Research on arrhythmia recognition by using convolutional neural network in ECG images.

Area of Science:

• Cardiology
• Biomedical Engineering
• Artificial Intelligence in Medicine

Background:

• Accurate arrhythmia identification is vital for preventing and diagnosing cardiovascular diseases.
• Preprocessing steps in existing methods can lead to information loss.
• A need exists for improved models to accurately classify multiple arrhythmia types.

Purpose of the Study:

• To enhance arrhythmia classification by minimizing information loss during preprocessing.
• To improve the performance of arrhythmia identification models.
• To accurately classify six distinct types of cardiac arrhythmias.

Main Methods:

• Wavelet transform denoising was applied for signal preprocessing.
• Electrocardiogram (ECG) signals were converted into 2D grayscale images.
• An improved 2D convolutional neural network (CNN) model was utilized for classification.

Main Results:

• The proposed method achieved a classification accuracy of 90.50%.
• Sensitivity and specificity reached 81.70% and 97.16%, respectively.
• Six types of arrhythmias were successfully and accurately identified.

Conclusions:

• Wavelet transform preprocessing effectively enhances the classification accuracy of multiple arrhythmias.
• The developed CNN model offers a promising new tool for clinical arrhythmia diagnosis.
• This method provides a valuable reference for clinicians in diagnosing cardiac arrhythmias.