An Objective and Repeatable Sac Isolation Technique for Comparing Biomechanical Metrics in Abdominal Aortic Aneurysms

Figure 1. Image segmentation pipeline utilizing an automated segmentation technique based on a U-NET convolutional neural network. After the initial segmentation is accurately performed, the automated pipeline creates a finite element model by converting the point clouds into a mesh, assigning boundary conditions, performing FEA, and post-processing for aneurysm sac isolation.

Figure 2. (A) 3D Scatter plot of an AAA model with the projected 2D boundary in red and a dividing blue line to split the left and right curves. A Fourier series (2) fit for the right projected boundary curve. (B) Subsequent first derivative of functionalized Fourier series. (C) The 2nd derivative of the Fourier series with the local extrema identified. (D) Sac isolated model in the same 2D projected view with local extrema assigning the superior and inferior boundaries on the right side of the aneurysm.

Figure 3. (A) Mesh independence study (data points were sampled) from ~600 k elements to 5.5 million elements. The points were spline fit to show a transition of a converging solution to report sac isolated peak wall stress values. (B) Total computational analysis time required to process all sac isolations (n = 245) plotted against the rotational increment. The increments evaluated were 0.5°, 1°, 7.5°, 15°, 22.5°, 45°, and 90°.

Figure 4. Comparison of MWS (A) and PWS (B) in sac and entire aneurysm regions, shows significant differences in stable versus unstable cases for the sac-isolated region but not for the entire aneurysm. The * symbol denotes a statistical difference (p < 0.05) and not significant is abbreviated as ‘ns’.