Improving the classification of cardiovascular diseases using statistical features obtained from empirical mode...

Area of Science:

• Cardiology
• Biomedical Engineering
• Signal Processing

Background:

• Cardiovascular diseases (CVDs) are a leading cause of global mortality.
• Early detection and treatment of CVDs are crucial for reducing mortality rates.
• Electrocardiograms (ECGs) are non-invasive tools for assessing cardiac health, but manual interpretation of subtle deviations for CVD diagnosis is challenging and time-consuming.

Purpose of the Study:

• To develop an automated system for identifying five specific cardiovascular diseases (CVDs) using ECG data.
• To leverage Empirical Mode Decomposition (EMD) and feature extraction for enhanced diagnostic accuracy.

Main Methods:

• Utilized Empirical Mode Decomposition (EMD) to decompose ECG signals into intrinsic mode functions (IMFs).
• Extracted temporal and spectral features from the IMFs.
• Ranked features using the one-way ANOVA test.
• Employed an extreme gradient boosting classifier for automated CVD identification.

Main Results:

• Achieved a maximum classification accuracy of 99.56% on a proposed dataset and 92.4% on the MIT-BIH database.
• Demonstrated strong generalizability across different datasets.
• Clinical validation on a large real-world dataset (RELA Hospital) yielded 81.3% accuracy, indicating deployment challenges.

Conclusions:

• The proposed EMD-based approach with feature extraction and extreme gradient boosting classifier shows high potential for automated CVD detection from ECGs.
• The method demonstrates robust performance on benchmark datasets but highlights the need for further refinement for real-world clinical application.
• Further research is needed to address the challenges identified during clinical validation for seamless integration into healthcare systems.