[Research on Clinical Electrocardiogram Classification Algorithm Based on Ensemble Learning].

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Imaging

Background:

• Increasing volumes of electrocardiogram (ECG) data necessitate advanced computer-aided analysis.
• Traditional ECG classification methods often rely on feature extraction, which can be limiting.

Purpose of the Study:

• To enhance the performance of clinical ECG classification using a Lead Convolutional Neural Network (LCNN).
• To develop a robust ECG classification algorithm by fusing multiple LCNN classifiers.

Main Methods:

• Developed two distinct classifiers using varied preprocessing and training strategies.
• Applied a multiple output prediction method to each classifier independently.
• Fused the predictions of the independent classifiers using a Bayesian approach.

Main Results:

• Achieved a classification accuracy of 85.04% on over 150,000 ECG records.
• Obtained an Area Under the Receiver Operating Characteristic Curve (AUC) of 0.9185.
• Demonstrated superior performance compared to traditional feature extraction-based techniques.

Conclusions:

• The proposed Bayesian fusion of LCNN classifiers significantly improves clinical ECG classification accuracy.
• This method offers a promising advancement for computer-aided ECG analysis in clinical settings.