Langevin equations for landmark image registration with uncertainty
For researchers in medical image analysis and shape modeling, this provides a principled way to quantify registration uncertainty, though the approach is incremental.
The paper addresses uncertainty in landmark-based image registration by formulating a Langevin equation to model small random perturbations. It introduces computationally efficient approximations and uses a Bayesian framework to compute posterior distributions and average multiple landmark sets.
Registration of images parameterised by landmarks provides a useful method of describing shape variations by computing the minimum-energy time-dependent deformation field that flows one landmark set to the other. This is sometimes known as the geodesic interpolating spline and can be solved via a Hamiltonian boundary-value problem to give a diffeomorphic registration between images. However, small changes in the positions of the landmarks can produce large changes in the resulting diffeomorphism. We formulate a Langevin equation for looking at small random perturbations of this registration. The Langevin equation and three computationally convenient approximations are introduced and used as prior distributions. A Bayesian framework is then used to compute a posterior distribution for the registration, and also to formulate an average of multiple sets of landmarks.