Proceedings - sci.utah.edu

Proceedings

ISBET 2003 – 14th World Congress of the InternationalSociety for Brain Electromagnetic Topography

November 19-23, 2003Organizer : Yoshio Okada

BESA: from raw data to source images and source co-herence. - Michael Scherg*^, Karsten Hoechstetter^,Dieter Weckesser^, Harald Bornfleth^, Nicole Ille^,Patrick Berg^+ (*University of Heidelberg, Germany;+University of Konstanz, Germany; ^MEGIS SoftwareGmbH, Munich, Germany).

The BESA (Brain Electrical Source Analysis) pro-gram provides a large variety of tools for the completeanalysis of EEG and MEG recordings. Using paradigmscripts and a 2D artifact scanning tool, optimized aver-ages can be created for multiple source analysis. A stan-dardized realistic head model, 3D whole head andcortical mapping, and various inverse source algorithmshave been implemented in a highly interactive sourceanalysis module. The latest BESA version 5 can trans-form the surface signals into brain activity using sourcemontages derived from multiple source models. This al-lows to display the on-going EEG, single epochs, and av-erages with much higher spatial resolution. The newSource Coherence Module provides an extremely fastand user-friendly implementation of time-frequencyanalysis based on complex demodulation. Users can cre-ate event-related time-frequency displays of power, am-plitude, event-related (de)synchronization, andcoherence for scalp or source montages. In an er-ror-related negativity EEG experiment a multiple dipolemodel was applied to transform the surface signals intobrain space. The activities of the anterior and posteriorcingulate gyrus, visual cortices, basal temporal lobes,and motor cortices could be separated. Time-frequency

analysis revealed induced gamma-band synchroniza-tion in the visual cortical areas. In a second example, anauditory reaction time experiment, similar multiplesource models could be created using EEG alone or loca-tion seeds based on fMRI BOLD clusters. The regionalsource waveforms revealed the sequential activation ofbilateral auditory cortices (N100 and N150), SMA, motorcortex, cerebellum, visual cortex, and cingulate gyrus.

Model building and hypothesis testing with EMSE®-R.E. Greenblatt (Source Signal Imaging, Inc., SanDiego CA 92102 USA).

Typical steps in the analysis of M/EEG data includedata conditioning, data exploration, model building andhypothesis testing. Source estimation (drawing infer-ences about brain source location, orientation and mag-nitude from extracranial M/EEG) is often an importantstep in model building. Source estimation requires solu-tion of forward (brain to scalp) and inverse (scalp tobrain) problems. Realistic forward models, such as finiteelement (FEM) analysis, may represent subject specificanatomy, derived from MRI, and are computationally ef-ficient. Inverse methods may be classified as local vs.global, overdetermined vs. underdetermined, and in-stantaneous vs. spatiotemporal. In addition, the sourcespace may be constrained using structural and/or func-tional MRI. Lastly, data may be in the time, frequency, ortime-frequency domains. Examples derived from EMSEEEG data analysis will illustrate signal and image pre-processing, source estimation, and statistical hypothesistesting. Strengths and weakness of several forward andinverse methods will be discussed.

Copyright © 2004 Human Sciences Press, Inc.

Brain Topography, Volume 16, Number 3, Spring 2004 (© 2004) 181

CURRY – Multimodal neuroimaging. - M. Wagner(Neuroscan Labs, Hamburg, Germany).

A high-performance, robust tool for electromagneticsource localization and visualization, Curry integrates mul-tiple, complementary modalities (EEG and MEG; MRI,fMRI, CT, PET or SPECT) in a single software package. Bycombining the latest techniques for measuring electrical ac-tivity in the brain with anatomical and functional imaging,Curry provides a powerful method for accurately localiz-ing electromagnetic sources. Curry uses the full physicalanatomy from MR and CT to segment individualizedthree-dimensional models of the skull and brain to be usedin source localization. Curry integrates functional imagingsuch as fMRI with EEG and MEG source reconstruction toallow the comparison of results and to enhance the validityof solutions. Specifications: Integration of EEG, MEG,ECoG with MRI and fMRI/CT/PET/SPECT. Completeanalysis of data leading to anatomically constrained sourcereconstruction. Independent Component Analysis (ICA)and Principal Component Analysis (PCA) for data analysisand filtering. Realistic volume conductor models using theBoundary Element Method (BEM). Finite Element Method(FEM) support. Dipole fits. Dipole confidence ellipsoids arecomputed. Dipole scans, extended source (patch) scans,and MUSIC scans. Current Density Reconstructions (CDR)based on dipolar or extended sources. Automated dataanalysis and source reconstruction using macros.

Comprehensive MEG software package for both clini-cal and research applications. - L. Parkkonen, S. Taulu,M. Kajola (Elekta Neuromag Oy, Helsinki, Finland).

With the advanced whole-head MEG systems, thefunctionality, usability and quality of the associated soft-ware has become increasingly important. The Neuromagsoftware suite is developed both for brain research andclinical work, incorporating both new analysis methodsand streamlined clinical protocols. Data acquisition mod-ule supports simultaneous preparation and measurementof consecutive patients, continuous head position track-ing, and multi-category on-line averaging with artefact re-jection. Signals are displayed in selectable layouts, andmultiple response categories can be overlayed and sub-jected to mathematical operations. Continuous data areprocessed offline with a programmable tool that includesall basic signal processing functions. Source estimationmodules feature time-varying multi-dipole modeling,EEG forward modeling, response simulator and mini-mum current estimation (MCE); all in a sphere or realisticmodel (BEM). MRIs can be overlayed by dipoles, MCEand electrode locations, and they can be segmented toyield BEM models and 3D renderings with dipoles.Finally, a report can be composed using drag-and-dropfrom the analysis modules. In addition to graphical userinterfaces, all modules are also scriptable and their setups

can be recalled, which facilitates routine work. Sig-nal-space projection (SSP) with reconstruction is includedfor noise compensation and source waveform extraction.Signal-space separation (SSS) is applied for head move-ment correction and noise compensation. For method de-velopers, MATLAB interface exists for MEG/EEG, MRI,MCE and dipole data. Interfaces to third party soft-ware/systems are implemented via standard formatssuch as DICOM and EPS. The package has demonstratedits applicability to both clinical and research work.

Cortically-constrained minimum-norm inverse solu-tions. - M.S. Hämäläinen, T. Witzel, F.H. Lin, B.R.Fischl, A.M. Dale (MGH/MIT/HMS Athinoula A.Martinos Center for Biomedical Imaging, Charles-town, Massachusetts, USA).

The FreeSurfer software developed in our center al-lows automatic segmentation of high-resolution ana-tomical MRI data. The software produces a topologicallycorrect tessellation of the pial surface as well as the bor-der between the white and the grey matter for each hemi-sphere. The latter cortical surface is employed as ananatomical constraint in our source modeling proce-dures. In addition to the locations vertices, estimates forthe local surface normals as well as curvature informa-tion are calculated. For visualization purposes, an "in-flated" surface can also be produced to reveal the corticalsulci. In the estimation of current sources from MEG andEEG data we employ L2 and L1-norm minimum-normsolutions. The current estimates are calculated in a gridwith typically 7-mm spacing between source spacepoints on the cortex. We prefer not to use an exact orien-tation constraint at each source space point, as given bythe cortical tessellation, but rather allow some "loose-ness" in the orientations to account for variation of thenormal direction within the cortical patches correspond-ing to the decimated points. The loose orientation con-straint also makes the estimates less sensitive toalignment errors between the MEG and MRI coordinateframes as well as inaccuracies in the forward model. Incase of L2-norm estimates we can transform the expectedvalue of the current into a statistical test variable by di-viding it with the current standard deviation. We canthus present the results as dynamical statistical parame-ter maps (dSPMs), similar to the corresponding staticmaps produced in fMRI and PET.

Spatial filtering in MEG: methods and challenges. -B.D. Van Veen (Department of Electrical and Com-puter Engineering, University of Wisconsin-Madison).

Several spatial filtering or beamforming methods,including linearly constrained minimum variance(LCMV) (Van Veen et al. 1997) and synthetic aperturemagnetometry (SAM) (Robinson and Vrba 1999) have

182 Organizer: Y. Okada

been proposed for analysis of MEG data. Spatial filteringemploys a weighted combination of the sensor data to es-timate the signal with a desired spatial amplitude patternat the sensor, while attenuating signals/noise with unde-sired amplitude patterns. The desired spatial amplitudepattern is usually determined from an equivalent currentdipole model. LCMV, SAM, and several other methodsdesign the weights, or spatial filter, to minimize the spa-tial filter output power subject to various constraints. Theconstraints are chosen to preserve the signal of interestand/or other properties of the data. The estimated spatialfilter output power or output signal to noise ratio, calcu-lated as a function of location, provides a metric for iden-tifying active regions of the brain. Two key challenges forspatial filtering methods are sensitivity to deviation be-tween the modeled spatial amplitude pattern and actualpattern associated with the underlying activity, and esti-mation of the spatial covariance matrix from a limitednumber of measurements. Current research directed to-ward mitigating these problems, including quadraticweight constraints and subspace mapping, are discussed.Subspace mapping involves passing the data through abank of fixed spatial filters, designed to pass signals froma region of interest and dramatically reduces the numberof data vectors required for covariance matrix estimation.

Exact localization with standardized Low ResolutionBrain Electromagnetic Tomography (sLORETA). -Roberto D. Pascual-Marqui (The KEY Institute forBrain-Mind Research, University Hospital of Psychia-try, Zurich, Switzerland).

Scalp electric potential differences and extracranialmagnetic fields originate from the primary current densitydistribution produced by neuronal post-synaptic pro-cesses. A solution to the inverse problem, i.e., the estimationof images of electric neuronal activity based on extracranialmeasurements, would provide important information onthe time course and localization of human brain function. Ingeneral, there is no unique solution to this problem. How-ever, in particular, a distributed linear solution capable ofexact localization of point sources is of great interest, sincethe principles of linearity and superposition would guaran-tee its trustworthiness as a functional imaging method,given that brain activity occurs in the form of a finite num-ber of distributed "hot spots". Despite all previous efforts,distributed linear solutions at best produced images withsystematic non-zero localization errors. We present a dis-tributed linear solution that yields images of statisticallystandardized current density (sLORETA) with zero local-ization error. This property holds for continuous and dis-crete distributions of the measurement space (with as fewas 25 sensors) and of the solution space (with as many as13,000 voxels), for both the electric and magnetic field cases.sLORETA has been compared to all other published 3D

discrete distributed instantaneous linear inverse solutions.Only sLORETA enjoys the "zero localization error" prop-erty. In particular sLORETA was compared to the recentlydeveloped family of so-called "noise-normalized" inversesolutions based on Dale et al. 2000 Neuron 26:55-67. Exceptfor sLORETA, all variants of "noise-normalization" havesystematic localization errors. In addition, sLORETA hashighest resolution.

Infomax ICA separates high-dimensional EEG datainto temporally independent and spatially compactsource projections. - S. Makeig and A. Delorme (SwartzCenter for Computational Neuroscience, Institute forNeural Computation, University of California SanDiego).

What are the major sources of scalp EEG signals? Theshort lateral extent of horizontal connections of most corti-cal neurons, particularly of inhibitory neurons, shouldproduce a strong tendency for near-synchronous patternsof local field activity to emerge in compact cortical do-mains. Biophysics dictates that the projection to the scalpsurface of coherent activity within each such EEG sourcedomain should strongly resemble the projection of anequivalent current dipole. Unfortunately, separating anEEG scalp map into the sum of dipolar projections frommultiple source domains is an underdetermined problem.A radically different approach to source separation usingindependent component analysis (ICA) (Makeig et al.NIPS, 1996) considers only the higher-order statistics ofthe recorded surface maps, rather than the biophysics ofvolume conduction, based on two assumptions: Over anexperimental session, these domains remain relativelyfixed spatially, and their time courses are relatively inde-pendent. We performed ICA decomposition of fourteen71-channel EEG data sets from a working memory experi-ment using over 20 ICA and related algorithms publiclyavailable in Matlab format on the Internet. Several ICA al-gorithms returned several to many components whosescalp projections were highly significantly more dipolarthan the raw EEG data. Most dipolar components wereobtained using the Runica algorithm from our EEGLABMatlab toolbox (sccn.ucsd.edu/eeglab/), which auto-mates extended natural gradient infomax ICA (Jung et al.Proceedings IEEE, 2001), and one other ICA algorithm.These results imply that scalp EEG signals predominantlysum partially coherent electrical activity occurring inde-pendently in multiple compact cortical source domains.

Introduction to TMS. - R.J. Ilmoniemi (Nexstim Ltd.,Elimäenkatu 22 B, FIN-00510 Helsinki, Finland,[email protected]).

The human brain can be stimulated non-invasively bytargeted magnetic field pulses that induce small electriccurrents to flow in the brain, thereby triggering neurons

Proceedings 183

into action. The popularity of this technique, transcranialmagnetic stimulation (TMS), has grown dramatically dur-ing the last few years. TMS shows great potential for brainresearch, diagnosis and therapy. Until now, most TMS ex-periments have been limited to the stimulation of the mo-tor cortex because the only immediately observable effectsof magnetic stimulation have been those reflected in mus-cle responses. Recently, however, it has been demon-strated that the cortical effects of TMS can also be observeddirectly, for example by means of electroencephalography(EEG), positron emission tomography (PET), functionalmagnetic resonance imaging (fMRI) or near infrared spec-troscopy (NIRS). To realize the full power of TMS, at-tempts are being made to develop stimulation systems thatcombine image-guided targeting (stereotactic navigation),sequential stimulation (scanning) and the recording anddisplay of brain activity evoked by TMS. It is expected thatTMS, in particula in combination with EEG or other func-tional brain mapping techniques, will have wide applica-tions both in basic research and in clinical applications,including diagnosis and therapy.

Combining transcranial magnetic stimulation (TMS)with PET and fMRI. - H.R. Siebner (Department ofNeurology, Christian-Albrechts-Universität, Kiel, Ger-many and Neuroimaging Institute NeuroImage-Nord,Hamburg, Germany).

Transcranial magnetic stimulation (TMS) combinedwith positron emission tomography (PET) and functionalmagnetic resonance imaging (fMRI) provides a behav-iour-independent assay of cortical excitability, connectiv-ity, and plasticity in the intact human brain. Bysystematically manipulating a specific variable of stimula-tion such as frequency or intensity, concurrent functionalbrain imaging can assess the regional excitability profileor the functional connectivity of a distinct cortical area.This opens up clinical applications such as the in-vivo as-sessment of disease-related or drug-induced changes incortical excitability. By applying regular trains of mag-netic pulses, repetitive TMS (rTMS) is capable of inducingenduring functional changes in a distributed functionalnetwork. The time course and the regional pattern of thesechanges can be readily mapped with PET and fMRI. In ad-dition, one can directly relate rTMS-induced changes intask performance with changes in task-related activationpattern. It will be demonstrated that this approach is suit-able to investigate the adaptability of cortical systems andto disclose abnormal patterns of reorganisation in patientswith neuropsychiatric disorders.

Neuromodulation of plasticity in human motor cortex.- U. Ziemann, A. Korchounov and F. Meintzschel.

We know from animal experiments thatneuromodulators (dopamine, norepinephrine, serotonin,

acetylcholin) may powerfully modify plasticity in cerebralcortex. Here we tested in a placebo-controlled cross-overdesign the effects of a single oral dose of variousneuromodulators on associative long-term potentiation(LTP)-like plasticity induced by paired associative stimu-lation (Stefan et al. 2000, Brain 123:572), anduse-dependent plasticity (Classen et al. 1998, J.Neurophysiol., 79: 1117) in eight healthy right-handedsubjects. Associative LTP-like plasticity was significantlyenhanced by 2 mg of cabergoline (D2 agonist) but sup-pressed by 2.5 mg of haloperidol (D2 antagonist), 1 mg ofprazosine (alpha-1 antagonist) and 8 mg of biperiden (M1antagonist). Forty milligrams of methylphenidate resultedin a trend towards an enhancement. Use-dependent plas-ticity was significantly enhanced by cabergoline,methylphenidate and 40 mg of tacrine (AChE inhibitor)but suppressed by haloperidol, prazosine and biperiden.One hundred milligrams of sertraline (selective serotoninre-uptake inhibitor) had no effect on either form of plastic-ity. These findings suggest that LTP-like plasticity anduse-dependent plasticity in human cortex can be predict-ably enhanced or suppressed by neuromodulation. It islikely that LTP like plasticity and use-dependent plasticityshare common mechanisms because a broad variety ofneuromodulators exerted very similar effects on bothforms of plasticity. Finally, these experiments may be im-portant for strategies to facilitate use-dependent plasticityin neurological patients after lesion.

Transcranial magnetic stimulation (TMS) in neurosur-gery: present status and future visions. - M. Fujiki andH. Kobayashi (Department of Neurosurgery, School ofMedicine, Oita University, Japan 1-1, Idaigaoka,Hasama-machi, Oita, Japan 879-5593. e-mail:[email protected]).

This study overviews the present status of transcranialmagnetic stimulation (TMS) in neurosurgery. After intro-ducing the technical and physiological principles and basicmechanisms of the technique, we demonstrate a represen-tative case of the use of TMS in neurosurgery. Preoperativemotor functional evaluation using a mapping technique(integrated with MRI and fMRI) and quantitative evalua-tion using MEPs are applied as the most important applica-tions of TMS in neurosurgery. We will then discuss howhuman cortical function can be modified by magnetic stim-ulation and how this might be used therapeutically in neu-rological disorders such as cerebral ischemia, epilepsy,chronic pain and Parkinson’s disease.

Cortical inhibition in the pathophysiology and treat-ment of schizophrenia. - J. Daskalakis (Centre for Ad-diction and Mental Health, University of Toronto,Toronto, Canada).

Background: Cortical inhibition (CI) deficits have


been proposed as a pathophysiologic mechanism inschizophrenia. Moreover, preliminary evidence suggeststhat antipsychotics, particularly atypical ones, increasecortical inhibition. Three transcranial magnetic stimula-tion (TMS) paradigms: short interval cortical inhibition(SICI), cortical silent period (CSP) and transcallosal inhibi-tion (TCI) are used to assess cortical inhibition. Methods:In study one, we assessed CI in 15 unmedicated and 15medicated patients with schizophrenia as well as 15healthy controls. These results have been previously pub-lished. In study two, CI was measured in 27 healthy con-trol subjects before, 6 and 24 hours after being randomlyassigned to receive a single dose of either 2 mg haloperidol(n=8), 10 mg olanzapine (n=10) or placebo (n=9). Results:In study one, unmedicated patients demonstrated signifi-cant CI deficits compared to normal controls across all in-hibitory paradigms whereas medicated patientsdemonstrated relatively greater CI compared to unmedi-cated patients. In study two there was no significant effectof olanzapine or haloperidol on CI at any time interval.Conclusions: These results suggest that schizophrenia isassociated with deficits in CI and that medications appearto attenuate these CI deficits. In contrast, a single dose ofantipsychotic medication does not have any effect on cor-tical inhibition in healthy control subjects. Potential expla-nations attempting to reconcile these findings inschizophrenia will be presented. Moreover, we will dis-cuss cortical inhibition as a potential therapeutic targetthrough which novel treatment strategies, such as repeti-tive TMS (rTMS), exert their antipsychotic effects.

Role of soft bone, CSF and gray matter in EEG simula-tions. - C. Ramon*, P. Schimpf+, J. Haueisen^, M.Holmes*, A. Ishimaru* (*Departments of Electrical En-gineering and Neurology, University of Washington,Seattle, WA, USA; +School of Electrical Engineeringand Computer Science, Washington State University,Spokane, WA, USA; ^Biomagnetics Center, Depart-ment of Neurology, F.S. University, Jena, Germany).

We examined the effect of the soft skull bone, graymatter and CSF on scalp potentials using a finite elementmodel of the head constructed from the segmented MRIimages with 11 different tissue types identified. The finiteelement model has a 1×1×3.2 mm voxel resolution and atotal of approximately 1.5 million hexahedral elements.We used a current dipole model in the motor cortex areaand computed the scalp potentials for head models withvarying complexities. We used four models: a full modelwith 11 tissue-types, a model in which the conductivity ofthe soft bone was set equal to the hard skull bone, a modelin which the conductivity of the gray matter was set equalto white matter, and a model in which the conductivity ofthe gray matter and CSF were set equal to the white mat-ter. The contour plots of the scalp potentials were made. It

was found that contour plots of scalp potentials were dif-ferent for all four models. The differences were the largestfor the model where the conductivities of the gray matterand CSF were set equal to the white matter. One possiblereason could be that by setting the conductivity of the CSFand white matter same removes the anatomical finestructure in the brain tissue and thereby influencing thevolume current flow patterns. These results suggest thatfor accurate modeling of the scalp potentials one shouldinclude anatomical details of the CSF geometry in thecomputer models of the head.

Cortically-constrained time-frequency analysis ofMEG and EEG data. - T. Witzel, S.M. Stufflebeam, F.-H.Lin and M.S. Hämäläinen (MGH/MIT/HMS AthinoulaA. Martinos Center for Biomedical Imaging, Charles-town, Massachusetts, USA).

Time-frequency analysis of MEG and EEG wave-forms has become a method of increasing interest duringthe recent years, because it has been shown to reflecthigher cognitive processing of sensory input, attention andworking memory as well as synchronization processes be-tween distant locations in the cortex. Here we present amethod for computing noise-normalized linear mini-mum-norm estimates in conjunction with time-frequencyanalysis. Initially, time-frequency analysis is performed byconvolution of the measured signals with a complexMorlet’s wavelet. The complex result of the convolution isthen projected into source space where the power andphase of the narrow-band signal is calculated. In order toyield induced oscillatory activity the solution is calculatedon a single trial basis and the across trial statistics is calcu-lated in source space. In order to penalize only environ-mental spatially correlated noise we use the empty roomnoise covariance for the regularization of the inverse oper-ator. The noise covariance needed for the noise normaliza-tion of the solution is calculated from the baseline of everysingle trial in source space. Furthermore, this method al-lows the use of anatomical orientation constraints and spa-tial priors, such as fMRI activation maps. In order toinvestigate phase-synchrony in the brain, we calculate thephase-locking statistics between a selected ROI on the cor-tex or a stimulus phase and all other source locations in thebrain. The results are generated in form of activation mov-ies displaying frequency power or phase locking signifi-cance. The surface representation of the source spaceadditionally allows cross-subject statistics.

