Q-AortaECG: A Quantum-Inspired Multi-Agent 12-Lead ECG System for Early Triage of Acute Aortic Syndrome.

Acute aortic syndrome (AAS) is a time-critical vascular emergency in which delayed recognition may occur because the presenting symptoms and electrocardiographic phenotypes often overlap with acute coronary syndrome, pulmonary embolism, pericarditis, and other emergency diagnoses. Computed tomographic angiography (CTA) remains the confirmatory standard, but the 12-lead electrocardiogram (ECG) is usually obtained earlier, is universally available at triage, and may carry distributed signatures of malperfusion, ischemia mimicry, autonomic stress, conduction delay, and hemodynamic compromise. We developed Q-AortaECG, a classically implemented quantum-inspired multi-agent ECG system that integrates multiscale waveform representations, specialist ECG agents, density-state uncertainty fusion, a variational-circuit style consensus layer, and a safety-gated multidisciplinary team-style report for early AAS triage. The system was evaluated in three suspected-AAS/CTA-evaluable cohorts comprising 1,251 patients, including 467 with CTA-, surgery-, or autopsy-confirmed AAS. Across 700 locked-test patients, Q-AortaECG achieved an area under the receiver operating characteristic curve (AUROC) of 0.881 (95% confidence interval (CI), 0.852-0.908), area under the precision-recall curve (AUPRC) of 0.743 (95% CI, 0.695-0.790), sensitivity of 0.955 (231/242; 95% CI, 0.920-0.977), specificity of 0.721 (330/458; 95% CI, 0.677-0.761), positive predictive value (PPV) of 0.643, and negative predictive value (NPV) of 0.968 at the prespecified high-sensitivity threshold. Performance remained consistent in external retrospective validation and prospective silent-run evaluation, although AUPRC decreased in the lower-prevalence prospective cohort. Ablation experiments supported the contributions of self-supervised ECG pretraining, agent specialization, density-matrix uncertainty fusion, and differential diagnosis modeling. In a counterbalanced paired reader study, Q-AortaECG assistance improved overall clinician sensitivity from 0.696 to 0.870 while preserving specificity (0.851 ver sus 0.846). In prospective silent-run analysis, counterfactual alerts would have advanced the median time to the correct diagnostic pathway from 187 to 61 minutes among initially unsuspected AAS presentations. These findings support ECG first, calibrated, safety-gated decision support as an early warning scaffold for AAS triage, while preserving CTA and expert imaging interpretation as the diagnostic standard.