Tetralogy of Fallot (TOF) is the most common form of cyanotic congenital heart disease (CHD), occurring in approximately 1 out of every 3,500 births. Innovative surgical strategies have significantly improved early survival, but patients with repaired TOF are at risk of late complications due to residual structural and electromechanical cardiac malformations. Chronic right-ventricular (RV) volume overload due to pulmonary valve incompetency is a common problem in repaired TOF, and pulmonary valve replacement can alleviate regurgitation and normalize RV volume and function. However, left-ventricular (LV) systolic dysfunction can develop in the long-term and its mechanisms are poorly understood. Traditional assessment of LV remodeling and dysfunction for prognostic purposes is limited due to the complex inter-patient variability of LV morphology and the compounding effects of RV dilatation and dysfunction.

In this thesis, we seek to help address the challenges with clinical management of repaired TOF by elucidating patterns of LV remodeling that associate with systolic dysfunction and can provide an indication of further LV deterioration. To this purpose, we use statistical shape analysis to quantify inter- and intra-population variation of LV morphology and systolic function. The aims of this dissertation are to:

1. Characterize the statistical variation of end-diastolic (ED) shape and systolic function in a large cohort of patients with repaired TOF using routinely acquired cardiac magnetic resonance image data;

2. identify novel features of ED morphology that associate with systolic dysfunction using computational models of LV mechanics informed by patterns discovered in patient data;

3. test the hypothesis that novel atlas-based features of remodeling are predictive of functional deterioration of the LV.

Here we demonstrated the first derivation and use of an atlas of LV shape of a large cohort of repaired TOF from multi-center cardiac magnetic resonance image data. Abnormal shape features from an asymptomatic reference population associated with global systolic function in our TOF cohort, but were inadequate for assessment of LV functional deterioration; conversely, we found novel LV shape features from the repaired TOF atlas, rather than standard CMR measures or shape features from the reference atlas, to be associated with the change in global systolic function in a longitudinal study. Mechanistic models of systolic mechanics informed by statistical atlases were capable of systematically testing for shape determinants of LV dysfunction, and were used to generate hypotheses of disease mechanisms that would otherwise need to be measured invasively. A more comprehensive assessment of LV morphology in repaired TOF using statistical shape atlas techniques has the potential to discover hidden patterns of adverse remodeling that contain mechanistic insight and prove valuable for clinical decision-making.