Auditory cortex modelled as a dynamical network of oscillators: understanding event-related fields and their adaptation.

Adaptation, the reduction of neuronal responses by repetitive stimulation, is a ubiquitous feature of auditory cortex (AC). It is not clear what causes adaptation, but short-term synaptic depression (STSD) is a potential candidate for the underlying mechanism. In such a case, adaptation can be directly linked with the way AC produces context-sensitive responses such as mismatch negativity and stimulus-specific adaptation observed on the single-unit level. We examined this hypothesis via a computational model based on AC anatomy, which includes serially connected core, belt, and parabelt areas. The model replicates the event-related field (ERF) of the magnetoencephalogram as well as ERF adaptation. The model dynamics are described by excitatory and inhibitory state variables of cell populations, with the excitatory connections modulated by STSD. We analysed the system dynamics by linearising the firing rates and solving the STSD equation using time-scale separation. This allows for characterisation of AC dynamics as a superposition of damped harmonic oscillators, so-called normal modes. We show that repetition suppression of the N1m is due to a mixture of causes, with stimulus repetition modifying both the amplitudes and the frequencies of the normal modes. In this view, adaptation results from a complete reorganisation of AC dynamics rather than a reduction of activity in discrete sources. Further, both the network structure and the balance between excitation and inhibition contribute significantly to the rate with which AC recovers from adaptation. This lifetime of adaptation is longer in the belt and parabelt than in the core area, despite the time constants of STSD being spatially homogeneous. Finally, we critically evaluate the use of a single exponential function to describe recovery from adaptation.

Absolute auditory threshold: testing the absolute.

The mechanisms underlying the detection of sounds in quiet, one of the simplest tasks for auditory systems, are debated. Several models proposed to explain the threshold for sounds in quiet and its dependence on sound parameters include a minimum sound intensity ('hard threshold'), below which sound has no effect on the ear. Also, many models are based on the assumption that threshold is mediated by integration of a neural response proportional to sound intensity. Here, we test these ideas. Using an adaptive forced choice procedure, we obtained thresholds of 95 normal-hearing human ears for 18 tones (3.125 kHz carrier) in quiet, each with a different temporal amplitude envelope. Grand-mean thresholds and standard deviations were well described by a probabilistic model according to which sensory events are generated by a Poisson point process with a low rate in the absence, and higher, time-varying rates in the presence, of stimulation. The subject actively evaluates the process and bases the decision on the number of events observed. The sound-driven rate of events is proportional to the temporal amplitude envelope of the bandpass-filtered sound raised to an exponent. We find no evidence for a hard threshold: When the model is extended to include such a threshold, the fit does not improve. Furthermore, we find an exponent of 3, consistent with our previous studies and further challenging models that are based on the assumption of the integration of a neural response that, at threshold sound levels, is directly proportional to sound amplitude or intensity.

Explaining event-related fields by a mechanistic model encapsulating the anatomical structure of auditory cortex.

Event-related fields of the magnetoencephalogram are triggered by sensory stimuli and appear as a series of waves extending hundreds of milliseconds after stimulus onset. They reflect the processing of the stimulus in cortex and have a highly subject-specific morphology. However, we still have an incomplete picture of how event-related fields are generated, what the various waves signify, and why they are so subject-specific. Here, we focus on this problem through the lens of a computational model which describes auditory cortex in terms of interconnected cortical columns as part of hierarchically placed fields of the core, belt, and parabelt areas. We develop an analytical approach arriving at solutions to the system dynamics in terms of normal modes: damped harmonic oscillators emerging out of the coupled excitation and inhibition in the system. Each normal mode is a global feature which depends on the anatomical structure of the entire auditory cortex. Further, normal modes are fundamental dynamical building blocks, in that the activity of each cortical column represents a combination of all normal modes. This approach allows us to replicate a typical auditory event-related response as a weighted sum of the single-column activities. Our work offers an alternative to the view that the event-related field arises out of spatially discrete, local generators. Rather, there is only a single generator process distributed over the entire network of the auditory cortex. We present predictions for testing to what degree subject-specificity is due to cross-subject variations in dynamical parameters rather than in the cortical surface morphology.

Simultaneous spatio-temporal matching pursuit decomposition of evoked brain responses in MEG.

