Conference Agenda

We present a multiphysics computational framework for simulating the fully coupled interplay of cardiac electrophysiology, active and passive muscular mechanics, blood dynamics and valve kinematics in the human heart [Bucelli et al., IJNMBE, 2023]. The framework supports relevant sources of feedback between the models, including electro-mechanical and mechano-electrical coupling and fluid-structure interaction (FSI). We account for the geometric coupling between fluid and solid through the body-fitted ALE approach. The fluid domain is displaced with a non-linear mesh motion operator, which provides the robustness needed to deal with the large displacements characterizing the heartbeat. We account for valves through the resistive immersed implicit surface (RIIS) method [Fedele et al., BMMB, 2017], adapted to account for the FSI model of cardiac walls and blood flow. The heart model is coupled at inlets and outlets to a closed-loop lumped parameter model of the circulatory system [Fedele et al., CMAME, 2023].

We solve the coupled problem through the finite element method, treating the multiphysics coupled system through a segregated-staggered time discretization, wherein different submodels are discretized with different temporal resolutions to account for their particular accuracy requirements. The FSI subproblem is addressed with a monolithic solver that balances computational efficiency, accuracy and robustness.

The resulting computational framework is highly modular, meaning that its complexity may be adjusted depending on the need of each application. We demonstrate this through applications to cardiac electro-fluid-structure interaction of a realistic left heart, whole-heart cardiac electromechanics, and standalone hemodynamics models. The latter can be driven by cardiac electromechanics (resulting in a one-way coupled FSI system) or by a patient-specific calibration of the circulatory system. This is applied to proof-of-concept simulations of atrial hemodynamics under fibrillation coupled to models of thrombus formation.

The EuroHPC JU project dealii-X grant no. 101172493, funded under the HORIZON-EUROHPC-JU-2023-COE-03-01 initiative, is acknowledged.

The Woven EndoBridge (WEB) is an endovascular device increasingly used for treating intracranial aneurysms. However, selecting the appropriate WEB size preoperatively remains challenging, as it deforms significantly upon deployment depending on the aneurysm morphology. Approximately 20% of cases require redeployment due to size mismatch. Furthermore, while hemodynamic alterations are considered key to aneurysm occlusion, the number of studies quantifying changes before and after WEB deployment is limited. This study aimed to predict the post-deployment geometry of the WEB using preoperative imaging and to evaluate hemodynamic changes induced by the device. A patient-specific model of unruptured intracranial aneurysm that achieved complete occlusion after WEB deployment was analyzed. The aneurysm and arteries were reconstructed to generate a computational grid. For the post-deployment analysis, a virtual WEB geometry was modeled using an original Virtual WEB Deployment Simulation technique, which accounts for the deformation of the WEB based on its geometrical configuration and preoperative intracranial aneurysm morphology. Computational fluid dynamics (CFD) analysis were conducted for both before and after WEB deployment. The flow field was assumed to be incompressible and laminar, with blood modeled as a Newtonian fluid having a density of 1100 kg/m³ and a viscosity of 0.0036 Pa·s. The arterial wall was assumed to be rigid. As boundary conditions, the average diastolic mass flow in a healthy adult’s internal carotid artery was used at the inlet, and fixed static pressure of 0 Pa was applied at the outlet. We evaluated the mean blood flow velocity within the aneurysm, mean pressure, and mean wall shear stress (WSS) on the aneurysm wall before and after WEB deployment. Additionally, the rate of change for these parameters were calculated. As a result, the simulation qualitatively reproduced the geometrical characteristics of the deployed WEB, with the wires expanded along the aneurysmal wall in a manner closely resembling clinical deployment. However, geometrical discrepancies were observed near the neck region, indicating the need for further refinement of the modeling approach. Furthermore, after the WEB deployment, the mean blood flow velocity within the aneurysm decreased from 0.342 m/s to 0.159 m/s (−53.5%), and the mean WSS decreased from 7.99 Pa to 2.76 Pa (−65.5%). A reduction in pressure concentration at the aneurysm dome was also observed. These findings suggest that our Virtual WEB Deployment Simulation may enable accurate preoperative prediction of post-deployment WEB geometry and assist in selecting optimal size before surgery. Furthermore, a reduction in mean blood velocity within the aneurysm, the disappearance of high-pressure regions, and a decrease in mean WSS were observed after WEB deployment. These findings suggest that the reduction in blood flow velocity induced by WEB deployment may contribute to aneurysm occlusion.

The complex motility patterns of the stomach, characterized by antral contractions and tonic contractions, play crucial roles in mixing, grinding, and propelling food through the gastrointestinal tract. These mechanical processes, combined with chemical digestion, break down ingested food into smaller particles, increasing surface area for enzymatic action and nutrient absorption. Computational Fluid Dynamics (CFD) has emerged as a powerful in silico tool to complement in vitro and in vivo studies, offering detailed insights into gastric flow patterns and mixing mechanisms that are difficult to measure experimentally. CFD models can simulate the fluid dynamics of gastric contents, track chemical species, and analyze the impact of varying physiological parameters on digestion efficiency.