Evaluation of sLORETA in the presence of noise andmultiple sources. - M. Wagner, M. Fuchs and J. Kastner(Neuroscan Labs, Hamburg, Germany).

Introduction: Standardized Low Resolution BrainElectromagnetic Tomography (sLORETA) [1] can beused to compute statistical maps from EEG and MEG

Proceedings 185

data that indicate the locations of the underlying sourceprocesses. These maps are derived by performing a loca-tion-wise inverse weighting of the results of a MinimumNorm Least Squares (MNLS) analysis with their esti-mated variances. We evaluate the performance of themethod under the presence of noise and with multiple,simultaneously active sources. Methods: Simulated datacontaining one and two sources at different locations andwith different strengths have been computed for an 81electrode EEG setup. Three spherical shells were used asthe volume conductor model. Data were disturbed usingGaussian noise of different levels. sLORETA analyseshave been performed using the CURRY software(Neuroscan, El Paso, TX, USA) and their results havebeen compared with other dipole and current density re-sults. Results: While the method is capable of localizingsingle sources even in the presence of noise, multiple si-multaneous sources can only be separated if they are nottoo close and their strengths are not too different.

Pascual-Marqui, A.D. Standardized low resolutionbrain electromagnetic tomography (sLORETA): techni-cal details. Methods and Findings in Experimental &Clinical Pharmacology, 2002, 24D: 5-12.

Extended region spatial-temporal MEG analysis usingbayesian inference. -D.M. Schmidt, J.S. George., S.C.Jun, J. Pare-Blagoev, S. Pliz, D.M. Ranken and C.C.Wood (Biological and Quantum Physics Group,MS-D454, Los Alamos National Laboratory, LosAlamos, New Mexico, USA).

We present progress made on extending our origi-nal spatial only, extended region Bayesian inferenceMEG analysis [1] to a spatial-temporal analysis. Begin-ning with the original formulation of the spa-tial-temporal analysis [2], we designed, coded and testedmultiple generations of analyses. Our testing and evalu-ation methods were systematic and thorough and witheach generation one or more issues were discovered thatrequired modifying the analysis and generating the sub-sequent generation. Many of the issues identified and themethods we developed for addressing them may be rele-vant to other MEG analysis approaches as well. Here wedescribe these issues and methods we developed for ad-dressing them. These include: 1) correlation betweensize of resulting regions of activity and prior values onthe strength of current density, 2) need to limit the tem-poral extent of active regions, 3) importance of modelingthe spatial-temporal correlation of the background noise,and 4) incorporate the uncertainties of the forwardmodel, of co-locating head and sensor coordinates, andof various motion effects.

[1] Schmidt, D.M., George, J.S. and Wood, C.C.Hum. Brain Mapp., 1999, 7: 195-212.

[2] Schmidt, D.M., George, J.S., Ranken D.M. and

Wood, C.C. Biomag2000, Proc. 12th Int. Conf. onBiomagnetism, 2001: 671-673.

Dipole analysis on spatiotemporal MEG signals usingBayesian inference. - S.C. Jun, J.S. George, J.Pare-Blagoev, S. Pliz, D.M. Ranken, D.M. Schmidt andC.C. Wood (Biological and Quantum Physics Group,MS-D454, Los Alamos National Laboratory, LosAlamos, New Mexico, USA).

New approaches to the MEG/EEG inverse problembased on Bayesian Inference have been recently described[1, 2, 3]. These approaches yield a probability distributionof solutions upon which all subsequent inferences arebased. References [1] and [3] focused on the analysis ofdata at a single point in time, and demonstrated the utilityof Bayesian inference both for including pertinent priorinformation and for yielding robust results. Reference [1]used an extended region model for neural activity whilereference [3] used a multi-dipole model. Reference [2] ex-tended their work to a spatiotemporal Bayesian inferenceanalysis of the full spatial-temporal MEG/EEG data set,using their extended region model for neural activity.Here we present a spatial-temporal Bayesian inferenceanalysis using a dipole model of neural activity that is rel-atively fast, does not require the use of the subject’s ana-tomical information, does not require prior determinationof the number of dipoles, and yields quantitative proba-bilistic inferences that are one of the hallmarks of theBayesian inference approach. We added a current timeprior probability distribution for the correlation of onetime point with another that allows us to include the tem-poral correlation at nearby latencies. In addition we haveincorporated the ability to handle much more complexand realistic estimates of the background noise [4] that re-duce the effects of under-modeling this noise. We usedMarkov Chain Monte Carlo (MCMC) to sample the manypossible likely solutions. The spatiotemporal Bayesiananalysis is demonstrated using both simulated and empir-ical whole-head MEG data.

[1] Schmidt, D.M., George, J.S. and Wood, C.C.Hum. Brain Mapp., 1999, 7: 195-212.

[2] Schmidt, D.M., George, J.S., Ranken, D.M. andWood, C.C., Biomag2000, Proc. 12th Int. Conf. onBiomagnetism, 2001: 671-673.

[3] Bertrand C. et al., IEEE Trans. Biomed. Engin.,2001, 48: 533-542.

[4] Pliz S. et al., Private Communication and abstractthis conference.

The signal space separation method. - S. Taulu, M.Kajola and J. Simola (Elekta Neuromag Oy).

Multichannel measurement with hundreds of chan-nels essentially covers all measurable degrees of freedomof a curl and source free vector field, like the magnetic


field in a volume free of sources. This fact is based on theconstraint caused by Laplace’s equation for the magneticscalar potential and that the measurable signals are spa-tially band limited. A functional expansion solution ofLaplace’s equation enables one to separate signals aris-ing from the sphere enclosing the interesting sources, e.g.the currents in the brain, from the rest of the signals. Thesignal space separation (SSS) is accomplished by calcu-lating individual basis vectors for each term of the func-tional expansion solution to create a signal basis coveringall measurable signal vectors. Because the SSS basis is lin-early independent for all practical sensor arrangements,any signal vector has a unique SSS decomposition withseparate coefficients for the interesting signals and sig-nals coming from outside the interesting volume. Thus,SSS basis provides an elegant method to remove externaldisturbances. The device-independent SSS coefficientscan be used in transforming the interesting signals to vir-tual sensor configurations. This can also be used in com-pensating for distortions caused by movement of theobject by modelling it as movement of the sensor arrayaround a static object. In practice, the SSS method hasbeen found to be a very effective method to remove exter-nal disturbances and to calculate various virtual signalswithout affecting the morphology or decreasing the sig-nal-to-noise ratio of the interesting signals.

Utilizing ICA and MR-FOCUSS to image spontaneousand stimulus synchronized activity. – John E. Moran*and Norman Tepley*# (*Henry Ford Hospital, Detroit,Michigan USA; #Oakland University, Rochester Michi-gan USA).

A typical current density imaging technique con-structs a time series estimate of cortical activation by ap-plying a single pass imaging filter to time series MEGdata. This is computationally efficient for imaging longtime sequences of MEG data. However, these techniquesgenerate low-resolution images of cortical activity. Incontrast, recursively applied imaging techniques suchFOCUSS and MR-FOCUSS generate high-resolution es-timates of cortical activity by adaptively imaging singletime slices of MEG data. However, long sequences ofMEG data cannot be rapidly imaged. Designing MEGstudies such that ICA techniques can be used to isolateMEG field patterns of spatially distinct sources elimi-nates this imaging efficiency problem. CombinedICA/MR-FOCUSS imaging of both spontaneous andtime synchronized cortical activity will be described.(Supported by NIH/NINDS R01-NS30914).

A combined multi-dipole analysis of simultaneouslyacquired MEG and EEG. - D. Ranken*, P. Volegov*, E.Best*, R. Kraus*, C.C. Wood* and D. Tucker+ (*Biologi-cal and Quantum Physics Group, Los Alamos National

Laboratory, Los Alamos, NM; + Electrical GeodesicsInc., Riverfront Research Park, Eugene Oregon, 97403).

As part of a collaboration between the LANL Bio-physics Group and Elecrical Geodesics Inc., the LANL SISMEG system and the EGI EEG system were used to collectdata simultaneously in a median nerve stimulation ex-periment. These data were analyzed using the corticalstart spatio-temporal MEG/EEG analysis program CSST(Ranken et al. Biomag 2002). CSST is a multi-startmulti-dipole analysis procedure based on MSST (Huanget al. Electro. Clinical Neurophys. 1998). CSST couples anonlinear simplex search of dipole spatial configurationswith a linear fitting to obtain dipole timecourses, using areduced Chi square as the minimization criterion. Addedfeatures in CSST include the ability to run on a computa-tional cluster, subject MRI based selection of starting loca-tions, a fast simplex search, and Monte Carlo analyses.For the combined MEG/EEG analysis, a weighted combi-nation of the MEG and EEG reduced Chi square valueswas used during the simplex search to obtain an optimalfit to the combined data. Spherical head models wereused for both modalities. We show results for the com-bined analysis, and compare these to results for the indi-vidual MEG and EEG analyses.

Realistic spatiotemporal noise covariance model forMEG/EEG source analysis. - S.M. Plis, J.S. George, S.C.Jun, J. Pare-Blagoev, D.M. Ranken, D.M. Schmidt andC.C. Wood (Biological and Quantum Physics Group,MS-D454, Los Alamos, New Mexico, USA).

Accurate characterization of the background noiseis important in the analysis of MEG and EEG data as it de-termines what will be designated as signal. The noise inempirical data typically has a large component frombackground brain activity and has thereforespatiotemporal correlation not dissimilar from the de-sired signal. Ignoring this correlation structure in theanalysis can lead to spurious sources that merely modelthe correlated noise or bias offsets in the location andtimecourse of true sources. There is generally not enoughdata available about the noise to determine the fullcovariance structure given the large number of sensors inmodern whole-head sensor systems. We have developeda method for modeling the noise covariance based on aseries of orthogonal products of spatial and temporalcovariances. By using a series we can model a more com-plex covariance structure than that from a single prod-uct, as in [1]. Furthermore, by using orthogonal productsthe inverse of the covariance, which is needed in the anal-ysis, can be easily calculated. We describe this approachand demonstrate some of its features using simulatedMEG data, and data from an empirical experiment [2]that had a novel stimulus protocol that yields better in-formation about the noise. We use a dipole based

Proceedings 187

spatiotemporal Bayesian inference analysis to demon-strate the use of this noise model [3]. We compare theseresults with those from another noise covariance model[1] and from assuming no covariance in the noise.

[1] Huizenga, H.M., de Munck, J.C., Waldorp, L.J.and Grasman, R.P. IEEE Trans. Biomed. Engin., 2002, 50:533-539.

[2] Pare-Blagoev J. et al., Abstract this conference.[3] Jun Sung Chan et al., Abstract this conference.

Dynamic neuroimaging with EEG: multi-scale currentsources, volume conduction, the inverse problem andthe dynamic behavior of macro-potentials. - P. Nunez.

The relationship between micro-current sources atcell membranes and the meso-sources or "dipoles" (moreaccurately dipole moments per unit volume of corticalcolumns) are considered. EEG dynamic behavior de-pends on just two underlying variables , themeso-source distribution and the Green’s function thatfully characterizes the head volume conductor. Thisphysical theory addresses the following contemporaryissues of interest in neuroimaging: Dynamic versusstatic views of brain science and possible physiologicalbases for dynamic behavior. Why do brains produce os-cillations and what physiological time scales underliethese oscillations? These ideas are illustrated with datashowing complex changes in coherence patterns duringtask performance, in which changes can occur in oppo-site directions in different frequency bands or in shortversus long range coherence.

Investigation of dipole localization accuracy in MEGusing the bootstrap. - Felix Darvas*, MikkoRautiainen+, Sylvain Baillet^, Alex Ossadtchi*, John C.Mosher~ and Richard M. Leahy* (*Signal and ImageProcessing Institute, University of Southern Califor-nia, Los Angeles, CA 90089,USA; +Ragnar Granit Insti-tute, Tampere University of Technology, Tampere,Finland; ^Cognitive Neuroscience and Brain ImagingLaboratory, Paris, France; ~Los Alamos National Labo-ratory, Los Alamos, NM 87545, USA).

We used the nonparametric bootstrap to investi-gate the accuracy of current dipole localization fromMEG studies of event related neural activity. The boot-strap is well suited to analysis of event-related MEGdata since the experiments are often repeated 100 ormore times and averaged to achieve acceptable SNRs.The set of repetitions can be viewed as a set of inde-pendent realizations of the brain’s response to the ex-periment. Bootstrap resamples are generated bysampling from these epochs and averaging. In thisstudy we applied the bootstrap resampling techniqueto MEG data from a somatotopic experiment. Four fin-gers of the right and left hand of a healthy subject were

electrically stimulated, and about 400 trials per stimu-lation were recorded and averaged to determine thesomatotopic mapping of the fingers in the S1 region.Using the single trial recordings for each finger, we per-formed 5000 bootstrap resamples. We reconstructed di-poles from the resampled averages using theRAP-MUSIC source localization algorithm. To find thecorrespondences between multiple sources in eachresample, dipoles were clustered using a GMM(Gaussian Mixture Model) clustering algorithm, usingtheir combined normalized time-series and topogra-phy as feature vectors. The mean and standard devia-tion of the dipole position and the dipole time-series ineach cluster were computed to provide estimates of theaccuracy of the reconstructed source locations andtime-series. For the 30-60ms time interval the bootstrapmethod revealed a stable somatotopic mapping on theleft and right hemisphere, with spatial standard devia-tions between 6 mm and 17 mm.

Visual fields of motion-defined stimuli analyzed by anovel blind maximum likelihood estimation algo-rithm. - B.V. Baryshnikov*, R.T. Wakai* and B.D. VanVeen+ (Departments of *Medical Physics and +Electri-cal and Computer Engineering, University of Wiscon-sin-Madison, Madison, WI, 53706, USA).

We present a novel covariance-based algorithm thatimproves the signal-to-noise ratio (SNR) of multi-epochdata, and we apply it to visual evoked MEG recordings.The algorithm is based on maximum likelihood signal es-timation. Its effectiveness lies in its ability to exploit dif-ferences in the covariance structure of the averaged andthe single-trial data while making very few assumptions.Simulations demonstrated the following advantages:improved SNR, simplicity of implementation, and appli-cability to recordings in which the signal and noise al-ways overlap in time. The algorithm was then applied to37-channel MEG data from a study of motion-defined vi-sual contrast. An object consisting of brush strokes (abird) or dots (a rectangle) was presented over a back-ground composed of identical, randomly distributedbrush strokes or dots. The object blended perfectly withthe background when stationary and was visible onlywhen moving. Two control recordings were made, firstwith every element of the image moving independentlyand randomly and second with the elements of the objectmoving independently and randomly over a static back-ground. Prominent responses at about 150-180 ms fol-lowing movement onset were seen for both themotion-defined and the control stimuli, while an addi-tional response in the range 250-450 ms was present forthe motion-defined stimuli. The SNRs of both the earlyand late responses were improved by blind maximumlikelihood estimation.


Dynamic spatial filter approach in deep biomagneticsources localization. - L. Marzetti, M. Demelis and S.N.Erné. (Division for Biosignals and Imaging Technol-ogies, ZIBMT, University of Ulm, Germany).

The reconstruction of deep biomagnetic sources is achallenging topic due the low signal-to-noise ratio of themeasured field which falls off at least as 1/r2. To obtain arobust estimate of the source parameters, many sourcesreconstruction techniques accomplish a restriction of thesource space over the volume conductor using physiolog-ical constraints [1] which can be of limited use in patho-logical cases, especially concerning brain plasticity. No apriori constraints are necessary in the Dynamic SpatialFilter (DSF) approach. To improve the signal-to-noise ra-tio the filtering here proposed takes into account onlygeneral properties of dipolar sources. Increasing thesource depth, the extension of the spatial frequency spec-trum decreases. Therefore, a spatial filter with a low passcharacteristic in a two dimensional frequency domain hasbeen designed. The filter transfer function is calculatedbefore performing dipole localization using the same firstguess. Practically, the procedure is repeated twice, a firstrun using an analytic volume conductor model and a heu-ristic first guess, a final run using a realistic model and theprevious result as first guess.

[1] Van Veen, B.D., Van Drongelen, W., YuchtmanM. and Suzuki, A. Localization of brain electrical activityvia linearly constrained minimum variance spatial filter.IEEE Trans. Biomed. Eng., 1997, 44: 867-880.

Dipole source models for visual ERPs elicited by firstand second targets in the pair. - E.V. Mnatsakanian*+

and I.M. Tarkka* (*Brain Research and RehabilitationCenter Neuron, Kuopio, Finland and +Institute ofHigher Nervous Activity and Neurophysiology RAS,Moscow, Russia).

Our purpose was to estimate the electrical sources ofthe evoked activity during the recognition and compari-son of pairs of familiar faces and abstract patterns. A totalof 18 healthy volunteers participated in the experiment.Each trial began with one of the two cues (S1) followed byconsecutive pictures (S2 and S3). Each picture was a pho-tograph of a familiar face on which an abstract dot patternwas superimposed. One cue directed attention to com-pare faces (Face Task) and another cue to compare pat-terns (Pattern Task). EEG was recorded using a128-channel net. Before source modeling, averaged ERPswere filtered (0.3-15 Hz). Spatiotemporal multiple dipolesource models were created in BESA2000 for the windowof 80-600 ms from S2 and S3 stimuli onsets. The modelswere initially developed for the Face Task and then ap-plied to Pattern Task data, first for ERPs elicited by S3.Dipolar sources were in the visual cortex, bilaterally inmedial temporal and inferotemporal regions, and in the

frontal area. The locations and orientations for dipolesfrom S3 segment were then applied to the S2-elicitedERPs. The residual variances were about 3% for grand av-erage data and around 10-13% for individual ERPs. Thetemporal dynamics of the source strengths (source wave-forms) differed between the models. In both segments, S2and S3, the task-specific differences were seen. The differ-ences between the models of the segments were observedmostly in frontal and inferotemporal sources.

Adaptability of the brain to TMS gives a direct evi-dence for a binding between cognitive and motor func-tions in humans. - J.B. de Graaf, M. Camus, J. Pailhousand M. Bonnard (UMR Movement and Perception,CNRS-University of the Mediterranean, 163 Av deLuminy, 13288 Marseille Cedex, France).

We used TMS both as a reliable tool for quantifyingthe corticospinal excitability, through the Motor EvokedPotentials (MEPs), and as a central perturbation evoking amovement (because the stimulation intensity was abovethreshold). The subjects were instructed to cognitively pre-pare (i.e. without changing their muscle activation) eitherto "let go" or to "resist to" this centrally evoked movement.We studied the simultaneous evolution of both the motorperformance and the Motor Evoked Potentials (MEPs) inthe wrist flexor and extensor, separately for the successfultrials (on average 66% of the trials whatever the condition)and the unsuccessful trials, allowing us to dissociate the in-tention-related and performance-related processes. Totheir own great surprise, the subjects were able tocognitively prepare themselves to resist to a TMS-inducedcentral perturbation. Moreover, because the TMS onlyevoked short-latency MEPs (and no long-latency compo-nents), the amplitude of these short-latency MEPs wasfound to be related in a continuous way to the actual move-ment whatever the prior intention. These results demon-strate that prior intention allows an anticipatorymodulation of the corticospinal excitability which is notonly selective (as already known) but also efficient, givingthe intended motor behaviour a real chance to be realized.This constitutes a direct evidence of the role of thecorticospinal excitability in the binding between cognitiveand motor processes in humans. Moreover, this resultshows that the cognitive state of the subject need to betaken into account while interpreting TMS experiments.

Experimental design issues in combinedhemodynamic and electrophysiological imaging. - E.J.Pare-Blagoev, J. George, S.C. Jung, S. Plis, D. Ranken,D. Schmidt, D. Sheltraw and C.C. Wood (Biological &Quantum Physics Group, MS-D454, Los Alamos Na-tional Laboratory, Los Alamos, New Mexico, USA).

A number of investigators have suggested thatcombining hemodynamic (fMRI/PET) and

Proceedings 189

electrophysiological (EEG/MEG) measures of brain ac-tivity can improve spatial and temporal resolution overeither type of imaging alone. A variety of specific tech-niques for integration have been proposed, includinguse of regions of activation in fMRI/PET to constrain di-pole solutions for EEG/MEG (e.g., Mangun et al. 1998),the use of fMRI/PET to weight linear inverse solutionsfor EEG/MEG e.g., Dale et al. 2000), and the use ofBayesian inference to combine fMRI/PET andEEG/MEG into a common estimate of neural activity(Schmidt et al. 1999). In order for both the hemodynamicand the electrophysiological measures to possess infor-mation about the same averaged neural activity it is de-sirable that the same stimulus paradigm be used for bothmodalities. However, the conventional experimentalparadigms used for hemodynamic andelectrophysiological are designed to achieve optimalsignal-to-noise ratios for each modality, often resultingin different or even contradictory paradigms. Here wepresent an experimental paradigm, using medial nervestimulation, that has been used without modification forboth fMRI and MEG with good signal-to-noise resultingfrom each. This makes it appropriate for testing differenttechniques for combining fMRI with MEG. We discussthe general requirements for such paradigms and de-scribe how this medial nerve stimulus paradigm satis-fies these criteria. We present preliminary fMRI analysisresults from this data as well.

