Sleep Quality Prediction From Wearable Data Using Deep Learning.

Area of Science:

• Sleep Science
• Digital Health
• Wearable Technology

Background:

• Sleep is crucial for overall health, impacting physical, emotional, and mental well-being.
• Physical activity and sleep quality are interconnected behaviors influencing each other.
• Wearable devices and actigraphy offer opportunities for advanced sleep data analytics and eHealth applications.

Purpose of the Study:

• To assess the feasibility of predicting sleep quality, specifically sleep efficiency (poor or adequate), using physical activity data from wearable devices during awake periods.
• To compare the predictive performance of deep learning models against traditional logistic regression for sleep quality assessment.

Main Methods:

• Utilized actigraphy data from 92 adolescents over one week, focusing on physical activity during awake time for prediction.
• Developed predictive models using traditional logistic regression and advanced deep learning techniques, including Multilayer Perceptron (MLP), Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), and Long Short-Term Memory (LSTM-RNN).

Main Results:

• Deep learning models successfully predicted sleep quality (sleep efficiency) based on awake-period physical activity data.
• Deep learning methods demonstrated superior performance compared to traditional logistic regression.
• The CNN model achieved the highest specificity and sensitivity, with an Area Under the ROC Curve (AUC) of 0.9449, significantly outperforming logistic regression (0.6463).

Conclusions:

• Deep learning models effectively predict sleep quality using actigraphy data from awake periods.
• These predictive models represent a valuable tool for advancing sleep research.
• The findings support the development of improved electronic health (eHealth) solutions for sleep management and monitoring.