Improved delineation model of a standard 12-lead electrocardiogram based on a deep learning algorithm

Area of Science:

• Cardiology
• Artificial Intelligence
• Signal Processing

Background:

• Manual delineation of 12-lead electrocardiogram (ECG) signals is challenging due to signal variations, noise, and irregular heart rhythms.
• Accurate ECG signal delineation is crucial for extracting comprehensive information and characteristics in clinical practice.

Purpose of the Study:

• To develop and evaluate a deep learning algorithm for automated delineation of 12-lead ECG signals.
• To classify ECG waveforms and boundaries, including P-wave, QRS-complex, and T-wave.

Main Methods:

• Implemented a deep learning model utilizing convolutional layers within convolutional neural networks (CNNs) for automated feature extraction.
• Employed a bidirectional long short-term memory (BiLSTM) network as a classifier.
• Experimented with both beat-based and patient-based approaches for ECG beat segmentation.

Main Results:

• The proposed deep learning model achieved excellent performance across all metrics.
• Achieved over 95% accuracy for beat-based segmentation and over 93% accuracy for patient-based segmentation.
• Evaluated on a dataset of 14,588 ECG beats.

Conclusions:

• The developed automated 12-lead ECG delineation model demonstrates high accuracy and efficiency.
• This deep learning approach represents a significant advancement towards clinical application in cardiology.
• The model's performance indicates its potential to aid clinicians in ECG interpretation.