organized by Maximilien Chaumon
Organizing committee: Sébastien Crouzet, Sophie Herbst, Simon Ludwig, Lyudmyla Kovalenko, Fosco Bernasconi, Felix Ball, Isabelle Bareither, Tessa Rusch, Anna Strasser
19–21 February 2014
Venue: Humboldt-Universität zu Berlin, Berlin School of Mind and Brain, Luisenstraße 56, Festsaal, 10117 Berlin
Electroencephalography (EEG) is one of the methodologically most rapidly expanding tools in cognitive sciences. Its exquisite temporal resolution, its ever-growing spatial precision, and its relatively modest price make it a method of choice for cognitive neuroscience labs.
However, the proliferation of new equipment, methods, software and analysis frameworks puts every EEG researcher in the middle of an immense field of possibilities that are rarely fully exploited, and can be very challenging without appropriate introduction.
Cutting EEG will bring together researchers who have recently developed or used new tools or frameworks for EEG data analysis and made use of these tools to advance our understanding of cognition. The attendance is open to experienced as well as novice EEG researchers seeking a thorough introduction to cutting-edge methods in EEG data analysis.
Cutting EEG is organized around six thematic half-day sessions with talks, structured discussions and technical tutorials. The first day will also end with a public lecture presenting exciting new developments in high frequency EEG.
Dynamic causal modelling of EEG data: recent developments
Among the methods of inferring effective connectivity from EEG data, dynamic causal modelling (DCM) has been successful at combining biological realism with mathematical rigour. DCM is based on a combined spatial and temporal forward model, modelling how the assumed neuronal sources, the connections between them, the sensory input, and experimental modulations generate the measured signal in terms of its topography and dynamics. Typically, several alternative models are inverted and their parameters optimised using a variational Bayes scheme. In case of EEG data, DCM can be applied to event-related potentials and induced time-frequency responses, allowing to infer the underlying effective extrinsic connectivity (such as the strength of bottom-up, top-down and lateral connections between the sources) and intrinsic connectivity (modelling the adaptation of neuronal responses to local influences) from both kinds of data. Furthermore, recent developments in DCMs - namely the canonical microcircuit model - enable a more detailed investigation of intrinsic connections between the neuronal layers of the modelled sources. In my presentation I will give an overview of DCM methods for EEG, including their implementation in SPM. I will also present the most recent methodological developments, as well as discuss the applications of DCM/EEG to studies on conscious perception.
Brain–computer interfaces and future industrial application of cognitive neuroscience
In this talk different applications of the brain computer interfaces (BCI) technology will be first presented. It will be described how by using the voluntary modulation of EEG activity normal subjects could control external devices such as a cursor on the screen, a mobile robot as well as a wheelchair. Successively, it will be illustrated how the BCI technology could be inserted within the rehabilitation path of the patients suffered of brain strokes. In particular, it will be showed how BCI technologies could enhance the rehabilitation exercise, by including the presentation of the attempt of the movement as early as possible to the patients, although they were not yet able to move their limbs. Successively, it will be showed different applications of the collection of brain activity in working contexts related to the airplanes pilots. It will be described as it is possible to detect the brain activity related to the insurgence of mental workload. It will be speculated that such detection could be employed in a short future to generate devices able to warn the operators about their perceived workload. Example of such detection of mental workload will be presented in three different conditions: on civil airline pilots, on military pilots and on car drivers. The possibility to detect in a reliable way the cerebral activity during “real-life” conditions and the possibility to detect brain activity with dry electrodes will be also discussed.
Using psychophysical modeling to study the interaction of spontaneous EEG oscillations and perception
The brain is never at rest – internally-generated, spontaneous neuronal activity is ever-present even in the absence of external stimulation. In EEG, spontaneous neuronal activity is visible as oscillations in different frequency bands. A longstanding research question concerns the interaction between the state of spontaneous oscillatory activity at the time of stimulus presentation and the subsequent perception of this stimulus. Specifically, oscillations in the alpha frequency-band (8-12 Hz) have been shown to interfere with processing of visual information. This interference can have a facilitatory effect, in that it suppresses perception of distracting stimuli in attention or memory tasks, yet it can also directly dampen performance on perceptual tasks. Currently, we have only a limited understanding of the nature of this interference: we know that prestimulus alpha oscillations interfere with perception, but we do not know how. While this question can be addressed at multiple levels, I am mostly interested in clarifying the functional level, namely how prestimulus alpha oscillations modulate the transformation of stimulus input into sensations and behavioral output. In my presentation, I will argue that the functional mechanisms of prestimulus oscillations can be revealed by psychophysical modeling of their effects on perception and performance. A key proposition in the modeling approach is that functional mechanisms are theoretical entities that cannot be observed directly in empirical data (e.g. in hit rates or microvolts). Rather, they can be identified by modeling empirical data with models that explicate the relationship between mechanisms, stimulus conditions and behavior. I will then present experiments, in which we used psychophysical modeling to reveal how prestimulus alpha oscillations modulate the gain of the visual system.
