Combining Rhythm Information between Heartbeats and BiLSTM-Treg Algorithm for Intelligent Beat Classification of...

Area of Science:

• Cardiology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Arrhythmia poses a significant threat to human health and necessitates accurate identification and diagnosis.
• Early detection of arrhythmia is crucial for preventing severe heart conditions.

Purpose of the Study:

• To develop and validate a BiLSTM-Treg algorithm for automatic arrhythmia classification.
• To enhance the accuracy and interpretability of arrhythmia diagnosis using deep learning.

Main Methods:

• ECG signals were denoised using discrete wavelet transform, followed by heartbeat segmentation preserving temporal relationships.
• A BiLSTM network was optimized through experiments with varying heartbeat segment lengths.
• Tree regularization was applied to the BiLSTM model for improved classification accuracy and interpretability.

Main Results:

• The proposed BiLSTM-Treg algorithm achieved an overall classification accuracy of 99.32% on the MIT-BIH arrhythmia database.
• The algorithm demonstrated superior sensitivity and positive predictive value compared to existing methods.
• The study successfully classified heartbeats into five categories: nonectopic (N), supraventricular ectopic (S), ventricular ectopic (V), fused heartbeats (F), and unknown (Q).

Conclusions:

• The BiLSTM-Treg algorithm offers a highly accurate and interpretable solution for automatic arrhythmia classification.
• This approach holds significant potential for improving the early diagnosis and management of cardiovascular diseases.
• The integration of rhythm information and advanced deep learning techniques advances the field of automated cardiac diagnostics.