Algorithmic Generation of Parameterized Geometric Models of the Aortic Valve and Left Ventricle

Area of Science:

• Cardiovascular modeling
• Computational fluid dynamics
• Biomedical engineering

Background:

• Simulating cardiac valves is complex and computationally intensive.
• Fluid-structure interaction simulations require extensive datasets.
• Machine learning offers a promising alternative for cardiovascular modeling.

Purpose of the Study:

• To develop a method for generating a synthetic database of aortic valve and left ventricular geometries.
• To create diverse geometric variations for training machine learning models.
• To enhance the predictive capabilities of machine learning in cardiovascular research.

Main Methods:

• Generation of 22 left ventricular geometric variations (original, varying thickness, height, shape).
• Verification of synthetic models using patient data for anatomical accuracy and physiological volumes.
• Numerical simulations to evaluate electro-physiological potential and wall shear stress.

Main Results:

• A comprehensive synthetic database of anatomically accurate left ventricular and aortic valve geometries was created.
• Simulations demonstrated the ability to assess electro-physiological potential and wall shear stress.
• The database supports the development of robust machine learning models for cardiovascular applications.

Conclusions:

• The proposed synthetic database facilitates the development of advanced machine learning models for cardiovascular research.
• This approach can lead to improved prediction of patient outcomes after aortic valve surgery.
• The method provides a scalable solution for generating diverse cardiac geometries for simulation and AI training.