Related Experiment Video
Updated: Jun 20, 2025
07:21
Magnetic Resonance Derived Myocardial Strain Assessment Using Feature Tracking
Published on: February 12, 2011
14.3K
An interpretable ensemble trees method with joint analysis of static and dynamic features for myocardial infarction
Chunmiao Liang1, Qinghua Sun1,2, Jiali Li1
1School of Control Science and Engineering, Shandong University, Jinan 250061, People's Republic of China.
Physiological Measurement
|July 18, 2024
Summary
This study introduces an AI method combining static and dynamic ECG features for accurate myocardial infarction (MI) detection. The novel StackTree approach enhances interpretability and achieves high accuracy, outperforming traditional single-feature methods.
Area of Science:
- Cardiology
- Artificial Intelligence
- Biomedical Signal Processing
Background:
- Artificial intelligence (AI) methods are increasingly applied to electrocardiogram (ECG) analysis for myocardial infarction (MI) detection.
- Existing AI-ECG methods often focus on static features, neglecting the joint analysis of static and dynamic ECG characteristics for improved MI detection accuracy and interpretability.
- A comprehensive approach integrating both static and dynamic ECG features is needed for robust and understandable MI diagnosis.
Purpose of the Study:
- To develop an AI-based method for accurate and interpretable myocardial infarction (MI) detection by jointly analyzing static and dynamic ECG features.
- To introduce a simplified ensemble tree method, StackTree, that incorporates a two-stage feature selection and a stacked ensemble scheme for enhanced MI detection.
- To evaluate the proposed method's performance and interpretability using public and clinical ECG databases.
Main Methods:
- Extraction of both classical static ECG features and dynamic features modeled via dynamic learning.
- Implementation of a two-stage feature selection strategy to identify significant static and dynamic features for ensemble tree construction.
- Development of the StackTree method, a simplified ensemble tree approach with a stacked ensemble scheme for interpretable MI classification.
Main Results:
- The proposed method achieved 97.1% accuracy on the PTB database and 84.5% on a clinical database, outperforming traditional single-feature methods.
- The algorithm demonstrated performance comparable to conventional random forest methods while offering enhanced interpretability.
- Selected important features confirmed the crucial roles of both static and dynamic ECG information in MI detection.
Conclusions:
- The joint analysis of static and dynamic ECG features using the proposed simplified ensemble tree method (StackTree) significantly improves MI detection accuracy and interpretability.
- The StackTree method provides a clear, visual understanding of its internal workings, facilitating clinical adoption.
- This approach represents a promising advancement in AI-driven cardiovascular diagnostics.
