Automated Patient-Specific Left Ventricular Simulations for Cardiac Function Evaluation Using Image-Based Computational Fluid Dynamics and Color Flow Imaging

Cardiovascular disease (CVD) is a leading cause of global mortality, necessitating improved diagnostic and treatment strategies. The integration of computational techniques and artificial intelligence holds promise for enhancing cardiovascular health assessment. Computational fluid dynamics (CFD) models offer detailed insights into heart function, particularly for patient-specific (PS) simulations. However, challenges in automation hinder its clinical applicability. We present an automated pipeline for modeling left ventricle (LV) flow patterns based on PS data. The pipeline involves preprocessing, CFD simulation, and postprocessing stages. Deep learning aids in LV and mitral valve segmentation, reducing manual effort. CFD simulations, using ANSYS FLUENT, capture intraventricular flow dynamics. Temporal evolution showcases vortex interactions and propagation, aligning with established patterns. However, limitations arise due to 2D constraints, impacting vortex dynamics. Velocity overestimation, attributed to the 2D assumption, underscores the need for improved segmentation and measurement accuracy. Despite challenges, this study signifies progress toward automated patient-specific 2D simulations for cardiovascular assessment. Further research should explore 2D ventricular wall movement comparisons and refine the pipeline for enhanced accuracy and clinical relevance.