Automatic cardiac arrhythmia detection and classification using vectorcardiograms and complex networks.

Area of Science:

• Biomedical Engineering
• Cardiology
• Signal Processing

Background:

• Cardiac arrhythmias require accurate detection and classification for effective treatment.
• Traditional electrocardiogram (ECG) analysis can be enhanced with additional signal information.
• Vectorcardiograms (VCG) offer a complementary perspective on cardiac electrical activity.

Purpose of the Study:

• To investigate novel feature extraction methods for cardiac arrhythmia detection and classification.
• To explore the utility of vectorcardiograms (VCG) in conjunction with electrocardiograms (ECG) for improved heart rhythm analysis.
• To apply complex network theory for extracting discriminative features from VCG signals.

Main Methods:

• Utilized the MIT-BIH arrhythmia database for analysis.
• Applied complex networks to extract features from VCG signals.
• Classified heartbeats into 5 standard classes (N, V, S, F, Q) following ANSI/AAMI EC57:1998.
• Employed a Support Vector Machine (SVM) classifier.
• Implemented a 22-fold cross-validation strategy based on de Chazal's dataset division scheme.

Main Results:

• Achieved a global accuracy of 84.1% in cardiac arrhythmia classification.
• Demonstrated high Sensitivity and Positive Predictive Value for beat classification.
• Reported low False Positive Rates compared to existing studies using the same methodology.
• The integration of VCG-derived features improved classification performance.

Conclusions:

• Complex network-based feature extraction from VCG signals is a promising approach for cardiac arrhythmia detection.
• Combining VCG and ECG data enhances the accuracy of arrhythmia classification systems.
• The proposed method offers a competitive performance against established techniques in the field.