Efficient Lightweight Multimodel Deep Fusion Based on ECG for Arrhythmia Classification.

Area of Science:

• Cardiology
• Artificial Intelligence
• Biomedical Engineering

Background:

• Arrhythmias result from irregular electrical signals in the heart, increasing risks of stroke and cardiac arrest.
• Early detection of arrhythmias is crucial for reducing mortality rates.

Purpose of the Study:

• To develop a lightweight multimodel using convolutional neural networks (CNNs) for arrhythmia classification from ECG signals.
• To enable knowledge transfer from multiple deep learning models into a single, efficient diagnostic tool.

Main Methods:

• A novel multimodel architecture based on CNNs was designed for knowledge consolidation.
• The model was trained and validated using a publicly accessible electrocardiogram (ECG) database.
• Performance was benchmarked against single-model approaches and existing arrhythmia detection methodologies.

Main Results:

• The developed multimodel demonstrated superior performance in classifying arrhythmias compared to single models.
• The system outperformed most previously reported arrhythmia detection methods.
• Accurate classification was achieved even with limited data collections.

Conclusions:

• The lightweight multimodel offers an effective and efficient solution for arrhythmia diagnosis using ECG data.
• This AI-driven tool can assist medical experts in decision-making and expedite the diagnostic process.
• The model's accuracy and efficiency hold significant potential for improving patient outcomes in arrhythmia management.