Interoperable atlases of the human brain.

The last two decades have seen an unprecedented development of human brain mapping approaches at various spatial and temporal scales. Together, these have provided a large fundus of information on many different aspects of the human brain including micro- and macrostructural segregation, regional specialization of function, connectivity, and temporal dynamics. Atlases are central in order to integrate such diverse information in a topographically meaningful way. It is noteworthy, that the brain mapping field has been developed along several major lines such as structure vs. function, postmortem vs. in vivo, individual features of the brain vs. population-based aspects, or slow vs. fast dynamics. In order to understand human brain organization, however, it seems inevitable that these different lines are integrated and combined into a multimodal human brain model. To this aim, we held a workshop to determine the constraints of a multi-modal human brain model that are needed to enable (i) an integration of different spatial and temporal scales and data modalities into a common reference system, and (ii) efficient data exchange and analysis. As detailed in this report, to arrive at fully interoperable atlases of the human brain will still require much work at the frontiers of data acquisition, analysis, and representation. Among them, the latter may provide the most challenging task, in particular when it comes to representing features of vastly different scales of space, time and abstraction. The potential benefits of such endeavor, however, clearly outweigh the problems, as only such kind of multi-modal human brain atlas may provide a starting point from which the complex relationships between structure, function, and connectivity may be explored.

Spatio-temporal correlations from fMRI time series based on the NN-ARx model.

For the purpose of statistical characterization of the spatio-temporal correlation structure of brain functioning from high-dimensional fMRI time series, we introduce an innovation approach. This is based on whitening the data by the Nearest-Neighbors AutoRegressive model with external inputs (NN-ARx). Correlations between the resulting innovations are an extension of the usual correlations, in which mean-correction is carried out by the dynamic NN-ARx model instead of the static, standard linear model for fMRI time series. Measures of dependencies between regions are defined by summarizing correlations among innovations at several time lags over pairs of voxels. Such summarization does not involve averaging the data over each region, which prevents loss of information in case of non-homogeneous regions. Statistical tests based on these measures are elaborated, which allow for assessing the correlation structure in search of connectivity. Results of application of the NN-ARx approach to fMRI data recorded in visual stimuli experiments are shown. Finally, a number of issues related with its potential and limitations are commented.

Latin American Brain Mapping Network (LABMAN).

On March 8, 2008 in Havana, the Latin American Network for Brain Mapping (LABMAN) was created with participants from Argentina, Brazil, Colombia, Cuba and Mexico. The focus of LABMAN is to promote neuroimaging and systems neuroscience in the region through the implementation of training and exchange programs, and to increase public awareness of the Latin American potential to contribute both to basic and applied research in human brain mapping.

Effects of spatial selective attention on the steady-state visual evoked potential in the 20-28 Hz range.

Steady-state visual evoked potentials (SSVEPs) were recorded from the scalp of subjects who attended to a flickering LED display in one visual field while ignoring a similar display (flickering at a different frequency) in the opposite visual field. The flicker frequencies were 20.8 Hz in the left-field display and 27.8 Hz in the right-field display. The SSVEP to the flicker in either field was enhanced in amplitude when attention was directed to its location. The scalp distribution of this SSVEP enhancement was narrowly focused over the posterior scalp contralateral to the visual field of stimulation. A source analysis using Variable Resolution Electromagnetic Tomography (VARETA) indicated that the source current densities for the SSVEP attention effect had a focal origin in the contralateral parieto-occipital cortex.

3D statistical parametric mapping of EEG source spectra by means of variable resolution electromagnetic tomography (VARETA).

