### Refine

#### Document Type

- Article (16)
- Conference Proceeding (4)

#### Keywords

- MEG (2)
- Causality (1)
- Crossmodal (1)
- DCM (1)
- Effective connectivity (1)
- Electroencephalography (1)
- Functional connectivity (1)
- Functional magnetic resonance imaging (1)
- Independent component analysis (1)
- Information theory (1)

#### Institute

- TRENTOOL: a Matlab open source toolbox to analyse information flow in time series data with transfer entropy (2011)
- Background: Transfer entropy (TE) is a measure for the detection of directed interactions. Transfer entropy is an information theoretic implementation of Wiener's principle of observational causality. It offers an approach to the detection of neuronal interactions that is free of an explicit model of the interactions. Hence, it offers the power to analyze linear and nonlinear interactions alike. This allows for example the comprehensive analysis of directed interactions in neural networks at various levels of description. Here we present the open-source MATLAB toolbox TRENTOOL that allows the user to handle the considerable complexity of this measure and to validate the obtained results using non-parametrical statistical testing. We demonstrate the use of the toolbox and the performance of the algorithm on simulated data with nonlinear (quadratic) coupling and on local field potentials (LFP) recorded from the retina and the optic tectum of the turtle (Pseudemys scripta elegans) where a neuronal one-way connection is likely present. Results: In simulated data TE detected information flow in the simulated direction reliably with false positives not exceeding the rates expected under the null hypothesis. In the LFP data we found directed interactions from the retina to the tectum, despite the complicated signal transformations between these stages. No false positive interactions in the reverse directions were detected. Conclusions: TRENTOOL is an implementation of transfer entropy and mutual information analysis that aims to support the user in the application of this information theoretic measure. TRENTOOL is implemented as a MATLAB toolbox and available under an open source license (GPL v3). For the use with neural data TRENTOOL seamlessly integrates with the popular FieldTrip toolbox.

- TRENTOOL: an open source toolbox to estimate neural directed interactions with transfer entropy (2011)
- Poster presentation from Twentieth Annual Computational Neuroscience Meeting: CNS*2011 Stockholm, Sweden. 23-28 July 2011. Poster presentation To investigate directed interactions in neural networks we often use Norbert Wiener's famous definition of observational causality. Wiener’s definition states that an improvement of the prediction of the future of a time series X from its own past by the incorporation of information from the past of a second time series Y is seen as an indication of a causal interaction from Y to X. Early implementations of Wiener's principle – such as Granger causality – modelled interacting systems by linear autoregressive processes and the interactions themselves were also assumed to be linear. However, in complex systems – such as the brain – nonlinear behaviour of its parts and nonlinear interactions between them have to be expected. In fact nonlinear power-to-power or phase-to-power interactions between frequencies are reported frequently. To cover all types of non-linear interactions in the brain, and thereby to fully chart the neural networks of interest, it is useful to implement Wiener's principle in a way that is free of a model of the interaction [1]. Indeed, it is possible to reformulate Wiener's principle based on information theoretic quantities to obtain the desired model-freeness. The resulting measure was originally formulated by Schreiber [2] and termed transfer entropy (TE). Shortly after its publication transfer entropy found applications to neurophysiological data. With the introduction of new, data efficient estimators (e.g. [3]) TE has experienced a rapid surge of interest (e.g. [4]). Applications of TE in neuroscience range from recordings in cultured neuronal populations to functional magnetic resonanace imaging (fMRI) signals. Despite widespread interest in TE, no publicly available toolbox exists that guides the user through the difficulties of this powerful technique. TRENTOOL (the TRansfer ENtropy TOOLbox) fills this gap for the neurosciences by bundling data efficient estimation algorithms with the necessary parameter estimation routines and nonparametric statistical testing procedures for comparison to surrogate data or between experimental conditions. TRENTOOL is an open source MATLAB toolbox based on the Fieldtrip data format. We evaluated the performance of the toolbox on simulation data and also a neuronal dataset that provides connections that are truly unidirectional to circumvent the following generic problem: typically, for any result of an analysis of directed interactions in the brain there will be a plausible explanation because of the combination of feedforward and feedback connectivity between any two measurement sites. Therefore, we estimated TE between the electroretinogram (ERG) and the LFP response in the tectum of the turtle (Chrysemys scripta elegans) under visual stimulation by random light pulses. In addition, we also investigated transfer entropy between the input to the light source (TTL pulse) and the ERG, to test the ability of TE to detect directed interactions between signals with vastly different properties. We found significant (p<0.0005) causal interactions from the TTL pulse to the ERG and from the ERG to the tectum – as expected. No significant TE was detected in the reverse direction. CONCLUSION: TRENTOOL is an easy to use implementation of transfer entropy estimation combined with statistical testing routines suitable for the analysis of directed interactions in neuronal data.

