Doctoral Defense by Bernhard Fuerst

Speaker: Bernhard Fuerst Date: Tuesday, October 25th, 2016 Time: 11:00 AM Location: FMI-Building, Room 00.05.035

Multi-modal Registration and Robotic Imaging for Computer Assisted Surgery

Abstract:

Imaging and visualization of anatomy and physiology are the very foundation of Computer Assisted Surgery, and crucially contribute to the success of interventions. The goal of the research presented in this thesis is the introduction of novel image registration techniques as means of transferring pre-operative imaging and planning to the surgical scenario, and intra-operative image acquisition to obtain images just in time and within the surgical site.

The contribution to image registration include the prediction of deformation, by generating a patient-specific anatomical biomechanical model for which spatial and temporal varying physiological properties are estimated based on observations from 4D Computed Tomography (CT) volumes. Another contribution to image registration are similarity measures for multi- modal image registration are proposed and evaluated. Linear Correlation for Linear Combinations (LC2) is used to enable the alignment of images from modalities that measure different physical properties. This allows the combination of imaging data containing complementing information, but the alignment fails when the initial alignment is poor.

The second part of this thesis focuses on the interventional imaging, which is of special importance when large deformations render pre-operative data of little use, real-time imaging is required for image-guided interventions, or when imaging is used to simplify surgeries. One of the contributions is the introduction of minimally invasive robotic Single-Photon Emission Computed Tomography (SPECT) imaging for robot-assisted surgery, for which vision based tracking of a miniaturized gamma probe enables the SPECT reconstruction using Maximum Likelihood Expectation Maximization (MLEM).

Finally, this thesis concludes with the presentation of augmented reality application for to interventional imaging during orthopedic surgeries. The technical contributions aim at fusing Cone-Beam Computed Tomography (CBCT) volumes and live 3D color images. The usability of this novel concept is evaluated in a multi-user study using objective and subjective measures.