This article describes a new method for 3D QEEG tomography in the frequency domain. A variant of Statistical Parametric Mapping is presented for source log spectra. Sources are estimated by means of a Discrete Spline EEG inverse solution known as Variable Resolution Electromagnetic Tomography (VARETA). Anatomical constraints are incorporated by the use of the Montreal Neurological Institute (MNI) probabilistic brain atlas. Efficient methods are developed for frequency domain VARETA in order to estimate the source spectra for the set of 10(3)-10(5) voxels that comprise an EEG/MEG inverse solution. High resolution source Z spectra are then defined with respect to the age dependent mean and standard deviations of each voxel, which are summarized as regression equations calculated from the Cuban EEG normative database. The statistical issues involved are addressed by the use of extreme value statistics. Examples are shown that illustrate the potential clinical utility of the methods herein developed.

A comparison of non-linear non-parametric models for epilepsy data.

EEG spike and wave (SW) activity has been described through a non-parametric stochastic model estimated by the Nadaraya-Watson (NW) method. In this paper the performance of the NW, the local linear polynomial regression and support vector machines (SVM) methods were compared. The noise-free realizations obtained by the NW and SVM methods reproduced SW better than as reported in previous works. The tuning parameters had to be estimated manually. Adding dynamical noise, only the NW method was capable of generating SW similar to training data. The standard deviation of the dynamical noise was estimated by means of the correlation dimension.

Multivariate Box-Cox transformations with applications to neurometric data.

The Box-Cox power transform methodology for achieving Gaussianity is developed for a variety of models useful in neurometric statistical analysis. Algorithms are proposed for estimating the optimal transformations in the univariate and multivariate cases. Their use is briefly illustrated with neurometric data.

Localization error in biomedical imaging.

Measures of false positive (FP) and false negative (FN) localization error are defined to assess the accuracy of diagnostic imaging procedures. These measures involve the weighting of FP and FN pixels in accordance to their distances from the true localization of the lesion and the region detected as abnormal by the classifier. The distance-based localization receiver operating characteristic (DL-ROC) curve is defined to describe the dependence of the FP and FN localization measures on the classifier's decision threshold. A computer system is presented for the analysis of localization experiments according to these concepts. As an illustration, the accuracy of two types of brain electric topographic montage is studied in the localization of brain tumors and sites of stroke.

Feature-selective attention enhances color signals in early visual areas of the human brain.

We used an electrophysiological measure of selective stimulus processing (the steady-state visual evoked potential, SSVEP) to investigate feature-specific attention to color cues. Subjects viewed a display consisting of spatially intermingled red and blue dots that continually shifted their positions at random. The red and blue dots flickered at different frequencies and thereby elicited distinguishable SSVEP signals in the visual cortex. Paying attention selectively to either the red or blue dot population produced an enhanced amplitude of its frequency-tagged SSVEP, which was localized by source modeling to early levels of the visual cortex. A control experiment showed that this selection was based on color rather than flicker frequency cues. This signal amplification of attended color items provides an empirical basis for the rapid identification of feature conjunctions during visual search, as proposed by "guided search" models.

Topographic time-frequency decomposition of the EEG.

Frequency-transformed EEG resting data has been widely used to describe normal and abnormal brain functional states as function of the spectral power in different frequency bands. This has yielded a series of clinically relevant findings. However, by transforming the EEG into the frequency domain, the initially excellent time resolution of time-domain EEG is lost. The topographic time-frequency decomposition is a novel computerized EEG analysis method that combines previously available techniques from time-domain spatial EEG analysis and time-frequency decomposition of single-channel time series. It yields a new, physiologically and statistically plausible topographic time-frequency representation of human multichannel EEG. The original EEG is accounted by the coefficients of a large set of user defined EEG like time-series, which are optimized for maximal spatial smoothness and minimal norm. These coefficients are then reduced to a small number of model scalp field configurations, which vary in intensity as a function of time and frequency. The result is thus a small number of EEG field configurations, each with a corresponding time-frequency (Wigner) plot. The method has several advantages: It does not assume that the data is composed of orthogonal elements, it does not assume stationarity, it produces topographical maps and it allows to include user-defined, specific EEG elements, such as spike and wave patterns. After a formal introduction of the method, several examples are given, which include artificial data and multichannel EEG during different physiological and pathological conditions.

