International Journal for Numerical Methods in Biomedical Engineering最新文献

Recent high-fidelity fluid–structure interaction (FSI) simulations of cerebral aneurysms have revealed flow-induced wall vibrations. However, those simulations were conducted under simplified conditions, and the robustness of the predicted vibrations remains unknown. This study aimed to advance the physiological accuracy of previous models and to investigate the sensitivity to parameter uncertainty. We compared the previously used near-linear St. Venant–Kirchhoff wall model with a three-term hyperelastic Mooney–Rivlin (MR3) model fitted to experimental data and also modeled effects of surrounding cerebrospinal fluid (CSF). We then varied flow rate (1.83 mL/s $��\pm �$ 25%), wall stiffness (soft, medium, stiff), and wall thickness (0.25 $��\pm �$ 0.1 mm). Our main findings for the four aneurysms considered were as follows: the MR3 model led to an average increase of 35% in pulsation and 240% in vibration amplitude, along with an 18% decrease in frequency. Viscous damping by the CSF reduced the vibration amplitude by 68% but did not affect the frequency or pulsation. Changes in flow rate had no effect on pulsation but increased vibration amplitude by 246%. Wall stiffness and thickness had a comparatively smaller impact on vibration, altering amplitude by 36% and 82% and frequency by 20% and 8%. In conclusion, the more advanced models led to a decrease of vibration amplitude and frequency during the cardiac cycle, consistent with clinical observations. Like computational fluid dynamics, FSI simulations can be sensitive to flow rates but are otherwise robust and can provide a fundamental understanding of aneurysm wall vibration without precise knowledge of wall properties.

Anterior cruciate ligament (ACL) reconstruction is a complex surgical procedure with high precision requirements. One of the most critical steps during the operation is the accurate localization of the insertion center and the proper placement of the graft. However, under limited visibility of the arthroscope, surgeons often face challenges in determining the correct position and angle. In this paper, we propose a novel framework for insertion center localization, incorporating two reconstruction approaches. First, an isometric reconstruction method based on 3D geometric processing algorithms is proposed, which leverages the feature lines and points of the geometric morphology of the actual bone model. This method ensures accurate alignment with bones of varying shapes. Second, a novel U-shaped network (EP-UNet), which integrates axial edge and coordinate features, is proposed for medical image segmentation to facilitate anatomical reconstruction. Experimental results demonstrate that EP-UNet exhibits superior performance in ACL segmentation tasks, achieving high accuracy and robustness. Compared to the baseline network, it improves the mean Intersection over Union (mIoU) by 8.16%, providing strong support for ligament image segmentation. This framework allows surgeons to efficiently and automatically determine the position of the patient's insertion center, addressing the challenge of clinical localization with high reliability and improving surgical success rates.

This study aims to develop an experimental platform that emulates the human cardiovascular system to investigate the effects of varying pulse rates and fluid properties on pressure drop, peristaltic pump output pressure, and power consumption. To support the experimental findings, computational fluid dynamics (CFD) simulations were conducted to analyze single-phase blood flow dynamics. Idealized arterial geometries representing healthy (0% stenosis) and diseased (65% stenosis) conditions were reconstructed from computed tomography (CT) images. A non-Newtonian blood-mimicking fluid (XSCN) was formulated to better replicate the rheological behavior of blood, while distilled water was used as the Newtonian reference fluid. Experiments were conducted at six different pulse rates: 72, 84, 96, 114, 132, and 156 beats per minute (bpm). The experimental setup was replicated in a virtual environment using ANSYS Fluent to simulate flow behavior under identical boundary conditions. The results demonstrate that increasing pulse rate leads to an increase in pressure drop (ΔP), pump output pressure, and power consumption for both arterial models. These effects were more pronounced in the stenosed artery due to flow constriction. Elevated turbulence intensity was observed at higher pulse rates, with notable differences between Newtonian and non-Newtonian fluids, particularly in terms of flow resistance and shear-dependent viscosity. Power consumption was found to be directly correlated with fluid viscosity, which varied with shear rate in the non-Newtonian fluid. The 65% stenosed model consistently exhibited higher pressure drops and flow irregularities. Fractional flow reserve (FFR) analysis confirmed that a 65% luminal narrowing poses significant hemodynamic risk. The highest wall shear stress (WSS) values were localized in the stenotic region, contributing to disturbed flow patterns and increased turbulence downstream. The non-Newtonian fluid model revealed that WSS was more sensitive to flow alterations, emphasizing the role of shear-dependent viscosity in vascular hemodynamics. These findings underscore the critical influence of hemodynamic parameters—such as pulse rate, viscosity, and arterial geometry—on cardiovascular performance. The study further highlights the detrimental impact of arterial stenosis on blood flow behavior and energy expenditure, with implications for clinical diagnosis and treatment planning in cardiovascular diseases.