Estimating the cortical activity from EEG recordings: what can be learnt from functional imaging?
Electroencephalography (EEG) has an excellent temporal resolution but the estimation of the cortical activity that generates the voltages recorded by the electrodes is a difficult inverse problem, mainly because of the dispersion of the electric field caused by the skull. In this talk, I will present different techniques that use anatomical and functional information of each individual subject to constrain the solution of this inverse problem. In particular, I will show how a segmentation of the cortex into meaningful areas from fMRI recordings can improve the spatial resolution of all the classical source reconstruction approaches. This point will be supported by both simulated and real data recorded during experiments on the visual system. Finally, I will show that at the group level, these techniques can exploit the inter-subject variability to provide improved estimation performance.
Recording spikes … non-invasively
The capability to detect spikes defines a striking contrast between invasive (microscopic) and non-invasive (macroscopic) recordings. While the latter are dominated by summed postsynaptic potentials reflecting neuronal input, invasive electrodes can provide direct access also to the very output of neuronal computation – spikes. This micro/macro gap, however, has been narrowed gradually over the last years by combining special physiological paradigms with neuro-technological advances. To convey the gist of this research agenda on high-frequency spike-related EEG/MEG the tripartite lecture will (i) address the basic neurophysics of near-field and far-field signals distinguishing slow from fast neuronal activities, (ii) elaborate on high-frequency (~600 Hz) somatosensory evoked burst responses serving as ‘workhorse’ to establish the feasibility of non-invasive spike-related recordings also in non-specialised labs, and (iii) report on recent neuro-technological progress providing a unique opportunity for high-resolution scalp mappings of EEG activities even above 1 kHz which reflect non-invasive correlates of human neocortical population spike responses. Critically, if recording conditions provide optimal SNR, event detection of even single-trial (i.e., unaveraged) 600 Hz bursts becomes feasible, offering a perspective to study the immediate impact of cognitive processes on the generation of human cortical population spikes.
Single-trial topographic analysis and its application in coma research
Cognitive and clinical research has benefit from a rising interest in the application of multivariate techniques for decoding brain states as measured by various neuroimaging techniques, including electroencephalography (EEG). Because decoding algorithms can typically take advantage of high-dimensional data such as EEG responses as measured at multiple electrode sites, they offer a valid alternative to the a priori selection of specific electrodes prior to statistics, a practice which can severely limit the possibility to reveal robust effects outside a region of interest and prevent the investigation of spatially distributed effects. In this talk I will outline the main steps of a multivariate decoding analysis based on voltage topographies measured at multiple electrode sites. The method is based on modelling the voltage topographies recorded at single-trial EEG level as a mixture of Gaussians. The posterior probabilities of each of the Gaussian in the mixture are used for evaluating time point-by-time point the intervals and the voltage topographies which best discriminate between experimental conditions. This information is then used for classifying new test trials as belonging to each condition of interest. This single-trial topographic analysis (“stta”) has turned out to be particularly robust in discovering spatio-temporal pattern that can discriminate EEG responses to sensory stimuli even in very noisy conditions or when the EEG response does not follow stereotypical components such as in the case of clinical data. In this context I will present recent results obtained by applying this analysis to single-trial EEG responses to auditory stimuli in comatose patients and how these results relate to patients’outcome.
Simultaneous eye-tracking and EEG: A tool to study active vision
To avoid measurement artifacts, event-related EEG data is typically recorded while subjects maintain a steady fixation. While this procedure has many advantages, it also differs in fundamental ways from everyday visual perception, which involves an active sampling of the environment with several saccadic eye movements per second. An alternative approach to EEG analysis, summarized in the present talk, is to align the EEG signal not to passive stimulations, but to the on- and offsets of eye movements under natural viewing conditions, yielding saccade- and fixation-related brain potentials (SRPs/FRPs). This technique requires simultaneous high-resolution eye tracking and the handling of technical and data-analytical problems. To exemplify that it is nevertheless a useful addition to standard EEG methodology, I will present data from two lines of research. In the first line, we have investigated how visual-cortical potentials from involuntary fixational eye movements influences EEG data even in traditional paradigms that require fixation. In a second line of research, the technique was used to compare the time course of cognition under traditional passive stimulation conditions to that observed during free viewing tasks such as reading or scene perception. I will discuss potential applications and limits for combining eye-tracking and EEG in natural vision.
