Premature Ventricular Contraction Recognition Based on a Deep Learning Approach

Area of Science:

• Cardiology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Electrocardiogram (ECG) signals are crucial for diagnosing heart diseases, with abnormal heartbeats like premature ventricular contractions (PVCs) indicating serious conditions.
• Automated computer-assisted techniques are increasingly vital for reducing the diagnostic burden of identifying heart arrhythmias.
• PVCs are a common arrhythmia originating in the heart's lower chambers, potentially leading to severe health issues.

Purpose of the Study:

• To propose and evaluate a deep learning-based approach for accurate recognition of premature ventricular contractions (PVCs).
• To leverage extracted ECG signal features for enhanced automatic arrhythmia detection.

Main Methods:

• Extracted 10 heartbeat and statistical features, including morphological (RS amplitude, QR amplitude, QRS width) and statistical measures, from 20-second ECG signal segments.
• Constructed feature vectors from the extracted ECG data.
• Utilized a convolutional neural network (CNN) to analyze feature patterns for effective PVC classification.

Main Results:

• The proposed pipeline demonstrated improved diagnostic performance in identifying PVCs.
• The deep learning approach effectively classified unique patterns within the ECG features.

Conclusions:

• The developed deep learning pipeline shows significant promise for improving the accuracy and efficiency of automatic PVC detection.
• This method can aid in the early diagnosis of heart conditions, reducing the workload on medical professionals.