X Xi Long

Single-accelerometer-based daily physical activity classification

In this study, a single tri-axial accelerometer placed on the waist was used to record the acceleration data for human physical activity classification. The data collection involved 24 subjects performing daily real-life activities in a naturalistic environment without researchers’ intervention. For the purpose of assessing customers’ daily energy expenditure, walking, running, cycling, driving, and sports were chosen as target activities for classification. This study compared a Bayesian classification with that of a Decision Tree based approach. A Bayes classifier has the advantage to be more extensible, requiring little effort in classifier retraining and software update upon further expansion or modification of the target activities. Principal components analysis was applied to remove the correlation among features and to reduce the feature vector dimension. Experiments using leave-one-subject-out and 10-fold cross validation protocols revealed a classification accuracy of ~80%, which was comparable with that obtained by a Decision Tree classifier.

Sleep and Wake Classification With Actigraphy and Respiratory Effort Using Dynamic Warping

This paper proposes the use of dynamic warping (DW) methods for improving automatic sleep and wake classification using actigraphy and respiratory effort. DW is an algorithm that finds an optimal nonlinear alignment between two series allowing scaling and shifting. It is widely used to quantify (dis)similarity between two series. To compare the respiratory effort between sleep and wake states by means of (dis)similarity, we constructed two novel features based on DW. For a given epoch of a respiratory effort recording, the features search for the optimally aligned epoch within the same recording in time and frequency domain. This is expected to yield a high (or low) similarity score when this epoch is sleep (or wake). Since the comparison occurs throughout the entire-night recording of a subject, it may reduce the effects of within- and between-subject variations of the respiratory effort, and thus help discriminate between sleep and wake states. The DW-based features were evaluated using a linear discriminant classifier on a dataset of 15 healthy subjects. Results show that the DW-based features can provide a Cohens Kappa coefficient of agreement κ = 0.59 which is significantly higher than the existing respiratory-based features and is comparable to actigraphy. After combining the actigraphy and the DW-based features, the classifier achieved a κ of 0.66 and an overall accuracy of 95.7%, outperforming an earlier actigraphy- and respiratory-based feature set ( κ = 0.62). The results are also comparable with those obtained using an actigraphy- and cardiorespiratory-based feature set but have the important advantage that they do not require an ECG signal to be recorded.

Analyzing respiratory effort amplitude for automated sleep stage classification

Abstract Respiratory effort has been widely used for objective analysis of human sleep during bedtime. Several features extracted from respiratory effort signal have succeeded in automated sleep stage classification throughout the night such as variability of respiratory frequency, spectral powers in different frequency bands, respiratory regularity and self-similarity. In regard to the respiratory amplitude, it has been found that the respiratory depth is more irregular and the tidal volume is smaller during rapid-eye-movement (REM) sleep than during non-REM (NREM) sleep. However, these physiological properties have not been explicitly elaborated for sleep stage classification. By analyzing the respiratory effort amplitude, we propose a set of 12 novel features that should reflect respiratory depth and volume, respectively. They are expected to help classify sleep stages. Experiments were conducted with a data set of 48 sleepers using a linear discriminant (LD) classifier and classification performance was evaluated by overall accuracy and Cohens Kappa coefficient of agreement. Cross validations (10-fold) show that adding the new features into the existing feature set achieved significantly improved results in classifying wake, REM sleep, light sleep and deep sleep (Kappa of 0.38 and accuracy of 63.8%) and in classifying wake, REM sleep and NREM sleep (Kappa of 0.45 and accuracy of 76.2%). In particular, the incorporation of these new features can help improve deep sleep detection to more extent (with a Kappa coefficient increasing from 0.33 to 0.43). We also revealed that calibrating the respiratory effort signals by means of body movements and performing subject-specific feature normalization can ultimately yield enhanced classification performance.

