2011 Winter Project Week:UncertaintyVisualization
Visualizing Registration Uncertainty
- SPL: Petter Risholm, William Wells
We have developed a probabilistic non-rigid registration framework where we characterize the posterior distribution over deformations with a MCMC method. In practice, a large number of deformations are drawn from the posterior distribution. From the set of deformation samples, we can estimate the most likely deformation as well as the uncertainty of the deformation.
The objective for the Project Week is to developing a command line module that can take the large set of deformation samples and generate useful marginal summarizes/visualizations of the registration uncertainty, e.g. marginal distribution of a biopsy point after registration.
Our approach for summarizing and visualizing registration uncertainty is described in [Risholm 2010].
Our plan for the project week is to incorporate the generation of marginal probability maps of points and surface models as a command line module in Slicer3.
We have developed a:
- A VTK class for generating marginal probability maps of deformed surface models.
- A command line module for running the code through Slicer3.
This work will be delivered to the NA-MIC Kit as a
- ITK Module
- Slicer Module
- Extension -- commandline Yes
- Extension -- loadable
- Risholm P, Samset E, Pieper S, Wells W Summarizing and Visualizing Uncertainty in Non-Rigid Registration. MICCAI 2010