Automatic classification of multivariate EEGs using an amount of information measure and the eigenvalues of parametric time series model features

Spontaneous variability in gamma dynamics described by a damped harmonic oscillator driven by noise

Circuits of excitatory and inhibitory neurons generate gamma-rhythmic activity (30-80 Hz). Gamma-cycles show spontaneous variability in amplitude and duration. To investigate the mechanisms underlying this variability, we recorded local-field-potentials (LFPs) and spikes from awake macaque V1. We developed a noise-robust method to detect gamma-cycle amplitudes and durations, which showed a weak but positive correlation. This correlation, and the joint amplitude-duration distribution, is well reproduced by a noise-driven damped harmonic oscillator. This model accurately fits LFP power-spectra, is equivalent to a linear, noise-driven E-I circuit, and recapitulates two additional features of gamma: (1) Amplitude-duration correlations decrease with oscillation strength; (2) amplitudes and durations exhibit strong and weak autocorrelations, respectively, depending on oscillation strength. Finally, longer gamma-cycles are associated with stronger spike-synchrony, but lower spike-rates in both (putative) excitatory and inhibitory neurons. In sum, V1 gamma-dynamics are well described by the simplest possible model of gamma: A damped harmonic oscillator driven by noise.

Asymptotic Information Measures Discrimination of Non-Stationary Time Series Based on Wavelet Domain

<p>This article is concerned with the problem of discrimination between two classes of locally stationary time series based on minimum discrimination information. We view the observed signals as realizations of Gaussian locally stationary wavelet (LSW) processes. The asymptotic Kullback - Leibler discrimination information and Chernoff discrimination information are developed as discriminant criteria for LSW processes. The simulation study showed that our procedure performs as well as other procedures and in some cases better than some other classification methods. Applications to classifying real data show the usefulness of our discriminant criteria.</p>

EEG-based investigation of brain connectivity changes in psychotic patients undergoing the primitive expression form of dance therapy: a methodological pilot study

Primitive expression (PE) is a form of dance therapy (DT) that involves an interaction of ethologically and socially based forms which are supplied for re-enactment. There exist very few studies of DT applications including in their protocol the measurement of neurophysiological parameters. The present pilot study investigates the use of the correlation coefficient (ρ) and mutual information (MI), and of novel measures extracted from ρ and MI, on electroencephalographic (EEG) data recorded in patients with schizophrenia while they undergo PE DT, in order to expand the set of neurophysiology-based approaches for quantifying possible DT effects, using parameters that might provide insights about any potential brain connectivity changes in these patients during the PE DT process. Indication is provided for an acute potentiation effect, apparent at late-stage PE DT, on the inter-hemispheric connectivity in frontal areas, as well as for attenuation of the inter-hemispheric connectivity of left frontal and right central areas and for potentiation of the intra-hemispheric connectivity of frontal and central areas, bilaterally, in the transition from early to late-stage PE DT. This pilot study indicates that by using EEG connectivity measures based on ρ and MI, the set of useful neurophysiology-based approaches for quantifying possible DT effects is expanded. In the framework of the present study, the causes of the observed connectivity changes cannot be attributed with certainty to PE DT, but indications are provided that these measures may contribute to a detailed assessment of neurophysiological mechanisms possibly being affected by this therapeutic process.

Quantitative multichannel EEG measure predicting the optimal weaning from ventilator in ICU patients with acute respiratory failure

OBJECTIVE

The objective of this study was to develop a novel quantitative multichannel EEG (qEEG) based analysis method, called Global Field Damping Time (GFDT), in order to detect potential EEG changes of patients admitted to the ICU with acute respiratory failure, and correlate them to the patients' recovery outcome predicting the optimal time-point to disconnect the patient from mechanical ventilation.

