B1588
Title: Functional data analysis for diagnosis of coronary artery disease
Authors: Yueyun Zhu - University of Galway (Ireland) [presenting]
Andrew Simpkin - University of Galway (Ireland)
Abstract: Coronary artery disease (CAD) diagnosis plays a pivotal role in guiding treatment decisions and improving patient outcomes. One emerging concept in CAD diagnosis is the recognition of different endotypes, which represent distinct physiological patterns of disease. Functional data analysis (FDA) has emerged as a powerful tool for analyzing such patterns, particularly in angiogram data, which captures the dynamic behaviour of blood vessels over length. Functional principal component analysis (FPCA) is employed on the quantitative flow ratio (QFR) and diameter of 344 vessels, and it is found that the FPC scores can capture the main characteristics of QFR and diameter curves. To predict the vessel endotype, these FPC scores (together with other angiogram indices) are used as predictors in a generalized linear model (GLM) with elastic net regularization, which helps to stabilize parameter estimates and prevent overfitting. The GLMs with elastic net provide accurate prediction results, which enable to quantification of the association between dynamic functional patterns and disease endotypes, and contribute to the advancement of cardiological decision-making.
