Heartbeat Classification by Random Forest With a Novel Context Feature: A Segment Label.

Area of Science:

• Cardiology
• Artificial Intelligence
• Biomedical Engineering

Background:

• Electrocardiograms (ECG) are vital for diagnosing cardiac abnormalities.
• Precise diagnosis of abnormal heartbeats often requires deeper analysis.
• Current methods may need enhancement for complex arrhythmia classification.

Purpose of the Study:

• To design a more accurate heartbeat classification algorithm.
• To assist physicians in identifying specific types of heartbeats.
• To improve the diagnosis of cardiac conditions.

Main Methods:

• Proposed a novel 'segment label' feature extracted using a Convolutional Neural Network.
• Trained a random forest classifier using the new feature alongside traditional ones.
• Validated the method on the MIT-BIH Arrhythmia dataset using inter-patient evaluation.

Main Results:

• Achieved an overall accuracy of 0.96.
• Obtained F1-scores of 0.98 for normal beats, 0.93 for ventricular ectopic beats, and 0.74 for Supra-Ventricular Ectopic Beats (SVEB).
• Demonstrated superior precision (0.76) and sensitivity (0.78) for SVEB compared to state-of-the-art methods.

Conclusions:

• The segment label feature enhances heartbeat classification accuracy, especially for arrhythmias needing contextual rhythm information like SVEB.
• Integrating this algorithm into medical devices can streamline cardiovascular disease diagnosis for physicians.
• The findings offer significant potential for improving clinical implementation in diagnosing conditions such as SVEB.