Search

We begin by reviewing the statistical framework of information theory as applicable to neuroimaging data analysis. A major factor hindering wider adoption of this framework in neuroimaging is the difficulty of estimating information theoretic quantities in practice. We present a novel estimation technique that combines the statistical theory of copulas with the closed form solution for the entropy of Gaussian variables. This results in a general, computationally efficient, flexible, and robust multivariate statistical framework that provides effect sizes on a common meaningful scale, allows for unified treatment of discrete, continuous, unidimensional and multidimensional variables, and enables direct comparisons of representations from behavioral and brain responses across any recording modality. We validate the use of this estimate as a statistical test within a neuroimaging context, considering both discrete stimulus classes and continuous stimulus features. We also present examples of analyses facilitated by these developments, including application of multivariate analyses to MEG planar magnetic field gradients, and pairwise temporal interactions in evoked EEG responses. We show the benefit of considering the instantaneous temporal derivative together with the raw values of M/EEG signals as a multivariate response, how we can separately quantify modulations of amplitude and direction for vector quantities, and how we can measure the emergence of novel information over time in evoked responses. Open-source Matlab and Python code implementing the new methods accompanies this article. Hum Brain Mapp 38:1541-1573, 2017. © 2016 Wiley Periodicals, Inc., (2016 The Authors Human Brain Mapping Published by Wiley Periodicals, Inc.)

Population coding is the quantitative study of which algorithms or representations are used by the brain to combine together and evaluate the messages carried by different neurons. Here, we review an information-theoretic approach to population coding. We first discuss how to compute the information carried by simultaneously recorded neural populations, and in particular how to reduce the limited sampling bias which affects the calculation of information from a limited amount of experimental data. We then discuss how to quantify the contribution of individual members of the population, or the interaction between them, to the overall information encoded by the considered group of neurons. We focus in particular on evaluating what is the contribution of interactions up to any given order to the total information. We illustrate this formalism with applications to simulated data with realistic neuronal statistics and to real simultaneous recordings of multiple spike trains., (Copyright (c) 2010 Elsevier Ltd. All rights reserved.)

In the search for the neural basis of conscious experience, perception and the cognitive processes associated with reporting perception are typically confounded as neural activity is recorded while participants explicitly report what they experience. Here, we present a novel way to disentangle perception from report using eye movement analysis techniques based on convolutional neural networks and neurodynamical analyses based on information theory. We use a bistable visual stimulus that instantiates two well-known properties of conscious perception: integration and differentiation. At any given moment, observers either perceive the stimulus as one integrated unitary object or as two differentiated objects that are clearly distinct from each other. Using electroencephalography, we show that measures of integration and differentiation based on information theory closely follow participants' perceptual experience of those contents when switches were reported. We observed increased information integration between anterior to posterior electrodes (front to back) prior to a switch to the integrated percept, and higher information differentiation of anterior signals leading up to reporting the differentiated percept. Crucially, information integration was closely linked to perception and even observed in a no-report condition when perceptual transitions were inferred from eye movements alone. In contrast, the link between neural differentiation and perception was observed solely in the active report condition. Our results, therefore, suggest that perception and the processes associated with report require distinct amounts of anterior–posterior network communication and anterior information differentiation. While front-to-back directed information is associated with changes in the content of perception when viewing bistable visual stimuli, regardless of report, frontal information differentiation was absent in the no-report condition and therefore is not directly linked to perception per se. [ABSTRACT FROM AUTHOR]

Information processing within neural systems often depends upon selective amplification of relevant signals and suppression of irrelevant signals. This has been shown many times by studies of contextual effects but there is as yet no consensus on how to interpret such studies. Some researchers interpret the effects of context as contributing to the selective receptive field (RF) input about which neurons transmit information. Others interpret context effects as affecting transmission of information about RF input without becoming part of the RF information transmitted. Here we use partial information decomposition (PID) and entropic information decomposition (EID) to study the properties of a form of modulation previously used in neurobiologically plausible neural nets. PID shows that this form of modulation can affect transmission of information in the RF input without the binary output transmitting any information unique to the modulator. EID produces similar decompositions, except that information unique to the modulator and the mechanistic shared component can be negative when modulating and modulated signals are correlated. Synergistic and source shared components were never negative in the conditions studied. Thus, both PID and EID show that modulatory inputs to a local processor can affect the transmission of information from other inputs. Contrary to what was previously assumed, this transmission can occur without the modulatory inputs becoming part of the information transmitted, as shown by the use of PID with the model we consider. Decompositions of psychophysical data from a visual contrast detection task with surrounding context suggest that a similar form of modulation may also occur in real neural systems. [ABSTRACT FROM AUTHOR]

