Correlation technique and least square support vector machine combine for frequency domain based ECG beat...

Area of Science:

• Biomedical Engineering
• Medical Informatics
• Cardiology

Background:

• Accurate and timely detection of cardiac arrhythmia is crucial for effective medical intervention.
• Existing methods for ECG beat classification may lack sufficient accuracy or generalization capabilities.

Purpose of the Study:

• To develop an automated medical diagnostic tool for classifying ECG beats.
• To classify ECG beats into three categories: normal, premature ventricular contraction (PVC), and other beats.
• To demonstrate the generalization capability of the proposed classification scheme.

Main Methods:

• Utilized a cross-correlation based approach for feature extraction using cross-spectral density information in the frequency domain.
• Developed a Least Square Support Vector Machine (LS-SVM) classifier.
• Trained the classifier on a small dataset and tested it on a large dataset from the MIT/BIH arrhythmia database.

Main Results:

• Achieved high classification accuracy ranging from 95.51% to 96.12% on 40 files from the MIT/BIH arrhythmia database.
• The proposed scheme demonstrated superior performance compared to several competing algorithms.
• Successfully classified ECG beats into normal, PVC, and other categories.

Conclusions:

• The developed automated tool provides an accurate and efficient method for ECG beat classification.
• The cross-correlation and LS-SVM approach shows significant potential for clinical application in arrhythmia detection.
• The high accuracy and generalization capability suggest the robustness of the proposed scheme for real-world diagnostic use.