The impact of model assumptions on results of computational mechanics in abdominal aortic aneurysm

Abstract

In principle, superiority of computational wall stress analyses compared with the maximum diameter criterion for rupture risk evaluation of abdominal aortic aneurysm (AAA) has been demonstrated. The results of finite element analyses should be evaluated carefully, however, because computational strains and stresses are highly dependent on the quality and complexity of each step of AAA simulation. Most clinically active vascular specialists are not familiar with the processes of computational mechanics to evaluate the quality of AAA simulations. For better understanding and to provide insights in computational biomechanics of AAA, the effect of different computational model assumptions on the results of simulation are explained and demonstrated. Four patients with asymptomatic (n = 3) and symptomatic (n = 1) infrarenal AAAs with distinctly different aneurysm morphologies were exemplarily studied. For segmentation and 3-dimensional (3D) reconstruction of AAA and thrombus, 3-mm computed tomography (CT) slices were used, and a high-density hexahedral element-dominated finite element mesh was generated. Subsequent AAAs were simulated on seven different levels, culminating in the most realistic ortho-pressure-finite element analyses simulations, including thrombus, wall calcifications, and prestress state of AAA geometry with nonlinear hyperelastic material and geometric model assumptions. Alterations in displacements due to model assumptions are up to 740% for a specific aneurysm. The average maximum discrepancy among the four morphologies between simple and advanced models is 607%. Differences in peak wall stress between simple and realistic models are up to 210% individually and 170% on average. Differences of model assumptions are more important for simulation results than differences between patient-specific morphologies. Because the biomechanical behavior of AAA is nonlinear in many senses, comparisons between individual morphologies and statistics are only valid when detailed information about preconditions and model assumptions is provided.