Fast clustering algorithm for large ECG data sets based on CS theory in combination with PCA and K-NN methods.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Machine Learning

Background:

• Long-term Electrocardiogram (ECG) monitoring is crucial for diagnosing heart conditions.
• Current ECG analysis methods face challenges in real-time processing, energy efficiency, and data sampling load.
• Efficiently analyzing large ECG datasets is essential for uncovering hidden patterns in P-QRS-T waves.

Purpose of the Study:

• To develop a novel, optimized clustering algorithm for low-power, long-term ECG recording and analysis.
• To enhance the real-time processing and energy efficiency of ECG data classification.
• To improve the accuracy of detecting abnormalities in long-term ECG signals.

Main Methods:

• An advanced K-means clustering algorithm incorporating Compressed Sensing (CS) for random sampling.
• Application of dimensionality reduction techniques: Principal Component Analysis (PCA) and Linear Correlation Coefficient (LCC).
• Classification using K-Nearest Neighbours (K-NN) and Probabilistic Neural Network (PNN) classifiers.

Main Results:

• The proposed algorithm, particularly PCA features with K-NN, demonstrated superior performance compared to existing methods.
• Achieved an 11% increase in classification accuracy.
• Reported high Receiver Operating Characteristics (ROC) areas: 99.98% for K-NN and 99.83% for PNN.

Conclusions:

• The developed algorithm offers a significant improvement in ECG data analysis efficiency and accuracy.
• The combination of CS, PCA, and K-NN provides a robust solution for low-power, long-term ECG monitoring.
• This approach enhances diagnostic capabilities for heart diseases through improved P-QRS-T wave analysis.