Performance of a fully automated ECG model

Despite guidelines, the standardized assessment of chest pain in the emergency department remains inconsistent, expensive and prone to misinterpretation. A retrospective two-center study approach published in Open Heart now presents a fully automated neural network (“Chest Pain-AI”, CP-AI) that links 12-channel ECG signals to age, gender and biomarker positivity and predicts a 7-day composite rate of severe cardiovascular diagnoses. In an external validation, CP-AI outperformed conventional models and reclassified a relevant proportion as “low risk” – with pre-fixed sensitivity of 98%.