From medical images to flow computations without user-generated meshes.

Area of Science:

• Biomedical Engineering
• Computational Fluid Dynamics (CFD)
• Medical Imaging

Background:

• Patient-specific biomedical flow computations traditionally require complex, multi-step image processing and meshing workflows.
• Existing methods demand significant user expertise and time for segmentation, surface mesh generation, and volumetric flow mesh creation.
• These sequential, separate software-dependent steps can be a bottleneck in clinical applications.

Purpose of the Study:

• To present an alternative, integrated framework for image-based modeling in biomedical flow computations.
• To streamline the process by performing all steps within a Cartesian domain using implicit surface representations.
• To demonstrate the framework's efficiency and accuracy compared to standard methods.

Main Methods:

• Developed a unified framework operating on a Cartesian domain, embedding patient images as implicit surfaces.
• Eliminated the need for explicit surface and body-fitted flow mesh generation.
• Employed Cartesian mesh pruning, local mesh refinement, and massive parallelization for computational efficiency on distributed memory architectures.

Main Results:

• Successfully computed fluid flow in two distinct 3D intracranial aneurysm reconstructions using the novel framework.
• Demonstrated computational efficiency through Cartesian mesh pruning and parallelization.
• Flow calculations showed comparable results to those obtained via standard, multi-step modeling routes.

Conclusions:

• The proposed integrated framework offers a more efficient and less labor-intensive approach to image-based biomedical flow simulations.
• Embedding images as implicit surfaces on Cartesian grids circumvents complex meshing requirements.
• This method is suitable for parallel computing architectures, paving the way for faster patient-specific simulations.