Functional correlations between resting EEG and fMRImetabolic changes. - Thomas Koenig*, Lester Melie+,Daniela Hubl*, Werner K. Strik* and Thomas Dierks*(*Department of Psychiatric Neurophysiology, Univer-sity Hospital of Clinical Psychiatry, Bern, Switzerland;+Cuban Neuroscience Center, CNIC, La Habana,Cuba).

EEG signals originate predominantly from the tran-sient, synchronized activity of cortical neuro-cognitivenetworks. EEG signals are characterized by their fre-quency spectrum, their spatial origin extent, and the se-quence of events. These variables fluctuatespontaneously, but in a well defined manner. This sug-gests that they code for a neural homeostasis that isadapted to the individual’s physiological state, whichdepends not only on the momentary external context butalso on individual past experience and genetics. In orderto further understand the brain state markers providedby EEG (i.e. the synchronization, activation and deacti-vation of cortical structures), we have begun to correlatethe spontaneous and induced fluctuations of EEG statemarkers with simultaneously acquired 3D measures ofblood oxygen level depended signals as measured byfMRI. This work presents on one side the technical as-pects of acquiring EEG data (96 channels) during an 1.5T

fMRI recording (subject safety, recording technique, arti-fact removal) and presents on the other side first results,correlating transient, spatially extended EEG events asidentified by topographic time-frequency analysis withfMRI bold response changes. Our preliminary resultsshows that in a given frequency band, typical EEG eventsalso have typical, extended and specific metabolic signa-tures, but all these events also share activated sites thatmay be the common pace makers. We conclude that si-multaneous EEG and fMRI may become essential tostudy the temporal, spatial and oscillatory properties oftransiently forming neural networks in the brain.

Detection of activation in the motor cortex with hightemporal and spatial resolution by multimodalMEG-fMRI recording. - S.N. Erné*, H.-P. Müller*, M.Demelis*, A. Pasquarelli*, E. Kraft+, A. Ludolph+ and J.Kassubek+ (*Division for Biosignals and ImagingTechnologies, ZIBMT, University of Ulm, Germany;+Department of Neurology, University of Ulm, Ger-many).

Functional regions of the human brain are usuallymapped by functional Magnetic Resonance Imaging(fMRI). FMRI measures the changes in blood oxygena-tion with high spatial accuracy in localizing activated re-gions but restricted temporal resolution. Comprehensivemapping is achieved by magnetoencephalography(MEG). MEG localizes the sources of intracellular electriccurrents, providing direct information of neural activitywith high temporal resolution in the range of millisec-onds. Ten subjects performed the identical stimulationparadigm both during MEG (55-channel magnetometersystem, AtB, Italy) and fMRI (1.5 T, Siemens, Germany)measurements. As the activation paradigm, simple mo-tor tasks (self-paced flexion of index finger and powergrip) were chosen. To pursue synergetic interactions be-tween these two non-invasive recording modalities, it isessential to analyze the data sets in one single softwareenvironment. The integrated bimodal imaging analysiswas performed using the OMEGA (Open Magnetic andElectric Graphic Analysis) software. Localizationsources of MEG and fMRI experiments were integratedand referred to the same individual 3-dimensional MRI(MP-RAGE), acquired during the MRI session: MRI dataused as morphological background for fMRI results arealso processed to build volume conductor modelsneeded in MEG source reconstruction. The fMRI analysisprovides the information for possible sites of sources,whereas MEG data complement the corresponding timecourse of the activation. Multimodal investigations yieldcomprehensive information about functional activationfrom different neuroimaging modalities with high spa-tial and temporal resolution.


Validation of the new SimBio electromagnetic sourcelocalization environment in an animal model. - J.Haueisen*, D. Güllmar*^, L. Flemming*, M. Eiselt+,J.R. Reichenbach^, A. Anwander~, M. Dümpelmann#,T.R. Knösche~ and C.H. Wolters~ (*Biomagnetic Cen-ter, +Institute of Pathophysiology, ^Institute of Diag-nostic and Interventional Radiology, Friedrich-Schiller University Jena, Germany; ~Max-Planck Insti-tute of Cognitive Neuroscience, Leipzig, Germany;#ANT Software b.v. Enschede, The Netherlands. Email:[email protected]).

The central objective of the SimBio project is the im-provement of clinical and medical practices by the use ofnumerical simulation for bio-medical problems. The pro-ject constructs a generic environment, running on paral-lel and distributed computing systems, capable ofhandling a range of important bio-electromagnetic andbiomechanical problems. The outstanding trait of theSimBio toolbox is its extremely modular structure that al-lows free extendibility without any change to the existingsoftware. This enables a wider community of researchersnot only to benefit from the system, but also to contributeto it. Here, we provide a validation of the new SimBio [1]electromagnetic source localization environment with ananimal model. We measured simultaneously theelectrocorticogram and the magnetoencephalogram af-ter peripheral nerve stimulation in the rabbit. FEM andthe BEM models were generated from segmented MRimages of the rabbit head. Conductivity tensor informa-tion was included into the FEM model. Both focal anddistributed source models were used. We found a verygood agreement between the results computed with theisotropic FEM and BEM model. The source localizationdifferences were below 1 mm for all source models. Forthe distributed source models we found the maximumsto be identically localized. The anisotropic FEM modelyielded slight differences in source localization as com-pared to using an isotropic model.

[1] Simbio, A. Generic Environment forbionumerical Simulation, Information Society Technol-ogies (IST) Programme, Framework V ProjectIST-1999-10378, http://www.simbio.de

Distributed current estimates with a cortical constraint.- F.-H. Lin, J.W. Belliveau, A.M. Dale and M.S.Hämäläinen (MGH/MIT/HMS Athinoula A. MartinosCenter for Biomedical Imaging, Building 149 13th St.Charlestown MA 02129).

A widely employed distributed source localizationapproach in MEG is based on the L2-norm prior (proba-bility distribution function, PDF, of normal distribution),resulting in diffuse minimum-norm estimates (MNE).More focal estimates can be obtained by using a mini-mum-current estimates (MCE), which is based on the

L1-norm prior (PDF of exponential distribution) instead.The implementation of MCE requires information aboutthe orientations of the sources, which can be estimated byfirst applying MNE. Here, we propose usinghigh-resolution anatomic magnetic resonance imaging(MRI) data to improve the estimation of sources in MEGdata using both MNE and MCE approaches. MRI pro-vides the local cortical anatomy including the variationof the cortex orientation in the neighborhood of eachsource location. Using the Dijkstra search algorithm, wewere able to define the spatial distribution of the corticalpatch associated with each decimated dipole source, theaveraged normal direction in the individual corticalpatch, the dispersion of the normal direction within acortical patch, and the area of each cortical patch. This al-lows one to either use a strict or a loose cortical orienta-tion constraint, both of which can be incorporated in theMNE and subsequently MCE of MEG data. In both audi-tory and somatosensory MEG experiments, we foundthat MRI constrained MCE with a loose cortical orienta-tion constraint produced more consistent localization(with respect to single equivalent dipole fitting) com-pared to either MNE-informed MCE or to a strict corticalorientation constrained MCE.

Dynamic sculpting of brain functional connectivity iscorrelated with performance measures. -R.B.Silberstein*+, J. Song*, P.L. Nunez*+ and W. Park*(*Brain Sciences Institute, Swinburne University ofTechnology, Melbourne Australia. +Brain PhysicsGroup, Tulane University, New Orleans, USA).

Rapid changes in the synchronization or functionalcoupling of neural systems is considered a key mecha-nism for the integration and segregation of neural sys-tems in cognition. In this study, we use Steady StateVisually Evoked Potential- Event Related Partial Coher-ence (SSVEP-ERPC) to investigate the relationship be-tween cortical functional coupling and speed of cognitiveprocessing while performing a computerized version ofRaven’s Progressive Matrices (RPM). Fifty-fiveright-handed subjects (29 males) performed the RPM a nda matched control task while brain electrical activity wasrecorded from 64 scalp sites. The SSVEP was elicited by adiffuse 13Hz visual flicker superimposed over the visualfields and the SSVEP-ERPC calculated for all pairs of elec-trodes. The correlation between SSVEP-ERPC and speedof processing was determined during the 3 sec epoch thatsubjects selected their task responses. Speed of process-ing was correlated with SSVEP-ERPC at right frontal andright prefrontal sites only at the time 0.8 msec prior to thetask presentation. These findings suggest that speed ofprocessing may depend on the ability to establish func-tional coupling between right frontal and prefrontal neu-ral networks at specific points in time.

Proceedings 191

Direct recording of auditory induced gamma/beta os-cillations in human temporal cortex. - O. Bertrand*, A.Bidet-Caulet*, F. Bauchet* and C. Fischer+ (*MentalProcesses and Brain Activation Lab., INSERM U280,Lyon, France, +Neurological Hospital, Lyon, France).

Neural oscillatory synchronization has been pro-posed as a dynamic link between the different brain areasengaged into the same perceptual or cognitive process.This hypothesis has mainly been supported by studies inthe visual modality, either at the unit level in animals orat a macroscopic level from human scalp EEG/MEG. Inthis latter case, induced gamma oscillations have beenproved to be associated with perception and rehearsal ofcoherent objects and to be modulated by attentiontop-down processes. Similar oscillations have been ob-served in other sensory modalities but have been muchless studied. We studied the temporo-spatial characteris-tics and the attentional modulation of inducedbeta/gamma oscillations in the auditory system. A pre-cise identification of multiple focal oscillatory sourceswas obtained by means of direct intracranial EEG record-ings in the temporal cortex of epileptic patients in severalsituations, including passive listening and active dis-crimination tasks. Focal sources of gamma oscillationshave been found in the 20-90 Hz range, between 150 and350 ms after stimulus onset in the primary auditory cor-tex and in higher level areas (superior temporal gyrus,planum temporale). These gamma sources showed dif-ferential modulations by attention. With similar timing,stimulus induced decrease of on-going beta activity(15-20 Hz) was also observed at different focal location ofthe supra-temporal cortex. The temporal characteristicsof these beta/gamma oscillatory activities and theirmodulation by attention are described and discussed.

The many faces of the intracranial gamma band re-sponse during visual perception in humans. - J.-P.Lachaux (Cognitive Neuroscience & Brain ImagingLaboratory, Paris, France).

While it is known that the perception of visual ob-jects translates in humans into a transient increase of en-ergy in their EEG above 30 Hz (the so-called visuallyinduced gamma response), the neural sources of this re-sponse are mostly unknown. We used depth electrodesin humans to record directly from some of those sourcesin response to face presentations. We found such gammainduced responses in selective occipital, parietal andtemporal sites, including the fusiform gyrus and theintra-parietal sulcus. They sometimes modulated fre-quencies up to 200 Hz and had highly anatomy-specificprofiles and latencies; in sharp contrast with the responsealready observed from the scalp. Such increases were si-multaneous with gamma energy decreases in the pri-mary visual cortex, suggesting that a balance between

regional cortical gamma activations and deactivations isnecessary for face perception.

Sub-millisecond synchronization of fast electrical os-cillations in somatosensory cortex. - Daniel S. Barth(Department of Psychology, University of Colorado,Boulder, CO 80309-0345, U.S.A.).

Fast electrical oscillations (FO; > 200 Hz) in sensoryneocortex can be recorded in a variety of species, includ-ing man, and may reflect extremely fast integration ofsensory information. The present work demonstratesthat in the whisker representation of rat cortex,multi-vibrissa stimulation produces propagating FOfield potential patterns and time-locked unit activity thatare sensitive to sub-millisecond delays in inter-stimulusintervals. We propose that FO may be produced by syn-chronized population spikes and their sub-thresholdsequelae in cortical pyramidal cells. FO serve to accu-rately mark stimulus onset as a phase-encoded excit-atory signal, producing phase sensitive interactions that,in the context of exploratory whisking, may extract fea-tures of an object under exploration.

Eye movement related activity and connectivity fromsingle trial tomography of awake and sleep MEG data.– A.A. Ioannides (Lab. for Human Brain Dynamics,RIKEN Brain Science Institute (BSI), 2-1 Hirosawa,Wakoshi, Saitama, 351-0198, Japan).

Three subjects were examined in three conditions.During awake state saccades were cued by an auditorytone in one condition and initiated voluntarily in theother. Saccades in each direction and condition were inseparate blocks, paced at 0.25 Hz. The third conditionused similar saccades during rapid eye movement(REM) sleep. Brain activity was reconstructed frommagnetoencephalographic data before, during and afterleft, right, up and down saccades, independently for eachtimeslice of each single trial (ST). The tomographic ST so-lutions were aligned to the EOG onset. Well timelockedactivations in each ST were identified in the pons, justpreceding saccade onset. These activations were as pre-dicted by the conjugate activation of the gaze centers. Ac-tivations in the cortex and pons was identifiedthroughout the one second pre-saccade analysis win-dow, and they were only loosely time-locked to thesaccade onset. The mutual information betweentime-lagged ST timeseries was computed to order in timethe activations relative to each other and EOG transients,reflecting ocular muscle activity. The results showedcomplex cascade of interactions with time lags betweenrelated activations well over one hundred milliseconds(Ioannides et al., Cerebral Cortex, 2004). REM saccadeonset was preceded by pontine and FEF activity that wasrelated to saccade initiation by 250 and 400 ms respec-


tively. Activity in the same areas in the last 200 ms beforesaccade onset was related to EOG transients after saccadeonset and hence they were likely to relate to fine adjust-ment and stopping of eye movement.

Corticospinal contributions to static and dynamic mo-tor acts studied by EEG-EMG coherence estimates. -B.A. Conway (Bioengineering Unit, University ofStrathclyde, Glasgow, G4 0NW, UK).

During the past ten years many studies have usedcoherence to investigate the electrophysiological rela-tionships that exist between rhythmic activity arising inthe motor areas of the brain and the EMG activity gener-ated by muscles participating in maintained voluntarymotor acts (e.g. precision grip tasks). In humans, thiswork has identified the existence of task related 15-30 Hzcoherence between localised recordings of EEG or MEGactivity and the EMG recorded during voluntary pos-tural motor acts. The 15-30 Hz coherence is considered toreflect activity transmitted to the spinal motor pools viacorticospinal pathways and estimates of the phase or lagtime between EEG/MEG signals and EMG activity sup-port this conclusion. Thus it can be argued that the as-sessment of coherence between cortical activity patternsand motor output provides an opportunity to study as-pects (spatial and temporal) of the cortical motor com-mand associated with voluntary movements in humans.Accordingly, we have adapted this approach to studywhat relationships can be established between the fre-quency components of EEG and EMG signals recordedduring dynamic motor behaviours. The behaviours inquestion are fast point-to-point wrist movements and lo-comotion. Although preliminary in nature, our recentdata suggest that coherent EEG/EMG activity can beidentified during dynamic behaviours and that coher-ence estimates can provide useful insights into the or-ganisation and temporal structure of some componentsof the motor command associated with static and dy-namic motor tasks.

Acknowledgements: This work has been supportedby The Wellcome Trust, MRC and The Royal Society.

Coherent brain oscillations and tremor. - A. Schnitzler(Department of Neurology, Heinrich-Heine--University Duesseldorf, Germany, Email :[email protected]).

Functional connectivity between brain areas mayappear as correlated time behavior of neural activity. Co-herence between signals of sensors covering differentscalp areas is commonly taken as a measure of functionalcoupling. However, this approach provides vague infor-mation on the actual areas involved. The analysis ap-proach DICS (Dynamic imaging of coherent sources)proceeds beyond the sensor level to estimate coherences

between brain areas [1]. DICS uses a spatial filter to local-ize coherent brain regions and provides the time coursesof their activity. We have applied DICS to explore physi-ological and pathological conditions in the humansensorimotor system. In general, in these studies simul-taneously recorded electromyographic activity serves asa reference signal to calculate cerebromuscular coher-ence which in turn provides cerebral reference points forthe computation of neural coupling with other brain ar-eas. Using this approach we have shown that tremor-likephysiological ~8 Hz movement discontinuities occuringduring slow finger movements emerge from coupledcerebellothalamo-cortical oscillations representing theneural mechanism of the intermittent control of continu-ous movements [2]. Likewise, oscillatory networks un-derlying tremor syndromes, such as parkinsonian,essential and hepatic tremors have been characterizedand will be presented [3, 4]. Taken together, the investi-gation of large scale oscillatory coupling in the humanbrain has yielded new insights into neurophysiologicalmechanisms of tremor syndromes.

[1] Gross, J. et al. Dynamic imaging of coherentsources: studying neural interactions in the humanbrain. Proc. Natl. Acad. Sci. USA, 2002, 98: 694-699.

[2] Gross, J. et al. The neural basis of intermittent mo-tor control in man. Proc. Natl. Acad. Sci. USA, 2001, 99:2299-2302.

[3] Timmermann, L. et al. The cerebral oscillatorynetwork of parkinsonian resting tremor. Brain, 2003, 126:199-212.

[4] Timmermann, L., Gross, J., Kircheis, G.,Häussinger, D. and Schnitzler, A. Cortical origin ofmini-asterixis in hepatic encephalopathy. Neurology,2002, 58: 295-298.

Non-invasive brain-computer interfaces (BCIs). - Jona-than R. Wolpaw (Laboratory of Nervous System Disor-ders, Wadsworth Center, New York State Departmentof Health and State University of New York, Albany,New York, U.S.A., [email protected]).

Brain activity produces electrical signals that can berecorded at the scalp, at the cortical surface, or fromwithin the brain. Brain-computer interfaces (BCIs)change these signals from mere reflections of brain activ-ity into outputs that convey the user’s intent to the out-side world (Clin. Neurophysiol., 2002, 113: 767-791 forreview). Because they do not depend on muscle control,BCIs can provide communication and control to peoplewith severe neuromuscular disorders such asamyotrophic lateral sclerosis (ALS), brainstem stroke, ce-rebral palsy, and spinal cord injury. BCIs can benon-invasive or invasive. Non-invasive BCIs derive theuser’s intent from specific features of scalp-recordedEEG activity. These features may be in the time-domain

Proceedings 193

(such as P300 evoked potentials or slow cortical poten-tials) or in the frequency-domain (such as mu or betarhythms recorded over sensorimotor cortex) .Non-invasive BCIs are already in clinical use for basiccommunication and control. Invasive BCIs derive theuser’s intent from neuronal action potentials or from lo-cal field potentials recorded within the cerebral cortex.They are being studied mainly in non-human primates.These invasive BCIs face substantial technical difficultiesand entail significant clinical risks: they require that re-cording electrodes be implanted in the cortex and func-tion well for long periods, and they risk infection andother damage. The current efforts to develop them, inspite of these disadvantages, are based on the wide-spread assumption that non-invasive BCIs will never beable to provide users with real-time multidimensionalcontrol of a robotic arm or a neuroprosthesis. New datademonstrate that this assumption is not correct. Theyshow in humans that a non-invasive BCI, usingsensorimotor rhythms recorded from the scalp, can pro-vide rapid multidimensional control of a movement sig-nal. The precision and speed of this control is similar tothat achieved in monkeys with invasive BCIs. The key el-ement in the achievement of this unprecented control bya non-invasive BCI is an adaptive training algorithm thatfocuses on those EEG features that the person can bestcontrol and encourages the user to improve that controlfurther. We are defining the salient characteristics of thisEEG control and the kinematics of the rapid and accuratemovements it provides. The presentation will include avideo illustrating the movement control achieved. Thiswork extends the potential applications of non-invasiveBCI technology to include real-time multidimensionalmovement control. It indicates that people with severemotor disabilities could use brain signals to operate a ro-botic arm or a neur.undation, and the James S. McdonnellFoundation).

Theories of working memory based on behavioral data.- P. Kyllonen (Educational Testing Service, Princeton,NJ).

Working memory may be defined as the system usedfor short-term storage and executive processes, such as at-tention and inhibition, task management, planning, mon-itoring, and coding (Smith and Jonides 1999). Workingmemory is exercised in virtually all cognitive activity, buttasks vary in the degree to which working memory (asopposed to other cognitive components, such as knowl-edge) is the critical performance determinant. Workingmemory’s importance increases to the extent that tasks re-quire the simultaneous processing and storage of infor-mation for their success, regardless of content (Baddeley1986). This means that such tasks may be used to engageor activate the working memory system. In this talk I re-

view evidence for the central role of working memory inall cognitive activity, highlighting reasoning, skill learn-ing, language, and psychomotor performance. I presentevidence that performance on working memory tasks isan extremely strong predictor of performance on reason-ing, learning, and psychomotor tasks (typically, r > .70).Working memory plays a central role in computationalmodels of cognition, such as ACT*, suggesting that a pa-rameter representing the capacity of the work-ing-memory system (e.g., the availability of activation)may be identical to Spearman’s g. I also review evidencefor the comparative goodness of tasks as measures ofworking memory, which suggests that a key feature is theminimization of specialized knowledge required.

Selective attention effects on anticipatory and stimu-lus-evoked activity across cortical and subcorticalstages of visual processing. - C.E. Schroeder (CognitiveNeuroscience and Schizophrenia Program, NathanKline Institute, and Department of Neuroscience, Al-bert Einstein College of Medicine).