We present a novel approach to the spatio-temporal decomposition of evoked brain responses in magnetoencephalography (MEG) aiming at a sparse representation of the underlying brain activity in terms of spatio-temporal atoms. Our approach is characterized by three attributes which constitute significant improvements with respect to existing approaches: (1) the spatial and temporal decomposition is addressed simultaneously rather than sequentially, with the benefit that source loci and corresponding waveforms can be unequivocally allocated to each other, and, hence, allow a plausible physiological interpretation of the parametrized data; (2) it is free from severe a priori assumptions about the solution space; (3) it comprises an optimization technique for the use of very large spatial and temporal subdirectories to greatly reduce the otherwise enormous computational cost by making use of the Cauchy-Schwarz inequality. We demonstrate the efficiency of the approach with simulations and real MEG data obtained from a subject exposed to a simple auditory stimulus.

Averaging auditory evoked magnetoencephalographic and electroencephalographic responses: a critical discussion.

In the analysis of data from magnetoencephalography (MEG) and electroencephalography (EEG), it is common practice to arithmetically average event-related magnetic fields (ERFs) or event-related electric potentials (ERPs) across single trials and subsequently across subjects to obtain the so-called grand mean. Comparisons of grand means, e.g. between conditions, are then often performed by subtraction. These operations, and their statistical evaluation with parametric tests such as ANOVA, tacitly rely on the assumption that the data follow the additive model, have a normal distribution, and have a homogeneous variance. This may be true for single trials, but these conditions are rarely met when ERFs/ERPs are compared between subjects, meaning that the additive model is seldom the correct model for computing grand mean waveforms. Here, we summarize some of our recent work and present new evidence, from auditory-evoked MEG and EEG results, that the non-normal distributions and the heteroscedasticity observed instead result because ERFs/ERPs follow a mixed model with additive and multiplicative components. For peak amplitudes, such as the auditory M100 and N100, the multiplicative component dominates. These findings emphasize that the common practice of simply subtracting arithmetic means of auditory-evoked ERFs or ERPs is problematic without prior adequate transformation of the data. Application of the area sinus hyperbolicus (asinh) transform to data following the mixed model transforms them into the requested additive model with its normal distribution and homogeneous variance. We therefore advise checking the data for compliance with the additive model and using the asinh transform if required.

Towards a unifying basis of auditory thresholds: Thresholds for multicomponent stimuli.

Sounds consisting of multiple simultaneous or consecutive components can be detected by listeners when the stimulus levels of the components are lower than those needed to detect the individual components alone. The mechanisms underlying such spectral, spectrotemporal, temporal, or across-ear integration are not completely understood. Here, we report threshold measurements from human subjects for multicomponent stimuli (tone complexes, tone sequences, diotic or dichotic tones) and for their individual sinusoidal components in quiet. We examine whether the data are compatible with the detection model developed by Heil, Matysiak, and Neubauer (HMN model) to account for temporal integration (Heil et al. 2017), and we compare its performance to that of the statistical summation model (Green 1958), the model commonly used to account for spectral and spectrotemporal integration. In addition, we compare the performance of both models with respect to previously published thresholds for sequences of identical tones and for diotic tones. The HMN model is similar to the statistical summation model but is based on the assumption that the decision variable is a number of sensory events generated by the components via independent Poisson point processes. The rate of events is low without stimulation and increases with stimulation. The increase is proportional to the time-varying amplitude envelope of the bandpass-filtered component(s) raised to an exponent of 3. For an ideal observer, the decision variable is the sum of the events from all channels carrying information, for as long as they carry information. We find that the HMN model provides a better account of the thresholds for multicomponent stimuli than the statistical summation model, and it offers a unifying account of spectral, spectrotemporal, temporal, and across-ear integration at threshold.

Reducing the Number of MEG/EEG Trials Needed for the Estimation of Brain Evoked Responses: A Bootstrap Approach.

