Novel artificial intelligence model using electrocardiogram for detecting acute myocardial infarction needing...

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Diagnostics

Background:

• Timely diagnosis of acute myocardial infarction (AMI) is crucial for effective myocardial revascularization and improved patient outcomes.
• Current diagnostic methods for AMI may still have limitations in speed and accuracy, necessitating advanced tools.
• Artificial intelligence (AI) offers potential for enhancing the interpretation of electrocardiograms (ECGs) for rapid disease detection.

Purpose of the Study:

• To develop and validate an AI model utilizing ECG data for the accurate detection of AMI requiring revascularization.
• To assess the performance of a transformer-based deep learning model in identifying AMI in large-scale patient cohorts.
• To evaluate the model's efficacy in both internal and external validation datasets for real-world applicability.

Main Methods:

• A transformer-based deep learning model was trained using a large dataset of 723,389 ECGs from 300,627 patients.
• The model underwent self-supervised learning on approximately one million unlabeled ECGs before fine-tuning for AMI detection.
• Performance was evaluated using internal testing and external validation on 261,429 ECGs from 259,454 patients at an independent center.

Main Results:

• The AI model demonstrated a significant improvement in AMI detection, with the area under the receiver operating characteristic curve (AUROC) increasing from 0.910 to 0.968 in the external validation set.
• Specific AUROCs for ST-elevation myocardial infarction and non-ST-elevation myocardial infarction were 0.991 and 0.947, respectively, in the external validation set.
• The integration of self-supervised learning enhanced the model's diagnostic performance for AMI.

Conclusions:

• The developed ECG-based AI model shows promise for the timely and accurate identification of patients with AMI who require revascularization.
• This AI tool has the potential to expedite clinical decision-making and improve management strategies for acute myocardial infarction.
• Further integration of such AI models into clinical workflows could enhance efficiency in cardiac care.