Abstract
BACKGROUND: The centroid radii model (CRM)-defined as the distance from aneurysm surface points to the centroid-has shown promise in distinguishing ruptured from unruptured intracranial aneurysms by capturing morphological characteristics. We propose a novel application of radiomics for analyzing CRM texture and patterns, extending radiomics beyond medical imaging-based analysis. METHODS: Three-dimensional rotational angiographic volumes from 187 aneurysms (49 ruptured) were analyzed. Aneurysm surfaces were segmented and converted to uniform triangular meshes. Established size and shape metrics (size, height, aspect ratio, height/width, size ratio, and nonsphericity) and CRM values were computed. CRM data were projected onto a unit sphere and mapped to grayscale images for radiomics analysis. Univariate and multivariate analyses were used to evaluate the established features and 93 radiomics features for the accuracy of rupture status discrimination. RESULTS: Ruptured aneurysms exhibited greater CRM texture complexity and heterogeneity, with higher entropy, variance, contrast, and gray-level nonuniformity. These aneurysms had localized high CRM intensities and widespread low CRM regions, as well as reduced pattern uniformity. All established morphological features were significantly elevated in ruptured aneurysms. Multivariate regression using radiomics features resulted in an area under the curve of 0.86 (specificity 0.81, sensitivity 0.78), compared with conventional features (area under the curve 0.82, specificity 0.75, sensitivity 0.81). CONCLUSION: Radiomics-based histogram and texture analysis of surface CRM offers strong rupture status discrimination power, which compares favorably with established size and shape features. This novel use of radiomics on surface-based features provides additional insight into the characteristics of ruptured aneurysms and may have potential utility in risk stratification.