Noncontrast CT AI more sensitive than expert readers for large vessel occlusion in Bayesian meta-analysis

Noncontrast CT-based unimodal imaging AI detected anterior circulation large vessel occlusion with sensitivity 0.78 (95% credible interval 0.68-0.86) and specificity 0.88, versus expert readers' sensitivity 0.62 and similar specificity; low-certainty retrospective evidence.

• Bayesian diagnostic test accuracy network meta-analysis of 10 retrospective validation studies (11 datasets, 28 node arms, 3632 patients) compared expert readers, nonexpert readers, unimodal imaging AI, and clinically informed multimodal AI.
• Unimodal AI sensitivity exceeded expert readers by 0.16 (95% credible interval 0.03-0.29) and nonexpert readers by 0.18 (95% credible interval 0.03-0.32), with no clear specificity separation.
• Clinically informed multimodal AI had sensitivity 0.81 (95% credible interval 0.64-0.92) and specificity 0.92 (95% credible interval 0.80-0.98), but this node was connected to human readers only through indirect evidence.

Evidence grade: Low (GRADE)

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.