Non-Inferiority Analysis of Electrocardiography Analysis Application vs. Point-of-Care Ultrasound for Screening Left...

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Point-of-care ultrasound (POCUS) is crucial for assessing heart function in emergency departments (EDs).
• POCUS requires specialized equipment and trained personnel, limiting its accessibility.
• There is a need for alternative, accessible methods to evaluate left ventricular (LV) systolic dysfunction.

Purpose of the Study:

• To evaluate the diagnostic accuracy of a mobile application using electrocardiography (ECG) for estimating LV systolic dysfunction (LVEF <40%).
• To compare the mobile app's performance against POCUS and NT-proBNP.
• To test the non-inferiority of the mobile app's LV dysfunction screening to POCUS.

Main Methods:

• 181 adult patients in the ED with POCUS and ECG within 24 hours and echocardiography within 2 weeks were included.
• A mobile app (ECG Buddy) estimated LV ejection fraction (EF) and LV dysfunction risk from ECG.
• App-derived metrics were compared to POCUS-EF, NT-proBNP levels, and echocardiography; a non-inferiority margin of AUC difference of 0.05 was set.

Main Results:

• The mobile app's EF estimation (EF from EB) achieved an AUC of 0.893, outperforming NT-proBNP (AUC 0.722) and demonstrating non-inferiority to POCUS-EF (AUC 0.885).
• The LV-Dysfunction score also showed high accuracy (AUC 0.884), narrowly missing the non-inferiority margin against POCUS.
• Subgroup analysis of sinus rhythm ECGs confirmed the non-inferiority of both app-derived metrics to POCUS-EF.

Conclusions:

• A smartphone application analyzing ECGs can effectively screen for LV dysfunction.
• The app demonstrates diagnostic accuracy comparable to POCUS for identifying LV systolic dysfunction.
• This ECG-based mobile application offers a promising, accessible alternative for cardiac assessment in emergency settings.