Not on HAIR - double check

A single-center prospective study in Singapore evaluated the LUNIT INSIGHT CXR Triage (Lunit) as a pre-triage tool for chest X-rays. The AI categorized 37,900 X-rays into normal, non-urgent, and urgent groups, prioritizing urgent cases (e.g., pneumothorax, pleural effusion) for faster attention while deferring non-urgent findings (e.g., cardiomegaly, fibrosis). The AI demonstrated high diagnostic accuracy, with sensitivity and specificity of 89% and 93% for normal, 93% and 91% for non-urgent, and 82% and 99% for urgent cases, achieving an overall AUC of 0.91.

Although radiologists were blinded to the AI’s classifications, the study projected a theoretical reduction in turnaround time by 818.9 minutes if the AI triage system were integrated into clinical workflows. This improvement would result from automated prioritization, where cases flagged as urgent by the AI could be fast-tracked for immediate review, enhancing efficiency without requiring manual triage. Consistent accuracy across diverse age, gender, and ethnic groups underscored the AI’s generalizability within a multi-ethnic population.

The study concluded that LUNIT INSIGHT CXR Triage could support faster, accurate prioritization of chest X-rays in high-demand settings and recommends further multicenter studies to confirm its broader applicability.