ECG Beats Classification Using Mixture of Features.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Cardiology

Background:

• Accurate classification of electrocardiogram (ECG) signals is crucial for diagnosing heart disease.
• Existing methods for ECG beat classification face challenges in distinguishing between various arrhythmias.

Purpose of the Study:

• To design and evaluate an efficient system for classifying five types of ECG beats: normal (N), ventricular ectopic (V), supraventricular ectopic (S), fusion (F), and unknown (Q).
• To compare the performance of novel feature extraction techniques with existing methods for ECG beat classification.

Main Methods:

• Proposed two feature extraction methods: S-transform based features with temporal features, and a combination of S-transform (ST) and Wavelet Transform (WT) based features with temporal features.
• Utilized a multilayer perceptron neural network (MLPNN) classifier for independent classification of the extracted features.
• Evaluated performance on the MIT-BIH arrhythmia database, comparing three feature extraction techniques against AAMI standards for five ECG beat classes.

Main Results:

• The proposed feature extraction techniques demonstrated superior performance compared to existing methods.
• Achieved average sensitivity of 95.70% for Normal (N), 78.05% for Supraventricular ectopic (S), 49.60% for Fusion (F), 89.68% for Ventricular ectopic (V), and 33.89% for Unknown (Q) beats.

Conclusions:

• The developed system offers an efficient and effective approach for ECG beat classification.
• The proposed feature extraction techniques show significant potential for improving the accuracy of automated cardiac arrhythmia diagnosis.