Search

Abstract Background The locked-in syndrome (LIS), due to a lesion in the pons, impedes communication. This situation can also be met after some severe brain injury or in advanced Amyotrophic Lateral Sclerosis (ALS). In the most severe condition, the persons cannot communicate at all because of a complete oculomotor paralysis (Complete LIS or CLIS). This even prevents the detection of consciousness. Some studies suggest that auditory brain–computer interface (BCI) could restore a communication through a « yes–no» code. Methods We developed an auditory EEG-based interface which makes use of voluntary modulations of attention, to restore a yes–no communication code in non-responding persons. This binary BCI uses repeated speech sounds (alternating “yes” on the right ear and “no” on the left ear) corresponding to either frequent (short) or rare (long) stimuli. Users are instructed to pay attention to the relevant stimuli only. We tested this BCI with 18 healthy subjects, and 7 people with severe motor disability (3 “classical” persons with locked-in syndrome and 4 persons with ALS). Results We report online BCI performance and offline event-related potential analysis. On average in healthy subjects, online BCI accuracy reached 86% based on 50 questions. Only one out of 18 subjects could not perform above chance level. Ten subjects had an accuracy above 90%. However, most patients could not produce online performance above chance level, except for two people with ALS who obtained 100% accuracy. We report individual event-related potentials and their modulation by attention. In addition to the classical P3b, we observed a signature of sustained attention on responses to frequent sounds, but in healthy subjects and patients with good BCI control only. Conclusions Auditory BCI can be very well controlled by healthy subjects, but it is not a guarantee that it can be readily used by the target population of persons in LIS or CLIS. A conclusion that is supported by a few previous findings in BCI and should now trigger research to assess the reasons of such a gap in order to propose new and efficient solutions. Clinical trial registrations: No. NCT02567201 (2015) and NCT03233282 (2013).

Abstract Predictive coding theories suggest that core symptoms in autism spectrum disorders (ASD) may stem from atypical mechanisms of perceptual inference (i.e., inferring the hidden causes of sensations). Specifically, there would be an imbalance in the precision or weight ascribed to sensory inputs relative to prior expectations. Using three tactile behavioral tasks and computational modeling, we specifically targeted the implicit dynamics of sensory adaptation and perceptual learning in ASD. Participants were neurotypical and autistic adults without intellectual disability. In Experiment I, tactile detection thresholds and adaptation effects were measured to assess sensory precision. Experiments II and III relied on two-alternative forced choice tasks designed to elicit a time-order effect, where prior knowledge biases perceptual decisions. Our results suggest a subtler explanation than a simple imbalance in the prior/sensory weights, having to do with the dynamic nature of perception, that is the adjustment of precision weights to context. Compared to neurotypicals, autistic adults showed no difference in average performance and sensory sensitivity. Both groups managed to implicitly learn and adjust a prior that biased their perception. However, depending on the context, autistic participants showed no, normal or slower adaptation, a phenomenon that computational modeling of trial-to-trial responses helped us to associate with a higher expectation for sameness in ASD, and to dissociate from another observed robust difference in terms of response bias. These results point to atypical perceptual learning rather than altered perceptual inference per se, calling for further empirical and computational studies to refine the current predictive coding theories of ASD.

V̇O2max is recognized as a key measure in exercise physiology and sports medicine. However, only 20–50% of maximal incremental exercise tests (IET) result in a plateau of V̇O2 (V̇O2pl). To our knowledge, no study has yet examined the possible difference in brain activity during an IET, in V̇O2pl and non-plateau athletes with the same V̇O2max and age. This study aimed to shed light on the central governor hypothesis, namely that the inability to reach a V̇O2pl may be dictated by the brain rather than by a peripheral physical limit. This hypothesis can now be explored using electroencephalography (EEG) during IET, measuring concomitant power in specific frequency bands. Forty-two athletes were divided into two groups: those who practiced endurance sports and those who did not, and were asked to perform an IET. EEG signals and gas exchange were recorded. A V̇O2pl was observed in twenty-two subjects (52%). EEG power increased in all subjects during IET, except in the alpha band, which showed variability, but not significantly (64% increase, 34% decrease, p = 0.07). No differences were found between endurance athletes and non-endurance athletes, except for V̇O2max (60.10 ± 6.16 vs. 51.77 ± 6.41, p < 0.001). However, the baseline-corrected ratio of EEG power to V̇O2 was found to decrease in all subjects during IET, in the alpha, beta and theta bands. In conclusion, the presence or absence of a V̇O2pl is not related to the type of EEG response during an IET. Nevertheless, the decline in brain and V̇O2 powers/ratios in all frequency bands suggests that aerobic power may be constrained by brain mobilization.

