[Insight-users] Multimodal atrium and lung vessels registration

[Andreas Hahn]

Hello users and developers,

i got 3D data of lungs and heart from MR and CT, wich i want to  
registrate in a way that the left atrium and the connected lung  
vessels are matched.

After a first look in the itkSoftwareGuide and the  
RegistrationMethodsOverview i am confused about the large number of  
transformers, optimizers and different measurements.

1. My first idea is to segment the blood pool in both datasets to get  
a binary vessel tree. I tried this one with otsu segmentation and  
confidence connectedness.

2. Then i would try first a 3D rigid registration. That might fail,  
because of the patients different lung filling or heart phase on the  
images.

3. Then i would try a 3D deformable registration. That might be a  
problem, because the left ventricel cannot be clearly separated from  
the atrium during segmentation and could cause a missalingment of the  
atrium borders.

4. Then i would try to use prior information of the ostia positions  
(where the lung vessels are "plugged" into the atriums body) and use  
them for better registration.

Is it a good idea to work here with segmented data?

Could you please give me some hints, which module combination is  
useful for my proposed registration approaches?

Could you please explain, how to put the prior information into the  
registration?

I would appreciate any ideas and discussions. Thanks in advance,
Andreas