Unsupervised EEG Artifact Detection and Correction

Abstract

Electroencephalography (EEG) is used in the diagnosis, monitoring, and prognostication of many neurological ailments including seizure, coma, sleep disorders, brain injury, and behavioral abnormalities. One of the primary challenges of EEG data is its sensitivity to a breadth of non-stationary noises caused by physiological-, movement-, and equipment-related artifacts. Existing solutions to artifact detection are deficient because they require experts to manually explore and annotate data for artifact segments. Existing solutions to artifact correction or removal are deficient because they assume that the incidence and specific characteristics of artifacts are similar across both subjects and tasks (i.e., “one-size-fits-all”). In this paper, we describe a novel EEG noise-reduction method that uses representation learning to perform patient- and task-specific artifact detection and correction. More specifically, our method extracts 58 clinically relevant features and applies an ensemble of unsupervised outlier detection algorithms to identify EEG artifacts that are unique to a given task and subject. The artifact segments are then passed to a deep encoder-decoder network for unsupervised artifact correction. We compared the performance of classification models trained with and without our method and observed a 10% relative improvement in performance when using our approach. Our method provides a flexible end-to-end unsupervised framework that can be applied to novel EEG data without the need for expert supervision and can be used for a variety of clinical decision tasks, including coma prognostication and degenerative illness detection. By making our method, code, and data publicly available, our work provides a tool that is of both immediate practical utility and may also serve as an important foundation for future efforts in this domain.

Feature Imitating Networks

“…A FIN is a neural network with weights that are initialized to approximate one or more closed-form statistical features. In this paper, we train FINs that approximate five commonly used features in biomedical signal processing: Shannon's Entropy, kurtosis, skewness, fundamental frequency, Mel-frequency cepstral coefficients (mfcc), and regularity [12]. We evaluate the utility of the FINs on three biomedical signal processing experiments, which we describe in Section 4, below.…”

“…Outcome The outcome data for the FINs consisted of closedform feature values calculated on the synthetic signals using SciPy and EEGExtract packages [12].…”

“…For instance, the fundamental frequency of the signal, defined as the lowest periodic frequency of the waveform should be irrelevant, while the Kurtosis is highly relevant to the task [16,17]. Similarly, we expect 'Complexity Features' such as The Cepstrum Coefficients and Shannon's entropy to outperform clinically grounded 'Continuity Features' such as the fundamental frequency or EEG regularity (burst suppression) [12]. Following these theoretical considerations we hypothesize that the MFCC, Entropy, and Kurtosis FINS, as well as a combination of these FINS will outperform the Regularity FINs.…”

“…The data used in this experiment is from an EEG artifact detection data-set [12]. The data contains EEG segments from a 1kHz recording made using 32 electrodes during a passive viewing task.…”

Decoding the Neural Signatures of Valence and Arousal From Portable EEG Headset

“…In this work, the following set of 36 features were extracted from the EEG signal data with the help of EEGExtract library [29] for all three datasets: These features were extracted with a 1s sliding window and no overlap. The extracted features can be categorized into two different groups based on the ability to measure the complexity and continuity of the EEG signal.…”