guest ::
| Home > Workflow collections > In process > Methodological approach for estimating nighttime sleep from smartphone activity data |
| Home > Workflow collections > In process > Methodological approach for estimating nighttime sleep from smartphone activity data |
| Abstract | FZJ-2026-05228 |
; ; ;
2026
This record in other databases:
Please use a persistent id in citations: doi:10.1016/j.nsa.2025.105878
Abstract: Sleep disturbances are increasingly recognized as a critical factor in the onset and progression of various neurological and psychiatric conditions. [1],[2]. Given the widespread prevalence of sleep-related disorders and their impact on overall health, continuous and non-invasive sleep monitoring has become a priority in clinical research and practice. In response to this need, our study explores the potential of leveraging everyday smartphone sensor data to estimate sleep duration and quality. We hypothesize a novel algorithm based on passively recorded motion and screen-off activity can estimate nighttime sleep parameters, providing reliable digital biomarkers for health monitoring.We conducted an observational, repeated measures investigation with a pilot cohort of 25 healthy adult participants recruited from the NRW region of Germany. Over a two-week period, participants’ sleep behaviour was continuously monitored using the JTrack Social and EMA+ applications [3]. Passive data capturing smartphone motion and screen-off events were collected alongside active self-reports via ecological momentary assessments (EMA), which included sleep start and end times, sleep quality, and related behavioral measures. The raw data underwent preprocessing and feature engineering to construct the algorithm. Parameter optimization was performed using cross-validation together with Pareto optimization to achieve an optimal balance between high correlation and low data loss. Three distinct parameter sets were evaluated: one dedicated to nightly sleep estimation, one for computing the average sleep duration per individual, and a third that provided a balanced approach between both metrics. Statistical analysis focused on Pearson correlation coefficients, mean absolute error, and root mean square error to quantify the relationship between the algorithm’s estimates and the self-reported sleep metrics.The algorithm demonstrated robust performance in estimating sleep duration. For nightly sleep estimates, the Pearson correlation coefficient was 0.71, while for average sleep duration it reached 0.79. The median data loss over the study period was limited to 60%. Quantitatively, for average sleep, the mean absolute error was 45 minutes, and the mean square error was 53 minutes, whereas for nightly sleep, the mean absolute error was 56 minutes with a root mean square error of 79 minutes. These metrics confirm that the algorithm meets our performance expectations when compared to conventional self-report measures. In addition to these findings, trend analysis revealed that the algorithm can effectively capture variability from night to night and reliably identify patterns of sleep fragmentation, providing further insight into sleep behaviour. The consistency and reliability of the selected parameter sets were confirmed through cross-validation, with the mixed configuration offering particularly balanced performance for continuous sleep monitoring in naturalistic settings.This study validates a novel, smartphone-based algorithm for non-invasive sleep monitoring, which exhibits high concordance with self-reported sleep measures. The findings underscore the potential of digital biomarkers derived from passive sensor data to not only track overall sleep patterns but also to capture meaningful night-to-night trends in sleep behaviour. Our work contributes to the development of scalable and objective methods for both momentary and longitudinal sleep assessments in clinical research, ultimately supporting more precise patient management and tailored intervention strategies.References[1] Rohani DA, Faurholt-Jepsen M, Kessing LV, Bardram JE. Correlations Between Objective Behavioral Features Collected From Mobile and Wearable Devices and Depressive Mood Symptoms in Patients With Affective Disorders: Systematic Review. JMIR MHealth UHealth. 2018 Aug 13;6(8):e165. [2] Wulff K, Gatti S, Wettstein JG, Foster RG. Sleep and circadian rhythm disruption in psychiatric and neurodegenerative disease. Nat Rev Neurosci. 2010 Aug;11(8):589–99. [3] Sahandi Far M, Stolz M, Fischer JM, Eickhoff SB, Dukart J. JTrack: A Digital Biomarker Platform for Remote Monitoring of Daily-Life Behaviour in Health and Disease. Front Public Health. 2021 Nov 19;9:763621.No conflict of interest
; 
; Article Processing Charges ; Clarivate Analytics Master Journal List ; DOAJ Seal ; Emerging Sources Citation Index ; Fees ; PubMed Central ; SCOPUS ; Web of Science Core Collection
|
The record appears in these collections: |
PDF