Artificial intelligence for the detection of large vessel occlusion stroke and workflow optimization: a systematic review and meta-analysis

BACKGROUND: To evaluate the impact of artificial intelligence (AI) tools on workflow efficiency, treatment times, and clinical outcomes in patients with large vessel occlusion (LVO) stroke. METHODS: We included observational studies that assessed the implementation of AI-based tools in the management of patients with suspected or confirmed LVO stroke. Eligible studies reported workflow, treatment, safety, or functional outcomes. Data were extracted on thrombolysis rates, CTA-to-arterial puncture time, door-to-puncture time, endovascular treatment (EVT) access, reperfusion success, symptomatic intracerebral hemorrhage (sICH), mortality, and functional independence (mRS 0-2) at 90 days. Random-effects models were applied to calculate pooled mean differences (MD) or risk ratios (RR) with 95% confidence intervals (CIs). RESULTS: Thirteen studies were included. AI implementation significantly reduced workflow delays: CTA-to-puncture time decreased by -40.20 minutes (95% CI -66.41 to -13.99; P<0.01) and door-to-puncture time by -23.17 minutes (95% CI -38.71 to -7.63; P<0.01). Thrombolysis administration showed no difference (RR 0.79; 95% CI 0.57-1.09; P=0.15). Similarly, no significant effects were observed for EVT access (RR 0.97; 95% CI 0.66-1.44; P=0.89), reperfusion success (RR 1.06; 95% CI 1.00-1.12; P=0.06), sICH (RR 0.63; 95% CI 0.30-1.35; P=0.23), or mortality (RR 0.72; 95% CI 0.38-1.37; P=0.32). However, AI-based workflows significantly improved functional independence at 90 days (RR 1.55; 95% CI 1.12-2.15; P=0.008). CONCLUSIONS: AI tools significantly improve treatment times and functional independence in acute LVO stroke without increasing adverse events.