An experimental and numerical investigation of the blood flow across the stenosis embedded in the elastic artery had been carried out within a framework of a project aimed at the development of the decision support system for the diagnostics of stable coronary artery disease, using a novel hybrid physics-informed machine learning (ML) and computational fluid dynamics (CFD) approach. Integration of CFD equations into a ML framework requires the development and validation of a CFD model optimized for this specific task. The values of empirical coefficients required for the implementation of the CFD component of the project were obtained by collecting experimental data on a pressure drop in elastic arterial models with stenoses in the physiologically relevant range of flow conditions, and the results were used for the validation of the numerical solver. Analysis of the experimental data also demonstrated a strong impact of the vessels' elasticity on the pressure drop across a stenosis, as well as a substantial role of the time-dependent (pulsative) flow parameters. It has been shown that fine-tuning of the values of the viscous and turbulent resistance coefficients to account for the elasticity of the vessels surrounding the stenosis can substantially improve the accuracy of the pressure drop prediction in the intermediate lesions, specifically relevant for clinical applications.

Total disc replacement (TDR) is an emerging technique for addressing degenerated intervertebral discs. However, the first generation of TDR has been associated with the generation of wear debris, which may adversely affect surrounding biological tissues, and they fail to fully replicate the range of motion (ROM) of a healthy intervertebral disc. This study aims to compare two generations of TDRs to determine which more effectively mimics the biomechanical behavior of a biological disc while minimizing associated complications. Four finite element models (healthy L4-L5, Prodisc-L, SB-Charité, and a second-generation TDR) were studied using Ansys under specific loads and moments: 7.5 Nm and 1175 N in flexion, 7.5 Nm and 500 N in extension, 7.8 Nm and 700 N in lateral bending, and 5.5 Nm and 720 N in axial rotation. First-generation TDRs reduce ROM in flexion (−61% for Prodisc-L, −65% for SB-Charité) and in extension (−59.37% and −79%). However, they increase ROM in lateral inclination (+121% and +100%) and in axial rotation (+129.41% and +111.76%). The second-generation TDR shows minimal deviations from the intact model, except in extension. First-generation of disc prostheses do not maintain 100% ROM of an intact intervertebral disc and generate wear debris during operation, potentially compromising surrounding biological tissues. In contrast, second-generation of disc prostheses closely mimic the ROM of an intact disc due to the hyperelastic properties of their core and eliminate wear debris production through to its monobloc design.

Despite being a common procedure, abdominal hernia treatments still require improvement due to the number of relapses and other postoperative issues. In silico testing can be employed to predict the behavior of the complex abdominal wall and implant systems. Here, uncertainty quantification and sensitivity analysis are required in order to optimize the parameters of hernia repair models. This paper concerns the modeling of an abdominal wall and implant using the finite element method. A Gasser-Ogden-Holzapfel (GOH) material model is used for the abdominal wall and an orthotropic material model for the implant. The parameters of the GOH model and the orientation of the implant are assumed to be uncertain. Regression-based polynomial chaos expansion is used as a meta-modeling method for uncertainty propagation and global sensitivity analysis. The maximum force in the connection between the implant and native tissue is considered as the quantity of interest. A failure risk criterion is also defined and presented. It has been found that the significance of the material parameters depends on the type of implant that is analyzed. Likewise, the risk of connection failure varies considerably depending on the implant used. Models with different types of implant produce very diverse results. Moreover, these differences also appear in the global sensitivity index and the risk of connection failure. This would indicate that specific implant designs and material properties are crucial to the success of hernia repair surgery.