Fast and effective removal of contamination from scalp electrical recordings.

OBJECTIVE: To present a new, automated and fast artefact-removal approach which significantly reduces the effect of contamination in scalp electrical recordings. METHOD: We used spectral and temporal characteristics of different sources recorded during a typical scalp electrical recording in order to improve a fast and effective artefact removal approach. Our experiments show that correlation coefficient and spectral gradient of brain components differ from artefactual components. We trained two binary support vector machine classifiers such that one separates brain components from muscle components, and the other separates brain components from mains power and environmental components. We compared the performance of the proposed approach with seven currently used alternatives on three datasets, measuring mains power artefact reduction, muscle artefact reduction and retention of brain neurophysiological responses. RESULTS: The proposed approach significantly reduces the main power and muscle contamination from scalp electrical recording without affecting brain neurophysiological responses. None of the competitors outperformed the new approach. CONCLUSIONS: The proposed approach is the best choice for artefact reduction of scalp electrical recordings. Further improvements are possible with improved component analysis algorithms. SIGNIFICANCE: This paper provides a definitive answer to an important question: Which artefact removal algorithm should be used on scalp electrical recordings?

Improved artefact removal from EEG using Canonical Correlation Analysis and spectral slope.

BACKGROUND: Contamination of scalp measurement by tonic muscle artefacts, even in resting positions, is an unavoidable issue in EEG recording. These artefacts add significant energy to the recorded signals, particularly at high frequencies. To enable reliable interpretation of subcortical brain activity, it is necessary to detect and discard this contamination. NEW METHOD: We introduce a new automatic muscle-removal approach based on the traditional Blind Source Separation-Canonical Correlation Analysis (BSS-CCA) method and the spectral slope of its components. We show that CCA-based muscle-removal methods can discriminate between signals with high correlation coefficients (brain, mains artefact) and signals with low correlation coefficients (white noise, muscle). We also show that typical BSS-CCA components are not purely from one source, but are mixtures from multiple sources, limiting the performance of BSS-CCA in artefact removal. We demonstrate, using our paralysis dataset, improved performance using BSS-CCA followed by spectral-slope rejection. RESULT: This muscle removal approach can reduce high-frequency muscle contamination of EEG, especially at peripheral channels, while preserving steady-state brain responses in cognitive tasks. COMPARISON WITH EXISTING METHODS: This approach is automatic and can be applied on any sample of data easily. The results show its performance is comparable with the ICA method in removing muscle contamination and has significantly lower computational complexity. CONCLUSION: We identify limitations of the traditional BSS-CCA approach to artefact removal in EEG, propose and test an extension based on spectral slope that makes it automatic and improves its performance, and results in performance comparable to competitors such as ICA-based artefact removal.

Evaluation of a minimum-norm based beamforming technique, sLORETA, for reducing tonic muscle contamination of EEG at sensor level.

BACKGROUND: Cranial and cervical muscle activity (electromyogram, EMG) contaminates the surface electroencephalogram (EEG) from frequencies below 20 through to frequencies above 100Hz. It is not possible to have a reliable measure of cognitive tasks expressed in EEG at gamma-band frequencies until the muscle contamination is removed. NEW METHOD: In the present work, we introduce a new approach of using a minimum-norm based beamforming technique (sLORETA) to reduce tonic muscle contamination at sensor level. Using a generic volume conduction model of the head, which includes three layers (brain, skull, and scalp), and sLORETA, we estimated time-series of sources distributed within the brain and scalp. The sources within the scalp were considered to be muscle and discarded in forward modelling. RESULT: (1) The method reduced EMG contamination, more strongly at peripheral channels; (2) task-induced cortical activity was retained or revealed after removing putative muscle activity. COMPARISON WITH EXISTING METHODS: This approach can decrease tonic muscle contamination in scalp measurements without relying on time-consuming processing of expensive MRI data. In addition, it is competitive to ICA in muscle reduction and can be reliably applied on any length of recorded data that captures the dynamics of the signals of interest. CONCLUSION: This study suggests that sLORETA can be used as a method to quantitate cranial muscle activity and reduce its contamination at sensor level.

