Random Walks For Deformable Image Registration



Karteek Popuri,
Dana Cobzas,University of Alberta
Abhishek Sen,University of Alberta


References

Popuri K., Cobzas D. and Jagersand M. A variational formulation for discrete registration International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI) 2013

Cobzas, D. and Sen, A. Random Walks For Deformable Image Registration International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI) 2011

Talk at FIELD 2011 Conference on Mathematics of Medical Imaging



Description



We introduce (MICCAI 2011) a novel discrete optimization method for nonrigid image registration based on the random walker algorithm. We discretize the space of deformations and formulate registration using a Gaussian MRF where continuous labels correspond to the probability of a point having a certain discrete deformation. The interaction (regularization) term of the corresponding MRF energy is convex and image dependent, thus being able to accommodate different types of tissue elasticity. This formulation results in a fast algorithm that can easily accommodate a large number of displacement labels, has provable robustness to noise and a close to global solution.

Later, we reformulated the same problem using finite elements, resulting in a more efficient method that the original one based on finite differences (MICCAI 2013).

Below are some promising results for the method on two difficult medical image registration tasks.


Results for the discrete registration method on a synthetic data set, and on two real datasets.