2025 AMA Research Challenge – Member Premier Access

October 22, 2025

Virtual only, United States

Would you like to see your presentation here, made available to a global audience of researchers?
Add your own presentation or have us affordably record your next conference.

Background: Neurointerventional procedures require rapid decision-making and continuous device manipulation under high cognitive load. Intraoperative distraction and poor visualization can delay recognition of unintentional device movement, leading to procedural errors or complications. Real-time artificial intelligence (AI) systems may mitigate these risks by enhancing situational awareness and alerting operators to potentially unsafe device behavior. This study presents the first clinical implementation of a real-time AI-based tracking platform in the United States for neurointervention.

Methods: A prospective single-center study was conducted at a tertiary neuroscience hospital from May to July 2025. Twenty-one adult patients undergoing neuroendovascular procedures were enrolled. A real-time AI system was integrated with the biplane fluoroscopy suite to track procedural devices including guidewires, microcatheters, coils, filters, and stents. The system issued audiovisual alerts when a tracked device entered or exited user-defined anatomical regions. Each notification was reviewed and categorized as true positive (TP), false positive (FP), or false negative (FN). Operator actions within 10 seconds of a TP were logged as clinically relevant responses.

Results: The AI system operated successfully in all 21 cases without technical failure or delay to clinical workflow. Across all cases, 245 notifications were recorded: 232 TPs, 13 FPs, and 17 FNs. The system achieved an overall precision of 94.7% and recall of 93.2%. Clinically relevant responses occurred in 25 of 59 evaluated TPs (42.4%), demonstrating real-time impact on operator behavior. Mean notifications per case were 11.7 ± 8.3. Alerts were most often triggered by guidewire advancement or filter migration.

Conclusion: This first-in-U.S. clinical study of real-time AI assistance in neurointervention demonstrated high technical accuracy, smooth integration into clinical workflow, and clinically actionable alerts in nearly half of evaluated events. AI-based intraoperative support tools may reduce cognitive burden, enhance detection of device migration, and improve procedural safety. These systems also show promise as future adjuncts in neurointerventional training, particularly in high-risk or mentor-limited settings.

Next from 2025 AMA Research Challenge – Member Premier Access

Hypoglycemia Enhances Risk and Progression of Diabetic Retinopathy and Nephropathy

Hypoglycemia Enhances Risk and Progression of Diabetic Retinopathy and Nephropathy

2025 AMA Research Challenge – Member Premier Access

Darren Jindal
Darren Jindal

22 October 2025