Intelligent Detection Method of Atrial Fibrillation by CEPNCC-BiLSTM Based on Long-Term Photoplethysmography Data

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Cardiovascular Diagnostics

Background:

• Atrial fibrillation (AF) is the most common heart arrhythmia, often asymptomatic, posing diagnostic challenges.
• Traditional AF detection methods struggle with the intermittent nature of AF, leading to potential misdiagnoses.
• Existing evaluation systems for AF detection lack comprehensive assessment of both speed and accuracy.

Purpose of the Study:

• To develop an intelligent method for accurate and efficient AF detection and diagnosis.
• To introduce a novel evaluation system, the ET-score, for holistic performance assessment.
• To enhance the detection of AF using photoplethysmography (PPG) signals.

Main Methods:

• Integration of Complementary Ensemble Empirical Mode Decomposition (CEEMD), Power-Normalized Cepstral Coefficients (PNCC), and Bi-directional Long Short-term Memory (BiLSTM) networks.
• Utilizing photoelectric volumetric pulse wave technology for signal acquisition.
• Development of the ET-score, incorporating F-measurement for efficiency and accuracy evaluation.

Main Results:

• The proposed CEPNCC-BiLSTM method demonstrated superior preprocessing efficiency and higher sensitivity in AF detection.
• The method effectively filtered false alarms from non-AF PPG recordings.
• Achieved up to 99.2% accuracy in AF detection within 5 seconds processing time.

Conclusions:

• The CEPNCC-BiLSTM method offers a significant advancement in AF detection accuracy and speed.
• The ET-score provides a more comprehensive evaluation of AF detection system performance.
• This approach holds great potential for real-time, long-term atrial fibrillation monitoring.