Roy Schestowitz Student Talks, 2006 Session

Model-based Assessment of Non-Rigid Registration

Abstract

Non-rigid registration is the task associated with alignment of images. It
is ubiquitously as a basis for medical image analysis and its applications
include  structural analysis, atlas matching and change analysis. The task
is  highly under-constrained, so a variety of different registration algo-
rithms produce different results for any given set of images.

We  compare two methods for assessing the performance of non-rigid  regis-
tration algorithms. One approach utilizes a measure of overlap between da-
ta  labels. Our new approach exploits the fact  that, given a set of  non-
rigidly  registered images, a generative statistical appearance model  can
be  constructed.  We observe that the quality of the model depends on  the
quality  of the registration, and can be evaluated by comparing  synthetic
images sampled from the model with the original image set.

Indices  of model specificity and generalisation are derived, which can be
used  to  assess model/registration quality. We show that both  approaches
detect the loss of registration as a set of correctly registered MR images
of  the brain is progressively perturbed. We compare the sensitivities  of
the different methods and show that, as well as requiring no ground truth,
our  new  specificity measure provides the most sensitive approach to  de-
tecting misregistration.