Sleep stage classification with ECG and respiratory effort

Automatic sleep stage classification with cardiorespiratory signals has attracted increasing attention. In contrast to the traditional manual scoring based on polysomnography, these signals can be measured using advanced unobtrusive techniques that are currently available, promising the application for personal and continuous home sleep monitoring. This paper describes a methodology for classifying wake, rapid-eye-movement (REM) sleep, and non-REM (NREM) light and deep sleep on a 30 s epoch basis. A total of 142 features were extracted from electrocardiogram and thoracic respiratory effort measured with respiratory inductance plethysmography. To improve the quality of these features, subject-specific Z-score normalization and spline smoothing were used to reduce between-subject and within-subject variability. A modified sequential forward selection feature selector procedure was applied, yielding 80 features while preventing the introduction of bias in the estimation of cross-validation performance. PSG data from 48 healthy adults were used to validate our methods. Using a linear discriminant classifier and a ten-fold cross-validation, we achieved a Cohens kappa coefficient of 0.49 and an accuracy of 69% in the classification of wake, REM, light, and deep sleep. These values increased to kappa = 0.56 and accuracy = 80% when the classification problem was reduced to three classes, wake, REM sleep, and NREM sleep.

Measuring dissimilarity between respiratory effort signals based on uniform scaling for sleep staging

Polysomnography (PSG) has been extensively studied for sleep staging, where sleep stages are usually classified as wake, rapid-eye-movement (REM) sleep, or non-REM (NREM) sleep (including light and deep sleep). Respiratory information has been proven to correlate with autonomic nervous activity that is related to sleep stages. For example, it is known that the breathing rate and amplitude during NREM sleep, in particular during deep sleep, are steadier and more regular compared to periods of wakefulness that can be influenced by body movements, conscious control, or other external factors. However, the respiratory morphology has not been well investigated across sleep stages. We thus explore the dissimilarity of respiratory effort with respect to its signal waveform or morphology. The dissimilarity measure is computed between two respiratory effort signal segments with the same number of consecutive breaths using a uniform scaling distance. To capture the property of signal morphological dissimilarity, we propose a novel window-based feature in a framework of sleep staging. Experiments were conducted with a data set of 48 healthy subjects using a linear discriminant classifier and a ten-fold cross validation. It is revealed that this feature can help discriminate between sleep stages, but with an exception of separating wake and REM sleep. When combining the new feature with 26 existing respiratory features, we achieved a Cohens Kappa coefficient of 0.48 for 3-stage classification (wake, REM sleep and NREM sleep) and of 0.41 for 4-stage classification (wake, REM sleep, light sleep and deep sleep), which outperform the results obtained without using this new feature.

Unobtrusive sleep state measurements in preterm infants – A review

Sleep is important for the development of preterm infants. During sleep, neural connections are formed and the development of brain regions is triggered. In general, various rudimentary sleep states can be identified in the preterm infant, namely active sleep (AS), quiet sleep (QS) and intermediate sleep (IS). As the infant develops, sleep states change in length and organization, with these changes as important indicators of brain development. As a result, several methods have been deployed to distinguish between the different preterm infant sleep states, among which polysomnography (PSG) is the most frequently used. However, this method is limited by the use of adhesive electrodes or patches that are attached to the body by numerous cables that can disturb sleep. Given the importance of sleep, this review explores more unobtrusive methods that can identify sleep states without disturbing the infant. To this end, after a brief introduction to preterm sleep states, an analysis of the physiological characteristics associated with the different sleep states is provided and various methods of measuring these physiological characteristics are explored. Finally, the advantages and disadvantages of each of these methods are evaluated and recommendations for neonatal sleep monitoring proposed.

