An Artificial Intelligence-Enabled Electrocardiogram to Evaluate Patients With Dyspnea in the Emergency Department

Objective: To evaluate whether an Artificial Intelligence-Enabled Electrocardiogram (AI-ECG) for diastolic function/filling pressure can determine whether dyspnea in emergency department (ED) patients is cardiac in origin. Patients and Methods: We identified 2412 patients aged 18 years or older presented with dyspnea/shortness of breath to the ED who had an ECG performed at the time of evaluation from January 2020 to December 2022. The AI-ECG for determining left ventricular diastolic function to identify the patients with cardiac cause of dyspnea was assessed, using the final diagnosis based on subsequent evaluation. Results: Of the 2412 patients, 966 (40%) were found to have cardiac dyspnea, and the remaining 1446 (60%) were noncardiac. The AI-ECG-estimated diastolic function was divided into 4 groups: 922 (38.2%) were normal, 245 (10.2%) grade 1, 1192 (49.4%) grade 2, and 53 (2.2%) grade 3. The probability of cardiac dyspnea was considerably higher in patients with grade 2 (62.2%±48.5%) and 3 (83%±37.9%) diastolic function compared with normal (14.1%±34.8%) and grade 1 (20.8%±40.7%). The incidence of cardiac dyspnea increased as the probability of increasing filling pressure increased on AI-ECG. Conclusion: Patients often present to the ED with undifferentiated dyspnea. It is important to promptly determine whether the symptoms have cardiac origin. Cardiac dyspnea often reflects elevated left ventricular filling pressures. Artificial intelligence-enhanced 12-lead electrocardiograms can precisely assess diastolic function and filling pressures. Among patients who presented to the ED with dyspnea/shortness of breath, AI-ECG assessing diastolic function strongly distinguished whether the cause was cardiac.