Technical Note Enhanced detection of artifacts in EEG data using higher-order statistics and independent component analysis

Computational Neurobiology Laboratory, Salk Institute for Biological Studies, 10010 N. Torrey Pines Road, La Jolla, CA 92107, USA.
NeuroImage (Impact Factor: 6.36). 03/2007; 34(4):1443-9. DOI: 10.1016/j.neuroimage.2006.11.004
Source: PubMed


Detecting artifacts produced in EEG data by muscle activity, eye blinks and electrical noise is a common and important problem in EEG research. It is now widely accepted that independent component analysis (ICA) may be a useful tool for isolating artifacts and/or cortical processes from electroencephalographic (EEG) data. We present results of simulations demonstrating that ICA decomposition, here tested using three popular ICA algorithms, Infomax, SOBI, and FastICA, can allow more sensitive automated detection of small non-brain artifacts than applying the same detection methods directly to the scalp channel data. We tested the upper bound performance of five methods for detecting various types of artifacts by separately optimizing and then applying them to artifact-free EEG data into which we had added simulated artifacts of several types, ranging in size from thirty times smaller (-50 dB) to the size of the EEG data themselves (0 dB). Of the methods tested, those involving spectral thresholding were most sensitive. Except for muscle artifact detection where we found no gain of using ICA, all methods proved more sensitive when applied to the ICA-decomposed data than applied to the raw scalp data: the mean performance for ICA was higher and situated at about two standard deviations away from the performance distribution obtained on raw data. We note that ICA decomposition also allows simple subtraction of artifacts accounted for by single independent components, and/or separate and direct examination of the decomposed non-artifact processes themselves.

Download full-text


Available from: Arnaud Delorme,
1 Follower
39 Reads
  • Source
    • "Infomax maximises joint entropy, i.e. it minimises information which is present at multiple recording sites[44] [39]. Where Infomax uses a parametric approach Fastica estimates the independent components by maximizing the neg-entropy of their distributions[28] [43]. Jade is based on the joint diagonalization of cumulant matrices[46]. "
    [Show description] [Hide description]
    DESCRIPTION: Sleep in ICU-patients is often deteriorated, which can impair their recovery. In order to determine which factors influence quality of sleep, it needs to be classified. Traditional methods for classification of sleep seem to be unreliable in ICU-data. The IDOS-index is a promising novel measure for classification of sleep using certain characteristics of the EEG. It is not known what the effect and extend of EMG-contamination on the IDOS-index is and whether classification of sleep improves when EMG is filtered. EEG-signal of PSG-recordings from the ICU and clinical neurophysiology department are filtered using several algorithms based on independent component analysis. Obtained independent components are selected manually in order to obtain purged EEG. Amuse seems to be the best ICA-algorithm for filtering EMG-contamination in EEG. It is not evident whether scoring sleep using the IDOS-index improves by filtering EMG-contamination. ICA is inconclusive in entirely separating EEG- components from EMG-components in some situations. Not all artefacts were removed, this could give a distorted image about the actual magnitude of the IDOS-index. Manual selection of components is very subjective and labour-intensive. Obtaining large amounts of purged epochs is therefore troublesome, which makes it difficult to successfully assess effects of filtering. Automated selection of components could provide a solution. Since manual selection of components is troublesome, the amount of obtained data was inadequate, therefore the effect of filtering contamination on classification of sleep using the IDOS-index is unknown.
  • Source
    • "Epochs with irregular noise were identified and rejected using a computer algorithm based on abnormal statistical distribution, as well as by inferences from visual inspection (Delorme et al., 2007). "
    [Show abstract] [Hide abstract]
    ABSTRACT: Recent studies have demonstrated neural overlap between resting state activity and self-referential processing. This "rest-self" overlap occurs especially in anterior cortical midline structures like the perigenual anterior cingulate cortex (PACC). However, the exact neuro-temporal and biochemical mechanisms remain to be identified. Therefore, we conducted a combined EEG-MRS study. EEG focused on pre-stimulus (e.g., prior to stimulus presentation or perception) power changes to assess the degree to which those changes can predict subjects' perception (and judgment) of subsequent stimuli as high or low self-related. Magnetic resonance spectroscopy (MRS) measured resting state concentration of glutamate, focusing on PACC. High pre-stimulus (e.g., prior to stimulus presentation or perception) alpha power significantly correlated with both perception of stimuli judged to be highly self-related, and with resting state glutamate concentrations in the PACC. In sum, our results show (i) pre-stimulus (e.g., prior to stimulus presentation or perception) alpha power and resting state glutamate concentration to mediate rest-self overlap which (ii) dispose or incline subjects to assign high degrees of self-relatedness to perceptual stimuli.
    Social neuroscience 07/2015; DOI:10.1080/17470919.2015.1072582 · 2.66 Impact Factor
    • "filtering to remove such unwanted components that are embedded within the EEG signal which otherwise may bias the analysis of the EEG and lead to wrong conclusions [1]. Suitable pre-processing within the BCI system leads to cleaner EEG signals, thereby enhancing the classification results. "
    [Show abstract] [Hide abstract]
    ABSTRACT: The raw EEG signal acquired non-invasively from the sensorimotor cortex during the motor imagery (MI) performed by a brain-computer interface (BCI) user is naturally embedded with noise while the actual noise-free EEG is still unattainable. This paper compares the enhancement in information when filtering these noisy EEG signals while using a Schrodinger wave equation (SWE) based recurrent quantum neural network (RQNN) model and a Savitzky-Golay (SG) filtering model, while investigating over multiple classification techniques on several datasets. The RQNN model is designed to efficiently capture the statistical behavior of the input signal using an unsupervised learning scheme. The algorithm is robust to parametric sensitivity and does not make any a priori assumption about the true signal type or the embedded noise. The performance of both the filtering approaches, investigated for the BCI competition IV 2b dataset as well as the offline datasets on subjects in the BCI laboratory, over multiple classifiers shows that the RQNN can potentially be a flexible technique that can suit different classifiers for real-time EEG signal filtering. The average classification accuracy performance across all the subjects with the RQNN technique is better than the SG (and the unfiltered signal) by approx. 5% (and 7%) and 1% (and 4%) during the training and the evaluation stages respectively.
    Neurocomputing 07/2015; 170. DOI:10.1016/j.neucom.2014.12.114 · 2.08 Impact Factor
Show more