Nonparametric statistical testing of EEG-and MEG-data

Kognitive Neurophysiologie des Menschen

In Event-Related Potential and Event-Related Field experiments, stimuli - often of several different types - are presented repeatedly, and the subject’s brain response is recorded using Electroencephalography (EEG) or, in the ERF case, Magnetoencephalography (MEG). After removing artifacts and epoching the data, many repetitions per stimulus type are available, which are later usually averaged and compared. At this stage, though, it is no longer possible to establish whether and for which latencies the averaged waveforms are significantly different between stimulus types, nor whether the epochs for a given stimulus type yield significant averages in the first place. A statistical analysis of all individual epochs can provide exactly this information. Topographic Analysis of Variance (TANOVA) and Statistical non-Parametric Mapping performed on the results of Current Density Reconstructions (CDR SnPM) are non-parametric permutation or randomization tests which have previously been pub...

2006

We present a method that enables the use of nonparametric EDF-like statistics for analysing fMRI data that is known to be autocorrelated over time. Analysis and comparison with existing methods like the common General Linear Model solution or a permutation test confirm its validity and usefulness. In addition, our method requires considerably less computation time than a permutation or Bayesian test.

When making statistical comparisons, the temporal dimension of the EEG signal introduces problems. proposed a formally correct statistical approach that deals with these problems: comparing waveforms by counting the number of successive significant univariate tests and then contrasting this number to a well-chosen critical value. However, in the literature, this method is often used inappropriately. Using real EEG data and Monte Carlo simulations, we examined the problems associated with the incorrect use of this approach under circumstances often encountered in the literature. Our results show inflated false-positive or falsenegative rates depending on parameters of the data, including filtering. Our findings suggest that most applications of this method result in an inappropriate family-wise error rate control. Solutions and alternative methods are discussed.

Journal of Neuroscience Methods, 2008

Frequency analyses of EEG data yield large data sets, which are high-dimensional and have to be evaluated statistically without a large number of false positive statements. There exist several methods to deal with this problem in multiple comparisons. Knowing the number of true hypotheses increases the power of some multiple test procedures, however the number of true hypotheses is unknown, in general, and must be estimated. In this paper, we derive two new multiple test procedures by using an upper bound for the number of true hypotheses. Our first procedure controls the generalized family-wise error rate, and thus is an improvement of the step-down procedure of Hommel and Hoffmann [Hommel G.,

Human Brain Mapping, 2005

We assess the suitability of conventional parametric statistics for analyzing oscillatory activity, as measured with electroencephalography/magnetoencephalography (EEG/MEG). The approach we consider is based on narrow-band power time-frequency decompositions of single-trial data. The ensuing power measures have a 2 -distribution. The use of the general linear model (GLM) under normal error assumptions is, therefore, difficult to motivate for these data. This is unfortunate because the GLM plays a central role in classical inference and is the standard estimation and inference framework for neuroimaging data. The key contribution of this work is to show that, in many circumstances, one can appeal to the central limit theorem and assume normality for generative models of power. If this is not appropriate, one can transform the data to render the error terms approximately normal. These considerations allow one to analyze induced and evoked oscillations using standard frameworks like statistical parametric mapping. We establish the validity of parametric tests using synthetic and real data and compare its performance to established nonparametric procedures. Hum Brain Mapp 26: 170 -177, 2005.

NeuroImage, 2005

Simultaneously acquired functional magnetic resonance imaging (fMRI) and electroencephalography (EEG) data hold great promise for localizing the spatial source of epileptiform events detected in the EEG trace. Despite a number of studies applying this method, there has been no independent and systematic validation of the approach. The present study uses a nonparametric method to show that interictal discharges lead to a blood oxygen level dependent (BOLD) response that is significantly different to that obtained by examining random deventsT. We also use this approach to examine the optimization of analysis strategy for detecting these BOLD responses. Two patients with frequent epileptiform events and a healthy control were studied. The fMRI data for each patient were analyzed using a model derived from the timings of the epileptiform events detected on EEG during fMRI scanning. Twenty sets of random pseudoevents were used to generate a null distribution representing the level of chance correlation between the EEG events and fMRI data. The same pseudoevents were applied to control data. We demonstrate that it is possible to detect blood oxygen leveldependent (BOLD) changes related to interictal discharges with specific and independent knowledge about the reliability of this activation. Biologically generated events complicate the fMRI-EEG experiment. Our proposed validation examines whether identified events have an associated BOLD response beyond chance and allows optimization of analysis strategies. This is an important step beyond standard analysis. It informs clinical interpretation because it permits assessment of the reliability of the connection between interictal EEG events and the BOLD response to those events.

