Can an AI-powered smartphone app that takes 12-lead ECGs help Japanese patients decide on strategies for coronary intervention?

The integration of artificial intelligence (AI) and 12-lead ECGs is an essential topic for digital cardiology, and evidence is growing. In recent years, there has been a smattering of smartphone apps that capture 12-lead ECGs. The AI-based app "ECG Buddy™" extracts ECG rhythms and digital biomarkers from 12-lead ECGs.

The purpose of this study is to evaluate the effectiveness of this app using 12-lead ECG imaging in determining emergency intervention strategies for coronary artery disease.

Cross-sectional study of patients with suspected ST-elevation myocardial infarction (STEMI) or non-ST-elevation myocardial infarction (NSTEMI)/unstable angina pectoris (UAP) who underwent emergency coronary angiography (eCAG) between 1/2024 and 12/2024 at St. Marianna University Hospital (Kawasaki, Japan). The application included ECG rhythm and 10 digital biomarkers, including "Critical Condition" score and "Acute Coronary Syndrome (ACS)" score (0–100 points each), which can estimate coronary artery occlusion and cardiac ischemic damage were extracted.

158 of 207 patients met the inclusion criteria (exclusions: out-of-hospital cardiac arrest, congenital heart disease, history of cardiac surgery, pacemaker rhythm, ECG not taken 48 hours before CAG, heart rate >= 150/min or < 40/min). At the time of the pre-eCAG diagnosis, there were 58 and 100 STEMI and NSTEMI/UAP patients, respectively. The mean age was 73 ± 13 years and 61% were male. 102 patients underwent ad-hoc percutaneous coronary intervention (PCI) and 31 patients required mechanical circulatory support (MCS). The AUC-ROC (receiver operating characteristic) of the app-calculated "critical" score for MCS use was 0.772 (95% CI: 0.682-0.862, p < 0.001) (Figure 1). The AUC-ROC of the app-calculated "ACS" score for ad-hoc PCI was 0.759 (95% CI: 0.686-0.831, p < 0.001) (Figure 2).

The app-calculated "critical" score and "ACS" score may be useful in determining the use of mechanical circulatory support and distinguishing the need for ad-hoc PCI.Figure 1

and 2