Detection of event related potentials.

A statistical approach is presented which provides efficient procedures to detect both Event Related Potential (ERP) and its spectral structure. Situations where undesirable signal or "artifact" is present, are considered. In these cases, a "noise" sample can be used which complements the insufficient knowledge given for the sample where we expect to detect the ERP. In this approach, Hotelling's T2 statistic for one and two samples arises as a natural detector of ERPs. Under the assumption of stationarity these statistics are calculated by approximate expressions in the frequency domain. For Brainstem Auditory Evoked Potentials, ROC curves confirm that the T2 statistic has higher detection rates than various indices proposed in the literature. A frequency decomposition of the T2 statistic yields a succession of complex versions of Student's t statistic that characterize the spectral structure of the ERP. Different assumptions about the recordings of ERP are discussed and several generalizations are suggested.

A parametric model for multichannel EEG spectra.

The EEG is modelled as the superposition of two component processes: the xi and the alpha processes. In the frequency domain, the xi process is always present and appears as a spectral peak with maximum amplitude at very low frequencies, while the alpha process is characterized by a spectral peak with its maximum located in the traditional alpha band (7-13 Hz), and is not necessarily always present. The multivariate properties of EEG spectra are adequately modelled with frequency independent coherence matrices for each process. Multichannel EEG studies reveal interesting properties: (1) the generalized coherence for alpha is much larger than for xi, indicating increased functional coupling for the alpha process; (2) the alpha coherence matrix has reduced dimensionality, possibly related to a small number of generators; (3) xi coherences are zero phase with magnitudes that decrease exponentially with interelectrode distance; and (4) alpha coherences have significant nonzero phase shifts.

Current source density estimation and interpolation based on the spherical harmonic Fourier expansion.

A method for the spatial analysis of EEG and EP data, based on the spherical harmonic Fourier expansion (SHE) of scalp potential measurements, is described. This model provides efficient and accurate formulas for: (1) the computation of the surface Laplacian and (2) the interpolation of electrical potentials, current source densities, test statistics and other derived variables. Physiologically based simulation experiments show that the SHE method gives better estimates of the surface Laplacian than the commonly used finite difference method. Cross-validation studies for the objective comparison of different interpolation methods demonstrate the superiority of the SHE over the commonly used methods based on the weighted (inverse distance) average of the nearest three and four neighbor values.

Model parameters estimation when the evoked potential recordings are affected by a random scale factor.

In many situations an important source of the average evoked potentials (EPs) variability is a random scale factor affecting each recording. As a result, the outcome of any EP detection method may be greatly affected. However, using an appropriate probabilistic model these scale factor can be estimated, and the performance of any available detection index improved by data rescaling. In this paper the Maximum Likelihood Estimators of the waveform of the response and the scale factor affecting both background noise and this waveform are obtained. Also, an iterative algorithm for model parameters estimation is presented and its convergence is examined in a simulation study. The Linear Discriminant function is computed using simulated test data in both situations, before and after rescaling of recordings. The performance of these statistics is evaluated by mean of ROC curves.

The estimation of event related potentials affected by random shifts and scalings.

There is considerable evidence for trial to trial variability of the event related potentials (ERPs) within a given subject's recording. This variability influences the outcome of usual procedures in ERP analysis. Better results may be obtained if the sources of variability are explicitly taken into account in an appropriate model. This paper considers a probabilistic model, the random shift and scaling (RSS) model, where the response is modified by a random time shift and a random scale factor. In addition to this, an additional random scale factor which affects both the response and the background noise is taken into account. This time shift and these scale factors are handled as nuisance parameters. Maximum likelihood and least squares estimators of these parameters and the waveform of response are derived for the RSS model. It is shown that the Woody estimate of the ERP reported in earlier work can be derived by restricting the assumptions for the RSS model. Test statistics for hypotheses on means are obtained for the RSS model and a new type of discriminant function. The usefulness of the method is illustrated by means of simulation studies. Receiver operating characteristic (ROC) curves are used to demonstrate that the new type of discriminant performs better than the usual Fisher's Linear Discriminant.

