ResNet-50 for 12-Lead Electrocardiogram Automated Diagnosis

Area of Science:

• Artificial Intelligence in Medicine
• Cardiology
• Deep Learning Applications

Background:

• Cardiovascular diseases (CVD) are a leading cause of mortality globally.
• Accurate and timely diagnosis of CVD is crucial for effective treatment.
• Deep learning (DL) models show significant potential in medical diagnosis.

Purpose of the Study:

• To develop an efficient DL model for the automatic diagnosis of 27 classes of 12-lead electrocardiogram (ECG) signals.
• To identify 26 types of CVD and normal sinus rhythm using ECG data.
• To validate the model's performance against existing literature.

Main Methods:

• Implementation of a Residual Neural Network (ResNet-50) model.
• Training and testing on a combined dataset from public databases in the USA, China, and Germany.
• Utilizing 12-lead ECG signals for classification.

Main Results:

• The proposed DL model achieved an accuracy of 97.63%.
• The model demonstrated a precision of 89.67%.
• Results were validated against current scientific literature, confirming the model's efficacy.

Conclusions:

• The developed DL model, ResNet-50, is effective for automatic CVD diagnosis from ECG signals.
• The high accuracy and precision indicate a significant advancement in AI-driven cardiology.
• This approach offers a scalable and reliable method for early detection of cardiovascular conditions.