A Multi-Domain Feature Fusion CNN for Myocardial Infarction Detection and Localization.

Area of Science:

• Cardiology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Myocardial infarction (MI) detection traditionally relies on single-domain electrocardiogram (ECG) features, which struggle to capture complex cardiac electrical activity.
• Limitations in single-domain analysis hinder accurate MI detection and localization.
• The variability and complexity of cardiac signals necessitate advanced analytical approaches.

Purpose of the Study:

• To develop and validate a Multi-Domain Feature Fusion Convolutional Neural Network (MFF-CNN) for enhanced MI detection and localization.
• To integrate time, frequency, and time-frequency domain ECG features for a comprehensive analysis.
• To overcome the limitations of traditional single-domain ECG analysis in cardiovascular disease diagnosis.

Main Methods:

• Generation of 2D frequency and time-frequency domain ECG images combined with 1D time-domain features.
• Implementation of a novel MFF-CNN architecture incorporating 1D and 2D Convolutional Neural Networks (CNNs).
• Multi-domain feature learning for automatic MI detection and localization using the proposed MFF-CNN.

Main Results:

• Achieved 99.98% detection accuracy and 84.86% localization accuracy in inter-patient validation.
• Demonstrated a 3.43% absolute improvement in detection accuracy over state-of-the-art methods.
• Showcased a significant 16.97% enhancement in localization accuracy compared to existing techniques.

Conclusions:

• The MFF-CNN effectively integrates multi-domain ECG features for superior MI detection and localization.
• This approach significantly advances the accuracy of automated cardiovascular disease analysis.
• The proposed method holds substantial promise for future research and clinical applications in diagnosing myocardial infarction.