Abstract 13949: Voronoi Characteristics for the Improvement of Middle Cerebral Aneurysm Rupture Differentiation

Introduction: Surgical intervention for detected intracranial aneurysms (IA) cary risks to patients. Analysis of IA geometry may help clinicians differentiate IAs at high rupture risk versus ones likely to remain stable. Yet traditional geometric indices lack sufficient accuracy, stressing a need for novel indices. Hypothesis: Adding novel Voronoi diagram-based indices (see Fig. 1) will improve IA differentiation. Methods: 3D computational models were taken from 47 subjects with one medium-sized (4-10mm) middle cerebral artery (MCA) IA via 3D digital subtraction angiography data. 7 geometric indices for IAs ( e.g. size, aspect ratio) were measured, alongside Voronoi analysis. In terms of the Voronoi diagram/sphere, each IA constitutes one or more spherical cores (large Inscribed sphere(s)) and many small protrusions. 2 variant curves denoting the inscribed spheres to small protrusion size’s relationships can be seen in Fig 1. Support vector machine with cross-validation (100 iterations) assessed Voronoi diagram-based curve’s strength for IA differentiation. The parsimonious model was determined as 3 geometric indices (volume, identified bulb, vessel diameter) and Voronoi characteristics. Results: Conventional IA geometric indices for ruptured vs unruptured differentiation showed limitations: 0.80 AUROC, 0.76 total accuracy. The addition of quantified Voronoi characteristics via both Voronoi curves augmented model strength for differentiation: 0.87 AUROC, 0.86 total accuracy. Conclusions: Preliminary analysis suggests including Voronoi characteristics for IA analyses improves rupture status prediction accuracy for medium-sized MCA aneurysms, which are more difficult to manage. Since Voronoi characteristics can be easily done in the clinical workflow, such an analysis may enrich the differentiation of IAs.