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

This paper proposes an approach to creating a synthetic database of geometric models based on multislice computed tomography (MSCT) data. Existing algorithms of MSCT data processing allow to create three-dimensional models by segmentation, but it is impractical to scale them to a large database. Therefore, we propose to use synthetic geometries that allow the creation of a variety of aortic valve and left ventricular models with different shapes and sizes. Specifically, we created 22 variants of left ventricular geometries, including one original model, seven models with different heights, seven with different wall thicknesses, and seven with different shapes. These models were validated on real patient data to ensure anatomical accuracy and physiologically normal fluid volumes. This comprehensive synthetic database is intended to enhance the predictive power of ML models in cardiovascular research and ultimately improve the outcomes of aortic valve replacement surgery in patients.