METHODS

Twenty-nine adult patients with acute respiratory failure out of 98 admitted to the Intensive Care Unit of Saint Paul General Hospital were enrolled, and among them only 15 completed the study. The patients were classified in 3 groups according to their outcome after 3 months follow-up. The patients were intubated with fraction of inspired oxygen (FiO2) of 100%. Neurological Deficit Scores (NDS) were measured 24 h after intubation to assess patients' neurological condition. Twenty-four hours after patient's intubation, FiO2 was decreased to 40% (weaning session), followed by a 5 min early recovery session, a 5 min recovery 1 session and a 5 min recovery 2 session. EEG recordings were performed during this experimental procedure. Multichannel EEG segments were processed and fitted into a multivariate autoregressive (mAR) model, and single channel EEG segments into a scalar autoregressive (sAR) model. The mAR and the sAR models of arbitrary order p were decomposed into mp and p oscillators and relaxators, respectively. Damping time of each oscillator and each relaxator, and the Global Field Damping Time (GFDT) as a weighted damping time were estimated for both mAR and sAR models.

RESULTS

A statistically significant increase of mAR model's GFDT during the weaning session was observed in the subjects of all groups. Comparing the 3 patients' groups, statistically significant differences for mAR model's GFDT were observed for the weaning and early recovery session. Linear regression analysis between NDS and mean mAR model's GFDT showed statistical significance during weaning session, early recovery session, and recovery 1 session. There was no statistical significance for SaO2 in the regression analysis with NDS. The sAR model's GFDT presented worst results in comparison with the mAR modelling GFDT in the identification of hypoxic conditions during weaning session and in the discrimination of patients with acute respiratory failure according to their neurological outcome.

CONCLUSIONS

Global Field Damping Time as correlated to the patients' neurological outcome appears to be a simple, compact, and substantial novel indicator of cerebral hypoxia and a potential predictor of the optimal time-point to disconnect the patient from the ventilator.

SIGNIFICANCE

Quantitative EEG seems to be an important tool for ICU clinicians assisting them to decide for the patients' optimal time-point to disconnect the patient from the ventilator.

Impulse response function based on multivariate AR model can differentiate focal hemisphere in temporal lobe epilepsy

The purpose of this study is to propose and investigate a new approach for discriminating between focal and non-focal hemispheres in intractable temporal lobe epilepsy, based on applying multivariate time series analysis to the discharge-free background brain activity observed in nocturnal electrocorticogram (ECoG) time series. Five unilateral focal patients and one bilateral focal patient were studied. In order to detect the location of epileptic foci, linear multivariate autoregressive (MAR) models were fitted to the ECoG data; as a new approach for the purpose of summarizing these models in a single relevant parameter, the behavior of the corresponding impulse response functions was studied and described by an attenuation coefficient. In the majority of unilateral focal patients, the averaged attenuation coefficient was found to be almost always significantly larger in the focal hemisphere, as compared to the non-focal hemisphere. Also the amplitude of the fluctuations of the attenuation coefficient was significantly larger in the focal hemisphere. Moreover, in one patient showing a typical regular sleep cycle, the attenuation coefficient in the focal hemisphere tended to be larger during REM sleep and smaller during Non-REM sleep. In the bilateral focal patient, no statistically significant distinction between the hemispheres was found. This study provides encouraging results for new investigations of brain dynamics by multivariate parametric modeling. It opens up the possibility of relating diseases like epilepsy to the properties of inconspicuous background brain dynamics, without the need to record and analyze epileptic seizures or other evidently pathological waveforms.

Estimation of the dynamics of event-related desynchronisation changes in electroencephalograms

A method for the estimation of medium rate transitions of non-stationary electroencephalograms (EEG) is proposed. The method is applicable to such EEG dynamics that are between (a) fast transitions for which segmentation procedures are used and (b) slow transitions for which adaptive filters work properly. The estimation of the transition dynamics is based on a novel time-varying autoregressive model. This model belongs to the class of deterministic regression time-varying autoregressive models and its parametrisation allows only simultaneous transitions in all coefficient evolutions. Data from 22 patients was analysed. The performance of the method is first evaluated with realistic simulations of known transition dynamics and it is shown to be able to track medium-rate transitions. The method is then applied to the estimation of the dynamics of event related desynchronisation. It is shown that the proposed method is able to estimate the transitions which are less apparent, such as from a multi-infarct patient.

