UC San Diego

Tetralogy of Fallot (TOF) is the most common cyanotic congenital heart disease accounting for about 10% of all congenital cardiac malformations. Due to improvements in surgical technique, individuals with repaired TOF (rTOF) are surviving into adulthood but face residual pulmonary regurgitation that can lead to adverse ventricular remodeling and, ultimately, heart failure. Cardiovascular magnetic resonance (CMR) imaging is the gold standard for evaluation in patients with rTOF, but the wealth of information available in CMR images is under-utilized. We sought to use computational tools to extract atlas-based biomarkers of regional biventricular shape and mechanics from standard of care CMR images to better aid clinical management of rTOF. Specifically, the aims of this dissertation were to 1) rapidly build biventricular shape models from CMR data using machine learning; 2) formulate shape atlases from these models to identify clinically meaningful shape variation within the population; and 3) construct biventricular biomechanics models to quantify shape marker relationships with differences in myocardial mechanical properties. Herein we demonstrate a fully automated, end-to-end pipeline that can robustly create biventricular shape models for the challenging anatomies present in rTOF. We also discovered specific markers of biventricular shape that are better discriminators of clinical prognosis than standard imaging indices. These markers of biventricular shape were also partial determinants of systolic function and may also be surrogate measures for altered myocardial contractility. Overall, patients with rTOF may benefit from routine atlas-based analyses of biventricular shape and mechanics that can supplement clinical decision making and provide insight into mechanisms underlying myocardial pathophysiology.