In a variety of paradigms selective attention: 1) mod-ulates visual responses in V4/MT, V2 and V1, 2) in-creases baseline activity in some locations, and 3)differentially depresses low frequency EEG oscillationswhile enhancing high frequency (γ) oscillations in V4.The generality of the oscillatory effects across areas hasnot been examined, and their physiological significanceis unclear. No electrophysiological effects have been seenin LGN. We examined intermodal selective attention ef-fects on processing in LGN, V1, V2, V4 and MT/MST in 3monkeys. Laminar EEG and action potential profileswere recorded with multielectrodes spanning the layersof each target structure. We used standard (at-tend-ignore) comparisons to detect attentional modula-tion of averaged responses in each area, and also appliedFourier and Wavelet analyses to single trial EEG. In theaveraged response analysis, attentional modulation wasdetected throughout the cortical pathways, but not in theLGN. In going from V1 to V4, effect magnitude in-creased, and latency decreased. These findings are con-sistent with a stimulus-driven feedback model ofattentional modulation. Concurrent surface ERP record-ings revealed a selection negativity, that appears to arisein part from feedback-mediated disinhibition ofextrastriate cortex. Our single trial analyses: 1) replicatethe earlier V4 finding, 2) find similar results in MT/MST,3) find more general enhancement in V2 and V1 (Thesepatterns are paralleled in prestimulus baseline activity),and 4) find an indication of attentional modulation inLGN: γ enhancement in P-laminae, and general suppres-sion in M-laminae. Prestimulus baseline effects through-out the system appear to reflect a tonic anticipatory biasof activity due to attention. Differential effects on low


and high frequencies in Area V4 may reflect modulationof inhibition versus excitation.

Neurophysiology of attending and ignoring. - G.V.Simpson (Dynamic Neuroimaging Laboratory, Depart-ment of Radiology, University of California San Fran-cisco).

We suggest that selective stimulus processing mayoperate by enhancing the cortical representations ofto-be-attended stimuli and suppressing the representa-tions of to-be-ignored stimuli (distractors). It can be ar-gued that ignoring distractors is a process thatautomatically accompanies attention, and is stimulusdriven. However, it is possible that ignoring may be un-der top-down control. We have investigated these issuesusing cued attention designs (S1-S2) in which a centrallypresented cue (S1) instructs the subject to deploy atten-tion, following a delay, the second stimulus (S2) is pre-sented. The S2 is either an attended stimulus or adistractor that must be ignored. The use of symboliccueing ensures that attentional deployment is undertop-down control. EEG measures following the cue re-veal several processes underlying attentional deploy-ment. The data support both rapid and slow deploymentmechanisms of attention under the same conditions.Slow deployment processes consist of frontal and pari-etal activity followed by sustained activity over the corti-cal representations of the to-be-attended location,consistent with a bias-signal. At the same time there is anincrease in alpha-band activity over the cortical repre-sentation of the to-be-ignored location (i.e. location of po-tential distractor stimuli), consistent with an activeprocess of ignoring. We have begun to examine the pro-cesses in common with sustaining attention during cuedelay periods and working memory in S1-S2 designs.Extrastriate intracranial recordings (in collaborationwith Gregor Rainer) explore the physiological signifi-cance of oscillatory activity by investigating the relationbetween oscillatory activity and single unit firing duringthese delay periods.

EEG and ERP signals of working memory. - A. Gevins(San Francisco Brain Research Institute and SAM Tech-nology, San Francisco, CA 94108; [email protected]).

Working memory, the conscious effortful processof maintaining and manipulating representations (in-formation) in one’s mind for several seconds, is argu-ably the most fundamental of higher cognitive brainfunctions. WM involves a hypothetical limited capacity"Central Executive" system and task-specific represen-tational processing systems. Neuroimaging studies ofWM tasks show activation of dorsolateral and medialprefrontal and parietal cortex, as well as task-specific ar-eas. Our lab has studied EEG and ERP signals of WM

since the mid "80s using the n-back task in which a re-membered stimulus, such as a letter or its position onthe screen, is compared with a new stimulus (Gevins etal. 1990). (The n in n-back refers to how many trials backthe remembered stimulus is.) This task is well con-trolled in that stimulus and response factors can be heldconstant while varying task difficulty and type of infor-mation. The n-back task evokes well-known EEG andERP signals (Gevins et al. 1990, 1996, 1997). The magni-tude of the continuous load on the attentional resourcesavailable for maintaining information in working mem-ory during the n-back task can be measured with fre-quency domain parameters of the ongoing EEG such asfrontal and parietal alpha band power and frontalmidline theta band power. ERP peak amplitudes and la-tencies are modulated by the transient allocation of at-tention to stimulus processing during the n-back task,in particular the competition between the concurrentdemands to maintain and update information whileevaluating and responding to new stimuli. The mainERP signals modulated by WM tasks are CNV, P250,P300 and Slow Waves (McEvoy, Smith and Gevins,1998). These EEG and ERP WM signals are stable afterpractice (Smith, McEvoy and Gevins 1999), have hightest-retest reliability (r>.90; p< .001) (McEvoy, Smithand Gevins 2000), and their modulation is consistentfrom childhood to old age (Pellouchoud, Smith,McEvoy and Gevins, 1999; McEvoy, Pellouchoud,Smith and Gevins 2001). The signals have high face va-lidity in that their modulation by variations in task diffi-culty can be accurately measured in individual subjects(>90%; p< .001) (Gevins, Smith, Leong, McEvoy et al.1998). They have high construct validity in that differ-ences between people in WM task performance andEEG and ERP signals are good predictors of awell-known measure of individual differences in cogni-tive ability, the Wechsler Adult Intelligence Scale IQscore (multiple R=.80; p < .001) (Gevins and Smith2000). Finally, they have high discriminative validity inthe sense that the n-back WM task and associated EEG &ERP signals are highly sensitive to transient and chronicchanges in neurocognitive function due to a variety ofstressors including: antihistamines, caffeine, alcoholand marijuana (Gevins and Smith 1999; Gevins et al.2001; Gevins, Smith and McEvoy 2002; Ilan and Gevins2001; Ilan, Smith and Gevins Submitted), anti-epilepticdrugs (Chung et al. 2002; Gevins, Meador, et al. InPrep.), and extended Wakefulness (Smith, McEvoy andGevins 2002). Together, this body of research resultssuggests that a clinical neurophysiological test of work-ing memory is scientifically feasible.

Supported by grants from NIMH, NINDS, NIA,NICHHD, NIAAA, NIDA & NHLBI.

Proceedings 195

Functional brain connectivity with high time resolu-tion sLORETA (Standardized Low Resolution BrainElectromagnetic Tomography). - Roberto D.Pascual-Marqui, Michaela Esslen, Kieko Kochi andDietrich Lehmann (The KEY Institute for Brain-MindResearch, University Hospital of Psychiatry, Zurich,Switzerland).

The human brain is comprised of a very large numberof highly interconnected neurons. Classical functional im-aging (PET and fMRI) has emphasized the localization offunction, but has practically neglected the role played byfunctional connections. Oscillations and synchronization,which emerge because of cortical interconnections, havebeen hypothesized to play an essential role in cognitionand consciousness. This work presents a method for thecomputation and imaging of cortical connectivity. 3Dfunctional imaging of electric neuronal activity is per-formed with standardized low resolution brain electro-magnetic tomography (sLORETA). This method isuniquely capable of exact localization. In addition, it hasthe lowest spatial dispersion as compared to other pub-lished 3D linear, discrete, distributed EEG/MEGtomographies (Pascual-Marqui, Method Find Exp. Clin.,2002, 24D: 5-12). sLORETA provides high time resolutionsignals of "virtually implanted electrodes" throughout thecortex. Computation of cortical connectivity is then basedon the following model: within short time windows, spa-tial clusters of cortical neurons have the same dynamics ofactivation, except for location-specific scale factors and forlocation-specific latency shifts. The method was tested in avisual event related potential experiment to righthemifield pattern reversal checkerboard stimulation,where time lagged connections from left to right visualcortices were demonstrated. In a single subject, eyesclosed alpha-EEG study, time lagged connections fromright to left posterior cortical areas were observed.

Time-frequency decomposition of ERPs from a visualoddball task: time-dependent delta, theta and alpha os-cillations. - E.M. Bernat*, S.M. Malone+, W.J. Wil-liams^, C.J. Patrick~ and W.G. Iacono# (*Department ofPsychology, University of Minnesota; + Department ofPsychology, University of Minnesota; ^Department ofEECS, University of Michigan; ~Department of Psy-chology, University of Minnesota; #Department of Psy-chology, University of Minnesota).

Recent work has shown coordinated delta, theta,and alpha activity during visual oddball tasks. To charac-terize this activity, single-trial time-frequency trans-forms (TFTs) were decomposed using PCA (Bernat,Williams, Gehring, Lorenz and Casey 2002). Data from1866 17yo participants in the Minnesota Twin FamilyStudy (979 males) from electrode Pz were used. Targetdata spanning 0-14 Hz and 1-1312.5 ms were chosen for

decomposition. Fifteen PCs evidenced coordinatedtime-dependent activity as follows. Prior to P300, alphadesynchronization was apparent in two alpha PCs (fastand slow) that transitioned to a single alpha PC endingjust before the P300 peak. Also, two delta PCs (100-400ms/1-2 Hz and 250-400 ms/2-3 Hz) and two interleavedtheta PCs occurred before P300. The P300 peak was asso-ciated with two delta PCs (375-475 ms/2-3 Hz and400-600 ms/1-2 Hz). After P300, two delta PCs were evi-dent (1 Hz/600-1000 ms and 1 Hz/1000 ms-end). Alpharesynchronization after P300 occurred in one alpha PCthat transitioned to fast and slow alpha PCs. Finally, atheta PC occurred after 900 ms. To assess stability of thesolution, odd and even halves of the dataset were decom-posed, producing structurally identical solutions. Targettask difficulty was assessed. Easy targets producedlarger and earlier alpha resynchronization, and simulta-neously greater delta and theta activity. Hard targetswere associated with greater delta activity precedingP300. In light of the sample size and solution stability, thedecomposition appears to offer a stable and rich descrip-tion of the coordinated TF ERP activity.

Mu-rhythm suppression and enhancement caused bytactile stimulation in healthy volunteers. - E.J.Marttinen-Rossi*+, H. Wikström*+, A. Korvenoja+ andS. Carlson^ (*BioMag Laboratory, Helsinki UniversityCentral Hospital, Helsinki Brain Research Center;+Functional Brain Imaging Unit, Helsinki UniversityCentral Hospital, Helsinki Brain Research Center; ^In-stitute of Biomedicine/physiology, University of Hel-sinki, Neuroscience Unit, Helsinki Brain ResearchCenter).

The mu-rhythm of the primary somatosensory cortex(S1) is momentarily suppressed (event-relateddesynchronisation (ERD)) and then enhanced(event-related synchronisation (ERS)) when sensory stimu-lation is applied to a peripheral site (1). A com-puter-controlled device, which mechanically stimulates theskin with a calibrated von Frey filament was recently devel-oped in our laboratory to study this phenomenon. The pro-duced stimulation is more natural than the commonly usedelectrical stimulation. Magnetoencephalograms from 13healthy right-handed volunteers were recorded during tac-tile stimulation (interstimulus interval 3 s) of the thenar ofthe right palm. The non-averaged signal from the sensorshowing maximum ERD or ERS in the contralateral cortexwas filtered using a wavelet centered at the peak frequencyof the mu-rhythm in two predefined frequency regions,8-14 Hz and 15-21 Hz. The average ERD in the first and sec-ond frequency regions were 28.3% and 25.7% respectivelyand the average latencies from the onset of the stimuli to thepeak response 313 ms and 219 ms respectively. Respectivevalues for ERS were 40.3% and 46.8% and for the latencies


833 ms and 689 ms. ERD and ERS were elicited in thecontralateral S1 and were identified in all subjects. ERD andERS are reliable indicators of activity in the S1 elicited bytactile stimulation. Compared to electrically induced ERD,ERD elicited by the von Frey filament appears to be shorterin duration (1). Stimulation with the von Frey filament haspotential applications in assessing normal and disturbedfunctions of the somatosensory cortex.

Nikouline et al. Neurosci. lett., 2000, 294: 163-166.

District-related frequency specificity in hand corticalrepresentation as revealed by somatosensory magneticfields. - Franca Tecchio*+, Claudio Babiloni^, FilippoZappasodi*+, Fabrizio Vecchio^, Carlo Salustri*,Vittorio Pizzella~, Gian Luca Romani~ and Paolo MariaRossini+ # % (*IFN-CNR, Roma; +AFaR-DipNeuroscienze, Osp. Fatebenefratelli, Roma; ^Ist.Fisiologia Umana, Univ. "La Sapienza", Roma; ~ITAB,Department of Clinical Sciences and Bioimaging , Uni-versity of Chieti, Chieti; #IRCCS "S. Giovanni diDio-Fatebenefratelli", Brescia; %Neurologia Clinica,Università Campus Biomedico, Roma, Italy).

Cortical sensory neurons synchronize their activity atmultiple frequency bands after an external stimulation. Inthe somatosensory cortical areas, previous reports de-scribe more discrete and somatotopically specific neuralsynchronization at the gamma band. Therefore, an effi-cient gamma synchronization of the neurons in primarysomatosensory cortex (S1) may be expected to characterizethe stimulus processing from the thumb, i.e., the hand’smost skillful area. To test this hypothesis, neuromagneticfields were evoked over human S1 by the electrical stimu-lation of the contralateral thumb or little finger. Neuronalsynchronization was indexed by the spectral coherence ofthe evoked neuromagnetic fields overlying S1. The fre-quencies of interest were the beta (16–32 Hz) and gamma(36–46 Hz) bands. The global amount of the coherence wasdefined as the total event-related coherence (ERCoh)among all magnetic sensors overlying the S1. Resultsshowed prevalent increment of beta ERCoh (20–32 Hz) af-ter the little finger stimulation and of gamma ERCoh(36–44 Hz) after the thumb stimulation. These results sug-gest that the neural synchronization in S1, as revealed bythe ERCoh, may vary in frequency as a function of the fin-ger stimulated. In this framework, the neural synchroniza-tion at gamma band may characterize the corticalrepresentation of thumb, functionally prevalent with re-spect to little finger in humans.

An ERP and scalp current density study of spatial orspectral processing of moving sounds. - A. Bidet-Cauletand O. Bertrand (Mental Processes and Brain Activa-tion Lab, INSERM U280, Lyon, France).

Human neuroimaging studies have suggested that

posterior temporo-parietal regions are involved in audi-tory spatial processing forming a putative dorsal"where" pathways. We used EEG to investigate the dy-namics of this network during the active or passive inte-gration of a varying acoustic stream in a two-factorparadigm: focused vs diverted attention, spatial vs spec-tral variations. A stream consisted of successive35-ms-noise bursts with varying location or pitch. Sub-jects had either to indicate the final direction of thestream (active integration) or to compare two additionalbursts at the end of the stream (diverted attention).Event-related potentials were recorded at 36 scalp elec-trodes in 14 subjects, and were analyzed by scalp currentdensity mapping (laplacian). During stream presenta-tion, a sustained bilateral temporal activity (polarity re-versals across Sylvian fissures) was present and morepronounced for pitch variations over the left hemi-sphere. When focusing on stream variations, a superiorparietal and two bilateral temporo-occipito-parietalcomponents were activated, showing no difference be-tween spatial and pitch processing. Thus, the active inte-gration of either spectrally or spatially moving soundsinvolved temporal and parietal cortices. These regionsare not uniquely implicated in auditory spatial process-ing but rather in the integration of rapidly varying acous-tic streams, thus questioning the actual nature of theauditory dorsal pathways.

Assessing cortical connectivity and information flowfrom scalp EEG data: a simulation study to address theproperties of Directed Transfer Function (DTF). - F.Cincotti*+, L. Astolfi+, F. Vecchio+, A. Basilisco+, C.Babiloni+^, F. Carducci+~, S. Salinari#, P.M. Rossini^~

and F. Babiloni+ (*Fondazione Santa Lucia IRCCS,Rome, Italy; +Dip. Fisiologia Umana e Farmacologia,Univ. "La Sapienza", Rome, Italy; ^AFaR, OspedaleFBF "S. Giovanni Calibita" CRCCS, Rome, Italy;~Ospedale FBF "S.Giovanni di Dio" IRCSS, Brescia, It-aly; #Dip. Informatica e Sistemistica, Univ. "LaSapienza", Rome, Italy).

Besides the research on "where" where sensorial orcognitive take place in the cerebral cortex, techniques thatmeasure the cooperation between cortical regions in carry-ing on these activities are recognized to be necessary to un-derstand the brain functions. The Directed TransferFunction (DTF) is a spectral index that measures the causalrelationship between pairs in a set of signals. Its use hasbeen proposed and applied to electroencephalographic(EEG) data. In this study the performances of this index ap-plied to cortical activity estimates obtained from EEG re-cordings are evaluated in a simulation setup. The accuracyof DTF estimation from finite and noisy recordings are firstinvestigated, evaluating the difference between estimatedvs. analytically computed values of DTF on data of differ-

Proceedings 197

ent length, corrupted by a variable amount of noise. Alsowe address the ability of (correctly computed) DTF of beingan index of functional coupling, in several cases in whichthe topology of the network ranges from simple cases oneto more complicated ones. The DTF is shown to be ratherrobust to noise and finiteness of data (8% RMSE in reason-able conditions). Thank to the factorial design of the simula-tion and to the ANOVA analysis of the results, we alsoclearly state the limits of the application of DTF the estima-tion of cortical connectivity in high resolution EEG studies.An application to real data is finally reported in order toshow an instance of connectivity estimation between corti-cal regions from non invasive EEG measurements.

Dynamic cortical patterns during conscious perceptionin binocular rivalry. - D. Cosmelli.*, O. David#, J.-P.Lachaux*, J. Martinerie*, L. Garnero*, B. Renault* andF. Varela*+ (+Deceased May 2001; *Cognitive Neurosci-ence and Brain Imaging Laboratory, CNRS UPR 640,Hôpital de La Salpêtrière, 47 bld de l’Hôpital, 75651Paris Cedex 13, France; #Wellcome Department of Im-aging Neuroscience, Functional Imaging Laboratory,12 Queen Square, London WC1N 3BG, UK).

In the last decades evidence has begun to converge onthe notion that the establishment of dynamic relations be-tween distant brain regions underlies conscious experi-ence. According to the synchronous ensembles hypothesissuch dynamical relations would be embodied by a tran-siently coactive network comprising distributed regions ofthe brain. Anyhow, the study of such integrative processesis far from achieved, specially in the case of spontaneouson-going perception. To be able to study this phenomenonin the human brain it is necessary to have both a good tem-poral and spatial resolution. The MEG/EEG inverse prob-lem represents a unique non-invasive approach towardsthis goal. We present an MEG study of binocular rivalry,an experimental paradigm that reveals the endogenousrhythmicity of perceptual experience under reproducibleconditions. Here, a subject is presented with non-fusibleimages, one to each eye, and his perception oscillates spon-taneously among both mutually exclusive possibilities.Our aim is the study of on-going neural patterns duringthe coming into consciousness of a given percept, from anintegrative and fundamentally dynamical point of view.Using an adapted inverse problem framework we revealspecific non-averaged patterns of synchronous brain activ-ity that follow dominance periods during visual rivalry inhuman beings. Such synchronous networks are centeredin visual cortex and extrastriate occipital dorsal areas, andcomprise parietal, inferotemporal and frontal regions.Moreover, the temporal evolution of the network suggestsa dynamical dialogue between the occipital, "visual" cortexand more frontal regions of the brain upon conscious vi-sual perception.

Learning by different strategies modulates prefrontaland temporal brain magnetic activity. - F. Maestú*+^, P.Campo*, P.G. Simos~, A. Capilla*+, N. Paul%, S.Fernandez^, A. Fernández*# , C. Amo*, J .González-Marqués+ and T. Ortiz*# (*Centro deMagnetoencefalografía Dr Pérez Modrego,Universidad Complutense Madrid; +DepartamentoPsicología Básica II (Procesos Cognitivos), Facultad dePsicología. UCM; ^Departamento de Psicología,Universidad Camilo José Cela, Madrid; ~University ofTexas-Houston Medical School, Vivian L. Smith Cen-ter for Neurologic Research, Department of Neurosur-gery; #Departamento de Psiquiatría, Facultad deMedicina, Universidad Complutense Madrid;%Unidad de daño cerebral, Hospital Beata María Anade Jesús, Madrid).

Three different verbal learning strategies (se-rial-order, phonological and semantic) were studied bymeans of magnetoencephalography. The course of acti-vation elicited by the serial-order strategy showed earlyactivation (between 200 and 400 ms after stimulus onset)of the inferior frontal, sensorimotor, and insular regionin the left hemisphere. Furthermore, activation was moreprominent in dorsolateral prefrontal cortices bilaterally.Conversely, activation profiles associated with the pho-nological strategy showed a preponderance activation ofthe superior temporal gyrus in the left hemisphere be-tween 500 and 600 ms. Finally, the semantic strategyshowed activation of the left middle temporal gyrus, be-tween 500 and 700 ms. As a conclusion, the level and typeof initial learning processes produce the engagement ofdifferent brain circuits.

Studying phase synchrony for classification of mentaltasks in brain machine interfaces. - E. Gysels*, J. del R.Millán+, S. Chiappa+, P. Celka* (*CSEM (Swiss Centrefor Electronics and Microtechnology), Neuchâtel, Swit-zerland; +IDIAP (Dalle Molle Institute for PerceptualArtificial Intelligence), Martigny, Switzerland).

Electroencephalogram recordings during imagina-tion of mental tasks allow for developing a new communi-cation device for, e.g., motor disabled people. 32-channelEEG was recorded from 5 healthy subjects while perform-ing, after instruction and in random order, repetitiveleft-hand movement imagination, right-hand movementimagination and "word generation". In 3 consecutivedays, 5 sessions of 4 minutes were acquired, without feed-back. Extracted features were based on the Phase LockingValue (PLV), coherence, and the spectral power in the α-,β1-, β2- and 8-30 Hz frequency bands. Different featuresubsets were considered for classification. The classifierconsisted of a combination of Support Vector Machines(SVMs), allowing for classifying the 3 mental tasks. Re-sults were obtained from 5-fold crossvalidation. We ana-


lyzed (offline) data from the last day of recording, whensubjects had a training experience of only 40 minutes fromthe two previous days. Selecting for every subject the bestfeature subset, often a combination of PLV and power fea-tures, we obtained for two subjects correct classificationrates of 67.81% and 62.58%, distinguishing the 3 mentaltasks. For the other subjects, classification rates were52.77%, 44.66% and 41.24% only, maybe due to notenough practice. The involvement of the motor cortexmanifested itself by a good performance of the features ex-tracted from electrodes located in the centro-parietal re-gion. Sole use of PLV or coherence features yielded goodclassification, proving phase synchronization appropriatefor classifying mental tasks. This shows the importance ofstudying functional relations between brain regions forfurther improvement of Brain-Machine Interfaces.

