An ECG Classification Method Based on Multi-Task Learning and CoT Attention Mechanism.

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Diagnostics

Background:

• Electrocardiograms (ECGs) are vital for diagnosing cardiovascular diseases.
• Accurate arrhythmia detection is crucial due to a shortage of cardiologists and widespread ECG use.

Purpose of the Study:

• To develop an efficient multi-task deep neural network for improved arrhythmia detection from ECGs.
• To dynamically model local and global ECG feature sequences for enhanced classification.

Main Methods:

• Proposed a multi-task deep neural network with a shared SE-ResNet for low-level feature extraction.
• Incorporated a Contextual Transformer (CoT) block for dynamic modeling of ECG feature sequences.
• Evaluated the model on the CPSC2018 and PTB-XL public datasets.

Main Results:

• Achieved an average F1 score of 0.827 on the CPSC2018 dataset.
• Achieved an average F1 score of 0.833 on the PTB-XL dataset.
• Demonstrated the effectiveness of the proposed deep learning approach in arrhythmia detection.

Conclusions:

• The novel multi-task deep neural network effectively detects arrhythmias from ECG data.
• The integration of CoT blocks enhances the modeling of ECG features for improved diagnostic accuracy.
• The proposed method shows significant potential for clinical application in automated cardiac diagnostics.