Abstract
In this paper, we use reverse modeling software to complete the reconstruction of the aortic valve based on the modeling principles of the human heart valve. By using cardiac CT scanning to obtain DICOM format data and then using the 3D software Mimics for segmentation, editing, and smoothing, we generate a realistic and precise structural heart valve model and peripheral blood surface model in STL format. These models are then integrated in Geomagic Studio for further optimization. Finally, the model is imported into ANSYS Workbench for finite element analysis preparation, focusing on the leaflets and the arterial wall.
Introduction
The human heart valve presents a significant geometric challenge for engineers due to its complex, highly curved, and dynamic surfaces. Creating digital models that are both anatomically accurate and mathematically suitable for computer simulations is difficult. Standard medical imaging provides raw data, but a significant gap exists between a stack of 2D CT slices and a functional 3D mesh used for simulating blood flow or structural stress. This paper addresses that gap by providing a step-by-step methodology to reconstruct the aortic valve and surrounding arterial walls with the precision required for engineering-grade analysis.
Methodology
The authors propose a "reverse modeling" pipeline that utilizes three distinct software environments to handle different stages of the reconstruction:
- Image Acquisition & Segmentation (Mimics): Raw DICOM data from a cardiac CT scan is imported. The authors use thresholding (setting a minimum value of 300) to isolate the valve tissue. They then employ regional growth algorithms to separate the valve from non-connected structures and image editing to fill anatomical "holes" that would otherwise break a simulation.
- Surface Optimization (Geomagic Studio): Because valve leaflets have high curvature, the initial mesh is often too noisy. The authors move the model into Geomagic to perform "reverse engineering" tasks: cleaning point clouds, repairing polygons, and smoothing surfaces to ensure the model meets the strict requirements of structural analysis.
- Leaflet Separation: Using "shear" and "shell" commands, the authors manually isolate the three individual valve leaflets and the arterial wall, saving them as separate STL files.
- Simulation Prep (ANSYS Workbench): The final optimized models are imported into ANSYS, where they are prepared for Finite Element Analysis (FEA).
Data Acquisition
The study relies on a single, high-resolution dataset from a 30-year-old healthy male with no history of cardiac disease. Scanning was performed using a GE Medical Systems Discovery CT750 HD, producing 209 slices with a thickness of 0.625mm and a pixel size of 0.488.
Results and Discussion
The primary result is a successful, reproducible pipeline for creating 3D bioprosthetic models. The process successfully converted 209 2D slices into a set of 3D STL files representing the aortic wall and three distinct valve leaflets.
Key insights from this workflow include:
- The Hybrid Software Approach: No single software tool is sufficient. Medical software (Mimics) is best for biological segmentation, but industrial reverse-engineering software (Geomagic) is necessary to handle the complex curvatures of heart leaflets for engineering purposes.
- Thresholding Optimization: The authors identify a specific threshold value (300) as the "sweet spot" for isolating aortic valve tissue from the surrounding blood and muscle in CT data.
- Mesh Quality: For FEA, the mesh must be optimized (smooth, hole-free, and logically divided) to prevent simulation errors, emphasizing quality over raw data fidelity.
Conclusion
While this study demonstrates a robust workflow, it was limited to a single healthy subject and requires significant manual intervention. Future work should focus on validating this workflow across different pathologies (such as calcified valves) and automating the segmentation steps. This methodology serves as a critical foundation for subsequent dynamic FEA simulations to study how these models behave under physiological blood pressure.
Summary updated