Tim Willemen

An Evaluation of Cardiorespiratory and Movement Features With Respect to Sleep-Stage Classification

Polysomnography (PSG) is considered the gold standard to assess sleep accurately, but it can be expensive, time-consuming, and uncomfortable, specifically in long-term sleep studies. Actigraphy, on the other hand, is both cheap and user-friendly, but depending on the application lacks detail and accuracy. Our aim was to evaluate cardiorespiratory and movement signals in discriminating between wake, rapid-eye-movement (REM), light (N1N2), and deep (N3) sleep. The dataset comprised 85 nights of PSG from a healthy population. Starting from a total of 750 characteristic variables (features), problem-specific subsets of 40 features were forwardly selected using the combination of a wrapper method (Cohens kappa statistic on radial basis function (RBF)-kernel support vector machine (SVM) classifier) and filter method (minimum redundancy maximum relevance criterion on mutual information). Final classification was performed using an RBF-kernel SVM. Non-subject-specific wake versus sleep classification resulted in a Cohens kappa value of 0.695, while REM versus NREM resulted in 0.558 and N3 versus N1N2 in 0.553. The broad pool of initial features gave insight in which features discriminated best between the different classes. The classification results demonstrate the possibility of making long-term sleep monitoring more widely available.

Sleep misperception, EEG characteristics and Autonomic Nervous System activity in primary insomnia: A retrospective study on polysomnographic data

Misperception of Sleep Onset Latency, often found in Primary Insomnia, has been cited to be influenced by hyperarousal, reflected in EEG- and ECG-related indices. The aim of this retrospective study was to examine the association between Central Nervous System (i.e. EEG) and Autonomic Nervous System activity in the Sleep Onset Period and the first NREM sleep cycle in Primary Insomnia (n=17) and healthy controls (n=11). Furthermore, the study examined the influence of elevated EEG and Autonomic Nervous System activity on Stage2 sleep-protective mechanisms (K-complexes and sleep spindles). Confirming previous findings, the Primary Insomnia-group overestimated Sleep Onset Latency and this overestimation was correlated with elevated EEG activity. A higher amount of beta EEG activity during the Sleep Onset Period was correlated with the appearance of K-complexes immediately followed by a sleep spindle in the Primary Insomnia-group. This can be interpreted as an extra attempt to protect sleep continuity or as a failure of the sleep-protective role of the K-complex by fast EEG frequencies following within one second. The strong association found between K-alpha (K-complex within one second followed by 8-12 Hz EEG activity) in Stage2 sleep and a lower parasympathetic Autonomic Nervous System dominance (less high frequency HR) in Slow-wave sleep, further assumes a state of hyperarousal continuing through sleep in Primary Insomnia.

Heart beat detection in multimodal data using automatic relevant signal detection

Accurate R peak detection in the electrocardiogram (ECG) is a well-known and highly explored problem in biomedical signal processing. Although a lot of progress has been made in this area, current methods are still insufficient in the presence of extreme noise and/or artifacts such as loose electrodes. Often, however, not only the ECG is recorded, but multiple signals are simultaneously acquired from the patient. Several of these signals, such as blood pressure, can help to improve the heart beat detection. These signals of interest can be detected automatically by analyzing their power spectral density or by using the available signal type identifiers. Individual peaks from the signals of interest are combined using majority voting, heart beat location estimation and Hjorths mobility of the resulting RR intervals. Both multimodal algorithms showed significant increases in performance of up to 8.65% for noisy multimodal datasets compared to when only the ECG signal is used. A maximal performance of 90.02% was obtained on the hidden test set of the Physionet/Computing in Cardiology Challenge 2014: Robust Detection of Heart Beats in Multimodal Data.

