Optimization of three-dimensional modeling for geometric precision and efficiency for healthy and diseased aortas

Abstract

The study purpose is to optimize modeling parameters, specifically segmentation spacing and centerline extraction, to efficiently construct accurate 3D aortic models. Models are constructed by centerline extraction and orthogonal 2D segmentations. We examine the effect of segmentation interval spacing (2, 1, 0.5, 0.25 cm) and orthogonal segmentation and centerline extraction iteration (one, two, three iterations) for constructing models of Healthy, Tortuous, Aneurysmal, and Dissected human thoracic aortas. Aortic arclength, curvature, and cross-sectional axis ratio were computed to compare variations in modeling parameters. Centerline arclength is precisely characterized for all aortas with a single iteration of centerline extraction (≤1% deviation), however, complex anatomies required 1 cm segmentation intervals whereas the Healthy aorta only required 2 cm intervals. Centerline curvature is more sensitive to modeling methods, requiring 1 cm intervals for ≤5% deviation in peak curvature for the three diseased anatomies, and two iterations of segmentation and centerline extraction for the Aneurysmal and Dissected aortas. Accurate lumen cross-sectional characterization required 1 or 0.5 cm segmentation intervals, and two or three segmentation and centerline iterations, with greater refinement needed for more complex geometries. Depending on the geometric characteristic and complexity of anatomy and pathology, different levels of segmentation interval refinement and iterations of segmentation and centerline extraction are required.

Keywords:

• 3D modeling
• image segmentation
• aorta
• arclength
• curvature
• geometry

Acknowledgements

The authors would like to thank all patients for their participation.