271 Sleep staging performance of a signal-agnostic cloud-based real-time sleep analytics platform

Introduction The coronavirus pandemic has brought unprecedented changes to the health care system, including sleep medicine. Remote monitoring and telemedicine played a significant role in this shift. We anticipate these changes to continue in the future with internet-connected wearables (ICWs) playing an important role in measuring and managing sleep remotely. As these ICWs measures a small subset of signals traditionally measured during polysomnography (PSG), manual sleep staging becomes non-trivial and sometimes impossible. The ability to do accurate and reliable automatic sleep staging using different modalities of physiological signals remotely is becoming ever more important. Methods The current work seeks to quantify the sleep staging performance of Z3Score-Neo (https://z3score.com, Neurobit Technologies, Singapore), a signal agnostic, cloud-based real-time sleep analytics platform. We tested its staging performance on the CINC open dataset with N=994 subjects using various combinations of signals including Electroencephalogram (EEG), Electrooculogram (EOG), Electromyogram (EMG), and Instantaneous Heart Rate (IHR) derived from Electrocardiogram (ECG). The staging was compared against manual scoring based on PSG. For IHR based staging, N1 and N2 were combined. Results We achieved substantial agreement (all Cohen’s Kappa > 0.7) between automatic and manual staging using various combinations of EEG, EOG and EMG channels with accuracies varying between 81.76% (two central EEGs, one EOG, one EMG), 79.31% (EEG+EOG), 78.73% (EEG only) and 78.09% (one EOG). We achieved moderate agreement (accuracy: 72.8% κ=0.54) with IHR derived from ECG. Conclusion Our results demonstrated the accuracy of a cloud-based sleep analytics platform on an open dataset, using various combinations of ecologically valid physiological signals. EOG and EMG channels can be easily self-administered using sticker-based electrodes and can be added to existing home sleep apnea test (HSAT) kits significantly improving their utility. ICWs are already capable of accurately measuring EEG/EOG (Muse, InteraXon Inc., Toronto, Canada; Dreem band, Dreem, USA) and IHR derived from ECG (Movesense, Suunto, Finland) or photoplethysmogram (Oura Ring, Oura Health Oy, Finland) or through non-contact ballistocardiogram/radio-based measurements (Dozee, Turtle Shell Technologies, India; Sleepiz, Sleepiz AG, Switzerland). Therefore, a well-validated cloud-based staging platform solves a major technological hurdle towards the proliferation of remote monitoring and telehealth in sleep medicine. Support (if any):

Comparative Measurement of the PPG Signal on Different Human Body Positions by Sensors Working in Reflection and Transmission Modes

The paper deals with the photoplethysmographic (PPG) optical sensor usage for non-invasive acquisition of vital information about the cardiovascular system from different parts of the human skin surface. Finger-ring and ear-clip realizations of the transmission-type PPG sensor were tested first. For the next PPG signal recording, the reflection PPG sensor was placed on fingers and on a wrist. PPG signal properties were described by energetic and temporal parameters and their statistical parameters together with determined instantaneous heart rate values. Our final aim was to find conditions, limitations, and recommendations for development of a wearable PPG sensor working in a magnetic field environment.

A novel fatigue detection method for rehabilitation training of upper limb exoskeleton robot using multi-information fusion

The utilization of upper extremity exoskeleton robots has been proved to be a scientifically effective approach for rehabilitation training. In the process of rehabilitation training, it is necessary to detect the fatigue degree during rehabilitation training in order to formulate a reasonable training plan and achieve better training efficiency. Based on the integral value of surface electromyography (sEMG), heart rate variability, and instantaneous heart rate, this article proposes a fatigue judgment method for multi-information fusion. Based on the integral value data, the feature extraction of the bioelectrical signals were implemented separately, then the fatigue recognition was conducted using the decision-level data fusion method. The bioelectrical signal acquisition system of electromyogram signals and electrocardiograph signals was developed for upper limb exoskeleton rehabilitation robot, and the acquisition and processing of electromyogram signals and electrocardiograph signals were completed. Finally, the fuzzy logic controller with instantaneous heart rate, heart rate variability, and surface electromyography signal was designed to judge fatigue degree, including the fuzzy device, fuzzy rule selector, and defuzzifier. The moderate fatigue state data were selected for testing, and the experimental results showed that the error of fatigue judgment is 4.3%, which satisfies the requirements of fatigue judgment.

Author Correction: Deep learning for automated sleep staging using instantaneous heart rate

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Deep learning for automated sleep staging using instantaneous heart rate

Clinical sleep evaluations currently require multimodal data collection and manual review by human experts, making them expensive and unsuitable for longer term studies. Sleep staging using cardiac rhythm is an active area of research because it can be measured much more easily using a wide variety of both medical and consumer-grade devices. In this study, we applied deep learning methods to create an algorithm for automated sleep stage scoring using the instantaneous heart rate (IHR) time series extracted from the electrocardiogram (ECG). We trained and validated an algorithm on over 10,000 nights of data from the Sleep Heart Health Study (SHHS) and Multi-Ethnic Study of Atherosclerosis (MESA). The algorithm has an overall performance of 0.77 accuracy and 0.66 kappa against the reference stages on a held-out portion of the SHHS dataset for classifying every 30 s of sleep into four classes: wake, light sleep, deep sleep, and rapid eye movement (REM). Moreover, we demonstrate that the algorithm generalizes well to an independent dataset of 993 subjects labeled by American Academy of Sleep Medicine (AASM) licensed clinical staff at Massachusetts General Hospital that was not used for training or validation. Finally, we demonstrate that the stages predicted by our algorithm can reproduce previous clinical studies correlating sleep stages with comorbidities such as sleep apnea and hypertension as well as demographics such as age and gender.