This study aimed to describe the relationship between blood and CSF volumes in different compartments on baseline CT after aSAH, assess if they independently predict long-term outcome, and explore their interaction with age. CT scans from patients participating in a prospective multicenter randomized controlled trial of patients with aSAH were segmented for blood and CSF volumes. The primary outcomes were the mRS, and the Subarachnoid Hemorrhage Outcome Tool (SAHOT) at day 28 and 180. Univariate regressions were conducted to identify significant predictors of poor outcomes, followed by principal component analysis to explore correlations between imaging variables and WFNS. A multivariate predictive model was then developed and optimized using stepwise regression. CT scans from 97 patients with a median delay from symptom onset of 271 min (131-547) were analyzed. Univariate analysis showed only WFNS, and total blood volume (TBV) were significant predictors of both short and long-term outcome with WFNS more predictive of mRS and TBV more predictive of SAHOT. Principal component analysis showed strong dependencies between the imaging predictors. Multivariate ordinal regression showed models with WFNS alone were most predictive of day 180 mRS and models with TBV alone were most predictive of SAHOT. TBV was the most significant measured imaging predictor of poor long-term outcome after aSAH. All these imaging predictors are correlated, however, and may have multiple complex interactions necessitating larger datasets to detect if they provide any additional predictive value for long-term outcome.
Keywords: Aneurysm; Image segmentation; Machine learning; Subarachnoid hemorrhage.
© 2024. The Author(s).