Enhancing P-wave localization for accurate detection of second-degree and third-degree atrioventricular conduction...

Area of Science:

• Cardiology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Accurate detection of second- and third-degree atrioventricular block (AVB) in electrocardiogram (ECG) signals is crucial for patient management.
• Automated detection algorithms often struggle with precise P-wave localization, impacting overall AVB diagnostic accuracy.

Purpose of the Study:

• To develop and validate a novel method, P-WaveNet, for accurate P-wave localization to improve automated detection of second-degree and third-degree AVB.
• To address data scarcity for rare AVB types through data augmentation.

Main Methods:

• P-WaveNet employs an attention mechanism for feature extraction and a bidirectional long short-term memory module for temporal dependency analysis.
• A mathematical approach was used to synthesize pseudo-data for training, addressing the scarcity of second- and third-degree AVB (2AVB, 3AVB) data.
• A classification rule was established using P-wave positions, RR interval rhythm, and PR intervals for automatic AVB detection.

Main Results:

• P-WaveNet achieved high F1 scores for P-wave localization (e.g., 93.62% on QT Dataset, 98.29% for 2AVB on BUTPDB).
• The AVB detection algorithm demonstrated F1 scores of 83.33% and 84.15% for 2AVB and 3AVB across independent datasets.
• The model showed promising performance even for challenging P-wave localization in 3AVB (62.65% F1 score on BUTPDB).

Conclusions:

• P-WaveNet accurately identifies P-waves in complex ECGs, significantly improving the efficacy of automated AVB detection.
• The fusion of medical expertise, data augmentation, and advanced AI techniques offers a robust solution for AVB diagnosis.
• The proposed method shows considerable potential for clinical applicability in real-world ECG analysis.