(163f) Chemical Engineering Principles Add Clinical Value to Coronary Thermodilution Curves

Heart disease is the leading cause of death in the developed world, and occlusive coronary artery disease (CAD) has long been at the center of cardiovascular research. Although fractional flow reserve (FFR) serves as the gold standard for assessing CAD, there exists a mismatch between intervention and improved long-term patient outcomes in specific demographics. In response, researchers have begun to investigate other comorbidities that may, in tandem with FFR, better inform patient care and improve outcomes. In recent years, coronary thermodilution has been investigated for its ability to assess microvascular dysfunction, a confounding factor in the use of FFR to determine cardiovascular health and risk. However, we hypothesize that current clinical practices do not fully exploit the information available from coronary thermodilution curves and overlook potentially critical insights into coronary health. To address this problem, this research project aimed to investigate if coronary thermodilution curves could be accurately modeled with simple, mathematical models borrowing from concepts in reactor engineering and if the parameters of these models correlated with clinically relevant hemodynamic information. To this end, nine model arteries were created that spanned the design space of 20%, 50%, and 70% diameter stenoses and 30°, 45°, and 60° stenotic ramp angles. Each arterial geometry was meshed using ANSYS Mesher. The flow through each arterial geometry was simulated at 15 different flow rates ranging from 0.5 to 4.0 mL/s in ANSYS Fluent. The resulting dataset consisted of 135 unique coronary thermodilution curves, which were processed using Python. The CFD model was validated against existing literature. This study has successfully deriveda framework to recast coronary thermodilution curves to be mathematically represented as probability density functions, equal to those in both Indicator-Dilution Theory and Residence Time Distribution Theory. We then successfully applied concepts from reactor design/diagnostics in chemical engineering to derive a mathematical expression that models the thermodilution curves in terms of ideal reactor flow patterns. The model was fitted to all 135 model artery simulations with a relative absolute error across the entire cohort of 0.081, 95% CI [0.090, 0.075]. We correlated one of the fitted model parameters to the size of stagnation zones which are known to be clinically significant for their atherogenerative environment and increased risk of clotting. The Pearson correlation coefficients of the model parameter with the relative volume of stagnant blood is 0.99 for laminar flow regimes and 0.89 for turbulent flow regimes. Furthermore, we are working to correlate the model parameters with flow rate and anatomical parameters to formulate the model using only measurable quantities. Lastly, we are constructing a framework to test the model on real patient arteries and find the optimal model, still in terms of ideal reactor flow patterns, using the Sparse Identification of Nonlinear Dynamics (SINDy) algorithm. Thus far, the study has established that coronary thermodilution curves of model coronary arteries can be described with simple mathematical models. The correlation between model parameters and major flow patterns in the CFD simulations establishes that these models can be fit to coronary thermodilution curves to give a coarse-grain view of the hemodynamic environment in the artery of interest. The framework will be further developed using machine learning to better represent the primary hemodynamic patterns in real patient arteries. Our results have shown that the current implementation of coronary thermodilution curves is likely overlooking valuable information about the hemodynamic environment in the artery of interest that is known to affect patient outcomes, such as the presence of low wall shear stress (WSS) stagnant zones.