Exploratory comparative analysis of automated morphologic and perfusion features derived from one-stop computed tomography angiography and computed tomography perfusion in large unruptured intracranial aneurysms.
Authors
Affiliations (2)
Affiliations (2)
- Medical College of Wuhan University of Science and Technology, Wuhan, Hubei 430065, China; Department of Radiology, General Hospital of Central Theater Command of People's Liberation Army, Wuhan, Hubei 430060, China.
- Medical College of Wuhan University of Science and Technology, Wuhan, Hubei 430065, China; Department of Radiology, General Hospital of Central Theater Command of People's Liberation Army, Wuhan, Hubei 430060, China. Electronic address: [email protected].
Abstract
To compare rupture-related signals captured by automatically extracted computed tomography angiography (CTA) morphologic features and threshold-based computed tomography perfusion (CTP) metrics in large unruptured intracranial aneurysms (UIAs), and to explore a parsimonious feature combination for rupture status. This retrospective cohort included 60 patients with UIAs who underwent CTA-CTP from 2013 to 2023. Patients without rupture completed 3-year follow-up. Morphologic and perfusion features were extracted automatically. Because only 8 ruptures occurred, analyses were exploratory and parsimonious. Univariable associations were assessed using Firth logistic regression, and variables showing signal were examined in bivariable combinations. Discrimination was evaluated by receiver operating characteristic analysis with bootstrap internal assessment. Among the 60 patients, 8 (13.3%) experienced aneurysm rupture during follow-up. Posterior circulation location, undulation index, and nonsphericity index were associated with increased rupture risk, whereas neck diameter and volume-to-neck area ratio were associated with lower rupture risk. No perfusion parameter showed a significant association with rupture. Among 10 bivariable models, neck diameter plus undulation index showed the best discrimination (AUC, 0.911; 95% CI, 0.834-0.988). In the final model, neck diameter remained inversely associated with rupture, whereas undulation index remained a risk factor. Bootstrap-corrected AUC was 0.888, and the corrected Brier score was 0.098. In this exploratory cohort, CTA-derived morphologic features showed stronger rupture-related signals than the evaluated CTP metrics. Neck diameter plus undulation index showed promising discrimination, but these findings are hypothesis-generating and not yet clinically deployable. External validation in larger cohorts with more rupture events is required.