Automated ASPECTS for multi-modality CT predict infarct extent and outcome in large-vessel occlusion stroke

Abstract

Purpose: This study aimed to use the automated Alberta Stroke Program Early CT Score (ASPECTS) software to assess the value of different CT modalities (non-contrast CT, CT angiography [CTA]-arterial, CTA-venous, and arterial- and venous-phase mismatch-ASPECTS) in predicting the final infarct extent and clinical outcome in large-vessel occlusion stroke.

Methods: This retrospective study included patients with large-vessel occlusion stroke who underwent reperfusion therapy during 2015 to 2019. Correlations between different CT-ASPECTS modalities and follow-up CT-ASPECTS and outcome were determined using Spearman rank correlation coefficient. Receiver operating characteristic curve analysis was used to assess the ability of different CT-ASPECTS modalities to identify patients with good outcomes.

Results: One hundred and thirty-five patients were included. We found almost-perfect correlation between CTA-venous-ASPECTS and follow-up CT-ASPECTS (r = 0.92; 95% CI: 0.89-0.95), better than that in other CT modalities. The 90-day modified Rankin scale (mRS) score substantially correlated with CTA-venous-ASPECTS (r = -0.64; 95% CI: -0.73 to -0.52). The ROC curve analysis showed CTA-venous-ASPECTS had the highest area under the curve (AUC: 0.82; 95% CI: 0.75-0.89; P < 0.001), followed by mismatch-ASPECTS (AUC: 0.75; 95% CI: 0.65-0.85; P < 0.001). When emphasizing the sensitivity for identifying patients with good outcomes, the best cut-off point of mismatch-ASPECTS was -3 with the highest sensitivity (91.30%).

Conclusions: CTA-venous-ASPECTS is a reliable tool to predict the infarct extent and outcome. Furthermore, mismatch-ASPECTS may represent images in different angiographic phases and was sensitive for prognosis prediction.

Keywords: Angiography; Artificial intelligence; Computed tomography; Middle cerebral artery occlusion; Stroke.