Beat-wise segmentation of electrocardiogram using adaptive windowing and deep neural network

Area of Science:

• Biomedical Engineering
• Signal Processing
• Artificial Intelligence in Healthcare

Background:

• Accurate electrocardiogram (ECG) analysis is vital for patient monitoring and post-treatment care.
• Beat-wise segmentation is a critical prerequisite for reliable automated ECG classification.
• Existing methods may struggle with the fidelity and confidence required for clinical applications.

Purpose of the Study:

• To develop and validate a robust ECG beat segmentation technique.
• To improve the accuracy and reliability of identifying cardiac cycle events and segmenting heartbeats.
• To enable more confident clinical applications of automated ECG analysis.

Main Methods:

• A Convolutional Neural Network (CNN) model integrated with an adaptive windowing algorithm was developed.
• The adaptive windowing algorithm identifies cardiac cycle events for precise segmentation.
• The technique segments both regular and irregular heartbeats from ECG signals.

Main Results:

• Achieved 99.08% accuracy and F1-score on the MIT-BIH dataset for heartbeat detection.
• Demonstrated 99.25% accuracy in boundary determination on the MIT-BIH dataset.
• Showcased high performance on European S-T (98.3% accuracy, 97.4% precision) and Fantasia (99.4% accuracy, 99.4% precision) databases.

Conclusions:

• The proposed ECG beat segmentation technique demonstrates high accuracy and reliability across multiple datasets.
• The method's ability to accurately segment regular and irregular beats supports its clinical utility.
• This algorithm offers a promising solution for enhanced confidence in various ECG analysis applications.