Artificial intelligence-based software to support mechanical thrombectomy transfer decision in low-volume primary...

Area of Science:

• Neurology
• Medical Imaging
• Artificial Intelligence

Background:

• Primary stroke centers (PSCs) often lack advanced imaging and expertise for mechanical thrombectomy (MT) decisions in acute ischemic stroke (AIS).
• Efficient interhospital transfer systems are crucial for MT-eligible patients managed initially in PSCs.
• Limited data exist on AI tool performance in low-volume PSC settings.

Purpose of the Study:

• To evaluate the benefit of AI-based imaging software in supporting MT transfer decisions for AIS patients from low-volume PSCs.
• To assess the impact of AI on workflow efficiency and transfer eligibility in a real-world, low-volume setting.

Main Methods:

• Multicenter retrospective analysis of 109 AIS patients transferred for MT from low-volume PSCs.
• Retrospective assessment of standard imaging using Brainomix 360 software for LVO, ischemic changes, and collaterals.
• Simulated transfer decisions by blinded neurologists, comparing AI-assisted workflow times with a comprehensive stroke center (CSC) cohort.

Main Results:

• AI demonstrated 83.5% sensitivity for anterior circulation LVO detection, higher for M1 occlusions.
• 78.9% of patients were deemed eligible for MT, potentially benefiting from reduced workflow times.
• AI potentially avoided futile transfers by identifying 4.6% of patients with extensive ischemic changes as ineligible.
• AI-assisted imaging in CSCs significantly reduced CTA to EVT notification time (11 min vs. 48 min).

Conclusions:

• AI-assisted imaging significantly improves transfer decisions and workflow efficiency in low-volume PSCs, especially without real-time interpretation.
• AI enhances MT eligibility assessment, strengthening regional stroke networks.
• Broader AI adoption can optimize patient pathways for acute ischemic stroke treatment.