Presentation: 2025 ND EPSCoR Annual conference 

October 21, 2025, NDSU Memorial Union, Fargo, North Dakota

Hemodynamic Modeling in Abdominal Aortic Aneurysm: A Comparative Study of Dimensionality Reduction Techniques

Abdominal aortic aneurysm (AAA) is characterized by localized aortic dilation that can lead to life-threatening rupture, with vortex ring formation during systole within AAA geometries established through experiments, simulations, and in-vivo measurements. However, the complex hemodynamic interactions between vortical structures and arterial walls remain challenging to investigate under physiological conditions. This study presents a computational framework for predicting hemodynamic structures within AAA bulge regions using computational fluid dynamics (CFD) followed by comparative analysis of dimensionality reduction techniques. Our simulations demonstrate the formation of vortex rings propagating along the aorta during systolic phases, with velocity field data serving as input for three approaches: Proper Orthogonal Decomposition (POD) as a traditional linear reduced-order modeling technique, Modal Decomposed Autoencoder based on conventional neural networks (MDAE-NN), and Modal Decomposed Autoencoder based on Kolmogorov-Arnold Networks (MDAE-KAN). The comparative analysis examines the capability of linear versus nonlinear approaches in capturing hemodynamic features and flow pattern extraction. While POD provides interpretable modal structures and computational efficiency, the AI-driven approaches (MDAE-NN and MDAE-KAN) are investigated for their potential to capture additional nonlinear features and complex flow structures within the aneurysm bulge region. This work contributes to understanding the comparative capabilities of linear and nonlinear dimensionality reduction techniques for AAA hemodynamic analysis, providing insights into feature extraction trade-offs and computational efficiency considerations for clinical applications.