Automatic artefact detection in single‐channel sleep EEG recordings

A Brief Summary of EEG Artifact Handling

There are various obstacles in the way of use of EEG. Among these, the major obstacles are the artifacts. While some artifacts are avoidable, due to the nature of the EEG techniques there are inevitable artifacts as well. Artifacts can be categorized as internal/physiological or external/non-physiological. The most common internal artifacts are ocular or muscular origins. Internal artifacts are difficult to detect and remove, because they contain signal information as well. For both resting state EEG and ERP studies, artifact handling needs to be carefully carried out in order to retain the maximal signal. Therefore, an effective management of these inevitable artifacts is critical for the EEG based researches. Many researchers from various fields studied this challenging phenomenon and came up with some solutions. However, the developed methods are not well known by the real practitioners of EEG as a tool because of their limited knowledge about these engineering approaches. They still use the traditional visual inspection of the EEG. This work aims to inform the researchers working in the field of EEG about the artifacts and artifact management options available in order to increase the awareness of the available tools such as EEG preprocessing pipelines.

Automated detection of artefacts in neonatal EEG with residual neural networks

BACKGROUND AND OBJECTIVE

To develop a computational algorithm that detects and identifies different artefact types in neonatal electroencephalography (EEG) signals.

METHODS

As part of a larger algorithm, we trained a Residual Deep Neural Network on expert human annotations of EEG recordings from 79 term infants recorded in a neonatal intensive care unit (112 h of 18-channel recording). The network was trained using 10 fold cross validation in Matlab. Artefact types included: device interference, EMG, movement, electrode pop, and non-cortical biological rhythms. Performance was assessed by prediction statistics and further validated on a separate independent dataset of 13 term infants (143 h of 3-channel recording). EEG pre-processing steps, and other post-processing steps such as averaging probability over a temporal window, were also included in the algorithm.

RESULTS

The Residual Deep Neural Network showed high accuracy (95%) when distinguishing periods of clean, artefact-free EEG from any kind of artefact, with a median accuracy for individual patient of 91% (IQR: 81%-96%). The accuracy in identifying the five different types of artefacts ranged from 57%-92%, with electrode pop being the hardest to detect and EMG being the easiest. This reflected the proportion of artefact available in the training dataset. Misclassification as clean was low for each artefact type, ranging from 1%-11%. The detection accuracy was lower on the validation set (87%). We used the algorithm to show that EEG channels located near the vertex were the least susceptible to artefact.

CONCLUSION

Artefacts can be accurately and reliably identified in the neonatal EEG using a deep learning algorithm. Artefact detection algorithms can provide continuous bedside quality assessment and support EEG review by clinicians or analysis algorithms.

An Unsupervised Multichannel Artifact Detection Method for Sleep EEG Based on Riemannian Geometry

In biomedical signal processing, we often face the problem of artifacts that distort the original signals. This concerns also sleep recordings, such as EEG. Artifacts may severely affect or even make impossible visual inspection, as well as automatic processing. Many proposed methods concentrate on certain artifact types. Therefore, artifact-free data are often obtained after sequential application of different methods. Moreover, single-channel approaches must be applied to all channels alternately. The aim of this study is to develop a multichannel artifact detection method for multichannel sleep EEG capable of rejecting different artifact types at once. The inspiration for the study is gained from recent advances in the field of Riemannian geometry. The method we propose is tested on real datasets. The performance of the proposed method is measured by comparing detection results with the expert labeling as a reference and evaluated against a simpler method based on Riemannian geometry that has previously been proposed, as well as against the state-of-the-art method FASTER. The obtained results prove the effectiveness of the proposed method.

Automatic Human Sleep Stage Scoring Using Deep Neural Networks

The classification of sleep stages is the first and an important step in the quantitative analysis of polysomnographic recordings. Sleep stage scoring relies heavily on visual pattern recognition by a human expert and is time consuming and subjective. Thus, there is a need for automatic classification. In this work we developed machine learning algorithms for sleep classification: random forest (RF) classification based on features and artificial neural networks (ANNs) working both with features and raw data. We tested our methods in healthy subjects and in patients. Most algorithms yielded good results comparable to human interrater agreement. Our study revealed that deep neural networks (DNNs) working with raw data performed better than feature-based methods. We also demonstrated that taking the local temporal structure of sleep into account a priori is important. Our results demonstrate the utility of neural network architectures for the classification of sleep.