Topology-preserving Scan-based Immersed Isogeometric Analysis

TL;DR

This paper introduces a topology-preserving segmentation method using refined B-splines to improve the construction of smooth, topologically accurate geometric domains from scan data for isogeometric analysis.

Contribution

It presents a novel THB-spline-based segmentation approach with a topological anomaly detection algorithm to maintain topology during domain smoothing.

Findings

Effective topological anomaly detection using Euler characteristic

Improved domain smoothness without topological errors

Validated on various test cases in immersed IGA

Abstract

To exploit the advantageous properties of isogeometric analysis (IGA) in a scan-based setting, it is important to extract a smooth geometric domain from the scan data (e.g., voxel data). IGA-suitable domains can be constructed by convoluting the grayscale data using B-splines. A negative side-effect of this convolution technique is, however, that it can induce topological changes in the process of smoothing when features with a size similar to that of the voxels are encountered. This manuscript presents an enhanced B-spline-based segmentation procedure using a refinement strategy based on truncated hierarchical (TH)B-splines. A Fourier analysis is presented to explain the effectiveness of local grayscale function refinement in repairing topological anomalies. A moving-window-based topological anomaly detection algorithm is proposed to identify regions in which the grayscale function refinements must be performed. The criterion to identify topological anomalies is based on the Euler characteristic, giving it the capability to distinguish between topological and shape changes. The proposed topology-preserving THB-spline image segmentation strategy is studied using a range of test cases. These tests pertain to both the segmentation procedure itself, and its application in an immersed IGA setting.