- Measuring information-transfer delays (2013)
- In complex networks such as gene networks, traffic systems or brain circuits it is important to understand how long it takes for the different parts of the network to effectively influence one another. In the brain, for example, axonal delays between brain areas can amount to several tens of milliseconds, adding an intrinsic component to any timing-based processing of information. Inferring neural interaction delays is thus needed to interpret the information transfer revealed by any analysis of directed interactions across brain structures. However, a robust estimation of interaction delays from neural activity faces several challenges if modeling assumptions on interaction mechanisms are wrong or cannot be made. Here, we propose a robust estimator for neuronal interaction delays rooted in an information-theoretic framework, which allows a model-free exploration of interactions. In particular, we extend transfer entropy to account for delayed source-target interactions, while crucially retaining the conditioning on the embedded target state at the immediately previous time step. We prove that this particular extension is indeed guaranteed to identify interaction delays between two coupled systems and is the only relevant option in keeping with Wiener’s principle of causality. We demonstrate the performance of our approach in detecting interaction delays on finite data by numerical simulations of stochastic and deterministic processes, as well as on local field potential recordings. We also show the ability of the extended transfer entropy to detect the presence of multiple delays, as well as feedback loops. While evaluated on neuroscience data, we expect the estimator to be useful in other fields dealing with network dynamics.

- Transfer entropy - a model-free measure of effective connectivity for the neurosciences (2010)
- Understanding causal relationships, or effective connectivity, between parts of the brain is of utmost importance because a large part of the brain’s activity is thought to be internally generated and, hence, quantifying stimulus response relationships alone does not fully describe brain dynamics. Past efforts to determine effective connectivity mostly relied on model based approaches such as Granger causality or dynamic causal modeling. Transfer entropy (TE) is an alternative measure of effective connectivity based on information theory. TE does not require a model of the interaction and is inherently non-linear. We investigated the applicability of TE as a metric in a test for effective connectivity to electrophysiological data based on simulations and magnetoencephalography (MEG) recordings in a simple motor task. In particular, we demonstrate that TE improved the detectability of effective connectivity for non-linear interactions, and for sensor level MEG signals where linear methods are hampered by signal-cross-talk due to volume conduction.

- Learning more by sampling less: subsampling effects are model specific (2013)
- Poster presentation: Twenty Second Annual Computational Neuroscience Meeting: CNS*2013. Paris, France. 13-18 July 2013. When studying real world complex networks, one rarely has full access to all their components. As an example, the central nervous system of the human consists of 1011 neurons which are each connected to thousands of other neurons [1]. Of these 100 billion neurons, at most a few hundred can be recorded in parallel. Thus observations are hampered by immense subsampling. While subsampling does not affect the observables of single neuron activity, it can heavily distort observables which characterize interactions between pairs or groups of neurons [2]. Without a precise understanding how subsampling affects these observables, inference on neural network dynamics from subsampled neural data remains limited. We systematically studied subsampling effects in three self-organized critical (SOC) models, since this class of models can reproduce the spatio-temporal activity of spontaneous activity observed in vivo [2,3]. The models differed in their topology and in their precise interaction rules. The first model consisted of locally connected integrate- and fire units, thereby resembling cortical activity propagation mechanisms [2]. The second model had the same interaction rules but random connectivity [4]. The third model had local connectivity but different activity propagation rules [5]. As a measure of network dynamics, we characterized the spatio-temporal waves of activity, called avalanches. Avalanches are characteristic for SOC models and neural tissue [6]. Avalanche measures A (e.g. size, duration, shape) were calculated for the fully sampled and the subsampled models. To mimic subsampling in the models, we considered the activity of a subset of units only, discarding the activity of all the other units. Under subsampling the avalanche measures A depended on three main factors: First, A depended on the interaction rules of the model and its topology, thus each model showed its own characteristic subsampling effects on A. Second, A depended on the number of sampled sites n. With small and intermediate n, the true A¬ could not be recovered in any of the models. Third, A depended on the distance d between sampled sites. With small d, A was overestimated, while with large d, A was underestimated. Since under subsampling, the observables depended on the model's topology and interaction mechanisms, we propose that systematic subsampling can be exploited to compare models with neural data: When changing the number and the distance between electrodes in neural tissue and sampled units in a model analogously, the observables in a correct model should behave the same as in the neural tissue. Thereby, incorrect models can easily be discarded. Thus, systematic subsampling offers a promising and unique approach to model selection, even if brain activity was far from being fully sampled.

