Exploring the association between coronary vascular anatomical features and future myocardial infarction through statistical shape modelling

Abstract

Considering the intricate interplay among coronary anatomy and hemodynamics in coronary artery disease (CAD), anatomy-based descriptors have been employed as surrogates of local aberrant hemodynamics and, ultimately, as clinical markers for diagnostic and predictive purposes. However, anatomical descriptors have demonstrated unsatisfactory accuracy, making their further investigation cogent in CAD applications. Therefore, this study investigates the presence of unexplored pathological shape features of left anterior descending (LAD) coronary arteries associated with myocardial infarction (MI) at 5 years using statistical shape modelling. A statistical shape modelling framework combining principal component analysis (PCA) and linear discriminant analysis (LDA), where PCA outputs served as inputs to LDA, was applied to: (i) a cohort of 69 patient-specific LAD geometries, including both future culprit (FCL) and controls, i.e., non-culprit lesions (NCL) of MI reconstructed from 3D quantitative coronary angiography; (ii) the same cohort after artificially removing the main lesion from each LAD model, aiming to isolate the contribution of the atherosclerotic burden beyond the main lesion severity, quantifiable using $\%AS$. Using LDA, the hyperplane with significant discriminant capacity (p < 0.0001) between NCL and FCL was identified for both cohorts. The combination of the statistical shape modelling-based representation accounting for the atherosclerotic burden exclusive of the main lesion severity with $\%AS$, accounting explicitly for the main lesion severity, exhibited notable discrimination capacity for future MI. This study supports the hypothesis that the overall atherosclerotic burden may predispose to future MI and highlights the potential of a statistical shape modelling-based approach for integration into current imaging-driven clinical decision-making.