Development of decision support tools by model order reduction for active endovascular navigation

Abstract

Endovascular therapies enable minimally invasive treatment of vascular pathologies by guiding long tools towards the target area. However, certain pathways, such as the Supra-Aortic Trunks (SATs), present complex trajectories that make navigation challenging. To improve catheterization access to these challenging targets, an active guidewire composed of Shape Memory Alloy has been developed. Our study focuses on navigating this device and associated catheters to reach neurovascular targets via the left carotid artery. In previous work, a finite element model was used to simulate the navigation of the active guidewire and catheters from the aortic arch to the branching of the left carotid artery in patient-specific aortas. However, these numerical simulations are computationally intensive, limiting their feasibility for real-time navigation assistance. To address this, we present the development of numerical charts that enable real-time computation based on high-fidelity FE simulations. These charts predict: (1) the behavior of the active guidewire, and (2) the navigation of the guidewire and catheters within specific anatomical configurations, based on guidewire and navigation parameters. Using the High Order Proper Generalized Decomposition (HOPGD) method, these charts achieve accurate real-time predictions with errors below 5 % and a response time of ${10}^{-3}$ seconds, based on a limited number of preliminary high-fidelity computations. These findings could significantly contribute to the development of clinically applicable methods to enhance endovascular procedures and the advance the broader field of neurovascular interventions.