[A DenseNet-based diagnosis algorithm for automated diagnosis using clinical ECG data].

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Imaging

Background:

• Electrocardiogram (ECG) interpretation is crucial for clinical diagnosis.
• Accurate and efficient ECG analysis remains a challenge in healthcare.
• Deep learning models offer potential for automated ECG interpretation.

Purpose of the Study:

• To develop and train a convolutional neural network (CNN) using multi-lead ECG data.
• To accurately classify new ECG data for reliable clinical diagnostic information.
• To enhance the detection rate of abnormal ECG samples through data augmentation.

Main Methods:

• ECG data pre-processing involved bandpass filtering and signal framing for consistent input size.
• A depth-wise separable convolution structure was employed for channel-specific feature extraction in twelve-lead ECG.
• An improved DenseNet architecture was utilized to train two classifiers for different diagnostic labels.

Main Results:

• The proposed model achieved 80.13% accuracy in distinguishing normal from abnormal ECG.
• Sensitivity, specificity, and F1 score were reported as 80.38%, 79.91%, and 79.35%, respectively.
• The model demonstrated a processing time of 33.59 ms per dataset on GPU.

Conclusions:

• The developed model provides rapid and effective classification of ECG data.
• Real-time prediction capabilities meet essential clinical diagnostic requirements.
• The approach shows promise for improving the efficiency and accuracy of automated ECG analysis.