Existing CFD models of gastric motility have primarily represented gastric contents as Newtonian fluids, with food often modelled as a chemical species rather than explicit solid particles. This simplification limits the ability to study the mechanical breakdown of solid foods, a critical aspect of gastric digestion. Furthermore, many models represent the opening and closing mechanisms of the pyloric sphincter, instead assuming a constantly open or closed pylorus. Additionally, most studies simulate only brief periods of gastric activity, typically minutes rather than the hours required for complete digestion, mainly due to the computational costs. These limitations highlight the need for more comprehensive models that capture the full complexity of digestion.

To address these limitations, we have developed a coupled particle-based model using Smoothed Particle Hydrodynamics (SPH) and Lattice Spring Model (LSM) techniques. This approach is better suited to simulate the large amplitude contractions of the stomach wall and pyloric sphincter, as well as the mechanical breakdown of solid foods. Our SPH model of gastric emptying was implemented using the open-source LAMMPS package on an idealized stomach and duodenum geometry. The model incorporates boundary deformation patterns derived from a previously developed model of gastric motility and pylorus sphincter activity, allowing for detailed simulation of gastric motility.

The SPH model successfully captured key gastric flow patterns, including retropulsive jets and recirculation regions, which are essential for effective mixing and emptying. The simulated half-emptying time for water (13 minutes) aligns closely with in-vivo human studies. The ability of the CFD model to represent pyloric sphincter dynamics provides a more accurate representation of gastric emptying regulation. The solid-phase of digesta is represented using LSM to explicitly model food particle breakdown.

Future work includes developing subject-specific models using MRI data from an ongoing study investigating the gastric digestion of halloumi cheese. This study simultaneously measures motility patterns and digestion kinetics, providing valuable validation data for our SPH-LSM coupled model. Our approach offers a significant advancement in computational modelling of gastric function, bridging the gap between micro-scale processes and whole-organ mechanics, and providing new insights into smooth muscle organ dynamics.

Accurate assessment of rupture risk is critical for the clinical management of intracranial aneurysms, as rupture can result in life-threatening subarachnoid hemorrhage with a high mortality rate. Although unruptured aneurysms are increasingly detected due to advances in diagnostic imaging, low annual rupture rate poses a clinical dilemma regarding surgical intervention. Previous studies have reported that thin-walled regions (TWRs)―areas of reduced wall thickness and structural fragility―are likely sites for rupture initiation. However, non-invasive identification of TWRs remains challenging, since conventional imaging modalities cannot directly evaluate aneurysm wall thickness. Recently, computational fluid dynamics (CFD) analysis has been employed to investigate hemodynamic factors associated with TWRs. This study aimed to construct a predictive model for TWRs using CFD analysis and to assess its practical applicability through validation.

A total of 124 aneurysm cases treated with surgical clipping at two institutions were analyzed. At Institution A, 109 cases before May 2021 were used to build the predictive model, and 13 subsequent cases were used for validation. Two additional cases from Institution B were used for external validation. Patient-specific three-dimensional arterial geometries were reconstructed from angiographic imaging. CFD simulations were performed assuming incompressible and laminar flow modeled as a Newtonian fluid. A rigid wall was assumed, and the no-slip boundary condition was applied to it. The inlet boundary condition reflected typical pulsatile blood flow in healthy adults, while the outlet pressure was set to 0 Pa. Hemodynamic parameters such as Pressure Difference (PD*) and Wall Shear Stress Divergence (WSSD*) were calculated on the aneurysm wall. Intraoperative microscopic images were used to quantify wall color via the comprehensive Red (cR) value, with the top 25% of pixels defined as TWRs. A prediction model for TWRs was formulated using logistic regression based on hemodynamic parameters. The Risk of TWR (RoT) was calculated for the aneurysm surface, and the top 25% RoT regions were defined as predicted TWRs. Prediction accuracy was evaluated using Inclusion Rate, which measures the extent to which the predicted region overlaps with the actual TWRs.
Logistic regression showed that PD* and WSSD* were significantly higher in TWRs than in thick-walled regions. A prediction formula for TWRs was constructed based on multivariate analysis. Application of this model to the training dataset yielded an average Inclusion Rate of 77.2%, while validation on cases from Institutions A and B resulted in rates of 77.9% and 75.2%, respectively. These results indicate comparable predictive performance across datasets. In high-Inclusion Rate cases, areas with high PD* and WSSD* qualitatively matched intraoperatively observed TWRs, reflecting localized stress concentrations in both perpendicular and tangential directions. In low-Inclusion Rate cases, often involving large aneurysms with stagnant flow, the prediction parameters did not exhibit high values, which likely made it difficult to identify TWRs. Incorporating morphological factors of intracranial aneurysms may improve the accuracy of TWR prediction.

In conclusion, this study demonstrates the feasibility of CFD-based prediction of TWRs in intracranial aneurysms across multiple datasets. The proposed method may support more accurate preoperative risk assessment and help guide treatment decisions to improve patient outcomes.