An interpretable framework for sleep posture change detection and postural inactivity segmentation using wrist...

Area of Science:

• Biomedical Engineering
• Neuroscience
• Sleep Medicine

Background:

• Sleep posture and movements are vital indicators of neurophysiological health and quality of life.
• Current clinical sleep assessment methods like polysomnography are invasive and resource-intensive.
• Wearable sensor technologies offer less invasive alternatives, but reliability and standardized algorithms remain challenges.

Purpose of the Study:

• To develop and evaluate a comprehensive framework for objective sleep posture change detection and postural inactivity segmentation.
• To utilize clinically relevant joint kinematics measured by a custom wearable sensor.
• To provide an explainable framework for potential clinical monitoring and diagnosis.

Main Methods:

• Development of a comprehensive framework using a custom-made wearable sensor to capture joint kinematics.
• Application of dimension reduction for intuitive 3D visualization of kinematic time series.
• Evaluation of the framework on wrist kinematic data from five healthy participants during simulated sleep.

Main Results:

• The proposed framework achieved up to 99.2% F1-score for sleep posture detection.
• A Pearson's correlation coefficient of 0.96 was achieved for temporal segmentation of postural inactivity.
• The framework demonstrated intuitive 3D visualizations and explainability through dimension reduction.

Conclusions:

• The developed framework enables reliable, objective analysis of sleep posture and movement.
• This technology supports home-based sleep movement analysis for patient-centred longitudinal care.
• The explainable nature of the framework may aid clinical monitoring and diagnosis.