Performance of Radiomics derived morphological features for prediction of aneurysm rupture status

• DOI: 10.1136/neurintsurg-2020-016808
• Publication Date: 2020-11-06T22:24:46Z
• Authors: Calvin Gerald Ludwig; Alexandra Lauric; Justin A Malek; Ryan Mulligan; Adel M Malek
• Keywords: Sphericity; Univariate; Univariate analysis; Flatness (cosmology)

Background Morphological differences between ruptured and unruptured cerebral aneurysms represent a focus of neuroimaging researchfor understanding the mechanisms aneurysmal rupture. We evaluated performance Radiomics derived morphological features, recently proposed for rupture status classification, against automatically measured shape size features previously established in literature. Methods 353 (123 ruptured) from three-dimensional rotational catheter angiography (3DRA) datasets were analyzed. Based on literature review, 13 descriptors extracted per aneurysm, prediction using univariate multivariate statistical analysis, yielding an area under curve (AUC) metric receiver operating characteristic. Results Validation overlapping size/volume both methods highly correlated (p<0.0001, R 2 =0.99). Univariate analysis selected AspectRatio AUC=0.75), Non-sphericity Index Height/Width AUC=0.73), SizeRatio AUC=0.73) as best among descriptors, Elongation AUC=0.71) Flatness AUC=0.72) features. with ( =0.52), whereas Ellipticity =0.54). Sphericity Undulation =0.65). Best performers, Flatness, =0.75). In (Height/Width, SizeRatio, Index; AUC=0.79) outperformed (Elongation, Maximum3Ddiameter; AUC=0.75). Conclusion Although introduced aneurysm evaluation has advantage being efficient operator independent methodology, it currently offers inferior discriminant compared descriptors. Future research is needed to extend current feature set better capture information.