Multivariate autoregressive models for classification of spontaneous electroencephalographic signals during mental tasks

This article explores the use of scalar and multivariate autoregressive (AR) models to extract features from the human electroencephalogram (EEG) with which mental tasks can be discriminated. This is part of a larger project to investigate the feasibility of using EEG to allow paralyzed persons to control a device such as a wheelchair. EEG signals from four subjects were recorded while they performed two mental tasks. Quarter-second windows of six-channel EEG were transformed into four different representations: scalar AR model coefficients, multivariate AR coefficients, eigenvalues of a correlation matrix, and the Karhunen-Loève transform of the multivariate AR coefficients. Feature vectors defined by these representations were classified with a standard, feedforward neural network trained via the error backpropagation algorithm. The four representations produced similar results, with the multivariate AR coefficients performing slightly better and more consistently with an average classification accuracy of 91.4% on novel, untrained, EEG signals.

Estimation of event-related synchronization changes by a new TVAR method

The modeling of nonstationary electroencephalogram (EEG) with time-varying autoregressive (TVAR) models is discussed. The classical least squares TVAR approach is modified so that prior assumptions about the signal can be taken into account in an optimal way. The method is then applied to the estimation of event-related synchronization changes in the EEG. The results show that the new approach enables effective estimation of the parameter evolution of the time-varying EEG with better time resolution compared to previous methods. The new method also allows single-trial analysis of the event-related synchronization.

A general statistical framework for frequency-domain analysis of EEG topographic structure

A wide variety of rhythmic electrophysiological phenomena--including driven, induced, and endogenous activities of cortical neuronal masses--lend themselves naturally to analysis using frequency--domain techniques applied to multichannel recordings that discretely sample the overall spatial pattern of the rhythmic activity. For such cases, a large but so far poorly utilized body of statistical theory supports a third major approach to topographic analysis, complementing the more familiar mapping and source-recovery techniques. These methods, many of which have only recently become computationally feasible, collectively provide general solutions to the problem of detecting and characterizing systematic differences that arise--not only in the spatial distribution of the activity, but also in its frequency-dependent between-channel covariance structure--as a function of multiple experimental conditions presented in conformity with any of the conventional experimental designs. This application-oriented tutorial review provides a comprehensive outline of these resources, including: (1) real multivariate analysis of single-channel spectral measures (and measures of between-channel relationships such as coherence and phase), (2) complex multivariate analysis based on multichannel Fourier transforms, and (3) complex multivariate analysis based on multichannel parametric models. Special emphasis is placed on the potential of the multichannel autoregressive model to support EEG (and MEG) studies of perceptual and cognitive processes.

An information flow technique for category related evoked potentials

We report a technique for studying interactions among many subsystems of a biological system. A general mathematical technique is developed for information flow among various subsystems of a system when two or more classes of stimuli are presented to the system. The technique is validated by various simulation studies and then applied to a brain system. The usefulness of the technique is demonstrated for visual event related potentials (ERP's) obtained from human subjects suffering from phobias. The stimuli are briefly flashed words and phrases. The word classes are pleasant, unpleasant, conscious, and unconscious. The conscious class consists of words known by the patient to relate to the problem, whereas the unconscious class of words consists of words related to deep conflicts which are not recognized by the patient. It is demonstrated that information flow is suppressed under supraliminal presentation of the unconscious class, but is strong under subliminal presentation. The technique has the potential of being an objective indicator of conflictual relationships in these patients. The principle of the technique can be applied to any system in which interactions among subsystems are to be analyzed.

Information statistics as applied to social interaction between mother and infant: synergic covariability vs. partial cross-covariability

Behaviour can, in general, not be quantified on an equal-interval scale; therefore, the description of sequential aspects of the behaviour of two interacting individuals can not be formulated on the basis of the usual cross-correlation techniques. If the information content of behaviour, however, is defined analogously to the entropy in a message, this allows a quantification of the contribution of either participant to the interaction process, and the identification of a non-additive (synergic) component in addition to the auto- and cross-covariability terms. Such a framework, based on the concept of transinformation or mutual information, was developed and realised in a suite of three FORTRAN programs. These were then applied in a longitudinal study on the development of early mother-infant interaction. The synergic term appears to increase with age in the first five months of life. This application illustrates the usefulness of information statistics for quantifying the ways in which participants contribute to a process of interaction.

