Conference Agenda

A cerebral aneurysm is a bulge that develops from a weak spot in the wall of a blood vessel in the brain. The rupture of such an aneurysm often leads to severe disability or the death of thousands of patients every year. However, several endovascular treatment options exist to prevent vessel wall rupture. In these minimally invasive surgical procedures, a medical device such as a coil or a WEB (Woven EndoBridge) is inserted into the bulge, or blood flow is diverted using a stent-like device, called flow diverter. The goal of all interventions is to reduce blood flow inside the aneurysm to trigger coagulation, thereby achieving complete occlusion of the aneurysm sac.

To support clinical planning and treatment, for each individual patient, numerical simulation can provide detailed insights into the patient-specific setting and serve as additional information for the attending neuroradiologist. Therefore, accurate models of all the various devices [2], such as flow diverters or contour neurovascular systems within cerebral aneurysms, are essential. Especially in complex, patient-specific cases, the combined application of different endovascular devices may offer improved treatment outcomes and long-term stability by accommodating diverse anatomical conditions up front into the decision process. Each of these devices consists of multiple thin nitinol wires, where their final placement significantly impacts the success of the intervention. To assist in this implantation process, we outline how a virtual deployment can be realized with the open-source multiphysics framework 4C[1]. The wire-like device will be modelled with geomtrically exact Simo-Reissner beam finite elements[3]. Different options are available to model the behavior at intersections due to interacting beam geometries. The deployment process consists of multiple steps: First, the flow diverter is compressed into a microcatheter. Then, the microcatheter moves the device into the vicinity of its target location. Finally, the device is deployed and expands inside the vessel. Throughout these steps, a consistent mixed-dimensional coupling approach [4] is employed to account for interactions between the device and the microcatheter or arterial wall. Within each intervention step a finite element approach is used to account for the mechanical behavior of the device, wall and catheter ensuring accurate representation of the deformations and interaction until the device reaches the desired position. Within this presentation, we will detail our modeling and simulation workflow and will compare our method to existing approaches in literature. Furthermore, we will discuss how this virtual deployment may increase the success rate of interventions through patient-specific treatment strategies.
Keywords:
finite element method, mixed-dimensional structural modeling, virtual endovascular deployment

References:
[1] 4C. 4C: A Comprehensive Multiphysics Simulation Framework. https://www.4c-multiphysics.org. Accessed: 12.05.2025. 2025.
[2] Martin Frank et al. “Numerical simulation of endovascular treatment options for cerebral aneurysms”. In: GAMM-Mitteilungen 47.3 (2024), e202370007.
[3] Christoph Meier, Alexander Popp, and Wolfgang A. Wall. “Geometrically Exact Finite Element Formulations for Slender Beams: Kirchhoff–Love Theory Versus Simo–Reissner Theory”. In: Archives of Computational Methods in Engineering 26.1 (2017), pp. 163–243. doi: 10.1007/s11831-017-9232-5.
[4] Ivo Steinbrecher et al. A consistent mixed-dimensional coupling approach for 1D Cosserat beams and 2D surfaces in 3D space. 2022. doi: 10.48550/ARXIV.2210.16010. url: https://arxiv.org/abs/2210.16010.

Introduction Minimally invasive cardiovascular interventions have significantly improved patient outcomes across various pathologies. However, post-operative complications remain a concern in every procedure, often influenced by device selection, performance, and procedural conditions. The use of finite element (FE) simulations (in-silico) and machine learning (ML) models trained for the prediction of interventional procedures, can facilitate the process of prosthesis selection, pre-procedural planning and post-operational risk assessment. The main bottle neck of training a Machine learning model on clinical data is the data accessibility. In this study, we present a comprehensive framework for assessing Thoracic Endovascular Aortic Repair (TEVAR) procedures by integrating structural simulations, synthetic population generation, and machine learning (ML) models. We use the relevant clinical data and imaging to augment the population, using statistical shape modelling (Anatomic analysis), create individual digital twins (high fidelity virtual replicates) of the TEVAR procedure, assess the post-operational complications and train machine learning models to predict the outcome of this procedure. This approach aims to provide additional information for clinicians in prosthesis selection and predicting procedural outcomes, ultimately reducing post-operative risks and enhancing intervention success rates.

Methods The framework utilizes statistical shape modeling to generate a synthetic population of thoracic aortas based on computed tomography (CT) scans from 25 patients who have undergone TEVAR. This original patient population is augmented by 100 synthetic anatomies, using statistical shape modelling. In this process, the geometrical characteristics of the real patients are assessed. Using principal component analysis, the most prominent shape modes are selected. Then assigning different weights to these shape modes new geometries, with similar geometric characteristics are created. Then statistical analysis on anatomical and pathological data of TEVAR patients are conducted to evaluate the representativeness of the synthetic geometries. In the second step, a simplified simulation model is employed to reduce computational costs while maintaining predictive accuracy. The results are compared with previously validated TEVAR simulation models. In the last step, different ML models are trained to determine the landing zone of the device for surgery planning.

Results The synthetic cohort of 100 virtual cases is analyzed, demonstrating that the generated geometries preserve key morphological characteristics of the initial patient-specific anatomies. The population exhibits favorable compactness, specificity, and generalization, enabling a thorough evaluation of post-operative device performance. The simplified simulation approach reduces computational time by 60% while maintaining prosthesis configuration deviations under 10% compared to standard models. Furthermore, the ML-based predictive model achieves 80% accuracy in determining the device landing zone.

Conclusions The integration of in-silico modeling with machine learning-driven intervention simulations provides clinicians with a powerful tool to refine prosthesis selection, streamline pre-procedural planning, and enhance post-operative risk assessment. By improving precision and reliability in cardiovascular interventions, these advancements ultimately lead to better patient outcomes.

