Convolutional neural network based automatic screening tool for cardiovascular diseases using different intervals of...

Area of Science:

• Cardiology
• Biomedical Engineering
• Artificial Intelligence in Medicine

Background:

• Cardiovascular diseases (CVDs) are a leading global cause of mortality.
• Electrocardiogram (ECG) analysis is a crucial non-invasive method for CVD detection.
• Developing automated screening tools for CVDs is essential for public health.

Purpose of the Study:

• To propose a deep convolutional neural network (CNN) for classifying five types of CVDs.
• To evaluate the performance of the CNN using standard 12-lead ECG signals.
• To assess the potential of the proposed method for real-time medical implementation.

Main Methods:

• Utilized the Physiobank (PTB) ECG database for the study.
• Segmented ECG signals into one, two, and three-second intervals without wave detection.
• Employed min-max normalization on raw ECG signals and a ten-fold cross-validation approach.

Main Results:

• The proposed CNN achieved high accuracy, sensitivity, and specificity, reaching up to 99.84%, 99.52%, and 99.95% respectively for three-second ECG signals.
• One-second ECG signals yielded an accuracy of 99.59%, sensitivity of 99.04%, and specificity of 99.87%.
• Pre-trained models on two-second signals demonstrated overall accuracy, sensitivity, and specificity of 99.80%, 99.48%, and 99.93%.

Conclusions:

• The developed CNN system demonstrates high performance in classifying CVDs from ECG signals.
• The system's efficiency and flexibility suggest its suitability for practical, real-time medical environments.
• This approach holds significant potential for improving automated CVD screening and diagnosis.