This article was originally published here

International Heart Journal. November 18, 2021: S0167-5273(21)01854-4. doi: 10.1016/j.ijcard.2021.11.039. Online before printing.

Background: Although the door-to-balloon time for ST-segment elevation myocardial infarction (STEMI) is usually less than 90 minutes, the time to door from symptoms is still very long, 2.5 hours, at least in part due to a delay in diagnosis.

Goal: Develop and validate a machine learning guided algorithm that uses a single-lead electrocardiogram (ECG) for STEMI detection to speed up diagnosis.

Method: The data comes from the Latin American Telemedicine Infarction Network (LATIN), a population-based acute myocardial infarction (AMI) program that provides care for patients in Brazil, Colombia, Mexico, and Argentina through telemedicine.

Sample: The first data set consists of 8511 ECGs, which are used in various machine learning experiments to test our deep learning methods for STEMI diagnosis. The second data set contains 2542 confirmed STEMI diagnostic ECG records, including specific ischemic heart wall information (anterior, inferior, and side walls), derived from the previous data set, used to test the STEMI localization model. Preprocessing: The QRS complex is detected by the wavelet system, and each ECG record is divided into individual heartbeats. The fixed window is 0.4 seconds to the left and 0.9 seconds to the right. Training and testing: The two models used 90% and 10% of the total data set, respectively.

Classification: Two one-dimensional convolutional neural networks are implemented, the first model (STEMI/Not-STEMI) considers two categories, and the second model (front/down/side) considers three categories, each category Corresponds to the affected heart wall. These individual probabilities are aggregated to generate the final label for each model.

Results: The single-lead ECG strategy can provide 90.5% accuracy for STEMI detection using lead V2, which also produces the best overall results in a single lead. The STEMI positioning model provides promising results for the front wall and bottom wall STEMI, but it is still not ideal for lateral STEMI.

Conclusion: AI-enhanced single-lead ECG is a promising screening tool. This technology provides an autonomous and accurate STEMI diagnostic alternative that can be integrated into wearable devices, which may provide patients with a reliable means to seek treatment early and provide the potential to improve STEMI results in the long run.