Patient-Specific Heartbeat Classification in Single-Lead ECG using Convolutional Neural Network.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Cardiovascular Health

Background:

• Arrhythmia detection from electrocardiograms (ECGs) is challenging for automated systems.
• Portable ECG devices generate vast amounts of data, necessitating efficient processing algorithms.
• Convolutional Neural Networks (CNNs) show potential for automated heartbeat classification.

Purpose of the Study:

• To develop an arrhythmia classification model using CNNs, adhering to Association for the Advancement of Medical Instruments (AAMI) standards.
• To evaluate static and dynamic heartbeat segmentation methods for ECG analysis.
• To create patient-specific CNN models for personalized, long-term health monitoring using single-lead ECG data.

Main Methods:

• Utilized the MIT-BIH Arrhythmia database for training and testing CNN models.
• Experimented with both static and dynamic heartbeat segmentation techniques.
• Evaluated patient-specific CNN models on data from a novel wireless single-lead ECG sensor.

Main Results:

• Developed CNN-based models for arrhythmia classification.
• Demonstrated the effectiveness of patient-specific models for personalized ECG analysis.
• Validated the models using data from a wireless single-lead ECG sensor, indicating potential for real-world application.

Conclusions:

• CNNs are effective tools for automated arrhythmia classification from ECG data.
• Patient-specific models are crucial for accurate long-term health monitoring.
• The developed approach shows promise for integration into wearable health monitoring platforms.