0321 Wearable Device ECG and G-Sensor-based Sleep Stage Evaluation using PSG-based Learning Signal. (27th April 2018)

• Record Type: Journal Article
• Authors: lin, Y; Wang, P; Lin, C; Sadrawi, M; Lin, C; Hsieh, Y; Kuo, C; Chien, J; Haraikawa, K; Abbod, M F; Shieh, J
• Is Part Of: Sleep. Volume 41(2018)Supplement 1
• Page Start: A123
• Page End: A123
• Publication Date: 2018-04-27
• Subjects: Sleep -- Physiological aspects -- Periodicals; Sleep disorders -- Periodicals; Sommeil -- Aspect physiologique -- Périodiques; Sommeil, Troubles du -- Périodiques; Sleep disorders; Sleep -- Physiological aspects; Sleep -- physiological aspects; Sleep Wake Disorders; Psychophysiology; Electronic journals; Periodicals; 616.8498
• DOI: 10.1093/sleep/zsy061.320
• Languages: English
• ISSNs: 0161-8105

Abstract

Abstract: Introduction: Sleep is essential for the human life. Insufficient sleep has negative effect on the cardiac system. Furthermore, a study revealed the correlation between sleep and diabetes. Meanwhile, sleep is also related memory loss. Autonomic nervous system (ANS) is an advanced part of the sleep quality measurement. Heart rate variability (HRV); collected through electrocardiography (ECG); is one of the substantial options for ANS. In order to have a better understanding of sleep behavior, this study aims to classify sleep stages by utilizing features extracted from ECG and G-Sensor signals utilizing wearable device that is based on PSG as the reference. Methods: The dataset was collected from 24 patients during whole sleep time. The PSG machine and single lead BC1 ECG with the 6-axis G-Sensor device (Bio Clothing One, XYZ life BC1, Kinpo Inc., Taipei, Taiwan) are simultaneously utilized for every patient to collect the data. Finally, the ECG and G-Sensor data from the BC1 ECG device are analyzed to evaluate the sleep stage with a reference generated by the PSG. Features extracted from the raw ECG and G-Sensor signals each 30 seconds with a 5-minute sliding window are used in the evaluation. In addition, R-R detection evaluation is performed for the HRV evaluation. Hence, a combination of RR-interval, ECG-derived respiration (EDR) and R-R amplitude differences from ECG with addition to features from the 6-axis G-Sensor are used as the ANN inputs. The output is the sleep stage level. Results: The evaluation of HRV from ECG with G-Sensor signal provides acceptable results. Data from 18 and 6 patients are used for training and testing of the ANN respectively. Results show the best achieved accuracy using 10-fold cross validation ANN are 76.68% and 75.38% for training and testing respectively. Conclusion: Sleep evaluation utilizing wearable device ECG and G-Sensor signals can be applied to train ANN with PSG signal as reference. However, this investigation needs further advanced evaluations for comparison to sleep study results utilizing EEG signal as the input evaluation signal. Support (If Any): Kinpo Electronics, Inc.