Complexity science for sleep stage classification from EEG

Abstract

Automatic sleep stage classification is an important paradigm in computational intelligence and promises considerable advantages to the health care. Most current automated methods require the multiple electroencephalogram (EEG) channels and typically cannot distinguish the S1 sleep stage from EEG. The aim of this study is to revisit automatic sleep stage classification from EEGs using complexity science methods. The proposed method applies fuzzy entropy and permutation entropy as kernels of multi-scale entropy analysis. To account for sleep transition, the preceding and following 30 seconds of epoch data were used for analysis as well as the current epoch. Combining the entropy and spectral edge frequency features extracted from one EEG channel, a multi-class support vector machine (SVM) was able to classify 93.8% of 5 sleep stages for the SleepEDF database [expanded], with the sensitivity of S1 stage was 49.1%. Also, the Kappa's coefficient yielded 0.90, which indicates almost perfect agreement.

Hearables: Automatic Overnight Sleep Monitoring With Standardized In-Ear EEG Sensor

“…We would like to highlight that the scalp-EEG montage (C3-M2 and C4-M1) is the gold standard for sleep medicine, and has been studied and validated over decades. Also, the algorithm applied in this study was originally tested and developed on a publicly available dataset of scalp-EEG [23]. In [23], the classification performance based on two scalp-EEG channels over 61 participants from a publicly available dataset was 88.6 % in accuracy with the corresponding κ of 0.84 in a 5-class sleep stage classification, which was similar to the results in this study -the accuracy and κ were respectively 85.9 % and 0.79 in a 5-class sleep staging using two channels of scalp-EEG over 17 participants.…”

“…These metrics were calculated for each epoch of both ear-EEG and scalp-EEG data. This combination of multi-scale permutation entropy (MSPE) and SEF was proven to be particularly successful in our previous automatic sleep staging work [23] which considered a publicly available overnight Sleep-EDF [expanded] dataset [39]. Based on two channels of scalp-EEGs from 61 participants, the achieved accuracy was 88.6 % with the corresponding kappa coefficient [40] of κ = 0.84 (Almost Perfect Agreement) in the 5-class sleep stage classification.…”

“…The details of MSPE can be found in our earlier related work [23]; the same parameters (i.e. the scale: τ = 20, the embedding dimension: d = 5, and the time delay: L = 1) were used for this study.…”

“…The corresponding automatic sleep stage classification approaches employ various machine learning algorithms and have been validated on both publicly available datasets [22] as well as on proprietary data recorded as part of various research projects [21]. Most studies have used sleep recordings of healthy participants, and the state-of-the-art [23] indicates the possibility of correctly discriminating five sleep stageswake (W), non-Rapid Eye Movements sleep (NREM1-3), and REM -based on a single EEG (channel).…”

Classification of Brainwaves for Sleep Stages by High-Dimensional FFT Features from EEG Signals

“…The dataset is open-source, and many previous researchers have utilized this dataset in sleep scoring research [15,[17][18][19][20][21][22]. Among three available versions of the dataset, we used an expanded version containing 61 recordings from 42 Caucasian male and female subjects.…”