Deep residual 2D convolutional neural network for cardiovascular disease classification.

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Imaging Analysis

Background:

• Cardiovascular disease (CVD) remains a significant global health issue.
• Manual interpretation of electrocardiograms (ECGs) limits widespread diagnostic accessibility.
• Automated ECG analysis offers potential for improved accuracy and efficiency.

Purpose of the Study:

• To implement and evaluate a novel deep two-dimensional convolutional neural network (2D-CNN) for cardiac disorder detection using ECG data.
• To assess the performance of the 2D-CNN across different classification complexities (2, 5, and 23 cardiovascular disease classes).

Main Methods:

• A deep two-dimensional convolutional neural network (2D-CNN) was developed and applied to the PTB-XL dataset.
• The model was trained and validated for classifying cardiovascular conditions into 2, 5, and 23 distinct classes.

Main Results:

• The 2D-CNN achieved an Area Under the Curve (AUC) of 95% and 87.85% average accuracy for healthy/sick patient classification.
• In a 5-class classification, the model reached an AUC of 93.46% and 89.87% average accuracy.
• For 23-class classification, the model demonstrated an AUC of 92.18% and 96.88% accuracy, outperforming other methods on the same dataset.

Conclusions:

• The developed 2D-CNN model shows strong performance in classifying various cardiovascular diseases from ECGs.
• This AI-driven approach can assist healthcare professionals in clinical ECG analysis and computer-aided diagnosis.
• The findings suggest a potential for enhanced accessibility and accuracy in cardiovascular care.