Visual versus automated collateral scoring on CT angiography for predicting outcome after endovascular therapy for...

Collateral circulation (CC) is a key determinant of tissue viability and clinical outcome in large-vessel occlusion (LVO) stroke. Whether automated collateral scoring on single-phase CT angiography (sCTA) provides prognostic information comparable to traditional visual assessment remains uncertain. This study aimed to compare the predictive performance of visual versus automated collateral evaluation for functional outcome after endovascular therapy (EVT).

We conducted a prospective multicentre cohort study including consecutive patients with anterior circulation LVO treated with EVT between 2020 and 2022. Collateral status on baseline sCTA was graded visually (Tan and ASITN/SIR scores) and automatically using Brainomix^® software. The primary endpoint was favourable 90-day outcome (modified Rankin Scale [mRS] 0-2). Multivariable logistic regression and ROC analyses compared the discriminative performance of visual and automated collateral assessments.

A total of 332 patients (mean age 72.6 ± 12.7 years; 48.5% female) were analysed. Good collateral status was independently associated with favourable outcome for both visual (Tan aOR 2.41, 95% CI 1.09-5.30; p = 0.029) and automated (aOR 2.45, 95% CI 1.11-5.24; p = 0.027) methods. Predictive performance was nearly identical (0.865 for visual and 0.867 for automated assessment; p = 0.777). Good collaterals were also associated with lower mortality and lower risk of symptomatic intracerebral haemorrhage.

Automated collateral scoring on baseline sCTA achieves equivalent predictive performance to expert visual assessment for functional outcome after EVT. Their added consistency and reproducibility support the integration of automated methods to standardise collateral evaluation across centres, particularly where immediate expert neuroradiological assessment is not available.