Statistical Analysis of Left Ventricular Remodeling Following a Myocardial Infarct

Myocardial infarct (MI) is a leading cause of mortality worldwide. It is known to cause left ventricular remodeling, characterized by changes in left ventricle myocardial size, shape, and function. Quantifying these changes is crucial for evaluating the progression of heart diseases following MI. Traditional clinical measurements primarily rely on the volume of the left ventricle (LV) from temporal imaging. In this work, we focus on detecting left ventricular remodeling from a single-phase Cardiac Computed Tomography Angiography (CCTA) using statistical shape analysis. Our pipeline consists of a template registration using implicit neural representations, followed by a statistical shape analysis on the mesh of the LV to classify between healthy individuals and infarct individuals with previous MI. Several methods for dimensionality reduction and classification are evaluated, with Partial Least Squares (PLS) regression achieving the highest classification accuracy of 96%. The PLS components can also be interpreted as directions of healthy vs. infarct LV shapes.