Maintenance of integrated and unintegrated informa-tion: a magnetoencephalography study. - P. Campo*, F.Maestú*+, A. Fernández*^, A. Capilla+, C. Amo* and T.Ortiz*^ (*Magnetoencephalography Center Dr. PérezModrego. University Complutense of Madrid; +De-partment of Basic Psychology. Cognitive Processes.University Complutense of Madrid; ^Department ofPsychiatry. University Complutense of Madrid).

Working memory (WM) is considered as the ability totemporary maintain and manipulate an active representa-tion of information. Sustained neural activity over mem-ory retention intervals in delayed tasks is generallyinterpreted as the neural basis of working memory. Evi-dence from studies with infrahuman primates as well ashumans revealed that working memory task performancerequired the activation of multiple widely distributed re-gions in posterior perceptual areas, motor cortices, andprefrontal cortex. Maintenance is considered a process ofkeeping information in mind in the absence of externalstimuli, and corresponds to the mnemonic aspect of theWM. Two basic processes are assumed to be involved inthe active maintenance of information: storage of materialand the rehearsal of material. In the present study partici-pants performed two different tasks in which they had tomaintain verbal (letters) and spatial (locations) informa-tion, either in an integrated or in unintegrated manner. Weinvestigated the spatiotemporal patterns that character-ized the maintenance of integrated information using aneuroimaging technique with a high temporal resolution,as is the magnetoencephalography (MEG). Results arediscussed in terms of the activation of different neural net-works. We also compared our results with those from afMRI study in terms of temporal resolution.

Hemispherical differences in middle latency auditoryevoked fields. - A. Kult*+, S. Supek+, P. Schneider*^,H.J. Specht^, H.G. Dosch^ and A. Rupp* (*Department

of Neurology, University Hospital Heidelberg, Heidel-berg, Germany; +Department of Physics, Faculty of Sci-ence, University of Zagreb, Zagreb, Croatia;^Department of Physics, University of Heidelberg,Heidelberg, Germany).

Middle latency auditory evoked fields (MAEFs) oc-cur about 20-80 ms after stimulus onset. Since the MAEFcomponents are relatively small, source identification inthat time interval is rather challenging. Some authors re-ported that the origin of the middle latency evoked activ-ity is located within the primary auditory cortex (PAC)[1,2] while other studies suggest that even the Pa peak,which has a latency of about 30 ms, is too late to be re-stricted to the activation from PAC [3] only. In this studywe tried to localize Pa and Pb sources (around 50 ms)separately and analyze their interhemispherical differ-ences. Stimuli were diotically presented, amplitudemodulated, sinusoids with carrier frequency of 1 kHzand a decaying exponential envelope of 4 ms half-lifetime. Evoked magnetic fields were recorded with awhole-head Neuromag-122 system from 10 normal hear-ing subjects. Spatio-temporal analysis confirmed that thePa source received contribution of more than one corticalarea, only one of which originated from the anteromedialportion of Heschl’s gyrus (HG) which is considered to bePAC. Pb sources were found on the anterolateral portionof HG. Interhemispherical amplitude and latency differ-ences were observed in the measured fields as well as theidentified sources. The MRI-based volumetry suggeststhat these neurophysiological differences might be ex-plained by anatomical differences in HG morphology.

Scherg, M. et al. Advances in Biomagnetism, NY:Plenum Press, 1989: 97-100.

Mäkelä, J. et al. Electroenceph. Clin. Neurophysiol.,1994, 96: 414-421.

Lütkenhöner et al. NeuroImage, 2003, 18: 58-66.

The human SI activity is related to the input stimulusfrequency and perception in a frequency discrimina-tion task. - Li Chan Liu , Peter B. C. Fenwick, NikolaosA. Laskaris, Marc Schellens, Vahe Poghosyan,Tadahiko Shibata and Andreas A. Ioannides (Labora-tory for Human Brain Dynamics, RIKEN Brain ScienceInstitute (BSI), 2-1 Hirosawa, Wakoshi, Saitama,351-0198, Japan).

We used magnetoencephalography (MEG) to inves-tigate non-invasively whether activity in primary sen-sory area (SI) of humans is shaped exclusively by theinput or it also reflects the subjective perception of thestimulus. Four right-handed males participated in a fre-quency discrimination task to detect frequency changesof an electrical stimulus applied to the right-hand digits2+3+4. Pairs of Stim1 (1s; 21 Hz) and Stim2 (1s; 21 Hz(50%) or 22 to 29 Hz) were used. We analyzed the single

Proceedings 199

trial MEG data using tomographic and pattern analysis,and correlated brain activations with the subjects’ re-sponses. The left-SI was the most consistently activatedarea and showed the first activation peak at 35-48 ms afterStim1 onset and sustained activity during both stimulusperiods. During the Stim2 period the left-SI activationstarted to differ significantly between two groups of trials(21 versus 26-29 Hz) within the first 100 ms and this differ-ence was sustained and enhanced thereafter (about 600ms). When only correct responses from the above twogroups were used, the difference was even higher at laterlatencies (>650 ms). For Stim2 delivered at 21 and 26-29Hz but both perceived as 21 Hz, the sustained differencewas present only before 650 ms. In addition to the ex-pected SI involvement with the analysis of input fre-quency, our results show a modulation of SI activity thatdepends on the subject’s perception and decision at latelatencies. This is the first evidence from neuroimagingshowing perception related correlates in the activity of SI.

Involuntary crossmodal enhancement of preattentiveauditory processing is selective in space. - K. Mathiak*,I. Hertrich*, W.E. Kincses+, S. Rothe+, H. Menning*, W.Lutzenberger* and H. Ackermann* (*MEG center,Dept. Neurology, Univ. Tübingen, Germany,+DaimlerChrysler AG, Stuttgart, Germany).

Different sensory pathways are known to interactwithin the central nervous system. Spatially divergentstimuli may delay responses, whereas a cohesive stimula-tion design can support correct responses and speed upreaction times. In animal studies, oculomotor, visual, andsomatosensory systems were shown to affect auditoryprocessing through distinct subcortical and early corticalpathways. The current study focused on cross-sensoryinfluences on preattentive auditory processing, which de-tects change and facilitates orienting reactions. We em-ployed an oddball paradigm to assess cortical processingusing whole-head magnetoencephalography (MEG) intwo series of 20 volunteers each. While subjects per-formed distraction tasks of varying difficulty, auditoryduration deviants occurred randomly at the left or theright ear preceded (200 – 400 ms) by oculomotor, visual,or tactile co-stimulation at either side. Mismatch fieldswere recorded over both hemispheres. Changes in gazedirection and static visual stimuli cued the most reliableenhancement of deviance detection at the same side, andmost strongly at the right auditory cortex. Tactile stimuliat either cheek enhanced the processing of auditoryevents from the respective ear by the ipsilateral hemi-sphere. In all three modalities, the lateralized unattendedand unpredictive pre-cues acted analogously to shifts inselective attention, but were not reduced by theparametrized cognitive load. Thus, early cognitive repre-sentation of localized sounds interferes with other modal-

ities. We suggest that preattentive multisensoryintegration provides a basis for orienting to objects inspace and, thus, for selective attention.

Comparison between tone and phonetic MMN usingdichotic listening task. - J. Shinoda, K. Nakagome, Y.Kuwakado, D. Muramatsu, T. Takahashi, S. Ikezawa,S. Kamio and M. Katoh (Showa University School ofMedicine, Tokyo Metropolitan Bokutoh Hospital).

It has been acknowledged that dichotic listening task(DLT) is well suited for detecting the interhemisphericdifference of brain activity. The aim of this study is to de-tect the difference in lateralization of automatic mismatchprocessing between tone and phonetic MMN as well asthe similarity of brain regional activities using DLT. Thetone MMN task comprised 3 sets of stimuli; 1 kHz stimulipresented to both ears (standard), 1 kHz to the left ear and2 kHz to the right (right deviant), 1 kHz to the right and 2kHz to the left (left deviant). In the phonetic MMN task, 1kHz and 2 kHz tones were substituted to syllables such as"ba" and "da", respectively. The subjects were engaged ina visual task ignoring all the acoustic stimuli throughoutthe recording. As a result, although there was no signifi-cant difference between the tones and phones in terms ofdiscriminability of deviants and standards behaviorallyconfirmed in the pre-test session, the phonetic MMN wassignificantly larger than the tone MMN. EEG topogra-phies indicated that the phonetic MMN was distributedpredominantly in the left hemiscalp, whereas the toneMMN was distributed bilaterally. SCD analyses revealeda prominent current sink in the left temporal region forthe phonetic MMN, whereas it was not evident for thetone MMN. The left hemispheric predominance of thephonetic MMN in DLT was in accordance with the previ-ous reports, suggesting that it may be a good biologicalmarker for left hemispheric dysfunction, which is fre-quently mentioned in schizophrenia.

Rapid change of tonotopic maps in the human auditorycortex during pitch discrimination. - I. Ozaki*, C.Y.Jin+, Y. Suzuki^, M. Baba~, M. Matsunaga~ and I.Hashimoto# (*Department of Physical Therapy, Fac-ulty of Health Sciences, Aomori University of Healthand Welfare, Aomori, Japan; +Department of Neuro-logical Science, Institute of Brain Science, HirosakiUniversity School of Medicine, Hirosaki, Japan; ^De-partment of Social Welfare, Faculty of Health Sciences,Aomori University of Health and Welfare, Aomori, Ja-pan; ~Department of Neurological Science, Institute ofBrain Science, Hirosaki University School of Medicine,Hirosaki, Japan; #Human Information Systems Labora-tory, Tokyo Office, Kanazawa Institute of Technology,Tokyo, Japan.

To study early cognitive processes and hemispheric


differences in the primary auditory cortex during selec-tive attention, we measured auditory evoked magneticfields (AEFs) to 400 Hz and 4000 Hz tone pips that wererandomly presented at right or left ear. Subjects paid at-tention to target stimuli during pitch (high or low) orlaterality (left or right) discrimination tasks. In controlsession, 400 Hz or 4000 Hz tone alone was presented atleft or right ear. We calculated location and strength ofN100m dipole for 400 Hz and 4000 Hz tones, based on theAEFs obtained from the hemisphere contralateral to thestimulated ear. N100m amplitude increased in bothhemispheres in pitch or laterality discriminating condi-tions. N100m latency shortened in the right auditory cor-tex during selective attention. The N100m dipoledistance between 400 Hz and 4000 Hz tones was enlargedin the right auditory cortex during pitch discriminationtask but was unchanged during laterality discriminationtask. We conclude that these dynamic changes in theN100m dipole reflect short-term plastic changes in theprimary auditory cortex and right hemisphere domi-nance of attention, supporting early selection models.

Change in visual brain responses during discrimina-tion of the differences in color or shape of colored let-ters. - I. Ozaki*, C.Y. Jin+, Y. Suzuki^, M. Baba~, M.Matsunaga~ and I. Hashimoto# (*Department of Physi-cal Therapy, Faculty of Health Sciences, Aomori Uni-versity of Health and Welfare, Aomori, Japan;+Department of Neurological Science, Institute ofBrain Science, Hirosaki University School of Medicine,Hirosaki, Japan; ^Department of Social Welfare, Fac-ulty of Health Sciences, Aomori University of Healthand Welfare, Aomori, Japan; ~Department of Neuro-logical Science, Institute of Brain Science, HirosakiUniversity School of Medicine, Hirosaki, Japan; #Hu-man Information Systems Laboratory, Tokyo Office,Kanazawa Institute of Technology, Tokyo, Japan).

To study early cognitive processes in the visual cor-tex during selective attention, we measured visualevoked magnetic fields (VEFs) to 4 colored letters (C+Gor G+C in green or red letters) that were randomly pre-sented on the screen at 2 sec interval (0.2 sec presenta-tion). Subjects paid attention to target stimuli duringcolor (red or green) or shape (G or C on the left side) dis-crimination tasks. Total number of the stimuli was 300:30×2 for target and 120×2 for non-target. In control ses-sion, each colored letter alone was presented. The meanreaction time for color discrimination task was 322-333msec and that for shape discrimination task, 392-411msec. We calculated location and strength of the dipolesthat were identified at 70-500 msec for each colored letteras control and non-target responses, based on the VEFsobtained from the occipital and/or temporal area. Forcolor discrimination, following V1/2 activation, activa-

tion of V4 area (the fusiform gyrus) was identified. Forshape discrimination, following V1/2 activation, the lat-eral occipital complex (the occipito-temporal area) wasactivated. On the other hand, in control session, the V4area or the lateral occipital complex was rarely activated.In addition, the strength of the dipoles at the V4 area orthe lateral occipital complex was much larger in selectiveattention than in control. We conclude that, even whensubjects look at the same colored letter, selective atten-tion to color or shape modifies the information process-ing pathways in the visual brain as the top-downmechanism, supporting early selection models.

Somatosensory Evoked Fields (SEFs) following fingerstimulation are changed by tongue movement or tactileinterference to cheek. - I. Ozaki*, C.Y. Jin+, M. Baba+,M. Matsunaga+ and I. Hashimoto^ (*Department ofPhysical Therapy, Faculty of Health Science, AomoriUniversity of Health and Welfare, Aomori, Japan; +De-partment of Neurological Science, Institute of BrainScience, Hirosaki University School of Medicine,Hirosaki, Japan; ^Human Information Systems Labo-ratory, Tokyo Office, Kanazawa Institute of Technol-ogy, Tokyo, Japan).

Phantom fingers following arm amputation areknown to be frequently represented on the cheekipsilateral to the amputated arm, suggesting close neuralconnections in cortical areas between the fingers andcheek. To test the hypothesis above, we analyzedsomatosensory evoked magnetic fields (SEFs) to fingerstimulation with or without tactile interference to thecheek ipsilateral to finger stimulation or the tonguemovement. To elicit SEFs, electrical stimulation at threetimes of the threshold was applied at the rate of about 1Hz to the fingers I, II, III and V independently. SEFs weremarkedly modified by tactile interference to the cheek orthe tongue movement. For the interference conditions,SEF power was significantly reduced at around 100 msecwhere the posterior parietal cortex or SII was activated incontrol condition. Especially, the response of SII that wasidentified in control condition diminished in the interfer-ence conditions. Furthermore, the reduction in SEFpower was of the same magnitude among the stimulatedfingers or between tactile interference to the cheek andthe tongue movement. On the other hand, the response ofSI (<50 msec) was sometimes unaltered in the interfer-ence conditions. We conclude that both tactile interfer-ence to the cheek and the tongue movement modify theresponse of SII or the posterior parietal cortex more thanthat of SI. We therefore speculate that phantom fingersfollowing arm amputation may be the results of alteredneural network activities at SII region where activation ofa cheek representing area spreads horizontally to the ad-jacent, formerly hand representing area.

Proceedings 201

The processing of semantic meaning in Chinese wordsand evoked brain topography. - W. Skrandies*, M.-J.Chiu+ and Y. Lin+ (*Institute of Physiology,Justus-Liebig University, Giessen, Germany; +Depart-ment of Neurology, NTUH, Taipei, Taiwan).

The connotative meaning of words can be quantifiedstatistically by the "semantic differential technique" re-sulting in statistically defined, independent dimensionswhere every word is uniquely located on the three di-mensions evaluation ("good - bad"), potency ("strong -weak"), and activity ("active - passive"). In an earlierstudy employing German subjects we demonstrated thatthere are electrophysiological correlates of these mean-ing dimensions. Here a group of 55 Chinese adults wasinvestigated in two experiments: first, 210 nouns wererated by 32 subjects, and factor analysis on the question-naire data yielded three independent semantic dimen-sions. Semantically unique words stemming from theseresults were then used as stimuli in electrophysiologicalexperiments in another group of 23 healthy right-handedadults. Words of similar physical appearance belongingto different semantic classes were presented visually inrandom order. The EEG was recorded from 32 channels,and evoked brain activity was computed for each seman-tic class. Significant differences in electrical brain activa-tion between semantic word classes were observed asearly as 80 ms after stimulus onset confirming earlier re-ports on similar findings in German subjects. These re-sults illustrate similar early neural activation in subjectgroups of different language and culture.

Supported by DAAD and NSC

EEG early evoked gamma-band synchronization re-flects object recognition in visual oddball tasks. -Gabor Stefanics*, Attila Jakab*, László Bernáth+,Lóránd Kellényi* and István Hernádi*^ (*Dept. ofGeneral Zoology and Neurobiology, Faculty of NaturalSciences, University of Pécs, Hungary, +Institute ofPsychology, Faculty of Humanities, University of Pécs,Hungary, ^Dept. of Anatomy, University of Cam-bridge, UK).

In a series of different visual EEG experiment theearly evoked gamma activity was found to be insensitiveto stimulus type and was suggested to reflect sensorybinding mechanisms. However, repetition of both targetand nontarget stimuli occured during stimulus presenta-tion in previous studies. Repetition of stimuli usuallyleads to a phenomenon often referred to as repetion prim-ing and may reflect a neural savings mechanism. Never-theless, the functional role of early evoked gammaactivity is still not clear. In our study EEG was recorded in3 visual oddball experiments during presentation of nat-ural photos of butterflies and plants in order to study theearly gamma activity evoked by familiar and novel stim-

uli during visual processing of biologically relevant com-plex objects. In all three experiment picture of one specificbutterfly served as target repeating 60 times and subjectstask was to silently count them. In Exp 1 neutral stimuliwere 180 individial pictures of butterflies from other spe-cies, in Exp 2 neutral stimuli were 180 indvidual picturesof plants and in Exp 3 both neutral stimuli were applied.Evoked synchronization (inter-trial coherence orphase-locking factor) was analyzed by sinusoid waveletbased time-frequency analysis. Consistent with otherstudies, significant phase-locked γ-synchronization wasfound from 80 to 140 ms postimulus in the 30-50 Hz rangeat parietal and occipital sites in response to the repeatingtarget stimulus followed by strong desynchronization inγ-band. Nontarget stimuli did not evoke similar activityin the γ−frequency range. The observed difference can beexplained if we assume that the repeated experience of anobject may lead to rapid formation of a neural assemblyrepresenting the object causing the repetition priming ef-fect. In our study the single target stimulus was intro-duced to the subjects before the expirement whereasindividual nontargets were unfamiliar images. Thus,subjects probably had a neural representation of the tar-get only. We suggest that the early phase-locked γ-activ-ity in the 30-50 Hz range might reflect the top downactivation of the neural representation of the familiar tar-get.We found strong evoked γ-band synchronizationwhich was significantly higher for target than nontargetstimuli. γ-synchrony was reported to reflect sensory andtask processing and was suggested to be involved intop-down processing. In line with previous findings theevoked γ-synchrony was sensitive to stimulus type prob-ably reflecting attentional mechanisms.

Neuromagnetic studies on emotional and invertedfaces in an oddball paradigm. - A. Susac*, R.J.Ilmoniemi+^, E. Pihko+ and S. Supek* (*Department ofPhysics, University of Zagreb, Zagreb, Croatia;+BioMag Laboratory, Helsinki University Central Hos-pital, Helsinki, Finland; ^Nexstim Ltd., Helsinki, Fin-land).

Faces are socially very important visual stimuli andthe detection of a change in face identity or facial expres-sion is an evolutionary very important skill. Non-targetneutral face presented as a deviant in an oddball paradigmwith happy face standard and face with glasses target de-viant elicited negative shift around 280 ms in EEG (electro-encephalography) an MEG (magnetoencephalography)recordings [1]. If the neutral face of another person wasused as the standard the amplitude of the observednegativity was smaller and at a later latency. Face inver-sion is known to disrupt face processing. The aim of thepresent study was to investigate the effect of face inversionon mismatch activity [1]. Neurodinamic measurements


were conducted at the BioMag laboratory using306-channel Vectorview system in the parallel with a60-channel EEG. MRI scans were available for all 8 youngmale subjects that participated in the study. Thespatio-temporal localization of neuromagnetic sources ofthe observed effect was carried out with standardNeuromag software. Four experimental conditions wereemployed: 1) happy standard, neutral deviant; 2) neutralstandard, neutral deviant; 3) inverted happy standard, in-verted neutral deviant; 4) inverted neutral standard, in-verted neutral deviant. The task was silent counting of theface with glasses deviant in all conditions. Mismatch activ-ity observed in condition 1 and 2 was attenuated or disap-peared in condition 3 and 4. This paradigm providesmeans for studying inversion effect on recognition of facialidentity and facial expression.

[1] Susac, A., Ilmoniemi, R.J. and Supek, S. Emo-tional faces in the visual oddball paradigm: A possiblemismatch negativity (in preparation)

Dynamic movement of N100m dipoles in evoked mag-netic field reflects sequential activation ofisofrequency bands in human auditory cortex. - Y.Suzuki*, I . Ozaki+, C.Y. Jin^, M. Baba^, M.Matsunaga^ and I. Hashimoto~ (*Department of SocialWelfare, Faculty of Health Sciences, Aomori Univer-sity of Health and Welfare, Aomori, Japan; +Depart-ment of Physical Therapy, Faculty of Health Sciences,Aomori University of Health and Welfare, Aomori, Ja-pan; ^Department of Neurological Science, Institute ofBrain Science, Hirosaki University School of Medicine,Hirosaki, Japan; ~Human Information Systems Labora-tory, Tokyo Office, Kanazawa Institute of Technology,Tokyo, Japan).

