Classification of Heartbeats based on Linear Discriminant Analysis and Artificial Neural Network

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Cardiovascular Signal Processing

Background:

• Accurate heartbeat classification is crucial for diagnosing cardiac conditions.
• Existing methods often involve complex feature extraction and high computational costs.
• The need for efficient and real-time algorithms for arrhythmia detection is significant.

Purpose of the Study:

• To develop and evaluate a novel heartbeat classification algorithm.
• To reduce feature dimensionality for improved computational efficiency.
• To compare the proposed algorithm's performance against a fuzzy inference system classifier.

Main Methods:

• Feature extraction from the first derivative of ECG signals and RR interval information.
• Dimensionality reduction using Linear Discriminant Analysis (LDA) from 275 to 6 features.
• Classification using an Artificial Neural Network (ANN).
• Validation using the MIT-BIH Arrhythmia database.

Main Results:

• The proposed algorithm achieved high performance metrics: 97.49% sensitivity, 97.91% specificity, and 96.36% accuracy.
• Feature extraction solely from the first derivative of ECG signals enables real-time implementation.
• Reduced feature dimensionality significantly decreases learning and testing time costs.

Conclusions:

• The LDA-ANN based heartbeat classification algorithm offers a computationally efficient and accurate solution for real-time arrhythmia detection.
• The method's reliance on the first derivative of ECG signals simplifies feature extraction for practical applications.
• This approach holds promise for improving automated cardiac monitoring systems.