Human-Specific Prediction, Training, and Visualization Tools for the Tricuspid Valve from Existing Data

点击翻译按钮获取中文摘要

ABSTRACT. Roughly 1.6 million Americans suffer from significant tricuspid valve regurgitation. The standard treatment for tricuspid valve regurgitation, surgery, is shockingly ineffective with recurrent regurgitation in up to 30% of patients within a few years. We posit that these poor outcomes are due to our very limited basic understanding of normal valve function and the valve's underlying, fundamental disease and failure mechanisms. Models for prediction, training, and visualization have been valuable in many other areas of surgery and could be critical tools toward overcoming our current knowledge gaps about the tricuspid valve. However, the few existing models of the tricuspid valve are simplified, population-averaged, not human-specific, or all of the above. To fill this clear gap in tools, our goal is to do a secondary analysis on previous data collected on isolated, beating, human hearts. Based on these data, we will create fully-detailed, human-specific tricuspid valve models and curate these models as an openly available portfolio of prediction, training, and visualization tools. To this end, we will first create detailed finite element models of the human tricuspid valve for predictive simulations. Specifically, we will establish and validate high-fidelity finite element models of eight subjects whose leaflet and annular geometries, leaflet and chordae mechanical/microstructural properties, and boundary conditions are all subject-specific. We will curate these models to be compatible with the widely popular, NIH-funded, and open- source finite element software FEBio. Because FEBio is a research code, we will make our models also compatible with one of the most widely-used commercial finite element software, Abaqus. Second, we will create eight responsive models of the human tricuspid valve for real-time training simulations. To this end, we will introduce low-fidelity, but high-efficiency computational models of the tricuspid valve that use linear material laws, dimensional reduction, simplified solution strategies, and GPU acceleration for speed-up. These models will be of lower fidelity than our finite element models, but fast enough to be integrated into a popular and open software suite for real-time surgical simulations: SOFA. Finally, we will also create eight augmented reality models of the human tricuspid valve for anatomic visualization. We will make these dynamic and labelled models openly accessible through an augmented reality web interface that hosts these models as a smart-phone accessible resource for clinicians, patients, and engineers. Once we have successfully concluded this work, we will have created tools for tricuspid valve predictive modeling, training, and visualization. All tools will be made openly available through the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. Together, these accessible outcomes will overcome our current tool gap for the tricuspid valve and address a highly prevalent, unmet clinical need.