Using phase space reconstruction for patient independent heartbeat classification in comparison with some benchmark...

Area of Science:

• Cardiology
• Biomedical Engineering
• Machine Learning

Background:

• Automatic heartbeat classification is crucial for diagnosing cardiac conditions.
• Existing methods often struggle with patient-independent classification, yielding suboptimal results.
• A need exists for robust algorithms that generalize across different individuals.

Purpose of the Study:

• To develop and evaluate novel methods for patient-independent heartbeat classification.
• To address the limitations of current techniques in generalizing across diverse patient populations.
• To improve the accuracy and reliability of automatic electrocardiogram (ECG) analysis.

Main Methods:

• Utilized phase space reconstruction (RPS) to model heartbeat dynamics.
• Applied Gaussian mixture models (GMM) and binning to RPS for classification with a Bayesian classifier.
• Employed time-delayed neural networks (TDNN) directly on RPS for classification based on prediction error.

Main Results:

• All three proposed methods demonstrated superior performance compared to existing patient-independent approaches.
• The Gaussian mixture model-Bayes (GMM-Bayes) method achieved the highest classification accuracy at 92.5%.
• Phase space reconstruction proved effective in enhancing the generalization capability of heartbeat classification models.

Conclusions:

• Phase space reconstruction offers a promising avenue for improving patient-independent heartbeat classification.
• The GMM-Bayes approach represents a significant advancement in the field, offering high accuracy.
• These findings pave the way for more reliable automated cardiac monitoring systems.