Identifying carotid atherosclerotic lesions at risk of plaque rupture, a key precursor to cerebral embolism and stroke, is of critical clinical importance. High stress within the fibrous plaque cap has been proposed as a risk factor; however, the influence of residual strains on stress distribution remains poorly understood.

We present an advanced computational framework integrating homeostasis-driven Growth and Remodeling (G&R) into patient-specific carotid geometries to predict residual strains and assess their impact on plaque stress. The method employs a multiplicative kinematics-based remodeling approach to homogenize tissue stress while accounting for heterogeneous plaque composition. Patient-specific vessel reconstructions were generated from Computed Tomography Angiography (CT-A) imaging, with tissue classifications obtained via an artificial intelligence (AI)-driven histology-based segmentation tool, distinguishing vascular matrix (MATX), calcifications (CALC), lipid-rich necrotic core (LRNC), and intra-plaque hemorrhages (IPH). The framework was applied to a cohort of 18 patient-specific cases, incorporating a total Lagrangian formulation and novel post-processing metrics for biomechanical evaluation.

The incorporation of residual strains reduced peak wall stress by up to 30%, though the extent of reduction varied depending on plaque morphology and tissue composition. While G&R consistently reduced stress peaks within matrix-rich regions, high calcification led to localized stress concentrations, restricting remodeling. Additionally, the effect of residual strains on stress distribution was found to be negligible in cases with either minimal or severe stenosis, underscoring the role of plaque heterogeneity in determining biomechanical risk.

Our findings highlight the importance of patient-specific biomechanical modeling in evaluating plaque rupture risk. By incorporating residual strains and remodeling mechanisms, this approach provides a more physiologically relevant assessment of fibrous plaque cap stress, offering enhanced predictive capabilities for stroke risk stratification and potential clinical decision-making applications in atherosclerotic disease management.

Articular cartilage serves an important mechanical function in the human body, i.e., load bearing and shock absorption. For the design of implants for cartilage repair after injury or disease, it is key to thoroughly understand the unique biomechanical properties of the native tissue and potential substitute materials.

Here, we characterize the macroscopic large-strain mechanical properties of healthy and osteoarthritic human articular cartilage and perform histological investigations to determine their relation to the underlying microstructure. Furthermore, we use multimodal mechanical testing methods combined with hyperelastic nonlinear continuum mechanics modeling, finite element simulations, and immunohistochemistry to correlate the macroscopic behavior to the underlying microstructural properties of bioprinted cartilaginous constructs generated from bone marrow-derived mesenchymal stromal cells (MSCs) embedded in thiolated hyaluronic acid (HA-SH) hydrogels.

We individually characterize human articular cartilage tissue from the medial and lateral sides of the femoral condyle and tibial plateau to determine region‐specific mechanical properties. Our results show that there are no significant differences between the medial and lateral sides, but tissue from the tibial plateau is slightly softer than tissue from the femoral condyle. Osteoarthritis leads to a significantly softened mechanical response, which is related to corresponding microstructural changes. Furthermore, we confirm that the cartilaginous tissues based on chondrogenically differentiated MSCs in HA-SH hydrogels exhibit similar mechanical characteristics including nonlinearity, hysteresis, conditioning, and stress relaxation as human articular cartilage tissue. Our study reveals a positive correlation between the amount of the main components of the extracellular matrix (ECM) of articular cartilage, collagen (COL) and glycosaminoglycans (GAG), and the classical shear modulus as a measure of stiffness.

The presented results will help guide the design of implants that are able to restore cartilage structure and function, bridging biomechanics and regenerative medicine for osteoarthritis treatment.

Diabetes is a chronic condition characterized by hyperglycemia resulting from insulin deficiency or some physiological dysfunction. Current therapeutic strategies predominantly rely on external insulin administration through active syringe injections (SIs). Frequently however, Continuous Subcutaneous Insulin Infusion (CSII) are resorted to if an automized administration is needed. Despite their theoretical efficiency in managing blood glucose levels, both these methods present counter effective drawbacks, including the development of lipodystrophies (LPDs). LPDs consist in a pathological reorganization of the adipose tissue in response to an external fluid in the subcutaneous tissue and to the injection pressure, either in the form of lipoatrophy (tissue loss) or lipohypertrophy (tissue accumulation).

Some currently commercial fixed-site devices, including Subcutaneous Insulin Injection Ports (SIIPs), mitigate certain drawbacks of SIs, such as needle phobia and distress, while facilitating clearer application sites. However, the limitations associated with LPDs remain, especially because there is an increased development risk with repeated injections at fixed locations. LPDs emerge therefore as a critical concern, leading to alterations in adipose tissue morphology, hindering drug diffusion and ultimately leading to metabolic imbalances and higher insulin usage.

This study addresses the need for advanced modeling frameworks to optimize insulin port designs and mitigate LPD-associated inefficiencies. A multi-field computational approach is developed, using fluid dynamics simulations for realistic insulin injection scenarios, pharmacokinetics coupling to analyze insulin diffusion pathways, and image-based histological reconstructions to incorporate tissue damage models. The spatiotemporal dynamics of insulin distribution are evaluated under three scenarios: a Nominal Case (NC), where the Worst Case (WC) involving obstructions, and Optimal Case (OC) with enhanced device design.

The presented results will show how our modeling framework is capable of assessing tissue damage and insulin absorption delays because of successive injections. Insights into this “wall effect” created by damaged adipose has highlighted the importance of optimizing the geometry of the device. By addressing several realistic complications, the optimized devices enable the redistribution of potential damage across a wider tissue volume and a minimization of adverse effects.

The proposed framework sets the stage for improved insulin port device engineering, advancing therapeutic efficacy and achieving more stable and effective treatment outcomes for diabetic patients.