Detection of large vessel occlusion using artificial intelligence tools: A systematic review and meta-analysis

Area of Science:

• Neuroscience
• Medical Imaging
• Artificial Intelligence

Background:

• Large vessel occlusion (LVO) constitutes a significant portion of ischemic strokes.
• Artificial intelligence (AI) demonstrates high accuracy in identifying LVOs on computed tomography angiograms (CTA).

Purpose of the Study:

• To evaluate if AI-adjudicated CTA enhances workflow efficiency and clinical outcomes in acute ischemic stroke patients with LVO.

Main Methods:

• A systematic literature search was conducted across PubMed, Embase, and Web of Science.
• Studies comparing AI-assisted radiological assessment with standard assessment for acute LVO strokes were included.
• Data were pooled using random-effects models to calculate mean differences and odds ratios with 95% confidence intervals.

Main Results:

• Nine studies involving 1,270 patients were analyzed, with 52.8% receiving AI-assisted assessment.
• AI significantly reduced door-to-groin time by 20.55 minutes and CTA-to-reperfusion time by 14.99 minutes.
• No significant differences were observed in functional independence (modified Rankin scale 0-2) or mortality rates between AI-assisted and standard assessment groups.

Conclusions:

• AI software integration for LVO detection in acute ischemic stroke improves workflow efficiency and reduces treatment times.
• Despite workflow improvements, AI did not demonstrate a significant impact on clinical outcomes such as functional independence or mortality.