Related Experiment Video
Updated: May 15, 2026
Optimized Management of Endovascular Treatment for Acute Ischemic Stroke
Published on: January 18, 2018
Real-Time Artificial Intelligence Assistance in Neuroendovascular Therapy: Comparative Analysis of Elective and
Ryo Aiura1, Yoshikazu Matsuda1, Hiroki Nagatsuka1
1Department of Neurosurgery, Showa Medical University Hospital, Shinagawa-ku , Tokyo , Japan.
Background And Objectives:
Artificial intelligence (AI) has emerged as an adjunct in neuroendovascular interventions; however, its clinical utility remains insufficiently characterized. This study aimed to delineate the clinical impact of a real-time AI guidance system during neuroendovascular therapy and to compare elective and emergency interventions, using device repositioning as a quantitative indicator of procedural guidance.
Methods:
A retrospective cohort of 130 neuroendovascular procedures performed utilizing an AI system (Neuro-Vascular Assist; iMed Technologies) was analyzed. The system generated automated notifications when the guidewire or guiding catheter (GC) exited the fluoroscopic field. Notifications were defined as "clinically useful" if the device exited biplane views and was repositioned within 10 seconds. The proportion of such events constituted the clinically useful notification rate. Univariate analyses compared elective and emergency procedures, and multivariate logistic regression identified independent correlates of emergency procedures.
Results:
Emergency procedures were more frequently performed by noncertified operators ( P = .002) with fewer years of neurosurgical experience ( P = .012) under local anesthesia ( P = .010) and exhibited shorter operative durations ( P = .001). Both the absolute number ( P = .021) and rate ( P = .027) of clinically useful GC notifications were significantly higher in the emergency group, with the rate being twice that of the elective group (26% vs 13%). Multivariate analysis revealed that limited neurosurgical experience (odds ratio: 0.88; 95% CI: 0.77-1.00, P = .039) and elevated GC notification rates (odds ratio: 1.02; 95% CI: 1.00-1.03, P = .018) were associated with emergency interventions. No adverse events were considered to be attributable to the AI system.
Conclusion:
Real-time AI guidance conferred greater procedural assistance during emergency neuroendovascular procedures, typically performed by less-experienced operators. These data suggest that AI integration may augment procedural safety in high-acuity cerebrovascular settings. Prospective studies are warranted to evaluate its impact on patient-centered and workflow-efficiency outcomes.
