New Guidance Strategies for Endovascular Catheterization: Exploring the Potential of Radiation-free Sensing

Coronary artery disease is the result of plaque build-up in the arteries supplying blood to the heart. Severe plaque accumulation can result in complete blockage of one or several coronary arteries for more than 3 months. Such blockage is referred to as a coronary Chronic Total Occlusions (CTO), which significantly increases the risk of adverse cardiovascular events. In order to treat CTOs, patients undergo Percutaneous Coronary Intervention (PCI). PCI is a minimally invasive procedure, involving the percutaneous insertion and proximal manipulation of catheters and guidewires towards a target site, under (two-dimensional) fluoroscopic guidance. Despite providing significant advantages for patients, PCIs subject interventionists to very high radiation doses, in fact possibly the highest among all medical disciplines, resulting in severe health effects. Efforts to mitigate radiation exposure have worsened the ergonomics of the procedure. In addition to health concerns, current PCIs still offer poor scene awareness (due to two-dimensional fluoroscopy) and a lack of accurate information of the catheter/surgical instruments within the vasculature. These characteristics hinder the clinicians’ ability to determine the exact vessel pathology to treat and to devise an optimal motion strategy. Current guidance and robotic catheterization systems have tried to address these shortcomings. However, they still rely on ionizing radiation for guidance, do not provide accurate intra-operative anatomical information and offer limited catheter shape and/or location feedback. Vessel morphology/pathological information is also lacking, and additional (tedious) setup steps such as co-registration are required.

This thesis makes contributions towards an innovative intra-operative guidance system for catheterization procedures, particularly (CTO-)PCI, employing radiation-free sensing. The thesis aims to expedite catheter navigation and reducing operator dependency in PCI. To this end, an easy to train network architecture for intravascular image segmentation is first presented. Proposing a novel encoding scheme, this network handles imbalanced lateral and axial image resolutions. In addition to segmentation, a novel encoder allows for the detection of vascular structures per image. Evaluation on three different datasets demonstrates that the proposed method outperforms alternative state-of the-art segmentation approaches, while efficiently predicting clean segmentation masks. Also, the method directly extracts abstract coordinate information at a fast speed. With the goal of further enhancing in-situ situational awareness and assisting catheter navigation, a three-dimensional (3D) vessel modeling framework is developed. The framework represents and estimates important features of the vessel geometry in the vicinity of the catheter tip in real time. Results show that the proposed framework could constitute a step towards providing navigation maps for (autonomous) catheter navigation. Next, a 3D vessel reconstruction strategy building up from the previous modeling framework is introduced. This method is capable of providing global and detailed intra-operative visual feedback while the catheter advances through the vessel. Finally, a constraint-based navigation strategy for safer, (semi-)autonomous robotic catheter steering is devised. Leveraging the intra-operative 3D vessel modeling framework, the proposed control scheme is tested using a robotic catheter with a distal active segment in a virtual aortic model and in an in vitro blood vessel phantom. Results confirm that intra-operative 3D vessel information can be used to formulate robot tasks automatically and that this technology can be integrated to improve the safety of catheter navigation.

This doctoral thesis demonstrates advancements on image segmentation, real-time 3D blood vessel modeling and reconstruction, and catheter navigation, using only non-harmful sensing technologies. In addition, even though conceived and tailored for PCI, the work hereby presented could be applied to other minimally invasive catheter-based procedures.