Multivariate analysis of diaphragm EMG power spectral moments

A single derived index of the power spectrum of the diaphragm electromyogram (EMG) has been used in detecting fatigue. Additional information in the EMG could be used to study diaphragm function in other respiratory conditions. Diaphragm EMGs and calculated power spectra at 12 frequencies were measured in normal subjects and patients with severe chronic obstructive pulmonary disease during several respiratory maneuvers both before and after treadmill exercise to dyspnea. The power spectra were characterized by the first five moments. Changes in the EMG were similar when assessed by multivariate analysis of variance of the spectral estimates or of the moments. Factor analysis provided two latent variables that correlated with the first and second moment respectively. The first moment was found to be the most sensitive single discriminant of fatigue and is only slightly improved by adding other information. It is concluded that the first and second moments of the EMG power spectra provide a concise, parsimonious description of the changes in the EMG.

EEG data reduction by means of autoregressive representation and discriminant analysis procedures

A program for automatic evaluation of EEG spectra, providing considerable reduction of data, was devised. Artefacts were eliminated in two steps: first, the longer duration eye movement artefacts were removed by a fast and simple 'moving integral' methods, then occasional spikes were identified by means of a detection function defined in the formalism of the autoregressive (AR) model. The evaluation of power spectra was performed by means of an FFT and autoregressive representation, which made possible the comparison of both methods. The spectra obtained by means of the AR model had much smaller statistical fluctuations and better resolution, enabling us to follow the time changes of the EEG pattern. Another advantage of the autoregressive approach was the parametric description of the signal. This last property appeared to be essential in distinguishing the changes in the EEG pattern. In a drug study the application of the coefficients of the AR model as input parameters in the discriminant analysis, instead of arbitrary chosen frequency bands, brought a significant improvement in distinguishing the effects of the medication. The favourable properties of the AR model are connected with the fact that the above approach fulfils the maximum entropy principle. This means that the method describes in a maximally consistent way the available information and is free from additional assumptions, which is not the case for the FFT estimate.

The dominant direction of interhemispheric EEG changes in the linguistic process

Hemispheric specialization for the linguistic process was studied by the 'entropy analysis' method which distinguishes the direction of information flow for mutually coupled time series. Meaningful and non-meaningful stimuli were presented to 14 subjects. EEGs were recorded from homologous Wernicke's and occipital areas in the two hemispheres. The amount of information flow both from the left to right hemisphere and from the right to left hemisphere was obtained under the 3 conditions (i.e., meaningful, non-meaningful and resting baseline conditions) in each area. The information flow from the left to right hemisphere was significantly larger under the meaningful condition than from the right to left hemisphere in Wernicke's area, but no significant differences in the information flow could be found between the two directions under the non-meaningful condition in Wernicke's or the occipital area. Therefore, the dominant direction of the left to right hemisphere was observed under the meaningful condition in Wernicke's area but not in the occipital (non-speech) area, while no dominant direction was observed under the two control (i.e., non-meaningful and baseline) conditions in either area. These results may suggest that the dominant direction of information flow with the meaningful stimuli is related to linguistic meaningfulness in Wernicke's area.

[Aid to the diagnosis of epilepsy. Codification of EEG tracings for the characterization of detectable non-stationary phenomena]

The aim of automatic EEG analysis is to detect and class the significant events of the signal. For the detection phase an autoregressive model is developed giving rise to non-stationary points. The model used to analyse information in each event goes further than the problem of existence or non-existence. It is possible to understand qualitative aspects regarding the recognition of typical patterns of epilepsy such as 'spike and waves' providing a preliminary classification of the pattern. Such a classification based on sentence coding established from quantities provided from the model leads to a qualitative understanding of the phenomenon. The first results are presented.

Automatic classification of electroencephalograms: Kullback-Leibler nearest neighbor rules

A prototypic problem in screening of electroencephalograms in the automatic classification of stationary electroencephalogram time series is treated here by the Kullback-Leibler nearest neighbor rule approach. In that problem, the category or state of an individual is classified by comparison of his or her electroencephalogram with those taken from other individuals in the alternative categories. The Kullback-Leibler nearest neighbor classification rules yield a statistically reliable estimate of the smallest possible probability of electroencephalogram misclassification with a relatively small number of labeled sample electroencephalograms. The automatic classification of anesthesia levels L1 and L3, respectively the anesthesia levels insufficient and sufficient for deep surgery, is treated by machine computation on the electroencephalogram alone.