Evolution of single-lead ECG for STEMI detection using a deep learning approach

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Diagnostics

Background:

• Symptom-to-door times for ST-Elevation Myocardial Infarction (STEMI) are prolonged due to diagnostic delays.
• Current diagnostic pathways for STEMI can be time-consuming, impacting patient outcomes.

Purpose of the Study:

• To develop and validate a machine learning (ML)-guided algorithm for rapid STEMI detection using single-lead electrocardiograms (ECGs).
• To enhance the speed and accuracy of STEMI diagnosis through an AI-powered approach.

Main Methods:

• Utilized a large dataset of 8,511 ECGs from the Latin America Telemedicine Infarct Network (LATIN) for model training and validation.
• Implemented 1-D convolutional neural networks for STEMI detection (STEMI/Not-STEMI) and localization (anterior, inferior, lateral walls).
• Preprocessed ECG data by detecting QRS complexes and segmenting individual heartbeats for analysis.

Main Results:

• The AI-guided single-lead ECG strategy achieved 90.5% accuracy for STEMI detection using Lead V2.
• The STEMI localization model showed promising results for anterior and inferior wall STEMIs, with areas for improvement in lateral wall detection.

Conclusions:

• AI-enhanced single-lead ECGs represent a viable and accurate screening tool for STEMI.
• This technology can be integrated into wearable devices, offering a potential pathway for earlier patient treatment and improved myocardial infarction outcomes.