The entrainment of slow rhythmic auditory cortical activity to the temporal regularities in speech is considered to be a central mechanism underlying auditory perception. Previous work has shown that entrainment is reduced when the quality of the acoustic input is degraded, but has also linked rhythmic activity at similar time scales to the encoding of temporal expectations. To understand these bottom-up and top-down contributions to rhythmic entrainment, we manipulated the temporal predictive structure of speech by parametrically altering the distribution of pauses between syllables or words, thereby rendering the local speech rate irregular while preserving intelligibility and the envelope fluctuations of the acoustic signal. Recording EEG activity in human participants, we found that this manipulation did not alter neural processes reflecting the encoding of individual sound transients, such as evoked potentials. However, the manipulation significantly reduced the fidelity of auditory delta (but not theta) band entrainment to the speech envelope. It also reduced left frontal alpha power and this alpha reduction was predictive of the reduced delta entrainment across participants. Our results show that rhythmic auditory entrainment in delta and theta bands reflect functionally distinct processes. Furthermore, they reveal that delta entrainment is under top-down control and likely reflects prefrontal processes that are sensitive to acoustical regularities rather than the bottom-up encoding of acoustic features. [ABSTRACT FROM AUTHOR]

The rodent whisker-associated thalamic nucleus (VPM) contains a somatotopic map where whisker representation is divided into distinct neuronal sub-populations, called "barreloids". Each barreloid projects to its associated cortical barrel column and soforms a gateway for incoming sensory stimuli to the barrel cortex. We aimed to determine how the population of neurons within one barreloid encodes naturalistic whisker motion. In rats, we recorded the extracellular activity of up to nine single neurons within a single barreloid, by implanting silicon probes parallel to the longitudinal axis of the barreloids. We found that play-back of texture-induced whisker motion evoked sparse responses, timed with millisecond precision. At the population level, there was synchronous activity: however, different subsets of neurons were synchronously active at different times. Mutual information between population responses and whisker motion increased near linearly with population size. When normalized to factor out firing rate differences, we found that texture was encoded with greater informational-efficiency than white noise. These results indicate that, within each VPM barreloid, there is a rich and efficient population code for naturalistic whisker motion based on precisely timed, population spike patterns. [ABSTRACT FROM AUTHOR]

The precise timing of action potentials of sensory neurons relative to the time of stimulus presentation carries substantial sensory information that is lost or degraded when these responses are summed over longer time windows. However, it is unclear whether and how downstream networks can access information in precise time-varying neural responses. Here, we review approaches to test the hypothesis that the activity of neural populations provides the temporal reference frames needed to decode temporal spike patterns. These approaches are based on comparing the single-trial stimulus discriminability obtained from neural codes defined with respect to network- intrinsic reference frames to the discriminability obtained from codes defined relative to the experimenter's computer clock. Application of this formalismto auditory, visual and somatosensory data shows that information carried by millisecond-scale spike times can be decoded robustly even with little or no independent external knowledge of stimulus time. In cortex, key components of such intrinsic temporal reference frames include dedicated neural populations that signal stimulus onset with reliable and precise latencies, and low-frequency oscillations that can serve as reference for partitioning extended neuronal responses into informative spike patterns. [ABSTRACT FROM AUTHOR]

When we listen to speech, we have to make sense of a waveform of sound pressure. Hierarchical models of speech perception assume that, to extract semantic meaning, the signal is transformed into unknown, intermediate neuronal representations. Traditionally, studies of such intermediate representations are guided by linguistically defined concepts, such as phonemes. Here, we argue that in order to arrive at an unbiased understanding of the neuronal responses to speech, we should focus instead on representations obtained directly from the stimulus. We illustrate our view with a data-driven, information theoretic analysis of a dataset of 24 young, healthy humans who listened to a 1 h narrative while their magnetoencephalogram (MEG) was recorded. We find that two recent results, the improved performance of an encoding model in which annotated linguistic and acoustic features were combined and the decoding of phoneme subgroups from phoneme-locked responses, can be explained by an encoding model that is based entirely on acoustic features. These acoustic features capitalize on acoustic edges and outperform Gabor-filtered spectrograms, which can explicitly describe the spectrotemporal characteristics of individual phonemes. By replicating our results in publicly available electroencephalography (EEG) data, we conclude that models of brain responses based on linguistic features can serve as excellent benchmarks. However, we believe that in order to further our understanding of human cortical responses to speech, we should also explore low-level and parsimonious explanations for apparent high-level phenomena. • 24 participants listened to a 1 h story in MEG • Encoding and decoding analyses are evaluated and merged using information theory • Simple acoustic predictors of MEG responses can explain apparent high-level phenomena • Results are repeated in openly available EEG dataset Daube et al. use a data-driven information theoretic analysis of auditory cortex MEG responses to speech to demonstrate that complex models of such responses relying on annotated linguistic features can be explained more parsimoniously with simple models relying on the acoustics only. [ABSTRACT FROM AUTHOR]

