Abstract
Background: Acute pulmonary embolism (PE) is a critical condition where the timely and accurate assessment of right ventricular (RV) dysfunction is important for patient management. Given the limited availability of echocardiography in emergency departments (EDs), an artificial intelligence (AI) application that can identify RV dysfunction from electrocardiograms (ECGs) could improve the treatment of acute PE. Methods: This retrospective study analyzed adult acute PE patients in an ED from January 2021 to December 2023. We evaluated a smartphone application which analyzes printed ECGs to generate digital biomarkers for various conditions, including RV dysfunction (QCG-RVDys). The biomarker's performance was compared with that of cardiologists and emergency physicians. Results: Among 116 included patients, 35 (30.2%) were diagnosed with RV dysfunction. The QCG-RVDys score demonstrated significant effectiveness in identifying RV dysfunction, with a receiver operating characteristic-area under the curve (AUC) of 0.895 (95% CI, 0.829-0.960), surpassing traditional biomarkers such as Troponin I (AUC: 0.692, 95% CI: 0.536-0.847) and ProBNP (AUC: 0.655, 95% CI: 0.532-0.778). Binarized based on the Youden Index, QCG-RVDys achieved an AUC of 0.845 (95% CI: 0.778-0.911), with a sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of 91.2% (95% CI: 82.4-100%), 77.8% (95% CI: 69.1-86.4%), 63.3% (95% CI: 54.4-73.9%), and 95.5% (95% CI: 90.8-100%), respectively, significantly outperforming all the expert clinicians, with their AUCs ranging from 0.628 to 0.683. Conclusions: The application demonstrates promise in rapidly assessing RV dysfunction in acute PE patients. Its high NPV could streamline patient management, potentially reducing the reliance on echocardiography in emergency settings.