# Brain Connectivity Workshop.

## Havana City, Cuba

### April 26 - 29, 2004

### Introductory Neuroinformatics Course : April 23 - 25

#### Organized by:

- Pedro A. Valdés-Sosa
- Rolf Kötter

#### Sponsored by:

#### Background:

Brain function is dependent on the interactions between specialized regions of cortex that
process information within local and global networks. Integration of information arises from
these interactions as a dynamic process on different time scales. Investigations of the physical
connections between neuronal structures and measurements of brain activity in vivo have given
rise to concepts of anatomical, functional and effective connectivity, which have been useful for
undestanding brain mechanisms and their plasticity. The First multi-disciplinary workshop on
*"Functional Brain Connectivity"* organized by
Rolf Kötter and Karl Friston in April 2002 in Düsseldorf, Germany, carefully defined
the concepts and explored the relationship between different conceptual approaches. Following
this successful event, the Second Workshop organized by Ed Bullmore and Lee Harrison was held in
*May 2003 in Cambridge,
England*, with a focus on complex analysis and dynamical systems theory. This year's
workshop will continue the multi-disciplinary discussion with a focus on the fusion of methods
with different spatial and temporal resolution.

The Third Workshop on Brain Connectivity will be held from April 26th to 30th, 2004 in Havana....

#### Aims:

The general aim of the meeting is to bring together experts from the fields of Computational and Experimental Neuroscience to review and advance recent work on structural, functional and effective connectivity. The specific focus of this workshop will be the fusion of different brain imaging approaches for measuring and explaining dynamic interactions between neuronal ensembles and their relation to information processing in the brain. For example, it will address questions that arise when interpreting functional imaging (fMRI and PET), electrophysiological (EEG, MEG, LFP and single/ multiple unit recordings) data and their fusion.

#### Workshop programme:

The workshop will be organized around seven general themes:

- Causal Inference: Graphical Models and Time Series
- Statistical Techniques for Measuring Connectivity
- Anatomical Connectivity
- Functional Connectivity
- Multimodal Neuroimages for Discovering Connectivity
- Interventional Studies of Neural Causal Systems
- Connectivity Changes in Pathology

As proven useful and popular in the past, the format of this workshop is special: Instead of lengthy slide presentations experts will give a brief (max. 15 min.) introduction of a topic of their choice and lead a discussion for up to one hour in interaction with questions and contributions from the audience.

#### Introductory Neuroinformatics Course:

This course provides an introduction to connectivity analyses in the context of functional imaging studies. Speakers will explain the conceptual issues and introduce relevant resources and software packages with practical examples. Participants are encouraged to try them using the computer infrastructure at the University of Computer Sciences.

**For further information on the Introductory Neuroinformatics Course and registration
click here**

#### Contributors and topics:

**Tim Behrens**(Oxford Centre for Functional Magnetic Resonance Imaging of the Brain):Connectivity-based parcellation of grey matter using diffusion tractography. The interpretation of cytoarchitectonically discrete brain units in terms of brain function is a major goal of Neuroimaging. Unfortunately, boundaries between such subunits do not correspond well to landmarks easily visible in vivo. However a brain region's function is constrained by its connectional anatomy - regions which are truly functionally discrete are expected to maintain markedly different connectivity patterns. Diffusion tractography can provide in-vivo information on anatomical connectivity in the human brain. Here, we test the the hypothesis that changes in circuitry revealed by diffusion tractography may be used to define the extent of, and boundaries between, functionally and cytoarchitectonically discrete brain regions without any prior knowledge of the regions' connectivity patterns.

