Agustín Lage-Castellanos

Estimating brain functional connectivity with sparse multivariate autoregression

There is much current interest in identifying the anatomical and functional circuits that are the basis of the brains computations, with hope that functional neuroimaging techniques will allow the in vivo study of these neural processes through the statistical analysis of the time-series they produce. Ideally, the use of techniques such as multivariate autoregressive (MAR) modelling should allow the identification of effective connectivity by combining graphical modelling methods with the concept of Granger causality. Unfortunately, current time-series methods perform well only for the case that the length of the time-series Nt is much larger than p, the number of brain sites studied, which is exactly the reverse of the situation in neuroimaging for which relatively short time-series are measured over thousands of voxels. Methods are introduced for dealing with this situation by using sparse MAR models. These can be estimated in a two-stage process involving (i) penalized regression and (ii) pruning of unlikely connections by means of the local false discovery rate developed by Efron. Extensive simulations were performed with idealized cortical networks having small world topologies and stable dynamics. These show that the detection efficiency of connections of the proposed procedure is quite high. Application of the method to real data was illustrated by the identification of neural circuitry related to emotional processing as measured by BOLD.

White matter architecture rather than cortical surface area correlates with the EEG alpha rhythm

There are few studies on the neuroanatomical determinants of EEG spectral properties that would explain its substantial inter-individual variability in spite of decades of biophysical modeling that predicts this type of relationship. An exception is the negative relation between head size and the spectral position of the alpha peak (P(alpha)) reported in Nunez et al. (1978)-proposed as evidence of the influence of global boundary conditions on slightly damped neocortical waves. Here, we attempt to reexamine this finding by computing the correlations of occipital P(alpha) with various measures of head size and cortical surface area, for 222 subjects from the EEG/MRI database of the Cuban Human Brain Mapping Project. No relation is found (p>0.05). On the other hand, biophysical models also predict that white matter architecture, determining time delays and connectivities, could have an important influence on P(alpha). This led us to explore relations between P(alpha) and DTI fractional anisotropy by means of a multivariate penalized regression. Clusters of voxels with highly significant relations were found. These were positive within the Posterior and Superior Corona Radiata for both hemispheres, supporting biophysical theories predicting that the period of cortico-thalamocortical cycles might be modulating the alpha frequency. Posterior commissural fibers of the Corpus Callosum present the strongest relationships, negative in the inferior part (Splenium), connecting the inferior occipital lobes and positive in the superior part (Isthmus and Tapetum), connecting the superior occipital cortices. We found that white matter architecture rather than neocortical area determines the dynamics of the alpha rhythm.

3D Statistical Parametric Mapping of quiet sleep EEG in the first year of life

This paper extends previously developed 3D SPM for Electrophysiological Source Imaging (Bosch et al., 2001) for neonate EEG. It builds on a prior paper by our group that established age dependent means and standard deviations for the scalp EEG Broad Band Spectral Parameters of children in the first year of life. We now present developmental equations for the narrow band log spectral power of EEG sources, obtained from a sample of 93 normal neonates from age 1 to 10 months in quiet sleep. The main finding from these regressions is that EEG power from 0.78 to 7.5 Hz decreases with age and also for 45-50 Hz. By contrast, there is an increase with age in the frequency band of 19-32 Hz localized to parietal, temporal and occipital areas. Deviations from the norm were analyzed for normal neonates and 17 with brain damage. The diagnostic accuracy (measured by the area under the ROC curve) of EEG source SPM is 0.80, 0.69 for average reference scalp EEG SPM, and 0.48 for Laplacian EEG SPM. This superior performance of 3D SPM over scalp qEEG suggests that it might be a promising approach for the evaluation of brain damage in the first year of life.

The Cuban Twin Registry: Initial Findings and Perspectives

The Cuban Twin Registry is a nation-wide, prospective, population-based twin registry comprising all zygosity types and ages. It was initiated in 2004 to study genetic and environmental contributions to complex diseases with high morbidity and mortality in the Cuban population. The database contains extensive information from 55,400 twin pairs enrolled in the period 2004-2006. Additionally, 2,600 new multiple births have been included from 2007 to date. In the past 4 years, more than 130 studies have been carried out using the registry with a classical genetic epidemiological approach in which concordance rates for monozygotic and dizygotic twins and heritability of various disease traits were estimated. This article summarizes the history, registrys methodology, recent research findings, and future directions of work.

A zero-training algorithm for EEG single-trial classification applied to a face recognition ERP experiment

This paper proposes a machine learning based approach to discriminate between EEG single trials of two experimental conditions in a face recognition experiment. The algorithm works using a single-trial EEG database of multiple subjects and thus does not require subject-specific training data. This approach supports the idea that zero-training classification and on-line detection Brain Computer Interface (BCI) systems are areas with a significant amount of potential.

Recurrent Activation of Neural Circuits during Attention to Global and Local Visual Information

The attentional selection of different hierarchical level within compound (Navon) figures has been studied with event related potentials (ERPs), by controlling the ERPs obtained during attention to the global or the local echelon. These studies, using the canonical Navon figures, have produced contradictory results, with doubts regarding the scalp distribution of the effects. Moreover, the evidence about the temporal evolution of the processing of these two levels is not clear. Here, we unveiled global and local letters at distinct times, which enabled separation of their ERP responses. We combine this approach with the temporal generalization methodology, a novel multivariate technique which facilitates exploring the temporal structure of these ERPs. Opposite lateralization patterns were obtained for the selection negativities generated when attending global and local distracters (D statistics, p < .005), with maxima in right and left occipito-temporal scalp regions, respectively (η2 = .111, p < .01; η2 = .042, p < .04). However, both discrimination negativities elicited when comparing targets and distractors at the global or the local level were lateralized to the left hemisphere (η2 = .25, p < .03 and η2 = .142, p < .05 respectively). Recurrent activation patterns were found for both global and local stimuli, with scalp topographies corresponding to early preparatory stages reemerging during the attentional selection process, thus indicating recursive attentional activation. This implies that selective attention to global and local hierarchical levels recycles similar neural correlates at different time points. These neural correlates appear to be mediated by visual extra-striate areas.

A Framework for Massive Searchlight MVPA Approach

Searchlight analysis (or information mapping) is a recent methodology based on multivoxel pattern analysis (MVPA) that has been used to analyze fMRI data. It consists in repeatedly training and testing the classification algorithm within small regions centered at many voxels in the brain, which implies a high computational cost when the implementation is based on using traditional algorithms in a loop over all searchlights. In this article we present a computational frame-work for developing massive classification algorithms that can be trained and evaluated at all searchlights simultaneously. We first address the Gaussian Naive Bayes (GNB) classifier which is equivalent to the Linear Discriminant Analysis (LDA) when the covariance matrix is diagonal and extend the approach to Quadratic Discriminant Analysis (QDA), also with a diagonal covariance matrix. Additionally, we describe a Logistic Regression (LR) classifier that uses a gradient descent optimization also applied to all searchlights massively. The results show how these algorithms reduce computational time and open new possibilities for searchlight analysis.