Robust electrocardiogram delineation model for automatic morphological abnormality interpretation

Area of Science:

• Biomedical Engineering
• Cardiology
• Artificial Intelligence

Background:

• Accurate electrocardiogram (ECG) wave delineation is crucial for diagnosing heart abnormalities.
• Existing methods struggle with noise interference and integrating medical knowledge for accurate P-QRS-T wave identification.
• Robust ECG signal processing is essential for reliable cardiac diagnostics.

Purpose of the Study:

• To develop a robust delineation model for precise P-QRS-T wave classification in ECG signals.
• To improve the accuracy and reliability of ECG delineation, especially in the presence of noise and artifacts.
• To integrate medical knowledge for interpreting ECG morphological abnormalities and detecting arrhythmias.

Main Methods:

• A convolutional recurrent network model was developed and optimized using grid search.
• The model was trained and validated on multiple diverse ECG datasets (LUDB, QTDB, PhysioNet, etc.).
• ECG waveform classification results were used to interpret morphological abnormalities and detect arrhythmias based on P-wave and RR interval analysis.

Main Results:

• The best performing model achieved 99.97% accuracy, 99.92% sensitivity, and 99.93% precision for ECG waveform classification.
• The model demonstrated robustness across seven different ECG datasets, handling artifact noise and baseline drift effectively.
• The proposed method successfully interpreted delineation results for arrhythmia detection, considering P-wave morphology and RR interval regularity.

Conclusions:

• The proposed convolutional recurrent network model offers a robust and accurate solution for ECG delineation.
• The model's ability to integrate medical knowledge enhances its capability in identifying cardiac abnormalities and arrhythmias.
• This approach provides a significant advancement in automated ECG analysis for clinical applications.