Example in medial area 6:

Medial area 6 consists of two cytoarchitectonically distinct regions in monkey and (depending on reports) two or three in human. In non-human primates there is a change in connectivity along medial frontal cortex: pre-SMA connects to prefrontal/anterior cingulate cortex whereas SMA proper connects to sensorimotor regions. We test the hypothesis that we can detect this change in connectivity in the human brain using diffusion tractography, and use it to define the boundary between SMA and preSMA. The resulting boundaries are compared with functionally defined boundaries from fmri experiments designed to activate the two areas.**Michael Breakspear**(School of Physics at the University of Sydney & Brain Dynamics Centre at Westmead Hospital, Sydney, Australia):Investigating dynamic correlations in a neural system with a multiscale architecture using wavelets. The architecture of the brain is characterized by a modular organization repeated across a hierarchy of spatial scales - neurons, cortical columns, Brodmann areas, etc. It is important to consider that the processes governing neural dynamics at any given scale are not only determined by the behavior of other neural structures at that scale, but also by the emergent behavior of smaller scales. Wavelets are natural basis functions for investigating such phenomena and we introduce two related applications: (1) A wavelet-based functional connectivity method which allows detection of correlations between brain regions within and between spatial scales ('information cascade'). The method is illustrated on human fMRI data and numerical data from a dynamical model of a neural system. (2) A general theoretical framework for neural systems in which the dynamics are nested within a multiscale architecture. Explicitly, a hierarchy of neural systems is modeled. The dynamics at any given spatial scale are coupled to the scale-congruent emergent dynamics of smaller scales. It is shown how synchronization in small-scale structures hence influences the dynamics in larger structures in an intuitive manner that cannot be captured by existing modeling approaches.

**Michael Eichler**(Department of Statistics, University of Chicago):Causal inference with graphical time series models. One major problem in the identification of causal relationships from observational data are the possible influences from latent variables which introduce so-called spurious causalities. This is true in particular for studies of functional connectivity in the brain where only a very restricted number of processes can be measured and analyzed. Recent advances in the understanding of such latent variable structures were based on graphical models which provides a general framework for describing and infering causal relations. In this talk, we present a graphical representation of the dependence structure of multivariate time series which is based on the concept of Granger causality. This graphical approach can be used for discussing spurious causality and leads to a new model for time series with latent variables.

**Karl Friston**(Functional Imaging Laboratory, Wellcome Department of Imaging Neuroscience. UCL):Learning and inference in the brain. There are several architectural principles of functional brain anatomy that have emerged from careful anatomic and physiologic studies over the past century. These principles are considered in the light of representational learning to see if they could have been predicted a priori on the basis of purely theoretical considerations. Specifically, I will focus on hierarchical dynamic models (HDMs) and expectation maximisation (EM) schemes for their estimation. These models are very general in the sense that they subsume many simpler variants, such as independent component analysis, through to generalised nonlinear convolution models. The generality of HDMs renders their EM a useful framework that covers procedures ranging from variance component estimation, in classical linear observation models, to blind deconvolution, using exactly the same formalism and operational equations. Critically, they may provide heuristics that inform our understanding of neuronal processing. For example, the central role of hierarchies in empirical Bayesian formulations of representational learning may provide an understanding of why sensory cortices in the brain are arranged hierarchically. A second example is the need for explicit generative and recognition models in the context of noninvertible processes generating auditory data. This dichotomy may be useful in understanding asymmetries between forward and backward connections of the brain in the context of predictive coding. The notion that the brain may use empirical Bayes for inference about its sensory input, given the hierarchical organisation of cortical systems, is compelling. Although it is fairly easy to develop this in the context of static observation models, it would be interesting to generalise the same idea to cover dynamical systems. This would enable us to model and understand evoked brain responses in a much more functionally informed fashion. Here predictive coding takes on a dual meaning in the sense that the prediction may involve, not only minimizing prediction error (to provide conditional estimators), but also a component of forecasting, to pursue conditional trajectories of dynamically evolving states.

