Validation of the Munich Actimetry Sleep Detection Algorithm for estimating sleep-wake patterns from activity recordings

Area of Science:

• Chronobiology
• Sleep Science
• Biomedical Engineering

Background:

• Estimating sleep-wake patterns is crucial for understanding circadian rhythms and sleep regularity.
• Wrist-actigraphy offers a non-invasive method for objective sleep monitoring.
• Existing algorithms require validation against gold-standard methods like polysomnography.

Purpose of the Study:

• To evaluate the performance of the Munich Actimetry Sleep Detection Algorithm.
• To validate the algorithm against sleep logs and polysomnography.
• To assess its utility for long-term sleep-wake pattern analysis in field settings.

Main Methods:

• The Munich Actimetry Sleep Detection Algorithm uses a 24-hour threshold and correlation for sleep-wake detection.
• Validation involved sleep logs from adolescents and young adults (N=62) over extended periods.
• Polysomnographic validation was conducted on a clinical sample (N=23) for one night.
• Epoch-by-epoch analyses, Bland-Altman plots, and correlations were used for comparisons.

Main Results:

• Against sleep logs, the algorithm showed 80% sensitivity and 91% specificity.
• Against polysomnography, sensitivity was high (92%) but specificity was low (33%) due to infrequent nighttime wake episodes.
• The algorithm showed high correlation for sleep onset/offset times (r=0.86-0.91).
• It overestimated sleep onset (~21 min) and underestimated wake after sleep onset (~26 min).

Conclusions:

• The Munich Actimetry Sleep Detection Algorithm demonstrates validity for estimating sleep-wake patterns in field studies.
• Its robust performance in daytime and nighttime recordings supports its use in assessing circadian and sleep regularity.
• It serves as an excellent objective alternative to sleep logs for long-term monitoring.