An Embedded System Using Convolutional Neural Network Model for Online and Real-Time ECG Signal Classification and...

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Signal Processing

Background:

• Electrocardiogram (ECG) analysis is crucial for diagnosing cardiac conditions.
• Accurate and automated ECG interpretation remains a significant challenge in clinical practice.
• Deep Learning (DL) offers promising avenues for improving ECG signal classification.

Purpose of the Study:

• To develop and evaluate an automated system for classifying four types of ECG signals using Deep Learning (DL).
• To investigate the performance of a Convolutional Neural Network (CNN) model for ECG signal prediction.
• To implement the predictive model on a hardware platform for real-time classification.

Main Methods:

• Utilized ECG signals from the PhysioNet open-source database, encompassing normal, sudden death, arrhythmia, and supraventricular arrhythmia classes.
• Developed a Deep Learning (DL) model, specifically a Convolutional Neural Network (CNN), involving pulse extraction, image reshaping, and dataset training/testing.
• Achieved up to 95% training accuracy after 100 epochs.

Main Results:

• The CNN model demonstrated high accuracy in classifying ECG signals, with 100% accuracy for sudden death ECG waveforms (80/80 samples).
• Normal sinus ECG waveforms showed the lowest prediction accuracy at 92.5% (74/80 samples), attributed to feature similarity with supraventricular arrhythmia.
• The predictive model was successfully embedded in an NVIDIA Jetson Nanoprocessor for online, real-time ECG waveform classification.

Conclusions:

• The developed DL-based system effectively classifies various ECG signal types with high accuracy.
• The CNN model shows strong potential for automated cardiac condition diagnosis.
• Hardware implementation enables real-time ECG analysis, paving the way for portable diagnostic tools.