**Lee Harrison**(Functional Imaging Laboratory, Wellcome Department of Imaging Neuroscience. UCL):Ensemble dynamics and the Fokker-Planck Equation. Cortical activity is the product of nonlinear electrodynamic interactions among neuronal populations in repsonse to exogenous input. Electrophysiological phenonmena are generated from these interactions, which forms the basis of a modelling strategy to estimate biologically plausible connectivity models of the underlying causes of our measurements, which are consistent with the data. Macroscopic features of cortical activity can be modelled in terms of the microscopic electrodynamics of individual neurons in the same way that pressure, volume and temperature can be formulated from the microscopic electro-mechanical properties of atoms within a gas. This leads to a formulation of population neurodynamics in terms of the Fokker-Planck equation. The solution of this equation is the temporal evolution of a probability density over state-space and represents the distribution of an ensemble of trajectories through state-space. A biologically motivated model of coupled populations of Hodgkin-Huxley like neurons connected through AMPA, GABA and NMDA synapses is proposed. Potential uses include using it as an observation model to estimate coupling among neuronal ensembles given data. The Fokker-Planck formalism will be motivated and illustrated through physical and neurobiological examples.

**Maciej Kaminski**(Faculty of Physics Warsaw University, Poland):Determination of transmission patterns in multichannel EEG. The methods of determination of causal influences (direction of flows) between signals became more and more popular in connection with the development of modern laboratory equipment and computers which allow recording and analyzing of many data channels. Most of the traditional methods of determination of directional relations are designed for pairs of channels. Unfortunately, bivariate methods which work well when only two channels are considered, will not necessarily give correct results when applied to multichannel sets of data. The advantages of multichannel way of causal influence estimation over pair-wise approach and possible cause of mistakes generated by a bivariate approach will be presented on the example of Directed Transfer Function and other methods.

**Rolf Kötter**(Computational | Systems | Neuroscience Group at the C. &. O. Vogt Brain Research Institute in Düsseldorf, Germany):Network Motifs. Complex brains have evolved a highly efficient network architecture whose structural connectivity is capable of generating a large repertoire of functional states. We detect characteristic network building blocks (structural and functional motifs) in neuroanatomical data sets and we identify a small set of structural motifs that occur in significantly increased numbers. Our analysis suggests the hypothesis that brain networks maximize both the number and the diversity of functional motifs, while the repertoire of structural motifs remains small. Using functional motif number as a cost function in an optimization algorithm, we obtain network topologies that resemble real brain networks across a broad spectrum of structural measures, including small-world attributes. These results are consistent with the hypothesis that highly evolved neural architectures are organized to maximize functional repertoires and to support highly efficient integration of information.

**Denis Lebihan**(Service Hospitalier Frédéric Joliot (SHFJ)):Brain connectivity seen by diffusion MRI. Functional Magnetic Resonance Imaging (fMRI) has appeared as a powerful new tool which offers the potential to look at the dynamics of cerebral processes underlying cognition, noninvasively and on an individual basis. Still, the real understanding of brain function requires direct access to the functional unit made of the neuron, so that one may look at the transient temporal relationships that exist between largely distributed groups of hundreds or thousands of neurons. Furthermore, communication pathways between networks, which are carried by brain white matter, must be identified to establish connectivity maps at the individual scale, taking into account individual variability. In this respect, MRI of molecular diffusion may play a significant role. During their random, diffusion-driven displacements water molecules probe tissue structure at a microscopic scale well beyond the usual image resolution. The observation of these displacements thus provides valuable information on the structure and the geometric organization of tissues. For instance, because diffusion is modulated by the spatial orientation of large bundles of myelinated axons running in parallel in brain white matter, an important potential application of diffusion MRI is the visualization of anatomical connections between different parts of the brain on an individual basis. This feature can be exploited to map out the orientation in space of the white matter tracks. Furthermore, recent data suggest that diffusion MRI could also be used to image brain activation by directly visualizing dynamic tissue changes associated with neuronal activation.