Spectral Boundary Adaptation on Heart Rate Variability for Sleep and Wake Classification

A method of adapting the boundaries when extracting the spectral features from heart rate variability (HRV) for sleep and wake classification is described. HRV series can be derived from electrocardiogram (ECG) signals obtained from single-night polysomnography (PSG) recordings. Conventionally, the HRV spectral features are extracted from the spectrum of an HRV series with fixed boundaries specifying bands of very low frequency (VLF), low frequency (LF), and high frequency (HF). However, because they are fixed, they may fail to accurately reflect certain aspects of autonomic nervous activity which in turn may limit their discriminative power, e.g. in sleep and wake classification. This is in part related to the fact that the sympathetic tone (partially reflected in the LF band) and the respiratory activity (modulated in the HF band) vary over time. In order to minimize the impact of these variations, we adapt the HRV spectral boundaries using time-frequency analysis. Experiments were conducted on a data set acquired from two groups with 15 healthy and 15 insomnia subjects each. Results show that adapting the HRV spectral features significantly increased their discriminative power when classifying sleep and wake. Additionally, this method also provided a significant improvement of the overall classification performance when used in combination with other HRV non-spectral features. Furthermore, compared with the use of actigraphy, the classification performed better when combining it with the HRV features.

Time-frequency analysis of heart rate variability for sleep and wake classification

This paper describes a method to adapt the spectral features extracted from heart rate variability (HRV) for sleep and wake classification. HRV series can be derived from electrocardiogram (ECG) signals obtained from single-night polysomnography (PSG) recordings. Traditionally, the HRV spectral features are extracted from the spectrum of an HRV series with fixed boundaries specifying bands of very low frequency (VLF), low frequency (LF), and high frequency (HF). However, because they are fixed, they may fail to accurately reflect certain aspects of autonomic nervous activity, which in turn may limit their discriminative power when using HRV spectral features, e.g., in sleep and wake classification. This is in part related to the fact that the sympathetic tone (partially reflected in the LF band) and the respiratory activity (modulated in the HF band) will vary over time. In order to minimize the impact of these differences, we adapt the HRV spectral boundaries using time-frequency analysis. Experiments conducted on a dataset acquired from 15 healthy subjects show that the discriminative power of the adapted HRV spectral features are significantly increased when classifying sleep and wake. Additionally, this method also provides a significant improvement of the overall classification performance when used in combination with some other (non-spectral) HRV features.

Using dynamic time warping for sleep and wake discrimination

In previous work, a Linear Discriminant (LD) classifier was used to classify sleep and wake states during single-night polysomnography recordings (PSG) of actigraphy, respiratory effort and electrocardiogram (ECG). In order to improve the sleep-wake discrimination performance and to reduce the number of modalities needed for class discrimination, this study incorporated Dynamic Time Warping (DTW) to help discriminate between sleep and wake states based on actigraphy and respiratory effort signal. DTW quantifies signal similarities manifested in the features extracted from the respiratory effort signal. Experiments were conducted on a dataset acquired from nine healthy subjects, using an LD-based classifier. Leave-one-out cross-validation shows that adding this DTW-based feature to the original actigraphy- and respiratory-based feature set results in an epoch-by-epoch Cohens Kappa agreement coefficient of κ = 0.69 (at an overall accuracy of 95.4%), which represents a significant improvement when compared with the performance obtained without using this feature. Furthermore it is comparable to the result obtained in the previous work which used additional ECG features (κ = 0.70).

Time delay between cardiac and brain activity during sleep transitions

Human sleep consists of wake, rapid-eye-movement (REM) sleep, and non-REM (NREM) sleep that includes light and deep sleep stages. This work investigated the time delay between changes of cardiac and brain activity for sleep transitions. Here, the brain activity was quantified by electroencephalographic (EEG) mean frequency and the cardiac parameters included heart rate, standard deviation of heartbeat intervals, and their low- and high-frequency spectral powers. Using a cross-correlation analysis, we found that the cardiac variations during wake-sleep and NREM sleep transitions preceded the EEG changes by 1–3 min but this was not the case for REM sleep transitions. These important findings can be further used to predict the onset and ending of some sleep stages in an early manner.