Paroxysmal atrial fibrillation recognition based on multi-scale wavelet α-entropy

Area of Science:

• Biomedical Engineering
• Cardiology
• Signal Processing

Background:

• Paroxysmal atrial fibrillation (PAF) poses diagnostic challenges.
• Existing methods for PAF recognition have limitations.
• Heart rate variability (HRV) analysis offers potential for improved detection.

Purpose of the Study:

• To develop and validate an effective method for paroxysmal atrial fibrillation (PAF) recognition.
• To utilize wavelet multi-scale α-entropy features of HRV for enhanced PAF detection.
• To distinguish between PAF, distant from PAF, and normal sinus rhythm signals.

Main Methods:

• ECG signal processing for QRS detection and HRV extraction.
• Feature vector extraction using wavelet multi-scale entropy from HRV signals.
• Classification of PAF using a support vector machine (SVM) model.

Main Results:

• The proposed method achieved high performance metrics on the MIT-BIT PAF prediction database.
• Correct rate: 92.18%
• Sensitivity: 94.88%
• Specificity: 89.48%

Conclusions:

• The wavelet multi-scale α-entropy method demonstrates superior performance compared to traditional time, frequency, and non-linear domain analyses.
• The proposed approach shows significant potential for clinical monitoring and treatment of PAF.
• This method offers a robust tool for accurate and reliable PAF detection.