Atrial Fibrillation Prediction Based on Recurrence Plot and ResNet

Area of Science:

• Cardiology
• Biomedical Engineering
• Signal Processing

Background:

• Atrial fibrillation (AF) is the most common heart arrhythmia, with increasing global prevalence and significant public health impact.
• Early and accurate detection of AF is crucial for effective patient management and preventing complications.

Purpose of the Study:

• To develop and validate a novel approach for predicting atrial fibrillation (AF) using electrocardiogram (ECG) signals.
• To combine Recurrence Plot (RP) techniques with a ResNet architecture for enhanced AF detection.

Main Methods:

• Wavelet filtering was applied to ECG signals for noise reduction.
• Recurrence Plots (RPs) were generated through phase space reconstruction.
• A multi-level chained residual network (ResNet) was employed for AF prediction.

Main Results:

• The proposed method achieved high performance metrics on a custom dataset, including 93.4% accuracy and 96% AUC.
• On a public AF dataset (AFPDB), the method demonstrated superior performance with 97.0% accuracy and 99.7% AUC.
• The approach effectively extracts subtle information from ECGs for accurate AF prediction.

Conclusions:

• The combined RP and ResNet method offers a highly effective and accurate approach for predicting atrial fibrillation.
• This technique shows potential for improving the early diagnosis and management of AF patients.
• The study highlights the capability of advanced signal processing and deep learning in analyzing complex biomedical data.