**Lucy Lee**(Functional Imaging Laboratory, Wellcome Department of Imaging Neuroscience. UCL):Using analyses of effective connectivity to explore the effects of rTMS on the motor system. The effects of repetitive transcranial magnetic stimulation (rTMS) on the excitability of the motor system are well characterised. However, very few effects of rTMS have been seen on measures performance during simple motor behaviour. In this talk I will present results from a number of studies where functional imaging was used to explore the effects of rTMS on the motor network during the performance of simple motor tasks. The combination of analyses examining changes in functional integration and segregation appears to offer useful insights into the mechanisms by which the motor system is able to maintain task performance during changes in cortical excitability.

**Randy McIntosh**(Rotman Research Institute of Baycrest Centre, Toronto, Canada):Causal Inference and the mind: how do we know when the math is right? The growth in neuroimaging over the past two decades has resulted in a growing interest in the development of new methods of image analysis. Many are focused on enabling inferences about causal influences among parts of the brain during different mental functions (e.g., structural equation modeling, multivariate autoregressive modeling, dynamic causal modeling). A common question arising in the implementation of such methods is the validity of the application in terms of the underlying neural and mental processes they are designed to reveal, and whether there are particular experimental paradigms that are better suited for the application of these methods. We will decide on the best answer to these questions in the course of my presentation.

**Jean F. Mangin**(Service Hospitalier Frédéric Joliot (SHFJ)):Inference of anatomical connectivity from diffusion weighted MR data: an inverse problem framework. A family of methods aiming at the reconstruction of a putative fascicle map from any diffusion-weighted dataset will be proposed. This fascicle map is defined as a trade-off between local information on voxel micro-structure provided by diffusion data and a priori information on the low curvature of plausible fascicles. The optimal fascicle map is the minimum energy configuration of a simulated spin glass in which each spin represents a fascicle piece. This spin glass is embedded into a simulated magnetic external field that tends to align the spins along the more probable fiber orientations according to diffusion models. A model of spin interactions related to the curvature of the underlying fascicles introduces a low bending potential constraint. The talk will end with a brief description of the inference of connectivity matrices from such fascicle maps and automatic parcellations of the cortical surface.

**Tohru Ozaki**(Department of Prediction Control Institute of Statistical Mathematics, Japan):Nearest Neighbour ARX (NN-ARX) modelling of spatial time series with application to localization & connectivity study of fMRI data. We would like to present a time series approach for the analysis of functional Magnetic Resonance Imaging (fMRI) data. Recently the MRI has become a vital tool for diagnosing brain tumers and other diseases of the central nervous system. We have been working on fMRI data analysis since 2000 when Prof.P.Valdes-Sosa and I visited Prof.N.Sadato in his lab. The most widely used standard method of fMRI data analysis is the SPM (Statistical Parametric Mapping) method developed by K.Friston and his group in 1994. We think the SPM method is useful in many ways, but it is not exploiting dynamic information involved in fMRI data properly. In the present talk, we would like to show how the useful spatio-temporal information, such as localization and connectivity, can be extracted from the data using our time series modeling approach with two types of experimental data: one is visual stimulus data from Prof.Sadato's lab, and another data is motor task data from Prof.Kawashima's lab in NICHe, Tohoku University.

**Geoffrey M. Parker**(Division of Imaging Science and Biomedical Engineering, University of Manchester, United Kingdom):Quantification of connectivity using diffusion weighted MRI: capabilities and challenges. Recent developments in fibre tracking using diffusion weighted MRI have raised the possibility of performing quantitative measurements to provide information on anatomical connectivity and the structural integrity of white matter tracts. Probabilistic connectivity methods allow the assignment of confidence to putative connections and the generation of anatomical brain connectivity matrices that may assist in system characterisation. Definition of the likely routes and volumes of connective pathways and their associated grey matter regions allows the relationships between abnormal function and possible causes related to tract damage (e.g. infarcts) to be defined. However, even with the recent rapid methodological developments, a number of fundamental problems remain in using fibre tracking as a quantitative scientific tool. The nature of the relationship between the measured diffusion weighted signal and fibre architecture is still incompletely understood; the models used to approximate this relationship are not universally agreed. Similarly, the models of fibre trajectories based on voxel-wise diffusion measurements often include heuristic assumptions that vary considerably between methods. These variations and uncertainties are in addition to the more mundane, but no less important, influence of basic EPI MR image quality, including the effects of voxel size, point spread function, noise, and distortion. The effects of these factors will be considered and their implications for the use of quantitative fibre tracking discussed.

