Learning Whole Heart Mesh Generation From Patient Images for Computational Simulations.

Area of Science:

• Computational biology
• Medical imaging analysis
• Biomedical engineering

Background:

• Patient-specific cardiac modeling is crucial for predicting heart function.
• Current methods for generating cardiac models from medical images are complex and labor-intensive.

Purpose of the Study:

• To present a fast and automated deep learning method for constructing simulation-suitable heart models from medical images.
• To improve the efficiency and accuracy of creating patient-specific cardiac geometries.

Main Methods:

• A deep learning approach that deforms a whole heart template using handles learned from 3D patient images (CT and MR).
• Mesh generation from 3D medical imaging data.

Main Results:

• The method accurately reconstructs whole heart geometries from both CT and MR data.
• It outperforms previous methods in creating simulation-suitable cardiac meshes.
• Evaluated on time-series CT data, it yields more anatomically and temporally consistent geometries suitable for cardiac flow simulations.

Conclusions:

• This automated deep learning method significantly simplifies and enhances the creation of patient-specific cardiac models.
• The generated models meet the requirements for advanced cardiac flow simulations, advancing personalized cardiac research.