DPSleep

DPSleep infers sleep onset, duration, and quality from raw accelerometer data collected by wearable devices for longitudinal sleep phenotyping.

Key Features:

• Raw accelerometer data processing: Processes raw accelerometer data from wearable devices to derive sleep timing and quality metrics.
• Stepwise missing-data detection: Employs a stepwise algorithm to identify and handle missing data points in longitudinal records.
• Minute-based spectral power percentiles: Calculates minute-based spectral power percentiles of activity within individuals.
• Iterative sliding windows for sleep episodes: Uses iterative forward- and backward-sliding windows to estimate onset and offset of major sleep episodes.
• Manual quality control adjustments: Provides modules for manual adjustments to derived sleep features for quality control.
• Time zone correction: Corrects for time zone changes to preserve temporal accuracy across longitudinal datasets.
• Actigraphy–smartphone–GPS integration: Integrates actigraphy with smartphone usage and GPS location data to validate sleep timing and detect discrepancies between phone-based and actigraphy measures.
• Support for deep phenotyping: Supports multi-dimensional deep phenotyping by combining actigraphy with other personal electronic device metrics.

Scientific Applications:

• Longitudinal sleep phenotyping in mental health: Enables extended measurement of sleep dynamics associated with mental illness.
• Behavioral variation and disease research: Facilitates exploration of individual differences in behavioral variation related to health and disease using longitudinal actigraphy.
• Validation of sleep timing measures: Validates actigraphy-derived sleep timing by comparison with smartphone usage and GPS data to identify discrepancies.
• Detailed sleep characterization: Provides metrics of sleep onset, duration, and quality for studies of sleep behavior and its correlates.

Methodology:

Computational methods explicitly include a stepwise algorithm for missing-data detection, calculation of minute-based spectral power percentiles, iterative forward- and backward-sliding windows for sleep episode onset/offset estimation, modules for manual quality-control adjustments, time zone correction, and integration of actigraphy with smartphone usage and GPS data to validate and compare sleep timing inferences.