Signal-quality-aware multisensor fusion for atrial fibrillation detection.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Wearable devices for healthcare monitoring struggle with inconsistent signal quality due to noise.
• Accurate detection of atrial fibrillation (AF) is crucial for patient management and preventing complications.

Purpose of the Study:

• To introduce a novel multimodal data fusion method for enhancing AF detection accuracy.
• To improve signal reliability and R-R interval precision in continuous monitoring using wearable sensors.

Main Methods:

• A multimodal data fusion technique integrating wavelet coefficients from electrocardiogram, photoplethysmogram, and arterial blood pressure signals.
• Signal quality weighting for integrating diverse physiological data streams.
• Development of a bi-directional long short-term memory network for AF detection based on heart rate and R-R intervals.

Main Results:

• The proposed method significantly improves AF detection accuracy, especially under noisy conditions.
• Accuracy increases by 4.51% with two sensor inputs and 10.92% with three inputs at a -10 dB signal-to-noise ratio compared to single-input methods.
• Demonstrated generalizability and improved performance with an increasing number of sensor inputs.

Conclusions:

• The multimodal data fusion approach enhances the reliability and accuracy of AF detection in wearable healthcare monitoring.
• This method offers a robust solution for continuous AF monitoring, outperforming single-channel techniques in challenging, noisy environments.