Abstract WP71: Multicenter ALADIN: Automated Large Artery Occlusion Detection in Stroke Imaging Using Artificial Intelligence

Abstract

Introduction: Large vessel occlusion (LVO) may account for up to 40% of all acute ischemic strokes (AIS) and may be associated with worse patient prognosis, being responsible for 60% of dependency and more than 90% of mortality after AIS. Therefore, accurate and rapid identification of LVO and notification of specialists is critical to maximizing the benefit of proven reperfusion therapies. Recent advances in artificial intelligence technology have facilitated the development of automated LVO detection on computed tomography angiography (CTA) imaging. This study evaluated the performance of the Viz LVO algorithm in AIS patients treated at four comprehensive stroke centers. Methods: We performed a multicenter retrospective analysis of 800 CTAs from 750 AIS patients admitted to three comprehensive stroke centers between 2014 and 2018. All studies were analyzed by the latest version of Viz LVO algorithm (version 4.1.2) and the performance was compared to expert neurointerventionalists’ reports. Algorithm run-time and notification time through the Viz platform times were also recorded for each CTA. Results: The Viz LVO algorithm demonstrated 92% sensitivity, 90% specificity in the analysis of 750 CTAs. The mean and maximum run time of the algorithm were 3 minutes and 6 minutes, respectively. Furthermore, mean time to notification was 6 minutes, with a maximum time to notification of 9 minutes. Conclusions: This multicenter retrospective study demonstrates fast and accurate performance of the Viz LVO algorithm in the detection and notification of LVOs on CTAs from three comprehensive stroke centers. Using artificial intelligence, this algorithm may permit early and accurate identification of LVO stroke patients and timely notification to emergency teams, enabling quick decision-making for reperfusion therapies or transfer to specialized centers if needed. Additional studies are required to demonstrate impact on the stroke workflow that result in improved patient outcomes, operational efficiencies, and cost reductions.