[Arrhythmia heartbeats classification based on neighborhood preserving embedding algorithm].

Area of Science:

• Cardiology
• Biomedical Engineering
• Machine Learning

Background:

• Arrhythmia, a common abnormality in cardiac electrical activity, necessitates accurate diagnosis through electrocardiogram (ECG) analysis.
• Automated heartbeats classification is crucial for the clinical diagnosis of various arrhythmias.

Purpose of the Study:

• To propose and evaluate a novel feature extraction method for automatic arrhythmia heartbeats classification.
• To leverage manifold learning for enhanced ECG signal analysis and diagnosis.

Main Methods:

• Utilized the Neighborhood Preserving Embedding (NPE) algorithm for manifold learning to extract low-dimensional features from high-dimensional ECG signals.
• Employed a Support Vector Machine (SVM) classifier for the diagnosis of heartbeats based on the extracted feature vectors.
• Experimentally validated the method on the MIT-BIH arrhythmia database, clustering 14 distinct classes of arrhythmia heartbeats.

Main Results:

• Achieved a high overall classification accuracy of 98.51% for arrhythmia heartbeats.
• Demonstrated the effectiveness of NPE in reducing ECG signal dimensionality while preserving crucial structural information.
• Successfully classified 14 different types of arrhythmia heartbeats.

Conclusions:

• The proposed feature extraction method based on manifold learning (NPE) is effective for automatic arrhythmia heartbeats classification.
• The combination of NPE and SVM offers a robust approach for ECG-based arrhythmia diagnosis.
• This method holds significant potential for improving the accuracy and efficiency of clinical arrhythmia detection.