Abstract:Background and objectiveThe recognition of many sleep related pathologies highly relies on an accurate classification of sleep stages. Clinically, sleep stages are usually labelled by sleep experts through visually inspecting the whole-night polysomnography (PSG) recording of patients, wherein electroencephalogram (EEG), electrooculogram (EOG) and electromyogram (EMG) play the dominant role. Developing an automatic sleep staging system based on multi-channel physiological signals could relieve the burden of manual labeling by experts, and obtain reliable and repeatable recognition results as well.MethodsIn this work, we find the correlation between the spatial covariance matrices of multi-channel signals and their corresponding sleep stages. Based on that, we propose two novel sleep stage classification methods based on the features extracted from the covariance matrices of multi-channel signals. Sleep stages are classified using a minimum distance classifier according to their corresponding covariance matrices mapped on Riemannian manifolds. An alternative way to classify these covariance matrices is to represent the features of covariance matrices on the tangent space of Riemannian manifolds and classify them with an ensemble learning classifier. After any of these classification methods, a rule-free refinement process is utilized to further optimize the classification results.ResultsOn the MASS dataset that includes 61 whole-night PSG recordings, both two methods provide satisfactory classification results while the one based on tangent space projection has better performance. On average, an accuracy of 0.812 and a Cohen's Kappa coefficient of 0.722 are obtained under leave-one-subject-out cross validation, using EEG, EOG and EMG signals. Meanwhile, the most effective combinations of EEG channels for sleep staging have been found in this work.ConclusionsThe correlation between spatial covariance matrices of multi-channel signals and their corresponding sleep stages have been found. Features based on that are used for sleep stage classification, and experimental results show the superior performance of proposed methods compared to state-of-the-art works. Results of this work are expected to provide a new vision for dealing with multi-channel or multi-modal signal processing tasks in various applications.

Classification of EEG sleep stage based on Bayesian relevance vector machine

Automatic Classification of Sleep Stages Based on Raw Single-Channel EEG

Robust Sleep Stage Classification with Single-Channel EEG Signals Using Multimodal Decomposition and HMM-based Refinement.

Fast Sleep Stage Classification Using Cascaded Support Vector Machines with Single-Channel EEG Signals

Study on Sleep Staging Based on Support Vector Machines and Feature Selection in Single Channel Electroencephalogram

Evaluation of a Single-Channel EEG-Based Sleep Staging Algorithm.

[Resting-State Electroencephalogram Classification of Patients with Schizophrenia or Depression].

Application of Machine Learning to Sleep Stage Classification

Sleep Staging from the EEG Signal Using Multi-Domain Feature Extraction.

Automatic Sleep Staging Method Based on EEG Signal and Its Optimized Feature Space

Single-Channel Sleep EEG Classification Method Based on LSTM and Hidden Markov Model

Automatic Sleep Staging Algorithm Based on Random Forest and Hidden Markov Model

A Hierarchical Classification Method for Automatic Sleep Scoring Using Multiscale Entropy Features and Proportion Information of Sleep Architecture

An automated sleep staging tool based on simple statistical features of mice electroencephalography (EEG) and electromyography (EMG) data

A comparative study on classification of sleep stage based on EEG signals using feature selection and classification algorithms

A Single-Channel EEG Based Automatic Sleep Stage Classification Method Leveraging Deep One-Dimensional Convolutional Neural Network and Hidden Markov Model.

Development of Automated Sleep Stage Classification System Using Multivariate Projection-Based Fixed Boundary Empirical Wavelet Transform and Entropy Features Extracted from Multichannel EEG Signals

Analysis and Classification of Sleep Stages Based on Common Frequency Pattern from a Single-Channel EEG Signal

Single-channel EEG sleep stage classification based on a streamlined set of statistical features in wavelet domain

Sleep Stage Classification Using Covariance Features of Multi-Channel Physiological Signals on Riemannian Manifolds

Automatic Wake and Deep-Sleep Stage Classification Based on Wigner–Ville Distribution Using a Single Electroencephalogram Signal