A New Multichannel Parallel Network Framework for the Special Structure of Multilead ECG

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Cardiology

Background:

• Electrocardiograms (ECGs) capture crucial heartbeat rhythm and waveform morphology, varying significantly across different cardiovascular diseases.
• Accurate interpretation of ECG signals is vital for timely clinical diagnosis and effective patient management.
• Existing diagnostic methods may benefit from advanced computational approaches for enhanced feature extraction and analysis.

Purpose of the Study:

• To develop and evaluate a novel deep learning model, the multichannel parallel neural network (MLCNN-BiLSTM), for comprehensive ECG analysis.
• To explore the combined utility of morphological and rhythmic ECG features for improved cardiovascular disease detection.
• To assess the potential of the proposed model as a preliminary screening tool in clinical settings.

Main Methods:

• Proposed a hybrid deep learning architecture, MLCNN-BiLSTM, integrating Multichannel Convolutional Neural Network (MLCNN) and Bidirectional Long Short-Term Memory (BiLSTM) channels.
• MLCNN channel focused on extracting morphological features from multilead ECG waveforms, adept at handling ECG-specific structures.
• BiLSTM channel focused on extracting rhythmic features from continuous ECG heartbeat data, with weighted fusion of temporal-spatial features for sensitivity analysis.

Main Results:

• The MLCNN-BiLSTM model achieved a high accuracy rate of 87.81% in identifying multiple cardiovascular diseases.
• The model demonstrated strong diagnostic performance with a sensitivity of 86.00% and a specificity of 87.76%.
• Experimental results validate the model's capability in discerning disease-specific patterns from ECG signals.

Conclusions:

• The developed MLCNN-BiLSTM neural network effectively integrates morphological and rhythmic ECG features for enhanced cardiovascular disease diagnosis.
• This model shows significant promise as an efficient first-round screening tool for clinical ECG interpretation.
• Further research can explore the integration of this AI tool into existing clinical workflows to aid cardiologists.