**Tomas Paus**(Cognitive Neuroscience Unit/Neuropsychology Department, Montreal Neurological Institute):Studies of cortical connectivity and oscillations in healthy and disordered brain.

**William Penny**(Functional Imaging Laboratory, Wellcome Department of Imaging Neuroscience. UCL):State-Space Modeling. The focus of this presentation is the use Bilinear Dynamical Systems (BDS) for model-based deconvolution of fMRI series. BDSs are a type of Dynamic Causal Model which comprise a stochastic bilinear neurodynamical model specified in discrete-time and a set of linear convolution kernels for the hemodynamics. I will also discuss how best to make inferences about large-scale functional integration.

**Silke Dodel**(Service Hospitalier Frédéric Joliot (SHFJ)):Condition dependent changes in functional connectivity and physiological confounds. Our contribution consists of two parts: Firstly we investigate how functional connectivity networks in fcMRI are affected by the subject's heart beat and respiration. The latter were measured simultaneously during high rate MRI data acquisition and the functional connectivity networks were determined in a data driven manner using graph theory. Physiology highly affects the data variance, globally in the case of respiration and locally in the case of heart beat. We found that linear removal of the physiological signal based on the simultaneous measurements does not fully remove spurious functional connectivity. We therefore use in addition various other methods for physiological effect removal and compare them with respect to the resulting functional connectivity networks. Secondly we investigate condition dependent changes in functional connectivity that should be in general insensitive to physiological artefacts. As a first approach one could compare the functional connectivity networks obtained using data only from the respective conditions. We chose, however, a more general framework, by introducing a weight function for the contribution of every time step to the correlation and test or links in a non parametric framework. The method is illustrated on an fMRI paradigm designed to study brain substrates of language.

**Jorge Riera**(Advanced Science and Technology of Materials NICHe, Tohoku University Aoba 10, Aramaki, Aobaku, Sendai 980-8579, Japan):Bottom-up vs. top-down strategies: modeling the fusion of multi-modality neuroimages, causality and connectivity patterns. Recent advances in neuroscience have allowed for a basic understanding of brain functions associated with cognitive events in humans. This improved comprehension of the underlying mechanisms associated with the brain activation, ranging from the emerging electrophysiological processes in the neural-circuitry to the imminent vascular changes induced by a metabolic/oxygen demand, has brought into light a tremendous opportunity to develop sophisticated biophysical models to elucidate the temporal dynamic and connectivity patterns of "activation" in specific brain areas. Innovative neuro-imaging techniques have been emerging over the last few years, providing neuroscientists with a powerful tool to assess the spatio-temporal variations of some interpretable physical magnitudes inside the brain (fMRI) as well as their external reflections (NIRs and EEG).

In this context, we would like to emphasize the role that bottom-up models based on physiology will play in solving the inverse problems related to top-down data analysis approach. We will discuss some future prospects for the integration of neuroimaging multi-modalities, which require the concept of connectivity and causality to be carefully revised. We would like to bring into debate some of the different tendencies, which have recently become apparent for the analysis of the structures of activation, focusing on the following aspects:

A)- Is connectivity and causality strictly associated with any of the following: sequential episodes (either forced by or emerging from spatial inter-relationships); or/and concomitant temporal modes of oscillations (where pure anatomical connections only facilitate patterns formation)?

B)- Is activation directly related to the electro-chemical process at the level of the synapses? How does vasculature temporally filter these fast signals?

