Automation of Fenestration Planning for Physician-Modified Endovascular Grafts and Development of a Patient-Specific 3D-Printed Guidance System

Fenestrated Endovascular Aneurysm Repair treats complex abdominal aortic aneurysms involving visceral branch vessels. Physician-Modified Endovascular Grafts (PMEGs) offer a practical alternative when custom-manufactured devices are unavailable. Fenestration placement depends on precise ostium localization and measurement relative to the graft. Currently these measurements are performed manually, a process that is time consuming and subject to interobserver variability, motivating the need for a reproducible and automated FEVAR approach. This thesis describes an automated framework for patient-specific PMEG planning from Computed Tomography Angiography (CTA) data. The workflow begins with automatic aortic segmentation using a modified CIS-UNet model, followed by centerline extraction, automatic branch ostium detection, and branch geometry analysis. From these steps, the system computes longitudinal distances, circumferential clock positions, and fenestration rings dimensions. These parameters feed into patient-specific 2D planning templates, clock diagrams, coordinate tables, and planning reports, alongside 3D graft models with integrated fenestrations and STL export for visualisation and additive manufacturing. To support different planning requirements, the framework combines patient-specific anatomical modelling with standardized graft representations suitable for template generation and device planning. A wire-overlap avoidance strategy was implemented to automatically identify fenestration locations that minimize interference with the graft wire structure while preserving the anatomical arrangement of target vessels. The framework was evaluated on patient-specific datasets and successfully generated complete planning outputs across all stages of the workflow. Results confirm automating principal FEVAR planning tasks is technically feasible with geometrically consistent outputs. Larger cohort studies and clinical comparison will be needed to establish accuracy.