OBJECTIVE: A common problem in magnetoencephalographic (MEG) and electroencephalographic (EEG) experimental paradigms relying on the estimation of brain evoked responses is the lengthy time of the experiment, which stems from the need to acquire a large number of repeated recordings. Using a bootstrap approach, we aim at reliably reducing the number of these repeated trials. METHODS: To this end, we assessed five variants of non-parametric bootstrapping based on the classical signal-plus-noise model constituting the foundation of signal averaging in MEG/EEG. We explain which of these approaches should and which should not be used for the aforementioned purpose, and why. RESULTS: We present results for two advocated bootstrap variants applied to auditory MEG data. The ensuing trial-averaged magnetic fields served as input to the estimation of cortical source generators, with spatio-temporal matching pursuit as an example of an inverse solution technique. We propose, for a wide range of trial numbers, a general framework to evaluate the statistical properties of the parameter estimates for source locations and related time courses. CONCLUSION: The proposed bootstrap framework offers a systematic approach to reduce the number of trials required to estimate the evoked response. The general validity of our findings is neither bound to any particular type of MEG/EEG data nor to any specific source localization method. SIGNIFICANCE: Practical implications of this work relate to the optimization of acquisition time of MEG/EEG experiments, thus reducing stress for the subjects (especially for patients) and minimizing related artifacts.

Why do humans have unique auditory event-related fields? Evidence from computational modeling and MEG experiments.

Auditory event-related fields (ERFs) measured with magnetoencephalography (MEG) are useful for studying the neuronal underpinnings of auditory cognition in human cortex. They have a highly subject-specific morphology, albeit certain characteristic deflections (e.g., P1m, N1m, and P2m) can be identified in most subjects. Here, we explore the reason for this subject-specificity through a combination of MEG measurements and computational modeling of auditory cortex. We test whether ERF subject-specificity can predominantly be explained in terms of each subject having an individual cortical gross anatomy, which modulates the MEG signal, or whether individual cortical dynamics is also at play. To our knowledge, this is the first time that tools to address this question are being presented. The effects of anatomical and dynamical variation on the MEG signal is simulated in a model describing the core-belt-parabelt structure of the auditory cortex, and with the dynamics based on the leaky-integrator neuron model. The experimental and simulated ERFs are characterized in terms of the N1m amplitude, latency, and width. Also, we examine the waveform grand-averaged across subjects, and the standard deviation of this grand average. The results show that the intersubject variability of the ERF arises out of both the anatomy and the dynamics of auditory cortex being specific to each subject. Moreover, our results suggest that the latency variation of the N1m is largely related to subject-specific dynamics. The findings are discussed in terms of how learning, plasticity, and sound detection are reflected in the auditory ERFs. The notion of the grand-averaged ERF is critically evaluated.

A probabilistic Poisson-based model accounts for an extensive set of absolute auditory threshold measurements.

Thresholds for detecting sounds in quiet decrease with increasing sound duration in every species studied. The neural mechanisms underlying this trade-off, often referred to as temporal integration, are not fully understood. Here, we probe the human auditory system with a large set of tone stimuli differing in duration, shape of the temporal amplitude envelope, duration of silent gaps between bursts, and frequency. Duration was varied by varying the plateau duration of plateau-burst (PB) stimuli, the duration of the onsets and offsets of onset-offset (OO) stimuli, and the number of identical bursts of multiple-burst (MB) stimuli. Absolute thresholds for a large number of ears (>230) were measured using a 3-interval-3-alternative forced choice (3I-3AFC) procedure. Thresholds decreased with increasing sound duration in a manner that depended on the temporal envelope. Most commonly, thresholds for MB stimuli were highest followed by thresholds for OO and PB stimuli of corresponding durations. Differences in the thresholds for MB and OO stimuli and in the thresholds for MB and PB stimuli, however, varied widely across ears, were negative in some ears, and were tightly correlated. We show that the variation and correlation of MB-OO and MB-PB threshold differences are linked to threshold microstructure, which affects the relative detectability of the sidebands of the MB stimuli and affects estimates of the bandwidth of auditory filters. We also found that thresholds for MB stimuli increased with increasing duration of the silent gaps between bursts. We propose a new model and show that it accurately accounts for our results and does so considerably better than a leaky-integrator-of-intensity model and a probabilistic model proposed by others. Our model is based on the assumption that sensory events are generated by a Poisson point process with a low rate in the absence of stimulation and higher, time-varying rates in the presence of stimulation. A subject in a 3I-3AFC task is assumed to choose the interval in which the greatest number of events occurred or randomly chooses among intervals which are tied for the greatest number of events. The subject is further assumed to count events over the duration of an evaluation interval that has the same timing and duration as the expected stimulus. The increase in the rate of the events caused by stimulation is proportional to the time-varying amplitude envelope of the bandpass-filtered signal raised to an exponent. We find the exponent to be about 3, consistent with our previous studies. This challenges models that are based on the assumption of the integration of a neural response that is directly proportional to the stimulus amplitude or proportional to its square (i.e., proportional to the stimulus intensity or power).