Electroencephalographic correlates of states of concentrative meditation.

Meditative techniques aim for and meditators report states of mental alertness and focus, concurrent with physical and emotional calm. We aimed to determine the electroencephalographic (EEG) correlates of five states of Buddhist concentrative meditation, particularly addressing a correlation with meditative level. We studied 12 meditators and 12 pair-matched meditation-naïve participants using high-resolution scalp-recorded EEG. To maximise reduction of EMG, data were pre-processed using independent component analysis and surface Laplacian transformed data. Two non-meditative and five meditative states were used: resting baseline, mind-wandering, absorptions 1, 2, 3, 4 and 5 (corresponding to four levels of absorption and an absorption with a different object of focus, otherwise equivalent to level 4; these five meditative states produce repeatable, distinctly different experiences for experienced meditators). The experimental protocol required participants to experience the states in the order listed above, followed immediately by the reverse. We then calculated EEG power in standard frequency bands from 1 to 80Hz. We observed decreases of central scalp beta (13-25Hz), and central low gamma (25-48Hz) power in meditators during deeper absorptions. In contrast, we identified increases in frontal midline and temporo-parietal theta power in meditators, again, during deeper absorptions. Alpha activity was increased over all meditative states, not depth-related. This study demonstrates that the subjective experiences of deepening meditation partially correspond to measures of EEG. Our results are in accord with prior studies on non-graded meditative states. These results are also consistent with increased theta correlating with tightness of focus, and reduced beta/gamma with the desynchronization associated with enhanced alertness.

A comparison of independent component analysis algorithms and measures to discriminate between EEG and artifact components.

Independent Component Analysis (ICA) is a powerful statistical tool capable of separating multivariate scalp electrical signals into their additive independent or source components, specifically EEG or electroencephalogram and artifacts. Although ICA is a widely accepted EEG signal processing technique, classification of the recovered independent components (ICs) is still flawed, as current practice still requires subjective human decisions. Here we build on the results from Fitzgibbon et al. [1] to compare three measures and three ICA algorithms. Using EEG data acquired during neuromuscular paralysis, we tested the ability of the measures (spectral slope, peripherality and spatial smoothness) and algorithms (FastICA, Infomax and JADE) to identify components containing EMG. Spatial smoothness showed differentiation between paralysis and pre-paralysis ICs comparable to spectral slope, whereas peripherality showed less differentiation. A combination of the measures showed better differentiation than any measure alone. Furthermore, FastICA provided the best discrimination between muscle-free and muscle-contaminated recordings in the shortest time, suggesting it may be the most suited to EEG applications of the considered algorithms. Spatial smoothness results suggest that a significant number of ICs are mixed, i.e. contain signals from more than one biological source, and so the development of an ICA algorithm that is optimised to produce ICs that are easily classifiable is warranted.

Constitutive spectral EEG peaks in the gamma range: suppressed by sleep, reduced by mental activity and resistant to sensory stimulation.

OBJECTIVE: In a systematic study of gamma activity in neuro-psychiatric disease, we unexpectedly observed distinctive, apparently persistent, electroencephalogram (EEG) spectral peaks in the gamma range (25-100 Hz). Our objective, therefore, was to examine the incidence, distribution and some of the characteristics of these peaks. METHODS: High sample-rate, 128-channel, EEG was recorded in 603 volunteers (510 with neuropsychiatric disorders, 93 controls), whilst performing cognitive tasks, and converted to power spectra. Peaks of spectral power, including in the gamma range, were determined algorithmically for all electrodes. To determine if peaks were stable, 24-h ambulatory recordings were obtained from 16 subjects with peaks. In 10 subjects, steady-state responses to stimuli at peak frequency were compared with off-peak-frequency stimulation to determine if peaks were a feature of underlying network resonances and peaks were evaluated with easy and hard versions of oddball tasks to determine if peaks might be influenced by mental effort. RESULTS: 57% of 603 subjects exhibited peaks >2 dB above trough power at or above 25 Hz. Larger peaks (>5 dB) were present in 13% of subjects. Peaks were distributed widely over the scalp, more frequent centrally. Peaks were present through the day and were suppressed by slow-wave-sleep. Steady-state responses were the same with on- or off-peak sensory stimulation. In contrast, mental effort resulted in reductions in power and frequency of gamma peaks, although the suppression did not correlate with level of effort. CONCLUSIONS: Gamma EEG can be expressed constitutively as concentrations of power in narrow or wide frequency bands that play an, as yet, unknown role in cognitive activity. SIGNIFICANCE: These findings expand the described range of rhythmic EEG phenomena. In particular, in addition to evoked, induced and sustained gamma band activity, gamma activity can be present constitutively in spectral peaks.

