Nav EVAR (Navigated Contrast-Agent and Radiation Sparing Endovascular Aortic Repair)

A fundamental problem of endovascular therapy of aortic aneurysms is the extraordinarily high radiation exposure - especially for the surgical team - caused by continuous X-ray imaging during the operation. In addition, this fluoroscopy requires the administration of high doses of nephrotoxic contrast medium to make the patient's blood vessels visible in the X-ray image. The BMBF-funded Nav EVAR project aims to solve these problems.

The Institute for Robotics and Cognitive Systems at the University of Lübeck (ROB Lübeck) is one of three technical partners in the project and has extensive expertise in the field of medical robotics and navigation. ROB Lübeck is involved in the subproject that deals with the following main points:

How can the position of an endovascular catheter be determined without X-ray imaging and contrast medium?

How can the surgeon understandably visualize the image and data acquired during the operation, together with the catheter position?

What accuracy can be achieved and can X-ray imaging be completely dispensed with?

There are approaches to all these questions which have been discussed and worked out jointly by the applicants and which are now to be developed into a demonstrator. The unique selling points of the project are in particular the use of state-of-the-art methods for the localisation of glass fibres with the aid of Bragg gratings in the fibre sheath and special augmented reality technology (Microsoft HoloLens). Robot-assisted 3D ultrasound is a pioneering technological alternative to X-ray imaging, which is also to be investigated in the project as a possible solution for the localization of the catheter.

The subproject's approach includes navigation, visualization, construction of endovascular hardware, intervention planning and modelling as well as validation and documentation. The research and realization of a combined Navigation solution (electromagnetic and glass fibre tracking), augmented reality visualization, patient localization by surface tracking, robot-assisted 3D ultrasound navigation as well as the validation of the accuracies of the individual approaches are the work steps.

The results will also be exploited jointly, through with different focal points. Commercial exploitation will take place in particular through the partners Fraunhofer-Institut für Bildgestützte Medizin MEVIS and the Medical Laser Center Lübeck and, in the medium term, a planned spin-off. The scientific exploitation (e.g. research into further areas of application, the integration of new technologies and the conception of follow-up projects) will be undertaken by the university partners ROB Lübeck, and the Department of Surgery and the Department of Radiology and Nuclear Medicine at University Hospital Schleswig-Holstein (UKSH).

We developed a new concept for rigid point cloud registration. In contrast to existing approaches, it is based on the physics laws of classical mechanics, electrostatics and thermodynamics. The registration problem is solved by modelling point clouds as manyparticle systems. Forces acting on the individual particles concentrate on the respective centre of mass of the cloud. Due to the application of force, a successive movement of the point cloud to be registered occurs in the direction of the static reference point cloud until both are aligned. The point cloud motion is derived from the physical model of rigid-body transformation. Force fields represent the metric of the registration and can be freely modeled. The sample metrics presented in this thesis are inspired by both, Newton’s law of gravitation and Coulomb’s law of electrostatics. Accordingly, the position of the point clouds can be described not only by their spatial point distribution, but also by additional characteristics such as color or intensity values.

Furthermore, an efficient algorithm is presented which implements the new registration approach. The regularization of point cloud movements is based on the method of simulated annealing. In order to keep the runtime complexity as low as possible even with very high-resolution point clouds, the Monte-Carlo-method is used. Furthermore, the algorithm can be parallelized and can be executed on multi-core processing architectures such as graphics processor units.

You can also find a free Matlab implementation here:

https://github.com/ROB-Uni-Luebeck/PIPL