Summary Over the past decade, extensive studies of the brain regions that support face, object, and scene recognition suggest that these regions have a hierarchically organized architecture that spans the occipital and temporal lobes [ 1–14 ], where visual categorizations unfold over the first 250 ms of processing [ 15–19 ]. This same architecture is flexibly involved in multiple tasks that require task-specific representations—e.g. categorizing the same object as "a car" or "a Porsche." While we partly understand where and when these categorizations happen in the occipito-ventral pathway, the next challenge is to unravel how these categorizations happen. That is, how does high-dimensional input collapse in the occipito-ventral pathway to become low dimensional representations that guide behavior? To address this, we investigated what information the brain processes in a visual perception task and visualized the dynamic representation of this information in brain activity. To do so, we developed stimulus information representation (SIR), an information theoretic framework, to tease apart stimulus information that supports behavior from that which does not. We then tracked the dynamic representations of both in magneto-encephalographic (MEG) activity. Using SIR, we demonstrate that a rapid (∼170 ms) reduction of behaviorally irrelevant information occurs in the occipital cortex and that representations of the information that supports distinct behaviors are constructed in the right fusiform gyrus (rFG). Our results thus highlight how SIR can be used to investigate the component processes of the brain by considering interactions between three variables (stimulus information, brain activity, behavior), rather than just two, as is the current norm. Highlights • We show how the brain reduces high-dimensional input into decision representations • Occipital cortex reduces features irrelevant for behavior ∼170 ms post stimulus • Past 170 ms, fusiform gyrus combines the features supporting behavioral decisions In a decision task, Zhan et al. visualize within a new information theoretic framework the dynamic representation of visual information in brain activity. They demonstrate rapid reduction of behaviorally irrelevant information in the occipital cortex and a combination of the features that supports distinct decisions in the right fusiform gyrus. [ABSTRACT FROM AUTHOR]

Human facial expressions are complex, multi-component signals that can communicate rich information about emotions, 1–5 including specific categories, such as "anger," and broader dimensions, such as "negative valence, high arousal." 6–8 An enduring question is how this complex signaling is achieved. Communication theory predicts that multi-component signals could transmit each type of emotion information—i.e., specific categories and broader dimensions—via the same or different facial signal components, with implications for elucidating the system and ontology of facial expression communication. 9 We addressed this question using a communication-systems-based method that agnostically generates facial expressions and uses the receiver's perceptions to model the specific facial signal components that represent emotion category and dimensional information to them. 10–12 First, we derived the facial expressions that elicit the perception of emotion categories (i.e., the six classic emotions 13 plus 19 complex emotions 3) and dimensions (i.e., valence and arousal) separately, in 60 individual participants. Comparison of these facial signals showed that they share subsets of components, suggesting that specific latent signals jointly represent—i.e., multiplex—categorical and dimensional information. Further examination revealed these specific latent signals and the joint information they represent. Our results—based on white Western participants, same-ethnicity face stimuli, and commonly used English emotion terms—show that facial expressions can jointly represent specific emotion categories and broad dimensions to perceivers via multiplexed facial signal components. Our results provide insights into the ontology and system of facial expression communication and a new information-theoretic framework that can characterize its complexities. [Display omitted] • Examined facial signals of broad-plus-specific emotion categories and dimensions • Used data-driven, perception-based modeling and information-theoretic analyses • Disentangled facial signals that multiplex broad-plus-specific emotion information • Provides insights into facial expression ontology and new methodological framework Liu et al. examine how facial expressions signal broad-plus-specific emotion category and dimensional information. Using a perception-based facial-signal-modeling technique and information-theoretic analyses, they find a latent set of facial signals that can multiplex categorical and dimensional information and a subset uniquely signaling either. [ABSTRACT FROM AUTHOR]