Persistent neural activity in auditory cortex is related to auditory working memory in humans and nonhuman primates.

Working memory is the cognitive capacity of short-term storage of information for goal-directed behaviors. Where and how this capacity is implemented in the brain are unresolved questions. We show that auditory cortex stores information by persistent changes of neural activity. We separated activity related to working memory from activity related to other mental processes by having humans and monkeys perform different tasks with varying working memory demands on the same sound sequences. Working memory was reflected in the spiking activity of individual neurons in auditory cortex and in the activity of neuronal populations, that is, in local field potentials and magnetic fields. Our results provide direct support for the idea that temporary storage of information recruits the same brain areas that also process the information. Because similar activity was observed in the two species, the cellular bases of some auditory working memory processes in humans can be studied in monkeys.

Multichannel matching pursuit and EEG inverse solutions.

We present a new approach to the preprocessing of the electroencephalographic time series for EEG inverse solutions. As the first step, EEG recordings are decomposed by multichannel matching pursuit algorithm--in this study we introduce a computationally efficient, suboptimal solution. Then, based upon the parameters of the waveforms fitted to the EEG (frequency, amplitude and duration), we choose those corresponding to the the phenomena of interest, like e.g. sleep spindles. For each structure, the corresponding weights of each channel define a topographic signature, which can be subject to an inverse solution procedure, like e.g. Loreta, used in this work. As an example, we present an automatic detection and parameterization of sleep spindles, appearing in overnight polysomnographic recordings. Inverse solutions obtained for single sleep spindles are coherent with the averages obtained for 20 overnight EEG recordings analyzed in this study, as well as with the results reported previously in literature as inter-subject averages of solutions for spectral integrals, computed on visually selected spindles.

Time-frequency-space localization of epileptic EEG oscillations.

This paper presents a hybrid method for localization of oscillatory EEG activity. It consists of two steps: multichannel matching pursuit with complex Gabor dictionary, and LORETA inverse solution. Proposed algorithm was successfully applied to the localization of epileptogenic EEG in a single patient.

Variance stabilization for computing and comparing grand mean waveforms in MEG and EEG.

Grand means of time-varying signals (waveforms) across subjects in magnetoencephalography (MEG) and electroencephalography (EEG) are commonly computed as arithmetic averages and compared between conditions, for example, by subtraction. However, the prerequisite for these operations, homogeneity of the variance of the waveforms in time, and for most common parametric statistical tests also between conditions, is rarely met. We suggest that the heteroscedasticity observed instead results because waveforms may differ by factors and additive terms and follow a mixed model. We propose to apply the asinh-transformation to stabilize the variance in such cases. We demonstrate the homogeneous variance and the normal distributions of data achieved by this transformation using simulated waveforms, and we apply it to real MEG data and show its benefits. The asinh-transformation is thus an essential and useful processing step prior to computing and comparing grand mean waveforms in MEG and EEG.

Single-trial evoked brain responses modeled by multivariate matching pursuit.

We present a new approach to the analysis of brain evoked electromagnetic potentials and fields. Multivariate version of the matching pursuit algorithm (MMP) performs an iterative, exhaustive search for waveforms, which optimally fit to signal structures, persistent in all the responses (trials) with the same time of occurrence, frequency, phase, and time width, but varying amplitude. The search is performed in a highly redundant time--frequency dictionary of Gabor functions, i.e., sines modulated by Gaussians. We present the feasibility of such a single-trial MMP analysis of the auditory M100 response, using an illustrative dataset acquired in a magnetoencephalographic (MEG) measurement with auditory stimulation with sinusoidal 1-kHz tones. We find that the morphology of the M100 estimate obtained from simple averaging of single trials can be very well explained by the average reconstruction with a few Gabor functions that parametrize those single trials. The M100 peak amplitude of single-trial reconstructions is observed to decrease with repetitions, which indicates habituation to the stimulus. This finding suggests that certain waveforms fitted by MMP could possibly be related to physiologically distinct components of evoked magnetic fields, which would allow tracing their dynamics on a single-trial level.