Privacy-Protected Sleep Staging Using Blurred Videos

Area of Science:

• Biomedical Engineering
• Sleep Science
• Computer Vision

Background:

• Cameras offer potential for unobtrusive, long-term home sleep monitoring.
• Privacy concerns limit camera use in sleep studies, necessitating solutions like defocused imaging.
• The efficacy of blurred video for sleep staging, particularly concerning physiological signal integrity, remains unexplored.

Purpose of the Study:

• To evaluate sleep staging performance using videos with varying levels of blurring for privacy protection.
• To investigate the impact of video blurring on motion cues and heart rate variability (HRV) for sleep staging.
• To introduce and assess Cardiopulmonary Coupling (CPC) as a method to enhance sleep staging accuracy with blurred videos.

Main Methods:

• Videos were processed with four levels of post-blurring to simulate privacy-preserving conditions.
• Sleep staging was performed on blurred videos, analyzing accuracy (ACC), kappa, and F1-score.
• Cardiopulmonary Coupling (CPC), utilizing photoplethysmography-derived respiration (PDR), was implemented to improve robustness against blurring effects.

Main Results:

• All blurring levels protected identity but degraded sleep staging accuracy, especially for light sleep stages.
• Video blurring distorted motion cues and high-frequency HRV components, impacting staging performance.
• CPC improved sleep staging accuracy under blurring, with ACC rising from 59.8% to 66.9% under severe blurring.
• CPC demonstrated superior robustness compared to inter-beat interval (IBI) analysis under video blurring.

Conclusions:

• Blurred videos offer a viable, privacy-preserving approach for sleep monitoring.
• Cardiopulmonary Coupling (CPC) effectively mitigates performance degradation caused by video blurring in sleep staging.
• The robustness of CPC, particularly PDR signals, highlights its potential for real-world, privacy-conscious sleep analysis.