C)- What role do the vascular and metabolic control mechanisms (i.e. sphincters and nitric oxide influence in the microvasculature) play?

D)- What kind of association exists between glucose and oxygen consumptions, the glycolysis and TCA cycle? What is the immediate consequence of the glycogen shunt model?

E)- How can the relationship between events related transients and spontaneous oscillatory activity be explored? How can we interpret the negativity correlation between EEG and fMRI/NIRs in the case of alpha rhythms?

F)- How could different levels of modeling simplify Inverse Problems (fMRI/NIRs/EEG/MEG) that share a common etiology at the physiological level?**Christian Beckman**(Oxford Centre for Functional Magnetic Resonance Imaging of the Brain):Investigations into Resting State Networks using FMRI and FMRI-EEG. Resting State Networks (RSNs) are self-coherent networks resembling plausible (grey-matter) activation networks'' found (for example) in resting BOLD FMRI data. A study of RSNs has been performed using probabilistic ICA, leading to the identification of 4 primary spatial patterns which are quite consistent across different subjects. A second study has investigated the link between the alpha-power time course and the BOLD RSN time course in simultaneous FMRI-EEG data. It is known that motor activity during relaxed, alpha-generating states enhance the alpha EEG rhythm. Therefore we investigated the modulation by a motor task of the alpha-correlated FMRI maps, finding different networks in the different states. The different maps could suggest multiple, state-dependent, global-network generators of the alpha rhythm.

**David S. Tuch**(Martinos Center for Biomedical Imaging

Massachusetts General Hospital):Diffusion MRI of neural circuitry. Magnetic resonance diffusion imaging may provide a technique for mapping neural connectivity in the human brain non-invasively. Current approaches for reconstructing neural connectivity patterns from the MR image are heavily model-dependent however. I will describe a novel approach to the connectivity reconstruction problem which resolves a number of open issues.

**Pedro A. Valdés-Sosa**(Cuban Neuroscience Center):Connectivity analysis via Bayesian EEG-fMRI fusion EEG/fMRI Neuroimage fusion promises simultaneous high spatial and temporal resolution, benefits that carry over to connectivity analysis. A principled fusion effort may be carried out via SYMMETRICAL hierarchical Bayesian modeling in which both EEG and fMRI are given equal status as data produced by unobserved state variables. Initial efforts at fusion have been only partially successful due to incomplete modeling/knowledge of the EEG/BOLD generation mechanism as is briefly outlined. Lack of progress may be explained using the state space formulation. While considerable progress has been made in modeling SYSTEM DYNAMICS very little has been achieved in terms of modeling the OBSERVATION EQUATION at a voxel level: A) the EEG is only partially observable spatially, B) There is a temporal mismatch between the EEG and the fMRI; and very importantly the process of neural synchronization has yet to be incorporated into forward EEG/fMRI modeling. Further empirical analysis of EEG/fMRI recordings is therefore needed to establish "neurodynamic" relations that may lead to more detailed modeling. We continue to develop tools for this task by expanding the Bayesian Multivariate Autoregressive Model using spatial priors that allow the analysis of massive datasets. In the previous connectivity workshop we introduced a spatial smoothness prior to analyze fMRI connectivity patterns over brain manifolds. In this presentation we show that the "lasso" sparse regression technique may be used to analyze jointly connectivity patterns efficiently