To investigate spatiotemporal features of theisofrequency bands for 400 Hz and 4000 Hz tones in hu-man auditory cortex and on the hemispheric differencesin the arrangement of the isofrequency bands, we re-corded auditory evoked magnetic fields (AEFs) to 400 Hzor 4000 Hz tone pips presented at right or left ear from 31normal subjects. The dipole location for the N100msources was successively calculated from the AEFs ob-tained from the hemisphere contralateral to the stimu-lated ear. In the right hemisphere, the current sources for400 Hz and 4000 Hz moved toward anterolateral direc-tion before the N100m peak, showing parallel arrange-ment of the isofrequency bands (4000 Hz in mediallocation). In the left hemisphere, the movement directionof 400 Hz dipoles was anterolateral, while that of 4000 Hzdipoles was lateral. This difference in the organization ofisofrequency bands between right and left auditory corti-ces reflects distinct functional roles in auditory informa-tion processing such as pitch vs. language discrimination.

Spatiotemporal dynamics of lexico-semantic process-ing in left perisylvian cortex. - T.W. Wilson*+^, P.J.Pardo*+~ and A.C. Leuthold*+# (*Brain Sciences Center,Veterans Affairs Medical Center, Minneapolis, Minne-sota; +The Domenici Research Center for Mental Ill-ness; Departments of ^Psychology, ~Psychiatry and#Neuroscience, University of Minnesota).

Numerous functional neuroimaging experimentshave investigated the neural substrates oflexico-semantic processing. These predominantly meta-bolically-based studies have greatly enhanced our un-derstanding of the neuroanatomy subserving linguisticfunctions, but have not provided insight into the tempo-ral dynamics that likely mediates our language process-ing capacities. Moreover, although much investigated,the precise neural correlates remain an area of contentiondue to the inconsistent results that have plagued this do-main of inquiry. The present study furthers this researchby focusing on the temporal behavior of neural re-sponses common across-condition in a within-subjectdesign. Using whole-head magnetoencephalography,we recorded the brain processes evoked when partici-pants view high-frequency concrete nouns, pronounce-able pseudowords, and consonant strings (i.e.,nonwords). In each participant (n=10), dipole-modelingtechniques were employed to characterize thespatiotemporal dynamics of word-level language pro-cessing. Within-subject, we observed substantial over-lap in the tissue underlying word and pseudowordprocessing; however, these left perisylvian neural re-gions were recruited much earlier (~100 msec) in the pro-cessing of words relative to pseudowords. In addition,the cell populations activated by pronounceable stimuliexhibited reasonable spatial variance when comparedacross-subject, which may reflect the origin of the repli-cation problems emerging in metabolic studies. Thenonword stimuli activated mostly right hemispherestructures, thus displaying little systematic relationshipwith the other stimuli classes. We conclude that all pro-nounceable stimuli activate a common neural circuit inleft perisylvian cortex dedicated to lexico-semantic pro-cessing, and that the time-course by which this circuit isrecruited may vary as a function of the linguistic stim-uli’s frequency.

Fetal and neonatal magnetic auditory evoked re-sponses. - R.T. Wakai (Department of Medical, Univer-sity of Wisconsin-Madison, Madison, WI, USA).

Over the last few years, there has been growing in-terest in recording fetal brain activity with MEG. In thefetus, magnetic detection provides much higher signalquality than electric detection. Furthermore, MEG is anideal technology in that it records neuronal activity di-rectly, has high temporal resolution, and is completely

Proceedings 203

safe and noninvasive. Our group has mainly attemptedto record fetal auditory evoked responses (AERs). Of thevarious AERs, long-latency cortical responses have beenthe easiest to record in the fetus. In one study, fetal AERrecordings were obtained from 23 fetuses at gestationalages 28-40 weeks, using a 37-channel biogradiometer.The stimuli were 1500 Hz tone bursts of duration 50 msecand intensity 100 dB. The fetal AERs had amplitudes onthe order of 25 fT and were recorded with a success rateof approximately 50%. The responses were dominatedby a component with approximate latency 250 msec. Thefetal AERs were consistent with AERs recordedpostnatally from the same subjects. Generally, AER la-tency decreased with increasing conceptional age; how-ever, the polarity of the dominant component appearedto reverse polarity from N to P during the last month ofpregnancy. Fetal MEG research is at a very early stage.Signal quality and detection rate need to be improved.We have shown, however, that under favorable circum-stances fetal MEG signals can be recorded with reason-able signal-to-noise, allowing initial studies of humanbrain activity throughout the perinatal period.

MEG recordings of fetal and infant auditory short-termmemory. - M. Huotilainen (Cognitive Brain ResearchUnit, Department of Psychology, University of Hel-sinki, Finland).

The functioning of the auditory short-term memoryis manifested in the mismatch negativity (MMN)event-related response in adults and children of variousages. MMN correlates well with behavioural measures ofauditory change detection, and measures of higher cogni-tive abilities needed in, for example, speech perception.MMN is also promising in studying defects in general cor-tical functions, decay of memory trace duration in, for ex-ample, Alzheimer’s disease, and problems of specificauditory perceptual abilities in, for example, autistic andAsperger children. MMN can be recorded also in infantsand even in prematurely born babies.Magnetoencephalography (MEG) is a non-invasivemethod to record the neuronal activity. We aimed at de-termining whether the magnetic counterpart of the MMN,the MMNm, could be obtained in infants of 1 to 5 days old,and in fetuses of 35 to 40 weeks GA. We were able to obtainstatistically significant MMNm responses in both subjectgroups. We regard this result as a promise for the future ofneonatal and fetal MEG since it proves that even such cog-nitive skills as change detection and short-term memorycan be studied in these subject groups.

Optical activation studies of newborn infants. - I.Nissilä, *, K. Kotilahti*, M. Huotilainen+, L. Lipiäinen*,T. Noponen*, V. Fellman^~ and T. Katila* (*Laboratoryof Biomedical Engineering, Helsinki University of

Technology, Finland; +Cognitive Brain Research Unit,Department of Psychology, University of Helsinki,Finland, ^Hospital for Children and Adolescents, Uni-versity of Helsinki, Finland, ~Department of Pediat-rics, Lund University, Sweden).

Imaging techniques based on near-infrared spectros-copy (NIRS) have been researched for the past decade foruse in functional brain imaging. Due to the different ab-sorption spectra of oxy- and deoxyhemoglobin, it is possi-ble to determine the local changes in these concentrationsusing at least two wavelengths in the near infrared. In ad-dition to changes in attenuation, frequency-domain opti-cal imaging devices can measure the mean opticalpathlength of the photons between the source and the de-tector. Optical imaging is at least in theory well suited forstudies on neonates, who can be investigated duringsleep. The optical attenuation of tissue is lower than withadults, and the cortex is closer to the surface of the head,allowing a higher signal-to-noise ratio. The compactnessof optical imaging technology may lead to bedside appli-cations. We have studied neonates using auditory,somatosensory, and visual stimuli. The idea is to comparethe block and event-related paradigms, and study the reli-ability and accuracy of optical imaging for the study ofcortical hemodynamics and brain activity in neonates. Weplan to compare responses of full-term neonates topre-term ones, as well as young children and adults. Si-multaneously recorded EEG will be used as a referenceduring the event-related optical studies.

Quantitative EEG norms for the first year of age. - T.Harmony*, L. Díaz-Comas+, E. Porras*, S. Ocampo*, E.Santiago*, L. Galán+ and T. Fernández* (*Institute ofNeurobiology, National Autonomous University ofMexico, +Cuban Neuroscience Center).

Quantitative EEG norms for school-age childrenhave shown to be very useful for the evaluation of manydisorders, as learning disabilities, attention deficit disor-ders, mental retardation, etc. We are involved in a projectfor early diagnosis and neurohabilitation of children withprenatal and perinatal risk factors of brain damage. Ourmain purpose is to try to reduce, as much as possible, neu-rological sequels in children. For this reason very earlytreatment should continue for at least one or two years.Therefore it is necessary the longitudinal evaluation of theinfants. The obtention of a normative data base might bevery useful for these evaluation. The EEG of normal in-fants between oneweek and one year of age wasrecordedin the 10/20 system. Children had no pathologi-cal antecedents and a normal neuropediatric examination.24 EEG segments of 2.56 s were edited and analyzed. Spec-tral parameters each 0.78 Hz were calculated.AbsolutePower, relative Power and Mean Frequency forthe classical EEG bands were obtained. Regression equa-


tions in function of age were obtained for the narrow spec-tral power and for the broad band parameters in thereferential (A1A2), Laplacian and Average montages.Topographic maps of the raw and Z values for absoluteand relative power and mean frequency were developed.

This project was partially supported by PAPIITIN231202 and CONACYT 36263M Grants. We acknowl-edge technical assistance of Ms Rosa Ma Rodriguez andEng. Hector Belmont.

Brain’s plastic changes caused by speech interventionin bilingual SLI children. - Elina Pihko*+, AnnikaMickos^, Teija Kujala~+, Paavo Alku#, Roger Byring^and Marit Korkman% (*BioMag Laboratory, HelsinkiUniversity Central Hospital; +Helsinki Brain ResearchCentre; ^Folkhälsan; ~CBRU, University of Helsinki;#Helsinki University of Technology; %Åbo Akademi).

We studied the effect of a new intervention programon the ability of bilingual children to discriminate speechsounds. In addition, we studied if the possible rehabilita-tion effect is reflected in the brain activity measured bymagnetoencephalography (MEG). Subjects were Finn-ish-Swedish preschool children with specific languageimpairment. Two sets of syllables (/su/- standard, /so/and /sy/-deviants, and /da/-standard, /ba/ and/ga/-deviants) were used as stimuli in the oddball para-digm. After the initial MEG measurement behavioral teston sound-pair discrimination was carried out. After an in-tervention period of 8 weeks MEG and behavioural testswere measured again. Half of the children had lan-guage-related intervention, the other half a physical exer-cise program. The language-group improved onbehavioural sound discrimination of one pair(/da/-/ba/). This improvement was accompanied by in-creased dipole moments to the /ba/-deviant on bothhemispheres (P1m at 210 ms). In addition, dipole mo-ments of P1m after /ga/, and of MMNm after /sy/ (atabout 520 ms) were increased in the right and left hemi-sphere, correspondingly. The results show that childrenreceiving language-related intervention improved on dis-criminating sounds according to the behavioural tests. Inaddition, this positive rehabilitation effect was reflected intheir brain activity. Both hemispheres seem to participatein the analysis of the speech sounds in these children.

Selective attention induces immediate changes in fin-ger representation in human area 3b. - Isao Hashimoto*and Yoshinobu Iguchi+ (*Human Information SystemsLaboratory (Tokyo Office), Kanazawa Institute ofTechnology, Tokyo, Japan; +Department of IntegratedNeuroscience, Tokyo Institute of Psychiatry, Tokyo, Ja-pan).

In order to clarify whether selective attention in-duces immediate changes in finger representation in hu-

man area 3b, somatosensory evoked magnetic fields(SEFs) were measured during selective attention tasks.In one experiment, we used vibratory stimuli to the indexor middle finger with a frequency of 100 or 400 Hz for se-lective discrimination of finger and frequency attributeof stimuli. Equivalent current dipole (ECD) for the firstM50 SEF component originating in area 3b indicated thatthe cortical regions for the fingers were specifically seg-regated in the finger discrimination task but not in thefrequency discrimination task or in the control condition.The task dependent and immediate switchover of corti-cal finger-representation demonstrates a dynamic SI ac-tivation for spatial information processing. In anotherexperiment, two tasks were performed under simulta-neous electric stimulation to the index and middle fin-gers. In the attention task, tactile stimuli were randomly(20%) presented to one of the two fingers simultaneouswith the two-finger electric stimuli and the subjects wererequested to respond to the tactile stimuli in order to in-duce selective attention to the target finger. In a controltask, the subjects ignored the tactile stimuli. In spite ofthe two-finger stimulation, the M50 ECD localizationshowed that the attention task induced a distinctive cor-tical somatotopy for the target finger, while the controltask did not. The M50 amplitude showed no change be-tween the attention and control task. These results sug-gest that selective attention not only enhances theactivity of the representation for the target finger but alsoactively suppresses that for the non-target finger.

Pharmaco-EEG of GABAA-receptor modulators and itsapplication in experimental epilepsy research. – RobA. Voskuyl*+ and Meindert Danhof* (*Leiden/Amster-dam Center for Drug Research, Leiden University;+Dutch Epilepsy Clinics Foundation, Heemstede, TheNetherlands).

Benzodiazepines are allosteric modulators of theGABAA receptor, exhibiting anxiolytic, anti-epileptic,hypnotic and anaesthetic properties. In humans and ani-mals they increase at moderate doses the β frequencyband (11.5-30 Hz). This EEG change serves as effect mea-sure in pharmacokinetic-pharmacodynamic (PK-PD)modelling, which is used to characterize the concentra-tion-pharmacological effect relation of drugs in terms ofreceptor activation and the ensuing transduction path-way. The in vivo concentration-EEG effect relation is sig-moid and can be described by the Hill equation.Investigation of full, partial, silent and inverse agonists,demonstrated good correlation between in vivo parame-ters and receptor affinity and intrinsic efficacy. Thus, theβ band is a good biomarker for enhancement ofGABAA-mediated inhibition. Neurosteroids are morepotent modulators, displaying a bi-phasic EEG effect,i.e. at high concentrations the amplitude decreases. This

Proceedings 205

was modelled by describing the transduction step by aparabolic function. This confirmed two predictions: (1)the in vivo potency of different neurosteroids correlatedwell with in vitro binding and (2) compounds with alower intrinsic efficacy, such as benzodiazepines, do ex-hibit a monophasic concentration-effect relation. Rela-tive to neurosteroids, benzodiazepines behave as partialagonists. Changes in concentration-effect relations mayalso reveal disease progression and associated pro-cesses. In kindled rats, which exhibit only increased sei-zure susceptibility, and in kainic acid treated rats, whichdevelop spontaneous seizures, the benzodiaz-epine-induced increase in β activity is reduced or virtu-ally eliminated respectively. This result suggests thatimpairment of GABAergic inhibition is in part responsi-ble for pharmacoresistance development in certaintypes of epilepsy.

QEEG predictors of treatment response. - L.S. Prichepand E.R. John (Brain Research Laboratories, Dept. Psy-chiatry, NYU School of Medicine, NY, NY, USA andNathan S. Kline Institute for Psychiatric Research,Orangeburg, NY, USA).

It has been demonstrated that clinically homoge-neous populations may contain heterogeneous quanti-tative electrophysiological profiles. Such findings aresuggestive of distinctive patterns of underlying brainpathophysiology, which could be expected to be relatedto differential treatment responsiveness. Using QEEGfeatures derived from 19 channels of eyes closed restingartifact-free EEG collected from drug-free states; wehave shown signif icant relat ionship betweenelectrophysiological subtype membership (multivariateQEEG profiles of abnormality) and prediction of treat-ment outcome for many diagnostic groups. In a study ofObsessive Compulsive Disorder (OCD), response toSSRI treatment was predicted with accuracy greaterthan 80% using pre-treatment QEEG subtype [1]. Thesefindings have been prospectively replicated [2] with sen-sitivity of 98% and specificity of 89%; and extended to in-clude follow-up QEEG evaluations on treatment.Responders belonged to the QEEG subtype whichwould be moved in the "normalizing" direction bySSRIs, as would be predicted from the work of Saletu etal. [3]. Using Variable Resolution Electrical Tomogra-phy (VARETA), clear differences were seen in very nar-row band theta and alpha sources for the responders andnon-responder subtypes, in the same regions identifiedby PET scans in these patients. Similarly, using this ap-proach for predicting relapse in cocaine dependence, forpredicting response to stimulants in attention deficit dis-order and for prospective prediction of conversion to de-mentia in normally elderly, has led to high predictiveaccuracy. Thus, these results suggest that QEEG repre-

sents a cost-effective, non-invasive evaluation tool thatappears highly sensitive to treatment responsiveness inpsychiatric patients.

Prichep, L.S., Mas, F., Hollander,E., Liebowitz, M.,John,E.R., Almas, M., DeCaria, C.M. and Levine, R.H.Quantitative electroencephalographic (QEEG)subtyping of obsessive compulsive disorder, Psychiat.Res., 1993, 50: 25-32.

Hansen, E.S., Prichep,L.S., Bolwig, T.G. and John,E.R. Quantitative electroencephalography in OCD- pa-tients treated with paroxetine. Clin. EEG, 2003, 34: 70-74.

Saletu,B., Anderer, P., Saletu-Zyhlarz, G., Arnold,O. and Pascual-Marqui, R.D. Classification and evalua-tion of the pharmacodynamics of psychotropic drugs bysingle-lead pharmaco-EEG, EEG mapping and tomogra-phy (LORETA). Methods, Find. Exp. Clin. Pharmacol.,2002, 24[Supplement C]: 97-120.

Somatosensory evoked potentials and magnetic fieldsduring active and quiet sleep in newborns. - L.Lauronen*+^, E. Pihko*+, H. Wikström+, S. Taulu~, J.Nurminen*+, S. Kivitie-Kallio# and Y. Okada%

(*BioMag Laboratory, Helsinki University CentralHospital, P.O. Box 340, FIN-00029 HUS, Finland; +Hel-sinki Brain Research Centre, ^Department of ClinicalNeurophysiology, Helsinki University Central Hospi-tal, Helsinki, Finland, ~Neuromag, Ltd., Helsinki, Fin-land, #Department of Social Pediatrics, Hospital forChildren and Adolescents, Helsinki, Finland, %Depart-ment of Neurology, University of New Mexico Schoolof Medicine, Albuquerque, NM, USA).

We investigated the effect of sleep stages on thesomatosensory evoked potentials (SEPs) and magneticfields (SEFs) of newborn babies. The subjects were 14full-term healthy newborns. The tip of the left index fin-ger and/or thenar eminence were stimulated by move-ment of a plastic membrane touching the skin. Theinter-stimulus interval was 2 seconds. MEG from thecontralateral hemisphere was measured simultaneouslywith eight-channel EEG-recording. The data were ana-lyzed off-line for active and quiet sleep stages. The re-sponses consisted of three main deflections. The firstdeflection, N1, peaked at about 30 ms. The following moreprominent deflections peaked at about 80 ms (P1) and 260ms (P2). The amplitudes of P1 and P2 were significantlyreduced in active sleep compared to quiet sleep (P1: p =0.003 and p= 0.02; P2: p= 0.018 and p = 0.000004, for EEGand MEG, T-test). In conclusion, our study shows thatMEG is feasible in studies of the newborn somatosensorysystem. Furthermore, the marked differences in responsestrengths in active and quiet sleep demonstrate the needto carefully monitor the newborn’s sleep stage in at leastthe studies of the somatosensory evoked responses.


Effect of skull holes on scalp potentials. - C. Ramon*, J.Haueisen+, P. Schimpf^, M. Holmes* and A. Ishimaru*(*Departments of Electrical Engineering and Neurol-ogy, University of Washington, Seattle, WA, USA;+Biomagnetics Center, Department of Neurology, F.S.University, Jena, Germany; ^School of Electrical Engi-neering and Computer Science, Washington State Uni-versity, Spokane, WA, USA).

The effects of a skull holes on the scalp potentialswas examined using a finite element model of the headconstructed from the segmented MRI images. The wholehead model has a 1×1×3.2 mm voxel resolution and weidentified 11 different tissue types in the model includ-ing, scalp, hard and soft skull bone, gray and white mat-ter, CSF, fat, muscle, cerebellum, etc. A square hole of asize of 20 mm filled with air on top of the head was used.The electric potentials and currents in the whole headwere computed using an adaptive finite element solver.We used a current dipole of 5 µA mm in the motor cortexarea. From the computed fields the scalp potentials wereextracted and contour plots were made. We used twomodels, one with a hole and the other one with no hole.The maximum and minimum of the contour vales for themodel with hole were -0.45 µV and 0.65 µV, respectively.Similarly, for the model with no hole the maximum andminimum contour values were -0.6 and 0.85, respec-tively. The dynamic range of the difference is about -0.15to 0.25 µV for a dipolar source in the motor cortex area.The spatial patterns of the contours were also differentfor the two models. This shows that the scalp potentialsare affected by the presence of a hole in the skull.

Evaluation of the distortion of EEG signals caused by ahole in the skull mimicking the fontanel in the skull ofhuman neonates. - L. Flemming, Y. Wang and Y. Okada(Biomagnetic Center Jena, Friedrich-Schiller-University, Jena, Germany, Department of Neurology,University of New Mexico, School of Medicine, Albu-querque, NM, USA).

Brain functions of human infants, includingpre-term and term babies, have been traditionally evalu-ated with EEG as part of the neurological examinationcarried out on infants with risk factors. EEG is also usedin staging the development of the nervous system to de-tect delays in the brain development. The interpretationof the EEG of newborns is very difficult because it isstrongly affected by the different conductivity offontanels and sutures compared to the skull. This effectwas studied in 10 farm swine. The fontanel was mim-icked by a hole, which was filled with materials of differ-ent conductivities (air, sucrose-agar, saline-agar).Cortical areas beneath the hole, at the edge of the holeand far away from the hole were electrically stimulated.30 stimulations were averaged and used for the analysis.

EEG signal of the cortical area beneath the hole wasstrongest affected by the different conductivities com-pared to the other areas. The further away from the holethe area was the smaller was the influence of the hole.The leakage of the current through the hole shown by thelaplacian transformation was strongest for the su-crose-agar conductivity in all cases. The conductivity ofsucrose-agar was in the middle between air and sa-line-agar. It was comparable with the conductivity of theskull. We also investigated the influence of the hole todifferent components of the EEG signal for the stimu-lated cortical area beneath the hole. The second compo-nent (p20) was strongest influenced and the thirdcomponent (n50) was weakest influenced.

