Abstract
Background/Objectives: The clinical impact of automated large vessel occlusion (LVO) detection tools using non-contrast CT (NCCT) is still unknown. We evaluated whether the implementation of Heuron ELVO, an artificial intelligence (AI)-driven software for triage and notification of LVO stroke using NCCT, can reduce treatment times and improve clinical outcomes in a real-world setting. Methods: We compared patients with LVO stroke before (pre-AI cohort, 84 patients) and after (post-AI cohort, 48 patients) the implementation of Heuron ELVO at a comprehensive stroke center. Primary outcomes included time-to-treatment initiation, including door-to-IV tPA and door-to-endovascular thrombectomy (EVT) times. Secondary outcomes measured changes in the National Institute of Health Stroke Scale (NIHSS) and modified Rankin Scale (mRS) scores. Statistical analyses involved multiple linear regression to adjust for confounders. Results: The implementation of Heuron ELVO significantly reduced the door-to-EVT time (30.2 min, 95% CI, -56. to -4.3), CT-to-neurologist examination time (16.4 min, 95% CI, -27.6 to -5.3), and CT-to-EVT time (29.4 min, 95% CI, -53.6 to -5.0). There was no statistical difference in the door-to-IV tPA time (8.9 min). The post-AI cohort exhibited a greater improvement in the NIHSS score compared to the pre-AI cohort, with a reduction of 4.3 points. While the post-AI cohort demonstrated a higher proportion of good outcomes (mRS 0-1, 26% vs. 40%) at the 3-month follow-up, there was no statistical significance. Conclusions: The implementation of Heuron ELVO demonstrated substantial improvements in the timeliness of stroke interventions and patient outcomes. These findings underscore the potential of AI-driven NCCT analysis in enhancing acute stroke workflows and expediting treatments in real-world settings.