Invited Talk: Random walker based discrete deformable registration
Abstract:
I will shortly introduce the medical imaging projects at University of Alberta. Next I will detailed one of our recent works on discrete deformable registration.
We proposed a novel discrete deformable registration method based on the random walker (RW) algorithm. Our proposed method has two different but equivalent formulations, namely, the Gaussian Markov random field (MRF) and the variational formulations. Both these formulations entail convex energies that involve diffusion regularization. However, the variational formulation facilitates a computationally efficient finite element method (FEM) based implementation of our proposed method. We empirically demonstrate the our method obtains competitive results in comparison to other popular deformable registration methods on the CUMC12 and POPI-model datasets.
Bio:
Dana Cobzas is a research associate in the Biomedical Engineering Department and adjunct professor in the Computing Science Department at University of Alberta, Canada.
Her areas of expertise are centered around imaging and computer vision, with particular interest in mathematical models for medical image processing. She has experience and interest in methods for medical image segmentation, registration, noise reduction, DTI processing.
Before, Dana was a PIMS postdoctoral fellow jointly between the Mathematics Department and the Computer Science Department at University of Alberta. She spent her first postdoctoral years in the MOVI team, at INRIA Rhone-Alpes. She got her doctoral degree at University of Alberta in 2004.