2010

This paper presents some considerations about the use of adequate statistical techniques in the framework of the neuroelectromagnetic brain mapping. With the use of advanced EEG/MEG recording setup involving hundred of sensors, the issue of the protection against the type I errors that could occur during the execution of hundred of univariate statistical tests, has gained interest. In the present experiment, we investigated the EEG signals from a mannequin acting as an experimental subject. Data have been collected while performing a neuromarketing experiment and analyzed with state of the art computational tools adopted in specialized literature. Results showed that electric data from the mannequin's head presents statistical significant differences in power spectra during the visualization of a commercial advertising when compared to the power spectra gathered during a documentary, when no adjustments were made on the alpha level of the multiple univariate tests performed. The use of the Bonferroni or Bonferroni-Holm adjustments returned correctly no differences between the signals gathered from the mannequin in the two experimental conditions. An partial sample of recently published literature on different neuroscience journals suggested that at least the 30% of the papers do not use statistical protection for the type I errors. While the occurrence of type I errors could be easily managed with appropriate statistical techniques, the use of such techniques is still not so largely adopted in the literature.

NeuroImage, 2003

Statistical parametric mapping (SPM) analysis of fMRI data requires specifying correctly the temporal and spatial noise covariance structure. This is a difficult if not impossible task. When these assumptions are not satisfied, statistical inference can be invalid or inefficient. Permutation tests are free of strong assumptions on the distribution of signal noise. We propose permutation tests of fMRI data based on experimental randomization of the stimulus sequences. Smooth hemodynamic response curves are estimated using quadratic B-splines. We study fMRI data obtained from event-related potential (ERP) oddball paradigm. Tests of two-tone and three-tone stimulus sequences are conducted to illustrate the application of the proposed method. Comparisons of the SPM and proposed method show that a permutation test is more specific and less susceptible to artifacts.

Human Brain Mapping, 2009

We describe a method to detect brain activation in cortically constrained maps of current density computed from magnetoencephalography (MEG) data using multivariate statistical inference. We apply time-frequency (wavelet) analysis to individual epochs to produce dynamic images of brain signal power on the cerebral cortex in multiple time-frequency bands. We form vector observations by concatenating the power in each frequency band, and fit them into separate multivariate linear models for each time band and cortical location with experimental conditions as predictor variables. The resulting Roy's maximum root statistic maps are thresholded for significance using permutation tests and the maximum statistic approach. A source is considered significant if it exceeds a statistical threshold, which is chosen to control the familywise error rate, or the probability of at least one false positive, across the cortical surface. We compare and evaluate the multivariate approach with existing univariate approaches to time-frequency MEG signal analysis, both on simulated data and experimental data from an MEG visuomotor task study. Our results indicate that the multivariate method is more powerful than the univariate approach in detecting experimental effects when correlations exist between power across frequency bands. We further describe protected F-tests and linear discriminant analysis to identify individual frequencies that contribute significantly to experimental effects. Hum

Neuroimage, 1996

This paper presents a multivariate analysis of evoked responses and their spatiotemporal dynamics as measured with electro-or magnetoencephalography. This analysis uses standard techniques (ManCova) to make possible statistical inference about differential responses, after the data have been transformed using singular value decomposition. The generality of this approach is limited only by the assumptions implicit in the general linear model and can range from simple analyses like Hotelling's T 2 test (in comparing evoked responses among different conditions) to complex analyses of a multivariate regression type (e.g., characterizing the response components associated with a behavioral or psychophysical parameter). To illustrate the technique we have characterized time-dependent changes (both within and between trials) in magnetic fields, evoked by self-paced movements. Our illustrative analysis showed that movement-evoked components were less prone to adaptation than premovement components, suggesting that functionally distinct (preparatory and early executive) biomagnetic signals show differential adaptation.