BAFNet: Bottleneck Attention Based Fusion Network for Sleep Apnea Detection

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Sleep Medicine

Background:

• Sleep apnea (SA) is a prevalent sleep disorder linked to severe health complications, necessitating early diagnosis.
• Portable monitoring (PM) devices offer convenient, at-home sleep condition assessment.
• Electrocardiogram (ECG) signals, easily acquired via PM, present a viable data source for SA detection.

Purpose of the Study:

• To develop and evaluate a novel deep learning model for accurate sleep apnea detection using single-lead ECG signals.
• To investigate the efficacy of a bottleneck attention-based fusion network (BAFNet) for enhanced SA detection performance.
• To assess the potential of BAFNet for application in home sleep apnea testing (HSAT).

Main Methods:

• Proposed a bottleneck attention based fusion network (BAFNet) integrating R-R interval (RRI) and R-peak amplitude (RPA) streams.
• Utilized fully convolutional networks (FCN) with cross-learning for feature representation of RRI/RPA segments.
• Implemented global query generation with bottleneck attention for controlled information flow and a hard sample scheme with k-means clustering for performance enhancement.

Main Results:

• BAFNet achieved competitive and superior results compared to state-of-the-art sleep apnea detection methods.
• The proposed network effectively learned feature representations from RRI and RPA segments.
• The bottleneck attention mechanism facilitated effective information fusion between different signal streams.

Conclusions:

• BAFNet demonstrates significant potential for accurate and non-invasive sleep apnea detection using ECG signals.
• The developed model is well-suited for integration into home sleep apnea testing (HSAT) devices.
• This approach offers a promising avenue for early diagnosis and management of sleep apnea and its associated complications.