AI Decision Support for Emergency Triage of Large Vessel Occlusion Using Noncontrast Computed Tomography: Systematic Review and Bayesian Diagnostic Test Accuracy Network Meta-Analysis.
Huiqing Zhou, Li Liu, Bolun Fu et al.
✦ AI-curated · Sources linked
In 30 seconds
This systematic review and Bayesian network meta-analysis evaluated the diagnostic accuracy of noncontrast computed tomography (NCCT)-based AI for detecting anterior circulation large vessel occlusion (LVO) compared to human readers. In 11 validation datasets involving 3632 patients, unimodal imaging AI demonstrated a sensitivity of 0.78 and specificity of 0.88, outperforming expert and nonexpert readers in sensitivity by 0.16 and 0.18, respectively.
Key findings
- Unimodal imaging AI had a sensitivity of 0.78 and specificity of 0.88.
- Expert readers had a sensitivity of 0.62, while nonexpert readers had 0.60.
- Unimodal imaging AI showed higher sensitivity than expert and nonexpert readers by 0.16 and 0.18, respectively.
- Clinically informed multimodal AI had a sensitivity of 0.81 and specificity of 0.92.
Why it matters
The findings suggest that NCCT-based AI could enhance early detection of LVO in emergency settings, potentially improving triage and treatment pathways. This may lead to faster decision-making and better patient outcomes in stroke management.