King's College London

Research portal

Non-Rigid 2D-3D Registration for Use in Computer-Assisted Abdominal Aortic Aneurysm Repair Procedures

Student thesis: Doctoral ThesisDoctor of Philosophy

This thesis contributes to knowledge by describing three methods to non-rigidly register 2D X-ray images acquired during a Complex Endovascular Aneurysm Repair (CEVAR) procedure to a 3D pre-operative CT scan.
The first part of the thesis presents an interpolation framework (thin-plate spline) that is tailored to accurately register 3D CT scan data to 2D X-ray projection data. Registering the 3D to the 2D images proves challenging, due to the lack of information perpendicular to the imaging plane. A method to interpolate manually selected displacements of 3D points located on the aorta surface has been tailored to model the known error distribution along the X-ray projection directions.
The second part of the thesis describes the intra-operative use of finite element based algorithms to deform the aorta surface based upon the positions of a guidewire during a CEVAR procedure. The aorta is sequentially deformed so as to encompass a simulated wire which was initialised at the centrelines of the aorta and which is dragged towards the guide-wire. Experiments were conducted on the mechanical parameters of the finite-element model and showed the influence of the Young Modulus and the Poisson’s ratio on registration accuracy.
The third part of the thesis focuses on the novel use of interventional digital tomosynthesis images to extract intra-operative information on the calcifications of the aorta and drive non-rigid registration of the aorta during CEVAR. Calciumbased correspondences were established between the pre-operative and the intraoperative scene. A similarity measure has been defined as a weighted sum of a bending energy term and a second term that estimates how well pre-operative and intra-operative patches of calcium match. Erroneous correspondences are corrected using a simulated annealing optimisation on this similarity measure.
Using the three methods, large rigid registration errors of 9 mm were brought down to 4 mm or below the clinical target of 3 mm (half the diameter of the renal arteries). The proposed methods fit well with current clinical workflows. The first method presented above requires little manual input during the operation and the two other methods are/can be fully automated. The work presented in this thesis has the potential to increase the availability of image guidance systems for CEVAR procedures and for minimally invasive surgery where soft tissues are involved.
Original languageEnglish
Awarding Institution
Award date2016


Download statistics

No data available

View graph of relations

© 2020 King's College London | Strand | London WC2R 2LS | England | United Kingdom | Tel +44 (0)20 7836 5454