Performance of a Multisensor Smart Ring to Evaluate Sleep: In-Lab and Home-Based Evaluation Relative to Polysomnography and Actigraphy: Importance of Generalized Versus Personalized Scoring

Study Objectives: Wearable sleep technology has rapidly expanded across the consumer market due to advances in technology and increased interest in personalized sleep assessment to improve health and performance. In this study, we tested the performance of a novel device, alongside other commercial wearables, against in-lab and at-home polysomnography (PSG). Methods: 36 healthy adults were assessed across 77 nights while wearing the Happy Ring, as well as the Actiwatch, Fitbit, Whoop, and Oura Ring devices. Subjects participated in a single night of in-lab PSG and 2 nights of at-home PSG. The Happy Ring includes sensors for skin conductance, movement, heart rate, and skin temperature. Epoch-by-epoch analyses compared the wearable de-vices to both in-lab and at-home PSG. The Happy Ring utilized two machine-learning derived scor-ing algorithms: a generalized algorithm that applied broadly to all users, and a personalized algorithm that adapted to the data of individual subjects. Results: Compared to in-lab PSG, the generalized and personalized algorithms demonstrated good sensitivity (94% and 93%, respectively) and specificity (70% and 83%, respectively). The other wearable devices also demonstrated good sensitivity (89%-94%) but lower specificity (19%-54%), relative to the Happy Ring. Accuracy was 91% for generalized and 92% for personalized algorithms, compared to other devices that ranged from 84%-88%. The generalized algorithm demonstrated an accuracy of 67%, 85%, and 85% for light, deep, and REM sleep, respectively. The personalized algorithm was 81%, 95%, and 92% accurate for light, deep, and REM sleep, re-spectively. Conclusions: The Happy Ring performed well at home and in the lab, especially regarding sleep-wake detection. The personalized algorithm demonstrated improved detection accuracy over the generalized approach and other devices, suggesting that adaptable, dynamic algorithms can enhance sleep detection accuracy.

Specific Windows Search for Multi-Ship and Multi-Scale Wake Detection in SAR Images

Traditional ship identification systems have difficulty in identifying illegal or broken ships, but the wakes generated by ships can be used as a major feature for identification. However, multi-ship and multi-scale wake detection is also a big challenge. This paper combines the geometric and pixel characteristics of ships and their wakes in Synthetic Aperture Radar (SAR) images and proposes a method for multi-ship and multi-scale wake detection. This method first detects the highlight pixel area in the image and then generates specific windows around the centroid, thereby detecting wakes of different sizes in different areas. In addition, all wake components can be located completely based on wake clustering, the statistical features of wake axis pixels can be used to determine the visible length of the wake. Test results on the Gaofen-3 SAR image show the special potential of the method for wake detection.

First Results on Wake Detection in SAR Images by Deep Learning

Spaceborne synthetic aperture radar (SAR) represents a powerful source of data for enhancing maritime domain awareness (MDA). Wakes generated by traveling vessels hold a crucial role in MDA since they can be exploited both for ship route and velocity estimation and as a marker of ship presence. Even if deep learning (DL) has led to an impressive performance boost on a variety of computer vision tasks, its usage for automatic target recognition (ATR) in SAR images to support MDA is still limited to the detection of ships rather than ship wakes. A dataset is presented in this paper and several state-of-the-art object detectors based on convolutional neural networks (CNNs) are tested with different backbones. The dataset, including more than 250 wake chips, is realized by visually inspecting Sentinel-1 images over highly trafficked maritime sites. Extensive experiments are shown to characterize CNNs for the wake detection task. For the first time, a deep-learning approach is implemented to specifically detect ship wakes without any a-priori knowledge or cuing about the location of the vessel that generated the wake. No annotated dataset was available to train deep-learning detectors on this task, which is instead presented in this paper. Moreover, the benchmarks achieved for different detectors point out promising features and weak points of the relevant approaches. Thus, the work also aims at stimulating more research in this promising, but still under-investigated, field.

P062 Sleep parameter validation of a home-based ballistocardiograph sleep tracker

Introduction Consumer home sleep trackers are gaining popularity for objective sleep monitoring. Amongst them, non-wearable devices have little disruption in daily routine and need little maintenance. However, the validity of their sleep outcomes needs further investigation. In this study, the accuracy of the sleep outcomes of EMFIT Quantified Sleep (QS), an unobtrusive and non-wearable ballistocardiograph sleep tracker, was evaluated by comparing it with polysomnography (PSG). Methods 62 sleep lab patients underwent a single clinical PSG and their sleep measures were simultaneously collected through PSG and EMFIT QS. Total Sleep Time (TST), Wake After Sleep Onset (WASO), Sleep Onset Latency (SOL) and average Heart Rate (HR) were compared using paired t-tests and agreement analysed using Bland-Altman plots. Results EMFIT QS data loss occurred in 47% of participants. In the remaining 33 participants (15 females, with mean age of 53.7±16.5), EMFIT QS overestimated TST by 177.5±119.4 minutes (p<0.001) and underestimated WASO by 44.74±68.81 minutes (p<0.001). It accurately measured average resting HR and was able to distinguish SOL with some accuracy. However, the agreement between EMFIT QS and PSG on sleep-wake detection was very low (kappa=0.13, p<0.001). Discussion A consensus between PSG and EMFIT QS was found in SOL and average HR. There was a significant discrepancy and lack of consensus between the two devices in other sleep outcomes. These findings indicate that while EMFIT QS is not a credible alternative to PSG for sleep monitoring in clinical and research settings, consumers may find some benefit from longitudinal monitoring of SOL and HR.

Development of an image processing method for wake meandering studies and its application on data sets from scanning wind lidar and large-eddy simulation

Wake meandering studies require knowledge of the instantaneous wake shape and its evolution. Scanning lidar data are used to identify the wake pattern behind offshore wind turbines but do not immediately reveal the wake shape. The precise detection of the wake shape and centerline helps to build models predicting wake behavior. The conventional Gaussian fit methods are reliable in the near-wake area but lose precision with the distance from the rotor and require good data resolution for an accurate fit. The thresholding methods usually imply a fixed value or manual selection of a threshold, which hinders the wake detection on a large data set. We propose an automatic thresholding method for the wake shape and centerline detection, which is less dependent on the data resolution and can also be applied to the image data. We show that the method performs reasonably well on large-eddy simulation data and apply it to the data set containing lidar measurements of the two wakes. Along with the wake detection method, we use image processing statistics, such as entropy analysis, to filter and classify lidar scans. The image processing method is developed to reduce dependency on the supplementary reference data such as wind speed and direction. We show that the centerline found with the image processing is in a good agreement with the manually detected centerline and the Gaussian fit method. We also discuss a potential application of the method to separate the near and far wakes and to estimate the wake direction.