KCL logo

Motion Modelling and Analysis Group

School of Biomedical Engineering and Imaging Sciences,

King's College London


Hierarchical Local Affine Registration

(Funded by EU HYPERIMAGE project)

In image registration the group has an interest in developing fast and efficient ways of estimating motion from imaging data. A hierarchical local affine approach to intensity-based registration was developed, in which the hierarchical divisions took into account the image content, in order to ensure that regions of approximately affine motion were formed. The hierarchical nature of the algorithm, which employed multiple simple registrations rather than a single complex nonlinear registration, resulted in a computationally less complex algorithm than other comparable techniques.

Example lreg registration
An example of hierarchical local affine registration.


Ultrasound/MR Registration Using an Ultrasound Imaging Model

(Funded by EPSRC grant EP/D061471/1)

In this work a novel registration scheme was developed for aligning preprocedure segmentations (i.e. 'roadmaps' used for guidance) to intraprocedure ultrasound data. The technique incorporated into the registration a model of how the ultrasound images were formed. The algorithm therefore attempted to find the registration that was most likely to have resulted in the observed ultrasound image. The technique was demonstrated for rigid-body registration and also for respiratory motion correction by integrating it with a motion model.

Rigid registration.
Respiratory motion correction.


Motion-Based Image Registration

(Funded by EPSRC grants EP/K030310/1 and EP/K030523/1)

As well as intensity information, often image sequences are available, and it can be useful to align such sequences. For example, multiple sequences from the same subject can be aligned and compounded for improved visualisation of anatomy, or sequences from different subjects can be aligned for comparative analysis. In this work the motion information provided by such sequences was exploited to facilitate robust registration of 4-D echocardiographic sequences. This was made possible by the use of a novel subspace error metric that encoded the dominant modes of the motion and permitted fast comparisons of the image seqences.

Motion-based image registration: (left) evolution of subspace error metric; (right) registered echo sequences.