BƯỚC ĐẦU XÂY DỰNG MÔ HÌNH TRÍ TUỆ NHÂN TẠO HỖ TRỢ CHẨN ĐOÁN NHỒI MÁU CƠ TIM CẤP TẠI BỆNH VIỆN NGUYỄN TRÃI

Abstract

Background: Acute myocardial infarction (AMI) is one of the most serious cardiovascular conditions and a leading cause of death worldwide. Early diagnosis of AMI remains a significant challenge. Artificial intelligence (AI) is a promising tool for assisting in the early detection of acute myocardial infarction. Objectives: To collect a complete dataset of cases with and without a diagnosis of acute myocardial infarction (AMI) and label them accurately. To develop an AI model capable of alerting for acute myocardial infarction. Methods: A retrospective review and labeling of 6,226 records, both with and without AMI, confirmed by a cardiology specialist as the final diagnosis. The LightGBM machine learning method was used to build the AI model. Results: Among the 6,226 labeled cases, there were 482 acute myocardial infarction cases, with 384 STEMI cases and 98 NSTEMI cases. The AI model built with and without using troponin T, sample 1, achieved the following results: Recall=96%, Precision=71%, F-score=84%, AUC=0.99, and Recall=97%, Precision=77%, F-score=86%, AUC=0.99, respectively. Conclusions: The AI model can effectively alert for acute myocardial infarction, and further applied research is needed to evaluate its effectiveness and safety in real-world settings.