Wearanize+: a multimodal dataset for evaluating wearable technologies in sleep research

Area of Science:

• Sleep science and biomedical engineering.
• Consumer sleep technology research.
• Data science in healthcare.

Background:

• Polysomnography (PSG) is the gold standard for sleep research but is resource-intensive.
• Wearable devices offer convenient sleep monitoring but often lack data quality and channel count for reliable analysis.
• Combining data from multiple wearables presents a potential solution to overcome individual device limitations.

Purpose of the Study:

• To introduce and describe the Wearanize+ dataset, a novel collection of concurrent polysomnography and multi-wearable device recordings.
• To facilitate the development of machine learning models for deriving PSG-grade sleep staging from wearable data.
• To explore alternative data modalities for sleep-stage scoring, especially with noisy electroencephalography (EEG) signals.

Main Methods:

• Collected concurrent overnight sleep recordings from 130 healthy participants using PSG and three wearable devices (Zmax headband, Empatica E4 wristband, ActivPAL leg patch).
• Acquired data in a home setting, ensuring ecological validity.
• Documented dataset setup, data collection procedures, and preprocessing steps for reproducibility.

Main Results:

• The Wearanize+ dataset provides a comprehensive resource for validating multi-wearable sleep monitoring approaches.
• The dataset enables research into machine learning algorithms for sleep stage classification using wearable sensor data.
• It supports investigations into alternative data sources for sleep analysis when traditional EEG is compromised.

Conclusions:

• The Wearanize+ dataset is a valuable resource for advancing sleep research using wearable technology.
• This dataset can accelerate the development of reliable, multi-device sleep monitoring systems.
• It paves the way for more accessible and scalable sleep studies and clinical applications.