**Thomas Koenig**(Department of Psychiatric Neurophysiology, University Hospital of Clinical Psychiatry Bern, Switzerland):Functional brain connectivity, transient microstates and combined EEG and fMRI. Abstract: Higher order brain information processing is assumed to require transient binding of extended neural patches, forming shortlasting neurocognitive networks. It has been proposed that this binding is achieved by the existence of electric oscillations that are synchronous over all regions involved in a processing step. Once that processing step terminates, the elements of the network disconnect and the following processing step is initiated, forming another transient network of synchronously oscillating brain regions. From that point of view, we propose that there are two types of connectivity that need to be considered separately: The first type is the transient binding of neural patches, the second type the preferential sequence of processing steps. In terms of analysis, the first 'binding' type of connectivity can only yields EEG patterns that are characterized by a single time-course with a stable spatial configuration (microstate). Across electrodes, no shift in time or phase can occur. The existence of such periods of stable EEG spatial configuration has indeed been shown repeatedly, and the predominant configurations appear consistently across subjects. By correlating the number or intensity of typical microstates with the fMRI BOLD signal acquired during simultaneous EEG fMRI recordings, the location of the neural patches that formed the transient network during those microstates can be identified. Once a set of such transient synchronous oscillations with stable configuration has been identified and their time-course has been established, the second type of connectivity, i.e. the preferred sequence of events can be investigated. Recent work in time-domain EEG has indeed shown that the sequence of microstates deviates from randomness and that these deviations are common across subjects. If the time-course of microstates is assessed in the time-frequency domain, other measures of connectivity and causality such as Granger-causality or cross-coherence can be employed to study the functional interrelations between microstates.

**Keith Worsley**(Department of Mathematics and Statistics Brain Imaging Centre, Montreal Neurological Institute McGill University):Detecting multivariate effective connectivity. Abstract: Univariate effective connectivity can be detected using the usual correlation coefficient between univariate data at two spatial locations, converted to a T-statistic, and thresholded. For multivariate data there are a number of choices, all functions of the canonical correlations between the multivariate measures at the two locations. Examples are the HRF sampled at 1s intervals, real and imaginary components of the complex fMRI signal or EEG data at a particular frequency, or, for anatomical data, vector deformations. We present simple extensions of random field theory that allows us to set a very precise threshold for a number of multivariate test statistics, all based on Roy's union-intersection principle.

#### Program details:

The workshop will commence on Monday morning, 26 April, and conclude on Thursday night, 29 April 2004. There will be held short "hands on" courses on the use of software in this field during the two days preceding the meeting.

#### Location and Directions:

The workshop will take place in a conference room inside Palco Hotel.

#### Costs and registration:

Registration is made electronically on the neuroinf.org server. More info can be found
**here**.

There is a charge of $250 to cover administration and catering (coffee break and lunch) and stationary costs.

On-site registration is accepted

For further information on the Introductory Neuroinformatics Course and registration
**click here**

#### Accommodation:

Once your registration fee is recieved, havanatur agency
office (**http://www.havanatur.cu**) in your country (if
exist) will contact you in order to offer you tour packages which includes **visa procedures,
airfare, internal tranfers in\out the airport and hotel acommodation** in two hotels of your
preference with which we have negotiated the the following rates:

**Package for Hotel Palco**

- Airfare for the specified country + $183 (single room with breakfast included for three nigths)
- Airfare for the specified country + $132 (double room with breakfast included for three nigths)
- Airfare for the specified country + $243 (single room with breakfast and a meal included for three nigths)
- Airfare for the specified country + $192 (double room with breakfast and a meal included for three nigths)

"Hotel Palco" is the only one included in those packages through a special arrangements of prices, other hotels near the place of the meeting are also available:

**Hotel Comodoro****Hotel Chateau****Hotel Copacabana****Hotel Neptuno****Novotel Miramar**

*Please take into account that we will start the meeting at 9am on Monday 26th
April.*

#### General Notes:

**($) means US dollar**

**Weather:**
www.cubaadvice.com/english/clima.asp

**Optional tour places to visit:**

- City Map
- Habana Vieja (Old Havana)
- Morro-Cabaña Fortress
- La Bodeguita del Medio
- El Floridita

#### Payment Options:

See also www.hirnforschung.net

and A dual Congress Psychiatry and the Neurosciences

#### Correspondence to:

**Organizing Committee:**

orgcommittee(at)cneuro.edu.cu

**Secretary of the Organizing Committee:**

Pedro A. Valdes Hernandez