Validation of a novel automated Alberta Stroke Program Early Computed Tomography Score (ASPECTS) software:...

The Alberta Stroke Program Early Computed Tomography Score (ASPECTS) on noncontrast computed tomography (NCCT) is widely used to assess early ischaemic changes, but its consistency is limited by inter-rater variability. We aimed to validate a novel automated software (JLK ASPECTS) by assessing its scoring agreement against expert readers and a validated commercial software (RAPID ASPECTS), and by correlating its scores with diffusion-weighted imaging (DWI) infarct volumes.

We retrospectively included patients with acute ischaemic stroke (AIS) admitted to three comprehensive stroke centres between June 2022 and December 2023. For agreement analysis, we included 74 patients with anterior circulation large vessel occlusion (LVO) who achieved successful endovascular thrombectomy (EVT) and underwent follow-up DWI within 24 hours, allowing comparison between pretreatment NCCT ASPECTS and posttreatment DWI ASPECTS. For correlation analysis, 306 AIS patients who underwent DWI within 30 minutes of NCCT were included, regardless of EVT. Agreement was assessed using weighted kappa and intraclass correlation coefficient (ICC), and infarct volume correlation was analysed using Spearman's coefficient and receiver operating characteristics (ROC) analysis.

JLK ASPECTS showed good agreement with both RAPID ASPECTS and experts (ICC range: 0.692-0.797), with higher concordance between software than between experts (0.797 vs 0.631; P = 0.039), particularly within 3 hours of onset (0.751 vs 0.434; P = 0.018). Automated ASPECTS correlated significantly with DWI infarct volume (P = -0.532, P < 0.001), effectively identifying infarcts ≥70 mL (cutoff = 5; sensitivity 88.2%, specificity 95.5%).

We validated the newly developed automated ASPECTS software, demonstrating higher inter-rater agreement between software than experts in early time windows. Furthermore, automated ASPECTS correlated with infarct volume on DWI.