- Setting up the speech production network: how oscillations contribute to lateralized information routing (2012)
- Speech production involves widely distributed brain regions. This MEG study focuses on the spectro-temporal dynamics that contribute to the setup of this network. In 21 participants performing a cue-target reading paradigm, we analyzed local oscillations during preparation for overt and covert reading in the time-frequency domain and localized sources using beamforming. Network dynamics were studied by comparing different dynamic causal models of beta phase coupling in and between hemispheres. While a broadband low frequency effect was found for any task preparation in bilateral prefrontal cortices, preparation for overt speech production was specifically associated with left-lateralized alpha and beta suppression in temporal cortices and beta suppression in motor-related brain regions. Beta phase coupling in the entire speech production network was modulated by anticipation of overt reading. We propose that the processes underlying the setup of the speech production network connect relevant brain regions by means of beta synchronization and prepare the network for left-lateralized information routing by suppression of inhibitory alpha and beta oscillations.

- The timing of feedback to early visual cortex in the perception of long-range apparent motion (2008)
- When 2 visual stimuli are presented one after another in different locations, they are often perceived as one, but moving object. Feedback from area human motion complex hMT/V5+ to V1 has been hypothesized to play an important role in this illusory perception of motion. We measured event-related responses to illusory motion stimuli of varying apparent motion (AM) content and retinal location using Electroencephalography. Detectable cortical stimulus processing started around 60-ms poststimulus in area V1. This component was insensitive to AM content and sequential stimulus presentation. Sensitivity to AM content was observed starting around 90 ms post the second stimulus of a sequence and most likely originated in area hMT/V5+. This AM sensitive response was insensitive to retinal stimulus position. The stimulus sequence related response started to be sensitive to retinal stimulus position at a longer latency of 110 ms. We interpret our findings as evidence for feedback from area hMT/V5+ or a related motion processing area to early visual cortices (V1, V2, V3).

- Subsampling effects in neuronal avalanche distributions recorded in vivo (2009)
- Background Many systems in nature are characterized by complex behaviour where large cascades of events, or avalanches, unpredictably alternate with periods of little activity. Snow avalanches are an example. Often the size distribution f(s) of a system's avalanches follows a power law, and the branching parameter sigma, the average number of events triggered by a single preceding event, is unity. A power law for f(s), and sigma=1, are hallmark features of self-organized critical (SOC) systems, and both have been found for neuronal activity in vitro. Therefore, and since SOC systems and neuronal activity both show large variability, long-term stability and memory capabilities, SOC has been proposed to govern neuronal dynamics in vivo. Testing this hypothesis is difficult because neuronal activity is spatially or temporally subsampled, while theories of SOC systems assume full sampling. To close this gap, we investigated how subsampling affects f(s) and sigma by imposing subsampling on three different SOC models. We then compared f(s) and sigma of the subsampled models with those of multielectrode local field potential (LFP) activity recorded in three macaque monkeys performing a short term memory task. Results Neither the LFP nor the subsampled SOC models showed a power law for f(s). Both, f(s) and sigma, depended sensitively on the subsampling geometry and the dynamics of the model. Only one of the SOC models, the Abelian Sandpile Model, exhibited f(s) and sigma similar to those calculated from LFP activity. Conclusions Since subsampling can prevent the observation of the characteristic power law and sigma in SOC systems, misclassifications of critical systems as sub- or supercritical are possible. Nevertheless, the system specific scaling of f(s) and sigma under subsampling conditions may prove useful to select physiologically motivated models of brain function. Models that better reproduce f(s) and sigma calculated from the physiological recordings may be selected over alternatives.

