Advanced Hybrid Tracking and Navigation for Computer-Assisted Interventions

Modern surgery is transitioning from traditional open surgery to minimally invasive procedures, and intra-operative imaging is increasingly used to refine surgical planning and to precisely target regions of interest. One fundamental requirement for image-guided interventions is the accurate knowledge about the spatial relations between all objects of interest. We critically review clinically established tracking technologies, including optical and electromagnetic tracking, and novel, upcoming technologies. Optical tracking offers high accuracy and reliability, but requires a constant line of sight. Electromagnetic tracking is susceptible to distortions of the electromagnetic field, but does not depend on line of sight. It is currently the dominant technology for tracking of internal organs, and of flexible instruments within the human body, but still poses challenges in terms of non-uniform accuracy and noise. This dissertation presents several new concepts for advanced hybrid tracking systems for computer-assisted interventions. First, the power of electromagnetic and optical tracking systems is amplified through mathematical modelling of a laparoscopic ultrasound transducer’s flexible parts. We visualise the ultrasound images in a tracked laparoscope’s video images, and via computer vision methods we close the loop to optimise the very same overlay accuracy, which is most important for live laparoscope video augmentation. Then we demonstrate a novel approach to hybrid image-based and electromagnetic tracking, by modelling the output as continuous. Compared to previous methods, smoothness in terms of inter-frame motion is greatly improved. Finally, we introduce a completely new tracking paradigm, by robotically keeping the electromagnetic tracking sensor always close to the centre of the tracking volume, in order to minimise electromagnetic tracking errors. This is the first general solution providing uniform tracking accuracy throughout the volume. In particular, this solution improves tracking accuracy by a factor of two, while at the same time extending the tracking volume up to the reach of the robot. The methods presented in this dissertation were developed in close collaboration with medical partners. In order to meet clinical requirements for use on real patients, they were thoroughly evaluated on phantoms and ex-vivo animal specimens. In a conclusion we demonstrate, how such advanced hybrid tracking solutions can be used to approach abdominal settings, where just a decade ago tracking might have been dismissed as impossible.