Computational Fluid Dynamics Simulation of Trans-Arterial Embolization for Liver Cancer: A Comparative Study of Patient-Specific Modeling Approaches

Hepatocellular carcinoma (HCC) is the most prevalent form of primary liver cancer. Since it is often diagnosed at a late stage, many patients are ineligible for potentially curative treatments. Therefore, developing innovative therapeutic strategies for advanced disease is of great importance. In patients with intermediate- or advanced-stage HCC, trans-arterial embolization (TAE) is a common palliative treatment. It aims to selectively obstruct the hepatic arteries supplying the tumor through the administration of embolizing microparticles via catheterization. Among its variants, trans-arterial radioembolization and chemoembolization are the most widely used, combining local delivery of radio- or chemo-therapeutic agents. The procedure success depends on achieving a selective and homogeneous microparticles distribution within the tumor vasculature, sparing the healthy liver parenchyma. In this context, computational models have been proposed to virtually reproduce TAE and to assess how clinical variables, such as catheter position or injection velocity, affect treatment outcomes. Computational fluid dynamics (CFD) has emerged as a valuable tool for simulating the complex hemodynamics involved in the procedure. Recent advances in computational resources and segmentation techniques now enable TAE simulation in patient-specific anatomical reconstructions, paving the way toward personalized treatment planning. One of the main drawbacks of in silico TAE models is the high computational cost associated with CFD simulations. The present thesis aims to provide a comparison of different particle tracking modeling strategies. The objective is to identify the modeling approach that offers the optimal trade-off between computational efficiency and accuracy in predicting particle distribution. The TAE procedure was virtually performed on a patient-specific model, adopting different CFD-based particle tracking methods. Therapeutic microspheres were injected through a standard catheter over one cardiac cycle within the hepatic arterial tree, which was simplified according to the cancer scenario to reduce computational time. Particles transport was modeled using three different approaches: i) Discrete Phase Modeling (DPM), based on an Euler-Lagrangian approach, either neglecting (one-way coupling) or accounting for (two-way coupling) the interaction between the dispersed and continuous phases; ii) a streamline-based approach, in which particles follow blood flow streamlines. Additionally, to assess the impact of arterial geometry simplification, the same procedure was simulated using two-way coupling DPM on three geometries with different truncation levels. Overall, the three particle tracking methods yielded comparable outlet-to-outlet particle distributions, with a maximum absolute difference of 2%. No significant differences were observed in the percentage of particles reaching the tumor-feeding branches. Regarding computational cost, the streamline-based approach proved to be the least demanding, as only hemodynamics needs to be explicitly simulated over a limited number of cardiac cycles, while particle trajectories are subsequently computed from the blood velocity field. This work represents a step forward in the development of standardized and reproducible in silico models of trans-arterial therapies for clinical decision support. Nevertheless, a more comprehensive analysis involving a broader set of patient-specific cases and TAE configurations, will be required to generalize these findings.