Heartbeat classification method combining multi-branch convolutional neural networks and transformer.

Area of Science:

• Cardiology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Arrhythmia detection and classification are vital for diagnosing cardiovascular diseases.
• Current deep learning methods often struggle to integrate both morphological and temporal ECG features.
• There is a need for advanced models that can simultaneously analyze diverse ECG signal characteristics.

Purpose of the Study:

• To propose a hybrid deep learning model for enhanced heartbeat classification.
• To combine Transformer and multi-branch Convolutional Neural Networks (CNNs) for ECG analysis.
• To validate the model's performance on SVEB and VEB arrhythmia classes using the MIT-BIH database.

Main Methods:

• Developed a hybrid model integrating Transformer and multi-branch CNN architectures.
• Implemented a fusion module to combine features from different classifiers.
• Conducted intra-patient and inter-patient classification protocols on the MIT-BIH arrhythmia database.

Main Results:

• Achieved 99.5% overall accuracy in intra-patient classification.
• Demonstrated high sensitivity (Sen) and specificity (Spe) for SVEB (92.4%, 99.9%) and VEB (98.2%, 99.9%) arrhythmias.
• Obtained strong results in inter-patient protocols with 98.8% and 97.2% overall accuracy.

Conclusions:

• The proposed hybrid model effectively classifies arrhythmias by integrating morphological and temporal ECG features.
• The method shows significant potential for improving the accuracy and reliability of automated arrhythmia detection.
• This approach offers a promising advancement in leveraging deep learning for cardiovascular disease diagnosis.