Related Experiment Video
Updated: Nov 22, 2025
Measuring Neural Mechanisms Underlying Sleep-Dependent Memory Consolidation During Naps in Early Childhood
Published on: October 2, 2019
Machine learning sleep duration classification in Preschoolers using waist-worn ActiGraphs
Nicholas Kuzik1, John C Spence1, Valerie Carson1
1Faculty of Kinesiology, Sport, and Recreation, University of Alberta, Edmonton, Alberta, Canada.
Insights
This study developed a machine learning technique to accurately classify sleep and wake in preschool children using waist-worn accelerometers. The method shows high accuracy for sleep duration and nap prediction, aiding in objective sleep assessment.
Area of Science:
- Biomedical Engineering
- Sleep Science
- Pediatric Research
Background:
- Objective sleep assessment in preschool children is challenging.
- Waist-worn accelerometers offer a non-invasive method for sleep monitoring.
- Accurate classification of sleep, nap, and wake states is crucial for pediatric sleep research.
Purpose of the Study:
- To develop and validate a sleep duration classification technique for preschool-aged children using waist-worn ActiGraph accelerometers.
- To compare machine learning models with a simplified formula for sleep prediction accuracy.
- To establish a reliable method for objective sleep analysis in young children.
Main Methods:
- Children (n=89) wore ActiGraph accelerometers for 7 days.
- Ground truth sleep/wake data were determined by visual inspection of accelerometer data and logs.
- Random Forest and Hidden Markov Models (HMM) were employed for classification, alongside a simplified formula.
Main Results:
- The Random Forest and HMM classifiers achieved 96.2% accuracy (Kappa=0.93).
- A simplified formula demonstrated 93.7% accuracy (Kappa=0.87), with high nap prediction (99.8%).
- Machine learning and simplified formulas showed minimal differences compared to ground truth for daily summaries.
Conclusions:
- Machine learning and simplified formulas derived from ActiGraph data exhibit high agreement with visual inspection for classifying sleep and wake in preschool children.
- These methods provide a promising approach for objective sleep assessment in pediatric populations.
- Further validation using polysomnography is recommended to confirm findings.
Objective:
To create a sleep duration classification technique for waist-worn ActiGraph accelerometers in preschool-aged children.
Methods:
Children wore ActiGraph wGT3X-BT accelerometers on their right hip for 7 days (24 h/day). Ground truth nap, sleep, and wake were estimated through visual inspection of accelerometer data, guided by sleep log-sheets and previously published visual inspection heuristics. Raw accelerometer data (30Hz) were used to generate 144 features aggregated to 1-min epochs. Machine learning classification (ie, Random Forest and Hidden Markov Modeling [HMM]) predicted nap, sleep, and wake. A simplified prediction formula was also created using features (n = 10) with the highest mean decrease in Gini index during training of Random Forests, and temporally smoothed with rolling median calculations.
Results:
Children (n = 89, mean age = 4.5 years, 67% boys) contributed >600,000 min of accelerometer data. Overall classification accuracy of the Random Forest and HMM classifier was 96.2% (95%CI: 96.1, 96.2%), with a Kappa score of 0.93. Additionally, overall classification accuracy for the temporally smoothed simplified formula was 93.7% (95%CI: 93.6, 93.7%) with Kappa = 0.87. Nap prediction accuracy was 99.8% for the final machine learning model, and 86.1% for the simplified formula. For participant-level daily summaries, generally small but statistically significant differences were found between machine learning and ground truth behaviour predictions, whereas non-significant differences were found between the simplified formulas and ground truth predictions.
Conclusions:
Predictions for both machine learning and the simplified formula had almost perfect agreement with visual inspection ground truth measurements. Future research is needed to confirm these findings using polysomnography ground truth sleep measurements.
