Ballistocardiogram of baby during sleep

Generalized Camera-Based Infant Sleep-Wake Monitoring in NICUs: A Multi-Center Clinical Trial

The infant sleep-wake behavior is an essential indicator of physiological and neurological system maturity, the circadian transition of which is important for evaluating the recovery of preterm infants from inadequate physiological function and cognitive disorders. Recently, camera-based infant sleep-wake monitoring has been investigated, but the challenges of generalization caused by variance in infants and clinical environments are not addressed for this application. In this paper, we conducted a multi-center clinical trial at four hospitals to improve the generalization of camera-based infant sleep-wake monitoring. Using the face videos of 64 term and 39 preterm infants recorded in NICUs, we proposed a novel sleep-wake classification strategy, called consistent deep representation constraint (CDRC), that forces the convolutional neural network (CNN) to make consistent predictions for the samples from different conditions but with the same label, to address the variances caused by infants and environments. The clinical validation shows that by using CDRC, all CNN backbones obtain over 85% accuracy, sensitivity, and specificity in both the cross-age and cross-environment experiments, improving the ones without CDRC by almost 15% in all metrics. This demonstrates that by improving the consistency of the deep representation of samples with the same state, we can significantly improve the generalization of infant sleep-wake classification.

Investigating Cardiorespiratory Interaction Using Ballistocardiography and Seismocardiography-A Narrative Review

Ballistocardiography (BCG) and seismocardiography (SCG) are non-invasive techniques used to record the micromovements induced by cardiovascular activity at the body's center of mass and on the chest, respectively. Since their inception, their potential for evaluating cardiovascular health has been studied. However, both BCG and SCG are impacted by respiration, leading to a periodic modulation of these signals. As a result, data processing algorithms have been developed to exclude the respiratory signals, or recording protocols have been designed to limit the respiratory bias. Reviewing the present status of the literature reveals an increasing interest in applying these techniques to extract respiratory information, as well as cardiac information. The possibility of simultaneous monitoring of respiratory and cardiovascular signals via BCG or SCG enables the monitoring of vital signs during activities that require considerable mental concentration, in extreme environments, or during sleep, where data acquisition must occur without introducing recording bias due to irritating monitoring equipment. This work aims to provide a theoretical and practical overview of cardiopulmonary interaction based on BCG and SCG signals. It covers the recent improvements in extracting respiratory signals, computing markers of the cardiorespiratory interaction with practical applications, and investigating sleep breathing disorders, as well as a comparison of different sensors used for these applications. According to the results of this review, recent studies have mainly concentrated on a few domains, especially sleep studies and heart rate variability computation. Even in those instances, the study population is not always large or diversified. Furthermore, BCG and SCG are prone to movement artifacts and are relatively subject dependent. However, the growing tendency toward artificial intelligence may help achieve a more accurate and efficient diagnosis. These encouraging results bring hope that, in the near future, such compact, lightweight BCG and SCG devices will offer a good proxy for the gold standard methods for assessing cardiorespiratory function, with the added benefit of being able to perform measurements in real-world situations, outside of the clinic, and thus decrease costs and time.

A behavioral approach to annotating sleep in infants: Building on the classic framework

In infants, monitoring and assessment of sleep can offer valuable insights into sleep problems and neuro-cognitive development. The gold standard for sleep measurements is polysomnography (PSG), but this is rather obtrusive, and unpractical in non-laboratory situations. Behavioral observations constitute a non-obtrusive, infant-friendly alternative. In the current methodological paper, we describe and validate a behavior-based framework for annotating infant sleep states. For development of the framework, we used existing sleep data from an in-home study with an unobtrusive test setup. Participants were 20 infants with a mean age of 180 days. Framework development was based on Prechtl's method. We added rules and guidelines based on discussions and consent among annotators. Key to using our framework is combining data from several modalities, for example, closely observing the frequency, type, and quality of movements, breaths, and sounds an infant makes, while taking the context into account. For a first validation of the framework, we set up a small study with 14 infants (mean age 171 days), in which they took their day-time nap in a laboratory setting. They were continuously monitored by means of PSG, as well as by the test setup from the in-home study. Recordings were annotated based both on PSG and our framework, and then compared. Data showed that for scoring wake vs. active sleep vs. quiet sleep the framework yields results comparable to PSG with a Cohen's Kappa agreement of ≥0.74. Future work with a larger cohort is necessary for further validating this framework, and with clinical populations for determining whether it can be generalized to these populations as well.

U-Net Neural Network for Heartbeat Detection in Ballistocardiography

Monitoring vital signs of neonates can be harmful and lead to developmental troubles. Ballistocardiography, a contactless heart rate monitoring method, has the potential to reduce this monitoring pain. However, signal processing is uneasy due to noise, inherent physiological variability and artifacts (e.g. respiratory amplitude modulation and body position shifts). We propose a new heartbeat detection method using neural networks to learn this variability. A U-Net model takes thirty-second-long records as inputs and acts like a nonlinear filter. For each record, it outputs the samples probabilities of belonging to IJK segments. A heartbeat detection algorithm finally detects heartbeats from those segments, based on a distance criterion. The U-Net has been trained on 30 healthy subjects and tested on 10 healthy subjects, from 8 to 74 years old. Heartbeats have been detected with 92% precision and 80% recall, with possible optimization in the future to achieve better performance.

Video-Based Actigraphy for Monitoring Wake and Sleep in Healthy Infants: A Laboratory Study

Prolonged monitoring of infant sleep is paramount for parents and healthcare professionals for interpreting and evaluating infants’ sleep quality. Wake-sleep patterns are often studied to assess this. Video cameras have received a lot of attention in infant sleep monitoring because they are unobtrusive and easy to use at home. In this paper, we propose a method using motion data detected from infrared video frames (video-based actigraphy) to identify wake and sleep states. The motion, mostly caused by infant body movement, is known to be substantially associated with infant wake and sleep states. Two features were calculated from the video-based actigraphy, and a Bayesian-based linear discriminant classification model was employed to classify the two states. Leave-one-subject-out cross validation was performed to validate our proposed wake and sleep classification model. From a total of 11.6 h of infrared video recordings of 10 healthy term infants in a laboratory pilot study, we achieved a reliable classification performance with a Cohen’s kappa coefficient of 0.733 ± 0.204 (mean ± standard deviation) and an overall accuracy of 92.0% ± 4.6%.