Neurofeedback effect on EEG current sources. - T.Fernández, T. Harmony, L. Díaz-Comas, E. Santiago, L.Sánchez, A. Fernández-Bouzas, W. Herrera, G.Aboytes, J. Bosch and S. Ocampo (Instituto deNeurobiología, Universidad Nacional Autónoma deMéxico and Centro de Neurociencias de Cuba).

Neurofeedback (NFB) is an operant conditioningprocedure, whereby an individual can learn to modifythe electrical activity of his/her own brain. In one hand, ithas been shown minimum alpha absolute power is nec-essary for adequate performance; in the other hand,Learning Disabled (LD) children have higher values oftheta EEG absolute and relative power than normal chil-dren do. Therefore, reduction of (theta/alpha) ratio mayimprove cognitive deficits. Eleven LD children were se-lected with higher than normal (theta/alpha) ratio atleast in one lead. Test Of Variables of Attention (TOVA)and WISC-R were applied. NFB was given in the regionwith highest ratio, triggering a sound each time the ratiofell below a threshold value. Twenty half-hour sessionswere applied. Two months after the end, TOVA, WISC-Rand EEG were obtained again. EEG current sources werecomputed. After NFB, current decreased at frequencies2.34-7.02 Hz, principally in occipital areas, and increasedat frequencies 8.58-18.72 Hz, more on the left hemi-sphere. Significant increase was observed in ADHDscore from TOVA and in verbal scale from WISC-R Theseresults suggest that NFB of (theta/alpha) ratio producesan important spurt on EEG maturation, and as a conse-quence, behavioral improvement in LD children. Al-though NFB treatment was applied according to theactivity of one lead, different for each child, EEG changeswere observed on wide regions, suggesting that thetreatment affects the whole cortex.

The authors acknowledge engineer Héctor Belmont,Ms. Rosa María Hernández and Ms. Pilar Galarza. Thisproject was partially supported by DGAPA (IN226001)and CONCYTEQ (2001).

Proceedings 207

Improving P300 latency determination by 128-channelEEG measurements: results in mild-to-moderate headinjury patients. - N.M. Maurits, J.W. Elting, J. van derNaalt and T.W. van Weerden (Groningen UniversityHospital, Department of Neurology, P.O. Box 30.001,9700 RB Groningen, The Netherlands).

The P300 event related potential is a measure of cog-nitive function and its latency and amplitude are thoughtto quantify information processing. P300 latency isknown to be increased in patients with dementia and se-vere head injury. Recently subcomponents of P300 havebeen defined; the more frontally distributed P3a (novelstimuli) and the centro-parietal P3b (stimulus process-ing). The large intra-individual inter-trial P300 latencyvariation in normal controls makes it hard to distinguishbetween normal and abnormal P300 results. One of thecauses of variation is the overlap in P3a and P3b poten-tials resulting in inaccurate latency determination of P3b(P300). Conventional P300 registration only allows forseparation of the two potentials in approximately 25% ofthe measurements. We have used 128-channel EEG and anew dipole source analysis technique for data reduction,to overcome this problem. We are able to separate the P3aand P3b components and thereby reduce the spread inP3b (P300) latencies compared to single-channel analysisin normal controls (n=16). Using these results, in a groupof mild-to-moderate head injury patients (n=23, EMV4-14, duration of PTA 1-60 days), we find that increased(conventional) P300 latency actually corresponds to di-minished P3a activity (pseudo-delay). The P3b latencyappears to be normal. We have also investigated the pre-dictive value of these results for outcome based on psy-chological test results (Stroop test and word-recall) afterthree months. These results indicate that our methodmay help interpret P300 results in head injury patientsand improve the prognostic value of abnormal P300.

Event-related correlations in learning impaired chil-dren during a hybrid go/no-go choice reaction vi-sual-motor task. - J.M. Peters*, D.P. Waber+,G.B.McAnulty^ and F.H. Duffy^ (*Department of Neurol-ogy, Leyenburg Hospital, The Hague, Netherlands; De-partments of Neurology^ and Psychiatry+, HarvardMedical School and Children’s Hospital, Boston, MA).169 learning impaired (LI) and 71 non-learning impaired(NLI) children underwent a hybrid go/no-go choice reac-tion time visual-motor task to study the behavioral andphysiological fundamentals of learning disorders. A leftbutton was pressed for Left Arrow (LA) stimuli, a rightfor Right Arrow (RA) stimuli, none (no-go) for anon-directional arrow. Stimulus specific visual evokedpotentials were formed and, with PZ as index electrode,were lag-correlated to frontal electrodes to formEvent-Related Correlations (ERC). Exploratory t-statistic

significant probability maps (t-SPM) were used to defineregions of interest (ROI). Behaviorally, there was a righthand advantage over the left in the NLI group, but less inthe LI group. Electrophysiologically, RA and LA condi-tions increased correlation between visual areas (PZ) andcontralateral frontal areas (F3 and F4). A unilateral ROI, atelectrode FC1, also preceded both left and right-handedresponses. Neurobehaviorally, increased visual-motorcorrelation was associated with better performance, espe-cially for the left hemisphere, at F3 and FC1. Surprisingly,visual-motor correlations were not associated with per-formance for the NLI group in the RA and no-go condi-tion. Our data support previously reported difficulties oflearning impaired children in low-level information pro-cessing. Furthermore, we hypothesize that LI, in contrastto NLI children, demonstrate difficulty in automatizingroutine tasks.

Aging effects on auditory processing in an oddballtask: an MEG study. - S. Kovacevic, J.M. Stephen+*,C.C. Woodruff*, C. Qualls+*, J.C. Adair+*, D. Hudson*,J. Knoefel+* and C.J. Aine+* (*University of New Mex-ico, Albuquerque, NM; +New Mexico VA Health CareSystem, Albuquerque, NM).

A possible explanation for age-related changes inperformance is that aging is accompanied by a generalslowing in cognitive processing. We examined changesin amplitude of the sources located in the superior tem-poral gyrus (STG) while subjects performed an oddballtask. Our hypothesis was that the attention effects onamplitude of the STG sources would be more prominentlater in elderly than in young subjects as a result of cogni-tive slowing. Fifteen healthy subjects (20 – 84 yrs) werepresented with frequent (1000 Hz) and rare (1050 Hz,20% probability) tones binaurally at a rate of 1 tone/s.MEG data was collected and online averaged while sub-jects silently counted the rare tones using 122-channelNeuromag magnetometer system. Active brain regionsand their associated timecourses were characterized byusing the multi-start spatial-temporal algorithm (MSST).In agreement with previous studies, both elderly andyoung subjects showed overall stronger STG responsesto the rare than to the frequent tones. The latency of theearly response (~ 100 ms) to the tones did not differ be-tween young and elderly. However, the peak latency inthe 180-600 ms time interval was longer for the elderly.These results indicate that although the timing of theearly (sensory) processing in the STG was not affected bynormal aging, the subsequent processing of the relevantstimuli occurred slightly later for elderly. Although STGactivity has traditionally been associated with early sen-sory processing, our results suggest that the STG activityis also important for later processing, as a part of theattentional network.


Age-related changes to median nerve stimulation mea-sured with MEG. - J.M. Stephen*+, D.M. Ranken^, C.C.Woodruff*+, S. Kovacevic*+, D. Hudson+, J.C. Adair+, J.Knoefel+, C. Qualls+ and C.J. Aine*+ (*University ofNew Mexico, Albuquerque, NM; +New Mexico VAHealthcare System, Albuquerque, NM; ^Los AlamosNational Laboratory, Los Alamos, NM).

Studying sensory processes provides a critical step tounderstanding age-related changes in cognitive function.Using multi-dipole analyses to analyze MEG data providesa unique opportunity to characterize age-related latencyand amplitude changes in the primary and secondary areasresponsible for the somatosensory response. Based on pre-vious EEG and ERP studies, we hypothesized increasedamplitude with age and increased latencies between pri-mary and secondary areas as a function of age due to in-creased central conduction times. Currently data from 9elderly (65 + years) and 7 young (20-29 years) healthy sub-jects have been collected. The left and right median nerveswere stimulated with a current pulse and the magnetic re-sponses were measured with a 122- channel Neuromagbiomagnetometer. The data were analyzed using amulti-dipole multi-start spatio-temporal modeling tech-nique (Ranken et al. 2002). The amplitude of the first peak ofthe primary somatosensory response was greater for theold than the young (p< 0.02). We also found that the vari-ability of the response from trial-to-trial was greater for theyoung for the first peak of the response (p< 0.02). There wasalso considerable inter-subject variability in responses insecondary areas. Drechsler (1978) and others suggestedthat the increased amplitude of the responses in older sub-jects could be related to a decrease in inhibition with age.An alternative is the larger variability in the trial-to-trial re-sponses in the young, which could lead to a lower ampli-tude averaged response due to poor time-locking.

MEG studies of memory in healthy aging. - C.J. Aine*+,J.C. Adair*+, J. Knoefel*+, D. Hudson^, C. Qualls*+, S.Kovacevic*+, C. Woodruff*+ and J.M. Stephen*+ (*NewMexico VA Healthcare Systems; +University of NewMexico SOM; ^BRINM, Albuquerque, NM).

Preliminary data from our laboratory and others sug-gest that lower-level auditory areas play a major role incognitive processing. A group of 17 healthy subjects (20-40and 65-85 years) completed two different verbal memorytasks while MEG recordings were obtained. It was pre-dicted that the late sustained activity localized to putativeauditory association cortex would reveal activity of greateramplitude and longer duration for younger subjects. Sub-jects decided whether the words, representing commonobjects, were larger than a television set. After a 15 minutedelay, subjects were asked whether or not each word, em-bedded within a larger list of words, was one they heardpreviously. The timecourses of activity from multiple cor-

tical regions were determined via multidipole,least-squares minimization procedures. A main effect ofage (F=4.54, p=.05) for the late sustained activity revealedearlier peak latencies for the elderly, compared to theyoung and poorer performance (i.e., fewer percent correctand longer RTs). An unexpected finding was greater peakamplitude and duration of activity around 60 ms whichdid not correlate with performance on the tasks. These re-sults suggest that activity in the superior temporal gyrus(in the vicinity of Heschl’s gyrus) is important for earlycomplex processing, beyond simple sensation. Supportedby VA MERIT Review, The MIND Institute, NIH COBRE.

Symmetry and non-symmetry in the brain. - M.J.A.M.van Putten (Department of Neurology and ClinicalNeurophysiology Ziekenhuis Leyenburg, Leyweg 275,2545 CH The Hague, The Netherlands. email:[email protected]).

The EEG of the physiological brain is characterizedby a relative strong symmetry between the two hemi-spheres. The EEG in the pathological brain, however, of-ten shows a particular non-symmetry. Variousparameters can be defined that characterize brain symme-try or non-symmetry. Measures range from parametersbased on, e.g., the spectral density, coherence and phasesynchronization to the recently proposed Link Rates [1].We evaluate the dynamic behavior of several symmetrymeasures in their application to controls and patientswith various neurological conditions, such as epilepsy,dementia and (threatening) brain infarctions. In addition,task specific asymmetry will be studied using a controlgroup of healthy medical students. We will show that thedynamic behaviour of symmetry measures shows charac-teristic differences in the application to the EEG of neuro-logical patients. This adds to the store of knowledge aboutphysiological and pathological brain function and allowsimproved differentiation between various EEG patternsof neurological patients. From a clinical point of view, weforesee additional applications in continuous EEG moni-toring in stroke and ICU patients.

[1] van Putten, M.J.A.M. Proposed Link Rates in theHuman Brain, J. Neurosci. Methods, 2003, 127(1): 1-10.

EEG global field power spectrum changes after a singledose of atypical antipsychotics in healthy volunteers. -Keizo Yamada*, Toshiaki Isotani*, Satoshi Irisawa*,Masafumi Yoshimura*, Aran Tajika+, Takami Yagyu*,Akemi Saito* and Toshihiko Kinoshita* (*Departmentof Neuropsychiatry, Kansai Medical University, Osaka570-8506 Japan and +Department of Public Health,Kansai Medical University, Osaka 570-8506 Japan).

The EEG effects of 4 different novel atypicalantipsychotics were compared with 2 conventional typicalantipsychotics and a placebo in 14 male, right-handed, nor-

Proceedings 209

mals (24.1±4.1 years). All subjects went through 7 sessions.In each session, a subject received orally either 0.5mgrisperidone, 4mg perospirone, 33mg quetiapine, 1.25mgolanzapine, 50mg chlorpromazine, 1mg haloperidol or pla-cebo according to a single-blind, crossover, Latin-squaredesign. EEG was recorded from 19 scalp-electrodes prior toas well as 2, 4 and 6 hours after drug administration.Twenty sec/subject/drug/time-point EEG were analyzedinto FFT Global Field Power spectra (GFP) (Lehmann andSkrandies 1980). GFP waveforms (one tracing measures thestrength of the whole brain electric field) were computedfor 7 frequency bands (Kubicki et al. 1979). Assuming themaximal effect would correspond to serum-concentration,2-hours after perospirone and quetiapine, 4-hours afterrisperidone, olanzapine, and chlorpromazine, and 6-hoursafter haloperidol were chosen. Difference from pre-drugGFP-values were compared using one-way repeated mea-sures ANOVA, then post-hoc multiple comparisons (LSD)and LORETA (Pascual-Marqui et al. 1999) were applied.Relative GFP in delta (1.5-6 Hz) increased (p<0.05), (a) afterquetiapine compared with perospirone, olanzapine,chlorpromazine, haloperidol and placebo, (b) afterrisperidone compared with haloperidol and placebo.LORETA showed (p<0.05) cortical posteriorization of deltaafter olanzapine and anteriozation of delta afterhaloperidol. Thus, quetiapine and risperidne showed seda-tive effects on the brain, which may reflect the strength ofboth receptor A1 and H1 bindings. The LORETA result ofolanzapine indicating posterior inhibition, i.e., frontal shiftof brain electric activity, might include a function againstthe negative symptoms in schizophrenia.

Topographical analysis of spectral EEG and late visu-ally evoked brain activity in patients with benignrolandic epilepsy of childhood. - W. Skrandies and D.Dralle (Institute of Physiology and Department ofNeuropediatrics, Justus-Liebig University, Giessen,Germany).

A group of 11 young patients (7-15 years of age) withbenign rolandic epilepsy of childhood was studied. We re-corded the interictal spontaneous background EEG andflash evoked potentials from 19 channels distributed ac-cording to the International 10/20 system. Following visualstimulation, five of the patients displayed a late sustainedcomponent over temporal and central scalp regions at amean peak latency of 194.8 ms. In the other patients thispattern of activity was absent while early VEP componentswere comparable in both groups. We related the occurrenceof this N200 component to the spectra of spontaneous EEGand clinical parameters. Significant topographical differ-ences were observed in the spectra of the spontaneous EEGof patients with and without a visual N200 component. Theeffects were most evident in the lower alpha (7.5-9.5 Hz)and beta band (13-15 Hz) when the spectral amplitudes ob-

tained over temporal, parietal and occipital regions of theleft and right hemisphere were compared. There was also atendency that children without N200 displayed morehypersynchroneous activity in the spontaneous EEG thanchildren with N200 activity. This suggests a protective roleof interictal spike activity, and probably coincides with abenign prognosis of epilepsy.

Propagation of epileptic activity in patients with juve-nile myoclonic epilepsy. - E. Santiago-Rodríguez, T.Harmony, A. Fernández-Bouzas and T. Fernández(Instituto de Neurobiología, UNAM).

Juvenile myoclonic epilepsy (JME) is a syndromebeginning in puberty, characterized by myoclonic jerksin the morning, tonic-clonic seizures (TCS), and, less fre-quently, absence seizures. Generalized spike-wave(SWC) and polyspike-wave complexes (PSWC), withfronto-central accentuation are the typical EEG patternshown in JME. Frontal alterations have been detectedwith anatomophatological studies, MRI voxel-basedmorphometric analysis and magnetic resonance spec-troscopy. In opposition to partial epilepsies, invasivemonitoring in JME is not indicated. The objective in thisstudy was to use a VARETA (Variable Resolution Elec-tromagnetic Tomography) for determining the electriccurrent sources in the propagation of epileptogenic ac-tivity in patients with JME. Five patients with all clinicalcriteria for JME were included. Four were female and onemale with mean age of 23 + 11 years. All patients hadmyoclonic and TCS with mean age of onset at 13 years.Nineteen referential leads were recorded using a linkedearlobe reference. One solution with VARETA each 5 msof all paroxysmal events was calculated. Initial localiza-tion of sources in medial frontal gyrus was observed, cor-responding to initial spikes of PSWC. In the next 5 to 10ms a cyclic propagation to right or left temporal lobe wasshowed. Always, the temporal sources were less intensethan frontal sources. The cyclic propagation of electricsources was observed independently of synchronic orasynchronic spikes in visual analysis. In synchronicspikes the time of cyclic alternation was shorter.

The use of hidden markov models for detection of par-oxysmal activity propagation in interictal MEG data. -Alexei Y. Ossadtchi*~, John C. Mosher+, William W.Sutherling^ and Richard M. Leahy* (*Electrical Engi-neering, University of Southern California, Los An-geles, CA, 90089; +Los Alamos National Laboratory, LosAlamos, NM, 87545; ^Epilepsy and Brain Mapping pro-gram, Huntington Medical Research Institute, Pasa-dena, CA, 91105; ~Source Signal Imaging Inc., 2323Broadway, San Diego, CA, 92102).

Automatic spike detection techniques applied tointerictal MEG data often discover several spike clusters


that represent potentially epileptogenic regions in thebrains of patients with partial epilepsy. For the purposesof treatment planning, it is important to determine whichof the detected regions are the most likely primary genera-tors of seizure activity. Analysis of the patterns of paroxys-mal activity propagation between the detected regionsmay allow for detection of these primary epileptic foci.We have developed a Hidden Markov Models (HMM) forestimation of the propagation patterns between severalspike clusters or regions. Analysis of the estimated transi-tion probability matrix allows us to determine the proba-bilities of a spike in each region being followed by a spikein each of the other regions. From these we can determinethe most likely primary foci as those which occur sponta-neously and have a relatively high probability of being fol-lowed by spikes in one or more of the other regions. Weapplied the HMM to four independent interictal datasetscollected from a patient with multi-focal epilepsy. Severalspike clusters were automatically detected and the propa-gation detection technique was applied to determine thepattern of propagation of abnormal interictal activity. Inall four datasets propagation from the temporal lobe tothe frontal lobe was detected. Statistical significance ofthese results was established using a permutation test.

Focality of extratemporal neocortical epileptiform EEGstudied with 256 scalp channels. - K.J. Eriksen*, M.D.Holmes+, S. Vanhatalo+ and D.M. Tucker* (*ElectricalGeodesics, Inc., Eugene, OR, USA;+Harborview Medi-cal Center, Seattle, WA, USA).

The appearance of diffuse scalp EEG suggests wide-spread simultaneous activity, but volume conduction canmask or smear focal activity. We wished to examinewhether the use of high density scalp EEG and corticallyconstrained source modeling in a realistic head modelcould demonstrate focality of extratemporal, neocorticalepileptiform activity. Five epilepsy patients were recruitedfor this study, selected for frequent interictal activity and adiagnosis of medically refractory extratemporal epilepsy.We recorded 30-90 minutes of 256 channel EEG at 500 sam-ples/second using a comfortable electrode net that al-lowed the subject to lie on a bed in a sleep lab. We acquiredhigh resolution structural MRI scans covering the wholehead and used then to build realistic boundary elementmodels of each subject’s head, including aparameterization of the individual cortical gray matter.Source modeling of interictal spike and sharp waves wasdone using a weighted minimum norm inversion onto thecortical surface. We repeated this modeling with subsets of128, 64, and 32 EEG channels. The visually observedfocality of the epileptiform activity increased as we in-creased the number of scalp electrodes from 32 through256, concentrating in smaller and smaller areas of cortex.We conclude that with 256 channels we are getting closer

to meeting the spatial Nyquist criterion and achieving ade-quate sampling of the spatial distribution of the underly-ing activity.

Validation with depth EEG recordings ofspatio-temporal algorithms for the localization of thesources of interictal activities with MEG. - J.-M. Badier,M. Gavaret, D. Schwartz, F. Wendling and P. Chauvel(UMIU9926 INSERM, Laboratoire deNeurophysiologie et Neuropsychologie, Faculté deMédecine, 27 bd Jean Moulin, 13005 Marseille, France).

Localization of the sources of the paroxysmal activi-ties in epileptic patients candidates for surgery is still achallenging task. A first series of 10 patients sufferingfrom various type of epilepsies, excluding the commonand considered as easier to cure mesial temporal lobeepilepsie, have been studied. They all had simultaneousEEG and MEG although MEG data are presented here.Various techniques had been applied to the data, includ-ing non linear dipole fi t and a more recentspatio-temporal algorithm developed in our laboratory.This algorithm uses a linear approach by calculating ametric, based on the decomposition of the signal, overthe cortical surface segmented from each individualMRI. This technique allows complex reconstruction ofsource distributions in time as well as in space. Results ofthe localisation are compared to the depth recordingsmade prior to surgery. This is done by the mean of auto-matic detection of the activities recorded in the depthand automatic extractions of typical patterns of activa-tion of the different brain structures involved. This al-lows not only to determine the distribution of theabnormal activity but also to reconstruct the dynamic ofthe activities. This study confirmed the power of theMEG to give a good image of the localization as well asthe dynamic of paroxysmal activity especially when us-ing the linear spatio-temporal fit algorithm.

Clinical contribution of interictal source localizationusing high-resolution EEG in 2 cases ofpharmacoresistant frontal lobe epilepsies without anyMRI structural lesion. - M. Gavaret, J.-M. Badier, P.Marquis, F. Bartolomei, J. Bourien, F. Wendling and P.Chauvel (Department of Clinical Neurophysiology,Timone Hospital, Marseille, France).