Imaging of brain dynamics underlying Natural Cognition
Imaging of brain dynamics underlying human cognitive processing restricts active and natural behavior of participants to avoid artifacts contaminating the recording. However, human cognition is based on and closely tight to active behavior in an ever-changing environment. Cognition thus is embodied in the sense that cognitive processes are based on and make use of our physical structure while being situated in a specific environment. Brain areas and activities that originally evolved to organize motor behavior of animals in their three-dimensional (3-D) environments also support human cognition (Rizzolatti, Fogassi, & Gallese, 2002), suggesting that joint imaging of human brain activity and motor behavior could be an invaluable resource for understanding the distributed brain dynamics of human cognition. However, due to technical constraints of traditional brain imaging methods there is a lack of studies investigating the brain dynamics underlying actively behaving subjects. This imposes a fundamental mismatch between the bandwidth of recorded brain dynamics (now up to 106 bits/second or more) and allowed behavior (typically, minimal button presses at ~1/second). To better understand the embodied aspect of human cognition, we have developed a mobile brain/body imaging (MoBI) modality to allow for synchronous recording of EEG, eye movement and body movements as subjects actively perform natural movements in 3-D environments (Makeig et al., 2009). Simultaneous recording of whole-body movements and brain dynamics during free and naturally motivated 3-D orienting actions, combined with data-driven analysis of brain dynamics, allows, for the first time, studies of distributed EEG dynamics, body movements, and eye, head and neck muscle activities during spatial cognition in situ. The new mobile brain/body imaging approach allows analysis of joint brain and body dynamics supporting and expressing natural cognition, including self-guided search for and processing of relevant information and motivated behavior in realistic environments.
Cross-frequency coupling between theta- and gamma-band oscillations in the human EEG: Methodological challenges and functional implications
Numerous studies suggest that synchronized neuronal activity in the gamma-band range (>30Hz) serve the activation and encoding of neuronal object representations. Additionally, various experiments show that theta-band oscillations play a crucial role during mnemonic processing.
In order to examine the relationship between oscillations in the theta- and gamma-band, we recorded high-density EEG during the encoding of new items into memory.
Methodological implications: (1) Cortical gamma-band oscillations might be overshadowed by an electromyogenic artifacts caused by miniature saccades. An ICA-based solution to deal with this problem will be offered (correction of saccade-related transient potentials - COSTRAP). (2) A method to quantify the interplay between gamma- and theta-band oscillations will be presented. In particular, it will be shown that the relationship between theta-band phase information and gamma-band amplitudes can be quantified by means of the so-called Modulation Index (MI).
Functional implications: Successful compared to unsuccessful encoding into memory was reflected in increased MI indices (i.e. frontal theta-phase information was stronger coupled to posterior gamma-band amplitudes in response to stimuli which participants later on remember versus items which participants subsequently forgot). These findings support the idea that during the formation of new memories executive frontal brain areas interact with cortical object representations in posterior regions.
Measuring cortical interactions with surface EEG, MEG and intracranial EEG
In this talk, I will briefly introduce the neuroscientific rationale for investigating brain connectivity as a key mechanism for healthy brain function. I will then provide an overview of various connectivity measures that can be used with electroencephalogrpahy (EEG), Magnetoencephalography (MEG) and intracranial EEG. The pitfalls and limitations of some of the methods will be addressed and some reccomendations and rules of good conduct will be provided. I will then describe a few illustrative examples of both invasive and non-invasive studies in which we assessed long-range cortico-cortical interactions both during goal-directed behavior and during the resting-state. Beyond unraveling the mysteries of the brainweb, analysis of large-scale network interactions in the brain will also be key to tackling a number of brain disorders. Clinical perspectives and future paths for bridging the gap between electropysiological and neuroimaging studies of brain connectivity will also be discussed.
Statistical testing in electrophysiological studies
In my talk I will briefly outline the rationale of four different approaches to the statistical testing of electrophysiological data: (1) the Neyman-Pearson approach, (2) the permutation-based approach, (3), the bootstrap-based approach, and (4) the Bayesian approach. I will describe in a bit more detail the mechanics of the Neyman-Pearson and the permutation-based approach.
These approaches are evaluated from the perspective of electrophysiological studies, which involve multivariate (i.e., spatiotemporal) observations in which source-level signals are picked up to a certain extent by all sensors. Besides formal statistical techniques, there are also techniques that do not involve probability calculations but are very useful in dealing with multivariate data (i.e., verification of data-based predictions, crossvalidation, and localizers). Moreover, data-based decision making can also be informed by mechanistic evidence that is provided by the structure in the data.