Visual Experiences during Paralysis.

RATIONALE: Paralyzed human volunteers (n = 6) participated in several studies the primary one of which required full neuromuscular paralysis while awake. After the primary experiment, while still paralyzed and awake, subjects undertook studies of humor and of attempted eye-movement. The attempted eye-movements tested a central, intentional component to one's internal visual model and are the subject of this report. METHODS: Subjects reclined in a supportive chair and were ventilated after paralysis (cisatracurium, 20 mg intravenously). In illumination, subjects were requested to focus alternately on the faces of investigators standing on the left and the right within peripheral vision. In darkness, subjects were instructed to look away from a point source of light. Subjects were to report their experiences after reversal of paralysis. RESULTS: During attempted eye-movement in illumination, one subject had an illusion of environmental movement but four subjects perceived faces as clearly as if they were in central vision. In darkness, four subjects reported movement of the target light in the direction of attempted eye-movements and three could control the movement of the light at will. CONCLUSION: The hypothesis that internal visual models receive intended ocular-movement-information directly from oculomotor centers is strengthened by this evidence.

Thinking activates EMG in scalp electrical recordings.

OBJECTIVE: Fast electrical rhythms in the gamma range (30-100Hz) in scalp (but not intracranial) recordings are predominantly due to electromyographic (EMG) activity. We hypothesized that increased EMG activity would be augmented by mental tasks in proportion to task difficulty and the requirement of these tasks for motor or visuo-motor output. METHODS: EEG was recorded in 98 subjects whilst performing cognitive tasks and analysed to generate power spectra. In four other subjects, neuromuscular blockade was achieved pharmacologically providing EMG-free spectra of EEG at rest and during mental tasks. RESULTS: In comparison to the paralysed condition, power of scalp electrical recordings in the gamma range varied in distribution, being maximal adjacent to cranial or cervical musculature. There were non-significant changes in mean gamma range activity due to mental tasks in paralysed subjects. In normal subjects, increases in scalp electrical activity were observed during tasks, without relationship to task difficulty, but with tasks involving limb- or eye-movement having higher power. CONCLUSIONS: Electrical rhythms in the gamma frequency range recorded from the scalp are inducible by mental activity and are largely due to EMG un-related to cognitive effort. EMG varies with requirements for somatic or ocular movement more than task difficulty. SIGNIFICANCE: Severe restrictions exist on utilizing scalp recordings for high frequency EEG.

Scalp electrical recording during paralysis: quantitative evidence that EEG frequencies above 20 Hz are contaminated by EMG.

OBJECTIVE: To identify the possible contribution of electromyogram (EMG) to scalp electroencephalogram (EEG) rhythms at rest and induced or evoked by cognitive tasks. METHODS: Scalp EEG recordings were made on two subjects in presence and absence of complete neuromuscular blockade, sparing the dominant arm. The subjects undertook cognitive tasks in both states to allow direct comparison of electrical recordings. RESULTS: EEG rhythms in the paralysed state differed significantly compared with the unparalysed state, with 10- to 200-fold differences in the power of frequencies above 20 Hz during paralysis. CONCLUSIONS: Most of the scalp EEG recording above 20 Hz is of EMG origin. Previous studies measuring gamma EEG need to be re-evaluated. SIGNIFICANCE: This has a significant impact on measurements of gamma rhythms from the scalp EEG in unparalysed humans. It is to be hoped that signal separation methods will be able to rectify this situation.