Biomechanics-based active control of bedding support properties and its influence on sleep

Proper body support plays an import role in the recuperation of our body during sleep. Therefore, this study uses an automatically adapting bedding system that optimises spinal alignment throughout the night by altering the stiffness of eight comfort zones. The aim is to investigate the influence of such a dynamic sleep environment on objective and subjective sleep parameters. The bedding system contains 165 sensors that measure mattress indentation. It also includes eight actuators that control the comfort zones. Based on the measured mattress indentation, body movements and posture changes are detected. Control of spinal alignment is established by fitting personalized human models in the measured indentation. A total of 11 normal sleepers participated in this study. Sleep experiments were performed in a sleep laboratory where subjects slept three nights: a first night for adaptation, a reference night and an active support night (in counterbalanced order). Polysomnographic measurements were recorded during the nights, combined with questionnaires aiming at assessing subjective information. Subjective information on sleep quality, daytime quality and perceived number of awakenings shows significant improvements during the active support (ACS) night. Objective results showed a trend towards increased slow wave sleep. On the other hand, it was noticed that % N1-sleep was significantly increased during ACS night, while % N2-sleep was significantly decreased. No prolonged N1 periods were found during or immediately after steering.

Automatic sleep stage classification based on easy to register signals as a validation tool for ergonomic steering in smart bedding systems

Ergonomic sleep studies benefit from long-term monitoring in the home environment to cope with daily variations and habituation effects. Polysomnography allows to asses sleep accurately, but is costly, time-consuming and possibly disturbing for the sleeper. Actigraphy is cheap and user friendly, but for many studies lacks accuracy and detailed information. This proof-of-concept study investigates Least-Squares Support Vector Machines as a tool for automatic sleep stage classification (Wake-N1-Rem to N2-N3 separation), using automatic trainingset-specific filtered features as derived from three easy to register signals, namely heart rate, breathing rate and movement. The algorithms are trained and validated using 20 nights out of a 600 night database from over 100 different healthy persons. Different training and test set strategies were analyzed leading to different results. The more person-specific the training nights to the test nights, the better the classification accuracy as validated against the hypnograms scored by experts from the full polysomnograms. In the limit of complete person-specific training, the accuracy of the algorithm on the test set reached 94%. This means that this algorithm could serve its use in long-term monitoring sleep studies in the home environment, especially when prior person-specific polysomnographic training is performed.

Probabilistic cardiac and respiratory based classification of sleep and apneic events in subjects with sleep apnea

Current clinical standards to assess sleep and its disorders lack either accuracy or user-friendliness. They are therefore difficult to use in cost-effective population-wide screening or long-term objective follow-up after diagnosis. In order to fill this gap, the use of cardiac and respiratory information was evaluated for discrimination between different sleep stages, and for detection of apneic breathing. Alternative probabilistic visual representations were also presented, referred to as the hypnocorrogram and apneacorrogram. Analysis was performed on the UCD sleep apnea database, available on Physionet. The presence of apneic events proved to have a significant impact on the performance of a cardiac and respiratory based algorithm for sleep stage classification. WAKE versus SLEEP discrimination resulted in a kappa value of κ = 0.0439, while REM versus NREM resulted in κ = 0.298 and light sleep (N1N2) versus deep sleep (N3) in κ = 0.339. The high proportion of hypopneic events led to poor detection of apneic breathing, resulting in a kappa value of κ = 0.272. While the probabilistic representations allow to put classifier output in perspective, further improvements would be necessary to make the classifier reliable for use on patients with sleep apnea.

Characterization of the respiratory and heart beat signal from an air pressure-based ballistocardiographic setup

Off-body detection of respiratory and cardiac activity presents an enormous opportunity for general health, stress and sleep quality monitoring. The presented setup detects the mechanical activity of both heart and lungs by measuring pressure difference fluctuations between two air volumes underneath the chest area of the subject. The registered signals were characterized over four different sleep postures, three different base air pressures within the air volumes and three different mattress top layer materials. Highest signal strength was detected in prone posture for both the respiratory and heart beat signal. Respiratory signal strength was the lowest in supine posture, while heart beat signal strength was lowest for right lateral. Heart beat cycle variability was highest in prone and lowest in supine posture. Increasing the base air pressure caused a reduction in signal amplitude for both the respiratory and the heart beat signal. A visco-elastic poly-urethane foam top layer had significantly higher respiration amplitude compared to high resilient poly-urethane foam and latex foam. For the heart beat signal, differences between the top layers were small. The authors conclude that, while the influence of the mattress top layer material is small, the base air pressure can be tuned for optimal mechanical transmission from heart and lungs towards the registration setup.