The Virtual Brain: Knowledge Inference and Application
Brain activity is characterized by metastable and itinerant dynamics yielding continuous reconfigurations of its space-time structure. Brain imaging methods such as fMRI and EEG derive detailed maps of functional network reconfigurations and link them to cognitive states. However, for some tasks the achieved level of detail is not sufficient, e.g. to infer precise cognitive states from imaging data, to control brain states or to build devices with brain-like capabilities. I will talk about full brain models that enable simulating brain activity at imaging resolution. By tuning the models on real functional imaging data, we reveal the behavior of control parameters and hidden states that are causal to the physiological dynamic repertoire and temporal trajectories of brain states. I will discuss the potential impact of this new development for new clinical and technical applications.
Tracking the Dynamics of Information Flow from EEG/MEG data
Reverse correlation is a powerful tool to extract the fine-grained information of a stimulus underlying its categorization. For example, the wide-opened eyes of a of face suggests that it is fearful. Likewise, reverse correlation is now applied to the voxels of a modelled brain to understand where, when and how different regions of the brain process the input information (e.g. the wide-opened eyes) that enables adapted behavior (e.g. "avoid"). But the quality of reverse correlation data critically depends on (a) the dimensions of the stimulus that are tested and (b) the behavioral and brain responses that are measured. Here, I will present recent developments of both (a) and (b) that enable direct comparisons of the information carried by different dimensions of the stimulus and comparisons of the information carried by different brain responses.
GLM-based single-trial modeling of EEG/MEG data
Traditional EEG/MEG analyses presume the averaging of data epochs over a substantial number of repetitions per condition, posing limitations on the complexity of experimental designs and on the range of phenomena being analyzed. I will present applications of more flexible analysis approaches to multidimensional data, in the framework of the general linear model (GLM). The GLM accommodates traditional factorial designs, but also parametric modulations and complex interactions, within a single statistical model suited for time-, frequency-, and source space. In particular, it will be shown how GLMs can be implemented on the single-trial level, circumventing data averaging and increasing experimenters’ freedom in designing EEG/MEG studies. In addition, I will consider a novel approach to analysis of time-frequency data based on convolution modeling. By this method, not only the aggregation, but also the epoching, and even artifact treatment of continuous recordings are passed on to the parameters of a GLM, enabling the analysis of temporally variable and overlapping events. Description of each technique will include application examples and practical notes on their implementation using SPM for M/EEG.
Filtering of electrophysiological data
Filtering is a ubiquitous step in the (pre-)processing of electrophysiological data. Besides the intended effect of the attenuation of signal components considered as noise, filtering can also result in various unintended adverse filter effects and filter artifacts such as smoothing (see e.g., Luck, 2005; van Rullen, 2011). The presentation aims at giving a practical guideline for the selection of filter types and filter parameters to optimize signal-to-noise ratio and avoid or reduce signal distortions for selected electrophysiological applications. Finite (windowed sinc) and infinite impulse response (Butterworth) filters and their parameters (high-/low-/bandpass/notch, cutoff, order, causality, window type) will be introduced and discussed. Resulting filter responses will be evaluated. In practical examples it will be demonstrated how to recognize and avoid common adverse filter effects and filter artifacts with optimized filter settings when filtering complex signals, in particular smoothing and ringing for low-pass filters and artificial peaks and oscillations for high-pass filters. Recommendations for the reporting, of filter settings, limitations and alternatives to filtering will be discussed.
Beyond event-related averaging: parametric model-based regression of EEG signals during perceptual decision-making
Behavioral and neuroimaging studies usually consider variability in neural and behavioral responses as a nuisance term that can be eliminated by presenting the same stimulus over and over again, and by averaging the resulting neural responses and behavior. However, recent studies have uncovered covariations between these two sources of variability, suggesting that they reflect variability in common underlying computations. In this talk, I will show how relating variability in stimulus, brain and behavior can be used to identify the computations underlying choice during perceptual decision-making. I will focus on a perceptual categorization task which requires the accumulation of multiple samples of sensory evidence delivered in rapid succession. Instead of computing event-related averages, I propose to study the neural correlates of evidence accumulation by regressing single-trial EEG signals against decision-theoretic quantities indexing the moment-to-moment information provided by each evidence sample on each trial. This neural 'encoding' approach has uncovered a cascade of overlapping neural events whereby successive samples of sensory evidence are first processed from lower to higher levels, and then integrated into an evolving motor plan (Wyart et al., 2012). I will then describe how the EEG residuals from this parametric regression can be related to variability in choice via psychophysiological interaction (PPI). This neural 'decoding' approach has revealed a clear neural distinction between two decision-theoretic computations: 1) a multiplicative weighting of each evidence sample according to the phase of low-frequency parietal signals, 2) an additive integration of the weighted evidence reflected in high-frequency motor preparation signals. I will conclude by showing how this regression-based framework is sufficiently general to be applied to other types of human decisions (e.g., economic, social) and task designs.
Wyart V, de Gardelle V, Scholl J, Summerfield C (2012) Rhythmic fluctuations in evidence accumulation during decision making in the human brain. Neuron 76, 847-858.