Construction and Analysis of a Risk Prediction Model for Acute Myocardial Infarction Based on Spatiotemporal Heterogeneous Data

Acute myocardial infarction (AMI), as the leading cause of death among cardiovascular diseases, has its diagnosis and treatment efficiency directly affecting survival. Although the current diagnosis and treatment system has significantly improved in-hospital outcomes, delays in seeking medical care due to patients' insufficient awareness and out-of-hospital deaths are common, representing the biggest bottleneck in improving diagnostic and treatment capabilities. This study takes intelligent-assisted diagnosis of AMI as the entry point and proposes a technical approach that combines a deep learning algorithm based on 12-lead electrocardiograms with wearable monitoring devices. By utilizing morphological feature extraction and deep learning models, it aims to achieve early identification and warning of AMI. The study plans to build a multi-center AMI long-term follow-up cohort covering the Beijing area based on spatiotemporal heterogeneous data. By integrating and forming a precise high-risk cohort of 3,000 acute myocardial infarction cases, it seeks to construct an AMI risk prediction model that combines deep learning with a retrieval-augmented generative expert system, breaking through bottlenecks in ECG recognition and temporal prediction, enhancing model generalization and transferability. Ultimately, it will support the application of wearable devices, shorten pre-hospital delays, achieve early warning and precise diagnosis of AMI, reduce reinfarction and cardiac-related mortality, and carry significant clinical and public health importance.