229 Quantitative Imaging Biomarkers for Risk Stratification of Intracranial Aneurysms

Abstract

INTRODUCTION: Aneurysm wall enhancement (AWE) has emerged as a potential imaging biomarker for risk stratification of Intracranial aneurysms (IAs). Herein, the contrast-enhanced MRI (CE-MRI) exhibits markedly higher intensity on the IA sac as compared to the non-enhanced MRI (NE-MRI). However, currently it is assessed subjectively by clinicians which can introduce subjectivity. METHODS: We retrospectively collected 105 IAs with their NE-MRI, CE-MRI and Time-of-flight angiography images. IAs were classified as high or low risk if they had PHASES >5 or ELAPSS >20. We then identified the significantly different RFs and used them to build a logistic regression model for identifying high-risk IAs. We also compared the performance of this model with traditional AWE metrics like CCRatio (Average intensity on the IA sac normalized by the corpus callosum intensity) and CRStalk (Average intensity on the IA sac normalized by the pituitary stalk intensity). RESULTS: We found 35 high-risk IAs in the cohort and identified 183 RFs that were significantly different between them. After eliminating collinear RFs, we used 52 RFs to build our final model. We observed that the RF based model vastly outperformed the conventional metrics. In the testing set, we observed that the LR model had a higher accuracy as compared to conventional metrics. We observed that non-uniformity in pre-contrast MRI, homogeneity in post-contrast MRI and the difference in gray level non-uniformity had the highest odds ratios in the LR model. CONCLUSIONS: RFs from pre- and post-contrast MRI could identify high-risk IAs better than conventional metrics. This model in combination with 3D visualization could aid in better visualization of VWE and better risk stratification of IAs.