Comparison of auditory-evoked potential detection methods using signal detection theory.

The performance of statistical evoked-potential detection methods was compared with that of human observers and among themselves by means of receiver operating characteristics (ROC) curves. The test material was a collection of brain stem auditory-evoked responses obtained from 98 infants with 60 and 30 dB nHL clicks. The observers and the statistical methods had to discriminate these responses from control recordings obtained without acoustic stimulation. Although the observers' criteria on different days varied considerably, the discrimination capacity was more stable. The discrimination capacity depended on the observers' experience. The statistical methods tested were the correlation coefficient (CCR), the standard deviation ratio (SDR) and a new method named T2R. The most efficient detection method was T2R. For false-alarm rates of 0.01 the statistical methods were more efficient than the human observers. Signal detection theory is useful for the evaluation of evoked-potential analysis methods.

Intracerebral sources of human auditory-evoked potentials.

Evoked potentials to brief 1,000-Hz tones presented to either the left or the right ear were recorded from 30 electrodes arrayed over the head. These recordings were submitted to two different forms of source analysis: brain electric source analysis (BESA) and variable-resolution electromagnetic tomography (VARETA). Both analyses showed that the dominant intracerebral sources for the late auditory-evoked potentials (50-300 ms) were in the supratemporal plane and lateral temporal lobe contralateral to the ear of stimulation. The analyses also suggested the possibility of additional sources in the frontal lobes.

Invariant reversible QEEG effects of anesthetics.

Continuous recordings of brain electrical activity were obtained from a group of 176 patients throughout surgical procedures using general anesthesia. Artifact-free data from the 19 electrodes of the International 10/20 System were subjected to quantitative analysis of the electroencephalogram (QEEG). Induction was variously accomplished with etomidate, propofol or thiopental. Anesthesia was maintained throughout the procedures by isoflurane, desflurane or sevoflurane (N = 68), total intravenous anesthesia using propofol (N = 49), or nitrous oxide plus narcotics (N = 59). A set of QEEG measures were found which reversibly displayed high heterogeneity of variance between four states as follows: (1) during induction; (2) just after loss of consciousness (LOC); (3) just before return of consciousness (ROC); (4) just after ROC. Homogeneity of variance across all agents within states was found. Topographic statistical probability images were compared between states. At LOC, power increased in all frequency bands in the power spectrum with the exception of a decrease in gamma activity, and there was a marked anteriorization of power. Additionally, a significant change occurred in hemispheric relationships, with prefrontal and frontal regions of each hemisphere becoming more closely coupled, and anterior and posterior regions on each hemisphere, as well as homologous regions between the two hemispheres, uncoupling. All of these changes reversed upon ROC. Variable resolution electromagnetic tomography (VARETA) was performed to localize salient features of power anteriorization in three dimensions. A common set of neuroanatomical regions appeared to be the locus of the most probable generators of the observed EEG changes.

Brain electrical field measurements unaffected by linked earlobes reference.

Recent theoretical analysis supports the possibility that using a linked earlobe reference in EEG studies might appreciatively distort the measured electrical field due to current flow over a low resistance path across the wire joining both ears. Such an effect would invalidate published quantitative EEG norms. Evidence for the balancing effect of this distortion was sought for in the EEG of 4 patients with well localized unilateral lesions, a situation in which this distortion would be most apparent. Statistical tests failed to reveal significant differences between EEGs recorded when ears were linked or unlinked. An analysis of the equivalent circuit reveals that a high skin/electrode impedance effectively makes the linked ear reference behave as an ordinary reference.