Snakes Isogeometric Analysis (SIGA): Towards accurate and flexible geometrical models of the respiratory airways

Abstract

The construction of accurate patient-specific geometries from medical images is critical to achieving predictive numerical simulations of the respiratory system. However, the generation of surface representations of the airway tree is very challenging due to their complex morphology. In this work, we present a novel framework for creating Non-Uniform Rational B-Splines surface models of the respiratory airways. Our method relies on solving the variational formulation of the Snakes segmentation problem using Isogeometric Analysis (SIGA). We validate the SIGA method by comparing the resulting surface mesh against those delivered by two benchmark surface-fitting methods based on control-point projection. Our results confirm that SIGA outperforms the benchmark methods in creating surface models from computed-tomography images of a normal and a diseased lung. Further, we show that SIGA meshes adapt well to pathological airways with non-convex cross-sections where traditional surface-fitting methods fail. We envision that the geometrical flexibility and accuracy of SIGA will enhance the creation of computational models of the respiratory system under healthy and diseased conditions.