A Novel, Fast and Efficient Single-Sensor Automatic Sleep-Stage Classification Based on Complementary Cross-Frequency Coupling Estimates

Abstract

ObjectiveLimitations of the manual scoring of polysomnograms, which include data from electroencephalogram (EEG), electrooculogram (EOG), electrocardiogram (ECG) and electromyogram (EMG) channels, have long been recognised. Manual staging is resource-intensive and time-consuming and considerable efforts have to be spent to ensure inter-rater reliability. There is thus great interest in techniques based on signal processing and machine learning for a completely Automatic Sleep Stage Classification (ASSC).MethodsIn this paper, we present a single EEG-sensor ASSC technique based on dynamic reconfiguration of different aspects of cross-frequency coupling (CFC) estimated between predefined frequency pairs over 5s epoch lengths. The proposed analytic scheme is demonstrated using the PhysioNet Sleep European Data Format (EDF) Database using 20 healthy young adults with repeat recordings.ResultsWe achieved very high classification sensitivity, specificity and accuracy of 96.2 ± 2.2%, 94.2 ± 2.3%, and 94.4 ± 2.2% across 20 folds, respectively and high mean F1-score (92%, range 90–94%) when multi-class Bayes Naive classifier was applied.ConclusionsOur method outperformed the accuracy of previous studies on different datasets but also on the same database.SignificanceSingle-sensor ASSC makes the whole methodology appropriate for longitudinal monitoring using wearable EEG in real world and lab-oriented environments.

Joint Classification and Prediction CNN Framework for Automatic Sleep Stage Classification

“…In addition to in-bed parts (i.e. from lights off time to lights on time [46]), many previous studies also included other parts, such as in Supratak et al [22], or even entire recordings, such as in Dimitriadis et al [55] and in Alickovic & Subasi [13], into their experiments. These addon data, which are mainly Wake epochs, often boost the performance as Wake, in general, is easier to be recognized than other sleep stages.…”

“…Notice the large variation in the accuracy rate due to the differences in experimental setup. Top accuracy rates, such as in Aboalayon et al[3], Alickovic & Subasi[13], and Dimitriadis et al[55], are likely biased by nonindependent testing and usage of entire recordings rather than only in-bed data (cf. V-D3 for further detail).…”

An Automatic Sleep Scoring Toolbox: Multi-modality of Polysomnography Signals’ Processing

“…Some studies (Hassan et al, 2015;Hassan and Bhuiyan, 2016b) tried to extract statistical and spectral features from these rhythm waves to perform an automatic sleep scoring. Cross frequency coupling estimated between rhythm waves also showed high classification accuracy (Dimitriadis et al, 2018). Instead of traditional linear features, multiscale entropy and autoregressive models for single-channel EEG were employed in Liang et.al' s study, obtaining a good scoring performance .…”

Automatic sleep scoring: A deep learning architecture for multi-modality time series

“…Scoring methods based on conventional machine-learning were prevalent, which usually included two main components: feature extraction and classification. There were wide varieties of techniques for feature extraction, including but not limited to statistic methods [9], Fourier transforms [10], wavelet analysis [11] and Hilbert transform [12]. These techniques were responsible for describing sleep signals from multiple aspects.…”