Pharmacoresistant epilepsies of extra-temporal origincan represent a difficult therapeutic challenge in particularin case of apparently non lesional focal epilepsies. We re-port here 2 cases of frontal lobe epilepsy without lesion visi-ble with the structural MRI which presented usually fulland abundant paroxysmal interictal events. These paroxys-mal interictal events were recorded in High ResolutionEEG (64 electrodes, high sampling rate) and source local-izations were calculated with several algorithms

Proceedings 211

(spatio-temporal dipoles, moving dipole, MUltiple SIgnalClassification) and a realistic head model of the electricalactivity propagation mediums. SEEG (ste-reo-electro-encephalography) showed abundant,sub-continuous paroxysmal interictal events with a spatialdistribution restricted. Source localizations highlightedwere confirmed by an automatic detection of paroxysmalinterictal events performed on SEEG signals. Theanatomopathologic examination demonstrated a dysplasiclesion which was not visible on MRI including T1, T2, IRand FLAIR sequences. Source localization tools can pro-vide spatial information in the pre-surgical assessment ofmedically intractable epilepsies specially in difficult casesof extra-temporal epilepsies apparently non lesional.

An MEG/EEG examination of the generality of thegating deficit in schizophrenia. - J.C. Edgar*+, S. Mo-ses+^, R.J. Thoma+^, M. Huang+~, F. Hanlon+^, M.P.Weisend+~, G.A. Miller*%, L.E. Adler$ and J.M.Canive+# (*Department of Psychology and Beckman In-stitute, University of Illinois at Urbana-Champaign,Urbana, IL, USA; +Center for Functional Brain Im-aging, New Mexico VA Healthcare System, Albuquer-que, NM, USA; ^Department of Psychology,University of New Mexico, Albuquerque, NM, USA;~Department of Radiology, University of New Mexico,Albuquerque, NM, USA; #Department of Psychiatry,University of New Mexico, Albuquerque, NM, USA;%Department of Psychiatry, University of Illinois at Ur-bana-Champaign, Urbana, IL, USA; $Department ofPsychiatry, University of Colorado Health SciencesCenter, Denver, CO, USA).

Auditory paired-click studies have established a clearrelationship between schizophrenia and impaired sen-sory gating using the P50 component of the event-relatedbrain potential (ERP). However, the presumedcross-modal generality of the gating deficit has not beensystematically tested. The current study did so by investi-gating the M20 component of the somatosensory ERP. Pa-tients were expected to show impaired auditory andsomatosensory gating, but positive correlations were ex-pected between auditory and somatosensory gating ra-tios. 122 channels of magnetoencephalographic (MEG)and 19 channels of electroencephalographic (EEG) datawere simultaneously collected from 27 subjects withchronic schizophrenia and 21 control subjects. Subjectswere administered: (1) the standard paired-click auditoryparadigm (500 ms ISI) to assess P50 gating and, (2) a re-cently developed paired-pulse somatosensory paradigmusing a 500 ms or a 75 ms ISI to assess M20 gating. UsingMagnetic Source Imaging (MEG + structural magneticresonance imaging), left and right S1 and S2 M20somatosensory responses were localized and the sensorygating ratio (S2/S1) was calculated. Groups did not differ

in M20 gating in either hemisphere or at either ISI. Patientsfailed to show normal P50 gating. Negative correlationswere observed between M20 and P50 gating ratios in con-trols but not in patients. Thus, neither gating nor theschizophrenia gating deficit appears to be cross-modal.

Fast ripples in hippocampal epileptogenicity. - J.Engel, A. Bragin, C.L. Wilson and I. Mody (UCLA).

Fast ripples (FR) are abnormal, brief, high frequency(200-600 Hz) oscillations recorded from epileptogenic hip-pocampus and parahippocampal structures in patientswith mesial temporal lobe epilepsy, and rat models of thiscondition. FR is usually associated with interictal EEGspikes, and preferentially occurs in structures capable ofgenerating spontaneous seizures. FR, therefore, permitsdistinction between interictal epileptiform events thatidentify the epileptogenic region and those which are non-specific propagated events, or indicative of epilepticallyabnormal tissue that does not have the capacity for sponta-neous seizure generation. Because the temporal develop-ment of FR following intrahippocampal kainate injectionin rats precedes the development of behavioral seizures,and because FR is seen at the onset of both low-voltage fastand hypersynchronous ictal events in these animals, it isbelieved that they represent the pathophysiological sub-strate of epileptogenesis, and epileptogenicity. The topo-graphical distribution of FR activity within epileptogenictissue is not homogeneous, but, rather, discretely localizedwithin small neuronal clusters imbedded withinhippocampal and parahippocampal tissue that does notexhibit FR activity. The size and/or density of FR clusters,as measured by the number of multiple microelectrodecontacts recording these events, correlates directly withseizure frequency in rats. Furthermore, although the loca-tion of microelectrode contacts recording FR is stable overweeks to months, hippocampal slice recordings demon-strate that treatment with small amounts of bicuculline in-creases the size of FR neuronal clusters. This observation isconsistent with the in vivo observation that paired-pulseperforant path stimulation, which produces suppressionin epileptic hippocampus, produces facilitation and FR ac-tivity in discrete FR neuronal clusters. These data suggestthat spontaneous seizure generation could be initiated byrelatively generalized nonspecific disinhibition, whichwould permit FR clusters to increase in size, coalesce,and/or synchronize to create the critical neuronal massnecessary for seizure initiation and propagation.

Spikes and slow waves recorded with MEG in patientswith epilepsy. - J.C. de Munck, A. de Jongh, D. van’tEnt, I. Manshanden, M. Pulligedhu, J.C. Baayen, C.J.Stam and F.H. Lopes da Silva.

Interictal spikes represent only a very small part ofthe MEG/EEG data. In some cases, the spike rate is less


than two per hour. It has been proposed in the literatureto make use of another pronounced property of theMEG/EEG data of patients with epilepsy: the enhanceddelta band activity. The purpose of this work was to in-vestigate whether pathological delta band activity can belocalized using simple dipole models and detection tech-niques, and whether its sources coincide with the dipolesources found with the analysis of inter ictal MEG spikes.To localise the delta band activity efficiently, MEG datawere band-pass filtered, the large amplitudes were de-tected and a global search algorithm was applied to find agood initial guess for the final non-linear dipole searchstep. Spikes were detected by a human observer and sub-mit to a clustering algorithm to group the spikes intogroups with high similarity. Spikes within each clusterwere averaged to reduce the noise, before dipole localisa-tion. The fully automatic delta band analysis was ap-plied on the MEG data of a group of patients withintracranial tumours and epilepsy. In 70% of the patientsclear dipole clusters were found adjacent to the tumour.These clusters were quite robust and in post-operativeMEG data it was found that these clusters are hardly in-fluenced by medical treatment. Only a few data sets ofthe tumour group showed both abnormal delta and epi-leptic spikes. In all these data sets the delta activityagreed on the lateralization and on the lobe. Further-more, most spikes sources (70%) were located at the sameside of the structural lesions as the delta sources. Thisstudy shows that the delta band sources carry pathologi-cal information that is of potential clinical applicability.Furthermore, it shows that slow waves and inter-ictalspikes are related, but more data is required to establishthe automatic delta analysis as a substitute for theinter-ictal spikes.

Challenging MEG source imaging with simultaneousdepth recordings in epilepsy. - S. Baillet*, C. Adamm*+,A. Ducorps^, D. Schwartz^, R.M. Leahy~, J.C. Mosher#,B. Renault*, M. Baulac+ and L. Garnero* (*CognitiveNeuroscience and Brain Imaging Laboratory, CNRSHôpital de la Salpêtrière, Paris, France; +Epilepsy Unit,Hôpital de la Salpêtrière, Paris, France; ^MEG Center,Hôpital de la Salpêtrière, Paris, France; ~Signal and Im-age Processing Institute, University of Southern Cali-fornia, Los Angeles, USA; #Design Technology Group,Los Alamos National Laboratory, USA).

Simultaneous MEG and depth electrode recordingsin patients under evaluation for epilepsy surgery will bepresented. We will specifically discuss how MEG sourceimaging at the individual cortical surface can predict thepotentials recorded from the depth electrodes. Sensitiv-ity to deeper structures such as the Hippocampus andAmygdala will also be discussed in this context.

A new look at the dynamics of the epileptic brain: pre-diction of epileptic seizures and resetting of the epilep-tic brain leads to intelligent brain pacemakers. - L.D.Iasemidis*, J.C. Sackellares+, D.S. Shiau+, P.R. Carney+

and P.M. Pardalos+ (Brain Dynamics Lab, The Harring-ton Department of Bioengineering, Arizona State Uni-versity, Tempe, AZ; +Brain Dynamics Lab,Departments of Neurology, Neuroscience, BiomedicalEngineering, Pediatrics, Industrial and Systems Engi-neering, University of Florida, Gainesville, FL).

Epilepsy is considered a dynamical disorder of thebrain. Epileptic seizures and spikes occur intermittently,and seemingly in a random and unpredictable fashion. Inthe last 15 years, we have provided evidence that seizuresmay be deterministic and predictable [1]. Recently, wepresented the first automatic, prospective seizure predic-tion system [2] on continuous long-term EEG data.Spatiotemporal analysis of a total of 2,000 hours of contin-uous EEG recorded from depth (hippocampal), subdural(temporal and frontal), and/or scalp electrodes in 10 pa-tients with focal epilepsy showed that a progressive dy-namical entrainment (convergence of measures of chaoslike the maximum Lyapunov exponents) of the focuswith critical normal brain sites occurs hours to days priorto seizures onset, and reverses after seizures end (braindynamical resetting). From 80% to 100% of a total of 160focal seizures of temporal lobe origin could be predictedat the 99% confidence level, approximately 70 minutes inadvance, with 0.2 to 0.1 false warnings per hour respec-tively. Detection of seizure susceptibility periods of daysin duration prior to sequences of seizures was also ob-served. The significance of these findings for brain re-search, understanding of the mechanisms, diagnosis andnovel methods of treatment of epilepsy will be discussed.

[1] Iasemidis, L.D. Epileptic seizure prediction andcontrol. IEEE Trans. Biomed. Eng., 2003, 50: 549-558.

[2] Iasemidis, L.D., Shiau, D.S., Chaovalitwongse,W., Sackellares, J.C., Pardalos, P.M., Principe J.C., Car-ney, P.R., Prasad, A., Veeramani B. and Tsakalis, K.Adaptive epileptic seizure prediction system. IEEETrans. Biomed. Eng., 2003, 50: 616-627.

Mapping distributed sources of cortical rhythms inmild Alzheimer’s disease. A multi-centric EEG study. -C. Babiloni*+, G. Binetti+, G. Dal Forno~#, R. Ferri%, B.Lanuzza%, C. Miniussi+, D.V. Moretti*, F. Nobili$, R.D.Pascual-Marqui&, G. Rodriguez$, G.L. Romani!, S.Salinari?, F. Tecchio@, P. Vitali% and P.M. Rossini+^~

(*Dip. Fisiologia Umana e Farmacologia, Univ. "LaSapienza" Rome, Italy; +IRCCS "S. Giovanni diDio-F.B.F.", Brescia, Italy; ^A.Fa.R. Osp. FBF; IsolaTiberina, Rome, Italy; ~University "CampusBiomedico" Rome, Italy, #Department of Neurology,Johns Hopkins University School of Medicine, Balti-

Proceedings 213

more, Maryland, %Dept of Neurology, Oasi Inst. for Re-search on Mental Retardation and Brain Aging(IRCCS), Troina-Italy; $Division of ClinicalNeurophysiology (DIMI), University of Genova, Italy;&The KEY Institute for Brain-Mind Research, Univer-sity Hospital of Psychiatry, Zurich, Switzerland; !ITABUniv. Chieti and INFN, UdR L’Aquila, Italy;?Dipartimento Informatica e Sistemistica Univ. "LaSapienza", Rome-Italy, Roma, Italy; @IFN-ConsiglioNazionale delle Ricerche (CNR) Unità MEG - Osp.Fatebenefratelli Isola Tiberina).

This study aimed at mapping (i) the distributed EEGsources specific for mild Alzheimer’s disease (AD) com-pared to vascular dementia (VaD) or normal elderly peo-ple (Nold) and (ii) the distributed EEG sources sensitiveto the mild AD at different stages of severity. RestingEEG (10-20 electrode montage) was recorded from 48mild AD, 20 VaD, and 38 Nold subjects. Both AD andVaD patients had 24-17 of mini mental state examination(MMSE). EEG rhythms were delta (2-4 Hz), theta (4-8Hz), alpha 1 (8-10.5 Hz), alpha 2 (10.5-13 Hz), beta 1(13-20 Hz), and beta 2 (20-30 Hz). Cortical EEG sourceswere modeled by "Low resolution brain electromagnetictomography" (LORETA). Sources of EEG rhythmicitywere defined as specific for mild AD group when theircurrent density values were different than those for VaDand normal groups. Regarding issue (i), There was a de-cline of central, parietal, temporal and limbic alpha 1(low alpha) sources specific for mild AD group respect toNold and VaD groups. On the other hand, occipital alpha1 sources showed a strong decline in mild AD comparedto VaD group. Finally, distributed theta sources werelargely abnormal in VaD but not in mild AD group. Re-garding issue (ii), there was a lower power of occipital al-pha 1 sources in mild AD sub-group having more severedisease. Future studies should evaluate the clinical use-fulness of this source mapping in early differential diag-nosis, disease staging, and therapy monitoring.

MEG-migraine studies: update. - Norman Tepley*+,Susan M. Bowyer*+ and John E. Moran* (*Departmentof Physics Oakland University, Rochester, MI 48309USA; +Department of Neurology Henry Ford Hospital,2799 West Grand Blvd, Detroit, MI 48202 USA).

The Spreading Cortical Depression (SCD) of Leãohas long been implicated in migraine headache. In 1987,K.M.A. Welch suggested MEG could be used to detectSCD during migraine, and soon after Okada used aSQUID magnetometer to detect SCD in turtle cerebel-lum. Since 1988 our laboratory has used MEG to studySCD. We have observed DC shifts and suppression ofspontaneous cortical activity in MEG signals from hu-mans during migraine aura, results similar to those ofLeão in animal models of SCD. We have used a whole

head Neuromagnetometer and visual stimulation to con-firm the hyperexcitability of widespread regionsthroughout occipital cortex in M+A and M-A patientsduring spontaneous and induced aura. Thishyperexcitability explains the susceptibility of these pa-tients for triggering migraine aura. We are studying theeffects of prophylactic medications on cortical activity inmigraine patients. We have monitored visually excitedcortex pre and post treatment with divalproex sodium, amedication known to suppress cortical excitability. Ourstudies suggest a relationship between this drug and un-derlying hyper-excitability of cortex in migraine pa-tients. MEG is uniquely suited to the study of Migraineand SCD. Unlike fMRI, MEG directly measures neuronalactivity occurring over millisecond intervals rather thanthe secondary changes in blood oxygenation taking placeover the course of minutes. Being electrodeless, MEG isbetter able to measure DC changes than EEG. Thesecharacteristics may make MEG useful for selecting clini-cally applicable prophylactic medications that will be ef-fective in reducing the hyper-excitability in particularpatients. (Supported by NIH/NINDS R01-NS30914.)

The neurophysics of consciousness. - E.R. John (BrainResearch Laboratories, NYU School of Medicine, NewYork, NY; Nathan Kline Institute for Psychiatric Re-search, Orangeburg, NY).

Consciousness combines information about attrib-utes of the present multimodal sensory environmentwith relevant elements of the past. Information fromeach modality is continuously fractionated into distinctfeatures, processed locally by different brain regions rel-atively specialized for extracting these disparate compo-nents and globally by interactions among these regions.Information is represented by levels of synchronizationwithin neuronal populations and of coherence amongmultiple brain regions that deviate from random fluctua-tions. Significant deviations constitute Local and GlobalNegative Entropy, or information. Local field potentialsreflect the degree of synchronization among the neuronsof the local ensembles. Large-scale integration, or "bind-ing", is proposed to involve oscillations of local field po-tentials that play an important role in facilitatingsynchronization and coherence, assessed by neuronalcoincidence detectors, and parsed into perceptual framesby cortico-thalamo-cortical loops. The most probablebaseline levels of local synchrony, coherent interactionsamong brain regions, and frame durations have beenquantitatively described in large studies of their age- ap-propriate normative distributions and are considered asan approximation to a conscious "ground state". Thelevel of consciousness during anesthesia can be accu-rately predicted by the magnitude and direction of re-versible multivariate deviations from this ground state.


An invariant set of changes takes place during anesthe-sia, independent of the particular anesthetic agent. Evi-dence from a variety of neuroscience areas supportingthese propositions, together with the invariant reversibleelectrophysiological changes observed with loss and re-turn of consciousness, is used to provide a foundation forthis theory of consciousness. This paper illustrates the in-creasingly recognized need to consider global as well aslocal processes in the search for better explanations ofhow the brain accomplishes the transformation fromsynchronous and distributed neuronal discharges toseamless global subjective awareness.

Theory of mind: the interface of emotion and languagein schizophrenia. - N.C. Andreasen.

The concept of "theory of mind" (TOM) refers to theability to infer and attribute mental states to one’s self andto others and to recognize that behaviors are guided bythese mental states. Examples of mental states include be-liefs, wishes, thoughts, goals, and knowledge. This abil-ity is also referred to as "mentalizing." TOM or"mentalizing" requires the understanding that those men-tal states reflect a subjective reality rather than the realworld. This capacity, which is related to the capacity toput oneself in another’s place, or to have empathy, is animportant component of social interactions. It appears tobe impaired in many individuals with schizophrenia. Be-cause of its philosophical and clinical importance, we un-dertook a study of TOM in a group of healthy volunteersand patients with schizophrenia, using PET to identifythe neural circuits used during a language task that re-quired subjects to attribute a mental state to another per-son. Specifically, they were asked to "imagine that you satnext to a woman on an park bench and you realized shewas crying. Make up a story about what led up to her cry-ing." (The gender of the person was changed to female ifthe subject was a male.) The comparison task consisted ofreading a neutral story aloud, in order to control for thespeech component of the task. In normal individuals thistask activated a distributed group of nodes that includedanterior cingulate and paracingulate regions, L anteriorfrontal regions, L anterior temporal lobe, and R cerebel-lum. Many of these regions are implicated in the identifi-cation of goals and associative memories. The largecerebellar activations add further evidence to the impor-tance of the cerebellum in a many types of mental activity.The patients with schizophrenia had decreased flow inmultiple regions (lateral cerebellum and vermis, visualassociation cortex, and a nonsignificant decrease in thethalamus) and increases in others (R inferior frontal, Rdorsolateral frontal, R parietal, and R putamen). The ar-eas of decreased flow are consistent with many previousstudies indicating problems in recruiting corti-cal-cerebellar circuits in schizophrenia. The areas of in-

crease may reflect a need to draw on right hemisphereregions to perform the task, in order to compensate fordeficits in left frontal and cingulate regions.

MEG evidence for cortical reorganization in schizo-phrenia. - M. Reite, P. Teale, D. Rojas, D. Abrams andA. Bennett (Neuromagnetic Imaging Laboratory, De-partment of Psychiatry, Univ. of Colo. Health SciencesCenter, Denver CO 80262).

MEG evoked field components generated in pri-mary and secondary sensory cortex in response to audi-tory and tactile stimuli in patients with schizophreniafrequently have source location estimates at variancewith those recorded from non-schizophrenic controlsubjects. Schizophrenic subjects tend to exhibit a de-crease in normal left-right brain asymmetry in source lo-cation, and sources often appear somewhat furtheranterior in the left hemisphere. When source locationsare compared with MRI based brain anatomy, auditorysources appear to be dissociated with expected locationsin Heschl’s gyri, and somatosensory sources, at least inthe left hemisphere, often appear to be generated in loca-tions anterior to the central sulcus in subjects withschizophrenia. These findings raise the possibility of sen-sory cortical disorganization, or possibly sensory corticalmosaicism, accompanying schizophrenia. Supported byUSPHS MH 47476.

Development of MEG neuroimaging phenotypes. - R.Coppola (Clinical Brain Disorders Branch, NIMH).

Advances in both genetics and functional imaginghave cross-fertilized into "imaging genomics". FMRI al-lows the investigation of individual variations in complexcognitive behaviors. These variations identify individualphenotypes that can be related to genetic polymorphisms.Magnetoencephalography can add fine temporal resolu-tion to the study of brain activation patterns. An adaptivebeamforming technique is used to examine and localizevariances between active and control states. Once in 3dbrain space, group maps can be formed by appropriatelyregistered structural images. Results with group data forseveral cognitive tasks will be presented. The time com-parison of block and event related designs shows re-sponse related maps that also suggest group differenceswith a COMT polymorphism that has been identified as apossible susceptibility allele for schizophrenia.

Neurophysiological hints to understand psychosis. –W. Strik (Department of Psychiatric Neurophysiology,University Hospital of Clinical Psychiatry, Bern, Swit-zerland).

Traditional event-related potential measures such asamplitudes and latencies are statistically sound, but the

Proceedings 215

changes are non-specific in psychotic groups. Topograph-ical analysis has added specific features forpsychopathological subgroups and indications for thepossible differential mechanisms of the underlying dys-function. Own results will be presented to demonstratethe functional difference of acute remittent psychosis(cycloid psychosis), chronic schizophrenia and manic dis-orders which indicate a distributed hyperarousal, re-gional left-hemispheric functional deficits and frontal

disinhibition, respectively. The results will be interpretedin the context of the respective clinical symptoms. It is con-cluded that a necessary strategy in the neurophysiologicalstudy of psychoses is the definition and understanding ofphysiological behavioural and perceptional complexes,and their central control, and to define psychotic sub-groups in terms of a partial system failure. As an outlook,finally, the example of a paradigm apt to study thetop-down control of perception is presented.


Proceedings - sci.utah.edu

Documents

Transcript of Proceedings - sci.utah.edu