Non-linear Registration with the Variable Viscosity Fluid Algorithm

Hava Lester1, Simon R Arridge1, Kalvis M Jansons2, Louis Lemieux3, Joseph V Hajnal4 and Anjela Oatridge4

Dept Computer Science, University College London, London WC1E 6BT, UK
- Dept Mathematics, University College London, London WC1E 6BT, UK
- Institute of Neurology, Queen Square, London WC1N 3BG, UK
- The MRC Cyclotron Unit, Hammersmith Hospital, London W12 OHS, UK

hava@bic.mni.mcgill.ca s.arridge@cs.ucl.ac.uk

         

Abstract:

In this paper we classify inhomogeneous non-linear registration algorithms into those of variable data influence, of variable deformability and of variable model type. As examples we introduce three modifications of the viscous fluid registration algorithm: passing a filter over the computed force field, adding boundary conditions onto the velocity field, and re-writing the viscous fluid PDE to accommodate a spatially-varying viscosity field. We demonstrate their application on artificial test data, on pre-/post-operative MR head slices and on MR neck volumes.

Contents