High-Performance Personalized Heartbeat Classification Model for Long-Term ECG Signal.

Area of Science:

• Cardiovascular Medicine
• Biomedical Engineering
• Artificial Intelligence in Healthcare

Background:

• Long-term electrocardiogram (ECG) is crucial for diagnosing cardiovascular diseases like arrhythmia, myocardial infarction, and myocarditis.
• Automatic heartbeat classification is challenging, especially for personalized, long-term data analysis from methods like Holter monitoring.
• There is a significant need for efficient, personalized automatic classification models to expedite diagnosis.

Purpose of the Study:

• To develop an efficient and accurate automatic heartbeat classification method using a parallel general regression neural network (GRNN).
• To create a personalized heartbeat classification model for individual patients through an online learning program.
• To evaluate the performance and efficiency of the proposed parallel GRNN model.

Main Methods:

• Implementation of a parallel general regression neural network (GRNN) for heartbeat classification.
• Development of an online learning program for creating patient-specific classification models.
• Performance evaluation using standard metrics and comparison against specific patient ECG data.

Main Results:

• The parallel GRNN achieved an overall accuracy of 95% based on Association for the Advancement of Medical Instrumentation standards.
• The personalized models, developed via online learning, demonstrated an 88% accuracy on specific patient ECG data.
• The parallel GRNN, utilizing a GTX780Ti, showed a 450-fold improvement in efficiency.

Conclusions:

• The parallel GRNN offers a highly accurate and efficient solution for automatic heartbeat classification.
• Online learning enables the development of effective personalized models for long-term ECG analysis.
• This approach significantly accelerates the analysis of long-term ECG data, improving diagnostic capabilities.