Abstract:Abstract Introduction There has been a growing interest in utilizing non-contact sensors such as photoplethysmogram (PPG), ultra-wideband (UWB), millimeter-wave (mmWave) or piezoelectric sensors to measure physiological signals, which are then processed through machine learning for sleep stage analysis. In this study, we conducted an analysis of sleep stages by leveraging data collected from multiple sensors on the subjects and employing machine learning algorithms to combine and interpret the information. Methods mmWave and UWB sensors were placed within approximately 500mm around the subject, and Piezo sensors were positioned inside the subject's mattress. Concurrently, physiological signals from those sensors and from polysomnography (PSG) were obtained. . Key indicators for machine learning were extracted from the sensor data collected from each sensor. Convolutional neural network (CNN) and Transformer machine learning methods were employed to derive sleep stages for each sensor. the sleep stages obtained from each sensor were used as indicators, combined to estimate the final sleep stage. The results were then compared and analyzed against PSG parameters to estimate the accuracy of the algorithm. Results A total of 340 participants (244 males, mean age 44.95± 12.40 y) were enrolled in this study. Data from 200 individuals were utilized to train sleep estimation algorithms for each sensor, while data from 110 individuals were employed to train algorithms that combine results from each sensor to determine sleep. Additional 30 datasets were used to validate these algorithms. The sleep stage accuracy (Accuracy = True epoch/ Total epoch) relative to PSG: mmWave (72.9%), UWB (75.5%), and Piezo (72.4%). Cohen’s Kappa values were mmWave (0.55), UWB (0.62), and Piezo (0.53). The fusion algorithm based on each sensor demonstrated an accuracy of 79.8% and a Cohen’s Kappa value of 0.69 compared to PSG. Conclusion In contrast to previous research relying on single-sensor predictions, our approach harnessed the diverse strengths of different sensor types, conducting a comprehensive analysis of sleep stages by combining data collected from multiple sensors. Support (if any)

Light-weight sleep monitoring: electrode distance matters more than placement for automatic scoring

Automatic Classification of Sleep Stages Based on Raw Single-Channel EEG

Accurate whole-night sleep monitoring with dry-contact ear-EEG

Automatic sleep scoring for real-time monitoring and stimulation in individuals with and without sleep apnea

Current status and prospects of automatic sleep stages scoring: Review

Automatic scoring of sleep stages and cortical arousals using two electrodes on the forehead: validation in healthy adults

Large-Scale Automated Sleep Staging

An ensemble method for improving robustness against the electrode contact problems in automated sleep stage scoring

DeepSleepNet: A Model for Automatic Sleep Stage Scoring Based on Raw Single-Channel EEG

DeepSleepNet-Lite: A Simplified Automatic Sleep Stage Scoring Model with Uncertainty Estimates

Automatic sleep staging based on 24/7 EEG SubQ (UNEEG medical) data displays strong agreement with polysomnography in healthy adults

Transfer Learning for Automatic Sleep Staging Using a Pre-Gelled Electrode Grid

1121 Multi-sensor Fusion Approach for Sleep Stage Analysis: A Non-contact Sensor Study

Automatic sleep staging of EEG signals: recent development, challenges, and future directions

Automatic sleep monitoring using ear-EEG

Automatic artefact detection in single-channel sleep EEG recordings

Expert-level sleep staging using an electrocardiography-only feed-forward neural network

Single-channel EEG sleep staging based on data augmentation and cross-subject discrepancy alleviation

Sleep assessment using EEG-based wearables - A systematic review

Single-channel EOG sleep staging on a heterogeneous cohort of subjects with sleep disorders

Personalizing deep learning models for automatic sleep staging