Difference between revisions of "2009 Summer Project Week Cardiac Blood Pool Registration"

From NAMIC Wiki
Jump to: navigation, search
Line 2: Line 2:
 
<gallery>
 
<gallery>
 
Image:PW2009-v3.png|[[2009_Summer_Project_Week|Project Week Main Page]]
 
Image:PW2009-v3.png|[[2009_Summer_Project_Week|Project Week Main Page]]
 +
Image:s1p1.png|Patient 1
 +
Image:s1p2.png|Patient 2
 +
Image:log_demons_avg.png|Intensity average of 16 patients registered together
 
</gallery>
 
</gallery>
  
Line 29: Line 32:
  
 
<h3>Progress</h3>
 
<h3>Progress</h3>
 +
Although we haven't had the opportunity to compare as many algorithms as we would have liked, we obtained what at first sight seem like very good results using a combination of Slicer's affine registration and a log-domain diffeomorphic demons deformable registration.
  
 +
We are now planning to further investigate these results and also try to improve them using different algorithms, especially in the second deformable registration stage.
  
 
</div>
 
</div>
 
</div>
 
</div>

Revision as of 04:51, 26 June 2009

Home < 2009 Summer Project Week Cardiac Blood Pool Registration


Key Investigators

  • MIT: Michal Depa, Polina Golland
  • BWH: Ron Kikinis, Ehud Schmidt

Objective

We would like to register cardiac MRA images with the objective of using atlas-based methods to segment the left atrium.

This segmentation could then be used as an input to an automated cardiac ablation scar segmentation algorithm.

Approach, Plan

There are numerous registration algorithms available and it is unclear which one is the best for cardiac blood pool registration. We would like to experiment with different ones and see which one can provide the best results. Since there are several people working on registration algorithms at this project week, this will be the ideal opportunity to do this.

Progress

Although we haven't had the opportunity to compare as many algorithms as we would have liked, we obtained what at first sight seem like very good results using a combination of Slicer's affine registration and a log-domain diffeomorphic demons deformable registration.

We are now planning to further investigate these results and also try to improve them using different algorithms, especially in the second deformable registration stage.