- Analyzing possible pitfalls of cross-frequency analysis : poster presentation from Twentieth Annual Computational Neuroscience Meeting CNS*2011 Stockholm, Sweden, 23 - 28 July 2011 (2011)
- Poster presentation from Twentieth Annual Computational Neuroscience Meeting: CNS*2011 Stockholm, Sweden. 23-28 July 2011. One of the central questions in neuroscience is how neural activity is organized across different spatial and temporal scales. As larger populations oscillate and synchronize at lower frequencies and smaller ensembles are active at higher frequencies, a cross-frequency coupling would facilitate flexible coordination of neural activity simultaneously in time and space. Although various experiments have revealed amplitude-to-amplitude and phase-to-phase coupling, the most common and most celebrated result is that the phase of the lower frequency component modulates the amplitude of the higher frequency component. Over the recent 5 years the amount of experimental works finding such phase-amplitude coupling in LFP, ECoG, EEG and MEG has been tremendous (summarized in [1]). We suggest that although the mechanism of cross-frequency-coupling (CFC) is theoretically very tempting, the current analysis methods might overestimate any physiological CFC actually evident in the signals of LFP, ECoG, EEG and MEG. In particular, we point out three conceptual problems in assessing the components and their correlations of a time series. Although we focus on phase-amplitude coupling, most of our argument is relevant for any type of coupling. 1) The first conceptual problem is related to isolating physiological frequency components of the recorded signal. The key point is to notice that there are many different mathematical representations for a time series but the physical interpretation we make out of them is dependent on the choice of the components to be analyzed. In particular, when one isolates the components by Fourier-representation based filtering, it is the width of the filtering bands what defines what we consider as our components and how their power or group phase change in time. We will discuss clear cut examples where the interpretation of the existence of CFC depends on the width of the filtering process. 2) A second problem deals with the origin of spectral correlations as detected by current cross-frequency analysis. It is known that non-stationarities are associated with spectral correlations in the Fourier space. Therefore, there are two possibilities regarding the interpretation of any observed CFC. One scenario is that basic neuronal mechanisms indeed generate an interaction across different time scales (or frequencies) resulting in processes with non-stationary features. The other and problematic possibility is that unspecific non-stationarities can also be associated with spectral correlations which in turn will be detected by cross frequency measures even if physiologically there is no causal interaction between the frequencies. 3) We discuss on the role of non-linearities as generators of cross frequency interactions. As an example we performed a phase-amplitude coupling analysis of two nonlinearly related signals: atmospheric noise and the square of it (Figure 1) observing an enhancement of phase-amplitude coupling in the second signal while no pattern is observed in the first. Finally, we discuss some minimal conditions need to be tested to solve some of the ambiguities here noted. In summary, we simply want to point out that finding a significant cross frequency pattern does not always have to imply that there indeed is physiological cross frequency interaction in the brain.

- Graphical analyses in delay interaction networks (2013)
- Poster presentation: Twenty Second Annual Computational Neuroscience Meeting: CNS*2013. Paris, France. 13-18 July 2013. Network or graph theory has become a popular tool to represent and analyze large-scale interaction patterns in the brain. To derive a functional network representation from experimentally recorded neural time series one has to identify the structure of the interactions between these time series. In neuroscience, this is often done by pairwise bivariate analysis because a fully multivariate treatment is typically not possible due to limited data and excessive computational cost. Furthermore, a true multivariate analysis would consist of the analysis of the combined effects, including information theoretic synergies and redundancies, of all possible subsets of network components. Since the number of these subsets is the power set of the network components, this leads to a combinatorial explosion (i.e. a problem that is computationally intractable). In contrast, a pairwise bivariate analysis of interactions is typically feasible but introduces the possibility of false detection of spurious interactions between network components, especially due to cascade and common drive effects. These spurious connections in a network representation may introduce a bias to subsequently computed graph theoretical measures (e.g. clustering coefficient or centrality) as these measures depend on the reliability of the graph representation from which they are computed. Strictly speaking, graph theoretical measures are meaningful only if the underlying graph structure can be guaranteed to consist of one type of connections only, i.e. connections in the graph are guaranteed to be non-spurious. We propose an approximate solution to improve this situation in the form of an algorithm that flags potentially spurious edges that are due to cascade effects and "three node" common drive effects in a network representation of bivariately analyzed interactions. As these two effects are responsible for a large part of spurious connections in bivariate analyses, their removal would mean a significant improvement of the network representation over existing bivariate solutions. Our approach is based on the detection of directed interactions and the weighting of these interactions by their reconstructed interaction delays. We demonstrate how both questions can be addressed using a modified estimator of transfer entropy (TE). TE is an implementation of Wiener's principle of observational causality based on information theory [1], and detects arbitrary linear and non-linear interactions. Using a modified TE estimator that uses delayed states of the driving system, one can mathematically prove that transfer entropy values peak if the delay of the state of the driving system equals the true interaction delay [2]. From this analysis, we derive a delay weighted network representation of directed interactions. On this network representation, potentially spurious interactions can be detected by analyzing sets of alternative paths between two endpoints in terms of their summed delays. The proposed algorithm may be used to prune spurious edges from the network, improving the reliability of the network representation itself and enhancing the applicability of subsequent graph theoretical measures. For the detection of "multi-node" common drive effects, that are not considered in this study, a theoretical solution exists as well, extending the power of the method, but this solution has not been implemented yet. We demonstrate the application of this algorithm to networks of interacting neural sources in magneto-encephalographic data, and show that roughly 30% of bivariate interactions in these data are potentially spurious, and thus alter graph properties. We conclude that the post hoc correction provided by our approach is a computationally less demanding alternative to a fully multivariate analysis of directed interactions, and preferable in cases were a multivariate treatment of the data is difficult due to the limited amount of data available.