi Yxliu,

1) If you need to extract the surface of the gray matter,
   you should look at the set of segmentation methods
   available in ITK. Probably good options to try first
   are the LevelSet methods and RegionGrowing methods.

   You are strongly encouraged to read the Chapter on
   image segmentation from the ITK Software Guide.

2) Yes, ITK provides methods for performing

        PointSet to Image     registration   and
        PointSet to PointSet  registration


3) Do you actually have a dataset acquired when the brain
   was deformed ?

   If you don't, then there is no reference information
   that you can use in order to find the boundary conditions.
   You will have to rely on a simulation of a material with
   physical characteristics similar to real brain tissues.


4) NOTE that the application of FEM for deformable registration
   is purely artificial. At the end the FEM framework plays the
   role of a regularization mechanism and  the role of interpolator
   for the deformation vector field.

5) Yes you can use BSplineDeformableTrnasform with both
   the PointSetToImage and the PointSet to PointSet
   registration methods.

   Please read the Chapter on Registration in the Software Guide.


Before you go further, it seems that it will be helpful to do a
global overview of your goals, because you may not be gathering
the real data needed for this simulation to be meaningful.




  Regards,




     Luis




-----------------
yxliu wrote:

> Thank your help very much!  I send a explanatory text about the problem.
> 
>  
> 
> Dear Luis,
> 
> I am doing the research about Brain shift. I build a linear elastic 
> model of the brain. The simulated boundary condition is shown below:
> 
>  
> 
>                                                 exposed surface 
> (assuming nodes displacement is 9mm along gravity direction  
> 
>                                                                    
>      middle surface(no constration  on the nodes)
> 
>                                                   bottom 
> surface(assuming fixed on the nodes)
> 
>  
> 
>  
> 
>  
> 
> The brain deformation based on the simulated boundary condition is shown 
> as below:
> 
>  
> 
>  
> 
>  
> 
>                                                                        
>                   the undeformed brain                    the deformed 
> brain(the mesh denote the undeformed brain,
> 
>                                                                                                                                                   
> the entity denote the deformed brain)
> 
>  
> 
>  
> 
> Now the question is how to obtain the actual boundary condition to make 
> the physical model applicable in the clinical. In another words, how to 
> obtain the nodes displacement of the exposed surface(shown as below) 
> between the undefromed surface(fixed surface) and the deformed 
> surface(moving surface).
> 
> The fixed surface can be extracted from the 3D MRI and represented by 
> the point set(I do not know if ITK provides some filters to extract the 
> surface from the 3D MRI. Could you give me some advice?). The moving 
> surface can also be represented by the point set which is obtain by the 
> LRS(Laser Range Scanner). Now, the question is how to register the two 
> surfaces to obtain the nodes displacement.
> 
>  
> 
> One method I intend to use is using the FEM-based deformable 
> registration. Firstly, take the surface as a linear elastic membrane. 
> Then, compute the image force based on the metric based on the MI. At 
> last, apply the boundary condition to obtain the unique solution. 
> However, two things I cannot be sure are:
> 
> 1.      Does ITK provide the metric based on the point set other than 
> image data object?
> 
> 2.      How to apply the boundary condition? In another words, how to 
> know the displacement of some particular nodes? It seems paradox because 
> the displacement is just what I manager to obtain.
> 
> So, in my opinion, the FEM-based deformable registration may not be 
> feasible.
> 
>  
> 
> Another method I intend to use is using the deformable registration 
> based on the ICP. The question is : Does the ITK provide the Nor-rigid 
> transform( e.g.  Bspline) and Metric based on the point set data object?
> 
>  
> 
> I do not know if other methods are available. Hope your reply and Thanks 
> advance!
> 
>  
> 
> Best regards,
> 
>  
> 
> Yixun Liu
> 
>  
> 
>  
> 
>  
> 
>  
> 
>  
> 
>  
> 
>  
> 
>  
> 
>  
> 
>  
>