Viscoelastic Registration of Medical Images

Since the physical behavior of many tissues is shown to be viscoelastic, we propose a novel registration technique for medical images via physical principles of viscoelasticity. It allows the template and the target images to differ by large local deformations like viscous fluid and small local deformations like elastic solid. With an extra elastic term introduced, the proposed model constrains the transformation by coupled viscoelastic behavior and therefore reduces the smearing effect caused by the fluid terms. We apply our approach to both artificial test data and MR brain slices yielding accurate registration results.