ABSTRACT NUMBER: ESOC2026A1407 THE EFFECT OF ARTIFICIAL INTELLIGENCE TOOLS ON ACUTE STROKE PATIENTS TREATMENT PATHWAYS: A SIMULATION STUDY

Abstract

Abstract Background and aims Artificial intelligence (AI) for CT angiography (CTA) interpretation may influence the detection of large vessel occlusion (LVO), the allocation of patients to intravenous thrombolysis (IVT) and/or endovascular thrombectomy (EVT), and in-hospital treatment delays. The aim was to evaluate how incorporating an AI tool into acute ischemic stroke workflows affects patient routing and door-to-groin times in a comprehensive stroke center (CSC), when used by clinicians with differing levels of experience (resident, general radiologist, neuroradiologist). Methods Monte Carlo simulation modelling was used to establish the patient pathways of a CSC and to investigate the effect of an AI tool in CT angiography (CTA) scan interpretation time and LVO detection rates. 783 patients of UMCG diagnosed with acute ischemic stroke of which 35 were treated with both IVT and EVT, and 42 with EVT only. The main outcome measures were the door-to-groin times and percentage of patients receiving reperfusion treatments. Results AI reduced door-to-groin time by 3 minutes for residents, 0.6 minutes for general radiologists, and 0.1 minutes for neuroradiologists. LVO detection improved across all clinician groups. Among residents, EVT-only cases increased by 9% and IVT+EVT cases by 4.8%. For general radiologists, EVT-only and IVT+EVT cases increased by 7.8% and 6.9%, respectively. For neuroradiologists, increases were 2.8% and 4.4% Conclusions AI-based LVO detection in acute ischemic stroke patients has a higher effect in the decision making of resident doctors, indicating AI tools can be used as an extra assistance when the level of clinicians’ experience is low. Conflict of interest Anastasis Alexopoulos: nothing to disclose, Reinoud Bokkers reports funding from The Netherlands Organization for Health Research and Development, the Dutch Ministry of Economic Affairs and Climate Policy and an unrestricted grant from Siemens Healthineers. Maarten Lahr: Nothing to disclose, Durk-Jouke van der Zee: Nothing to disclose, Maarten Uyteenboogaart: Nothing to disclose, Saloua Akoudad: Nothing to disclose. Mathias Prokop: Nothing to disclose

Ähnliche Arbeiten