Related Experiment Videos
Using wavelet transform and fuzzy neural network for VPC detection from the Holter ECG
Liang-Yu Shyu1, Ying-Hsuan Wu, Weichih Hu
1Department of Biomedical Engineering, Chung Yuan Christian University, 22 Pu-Jen, Pu-chung Li, Chung Li 32023, Taiwan, ROC. lshyu@be.cycu.edu.tw
IEEE Transactions on Bio-Medical Engineering
|July 14, 2004
Summary
This study introduces a novel method for detecting ventricular premature contractions (VPCs) using wavelet transform and fuzzy neural networks. The approach efficiently reuses QRS detection information, achieving 99.79% accuracy for reliable VPC classification.
Area of Science:
- Cardiology
- Biomedical Engineering
- Signal Processing
Background:
- Ventricular premature contractions (VPCs) are common arrhythmias requiring accurate detection.
- Existing ECG classification algorithms often involve a QRS detection step.
- Artifacts and R-wave amplitude variations can challenge reliable VPC detection.
Purpose of the Study:
- To propose a novel, efficient method for VPC detection from Holter ECG data.
- To leverage information from QRS detection for improved VPC classification.
- To develop a robust method minimizing the impact of noise and amplitude variations.
Main Methods:
- Utilizing quadratic spline wavelet transform (WT) for feature extraction from ECG signals.
- Employing fuzzy neural networks (FNN) for classifying VPCs.
- Selecting QRS duration (scale 3) and QRS area (scale 4) as characteristic features.
- Implementing feature normalization to mitigate R-wave amplitude influence.
Main Results:
- The proposed method achieved a high accuracy of 99.79% for VPC classification.
- Feature extraction using WT effectively eliminated high and low-frequency noise.
- Normalization successfully reduced the impact of alternating R-wave amplitudes.
- Exclusion of left bundle branch block beats further improved classification reliability.
Conclusions:
- The novel WT and FNN method provides a highly accurate and reliable approach for VPC detection.
- Reusing QRS detection features simplifies implementation and reduces computational complexity.
- The method demonstrates robustness against common ECG signal artifacts and variations.