Despite attempts to unify the different theoretical accounts of the mismatch negativity (MMN), there is still an ongoing debate on the neurophysiological mechanisms underlying this complex brain response. On one hand, neuronal adaptation to recurrent stimuli is able to explain many of the observed properties of the MMN, such as its sensitivity to controlled experimental parameters. On the other hand, several modeling studies reported evidence in favor of Bayesian learning models for explaining the trial-to-trial dynamics of the human MMN. However, direct comparisons of these two main hypotheses are scarce, and previous modeling studies suffered from methodological limitations. Based on reports indicating spatial and temporal dissociation of physiological mechanisms within the timecourse of mismatch responses in animals, we hypothesized that different computational models would best fit different temporal phases of the human MMN. Using electroencephalographic data from two independent studies of a simple auditory oddball task (n = 82), we compared adaptation and Bayesian learning models' ability to explain the sequential dynamics of auditory deviance detection in a time-resolved fashion. We first ran simulations to evaluate the capacity of our design to dissociate the tested models and found that they were sufficiently distinguishable above a certain level of signal-to-noise ratio (SNR). In subjects with a sufficient SNR, our time-resolved approach revealed a temporal dissociation between the two model families, with high evidence for adaptation during the early MMN window (from 90 to 150-190 ms post-stimulus depending on the dataset) and for Bayesian learning later in time (170-180 ms or 200-220ms). In addition, Bayesian model averaging of fixed-parameter models within the adaptation family revealed a gradient of adaptation rates, resembling the anatomical gradient in the auditory cortical hierarchy reported in animal studies.

Statistical variability of electroencephalography (EEG) between subjects and between sessions is a common problem faced in the field of Brain-Computer Interface (BCI). Such variability prevents the usage of pre-trained machine learning models and requires the use of a calibration for every new session. This paper presents a new transfer learning (TL) method that deals with this variability. This method aims to reduce calibration time and even improve accuracy of BCI systems by aligning EEG data from one subject to the other in the tangent space of the positive definite matrices Riemannian manifold. We tested the method on 18 BCI databases comprising a total of 349 subjects pertaining to three BCI paradigms, namely, event related potentials (ERP), motor imagery (MI), and steady state visually evoked potentials (SSVEP). We employ a support vector classifier for feature classification. The results demonstrate a significant improvement of classification accuracy, as compared to a classical training-test pipeline, in the case of the ERP paradigm, whereas for both the MI and SSVEP paradigm no deterioration of performance is observed. A global 2.7% accuracy improvement is obtained compared to a previously published Riemannian method, Riemannian Procrustes Analysis (RPA). Interestingly, tangent space alignment has an intrinsic ability to deal with transfer learning for sets of data that have different number of channels, naturally applying to inter-dataset transfer learning.

Recent computational models of perception conceptualize auditory oddball responses as signatures of a (Bayesian) learning process, in line with the influential view of the mismatch negativity (MMN) as a prediction error signal. Novel MMN experimental paradigms have put an emphasis on neurophysiological effects of manipulating regularity and predictability in sound sequences. This raises the question of the contextual adaptation of the learning process itself, which on the computational side speaks to the mechanisms of gain-modulated (or precision-weighted) prediction error. In this study using electrocorticographic (ECoG) signals, we manipulated the predictability of oddball sound sequences with two objectives: (i) Uncovering the computational process underlying trial-by-trial variations of the cortical responses. The fluctuations between trials, generally ignored by approaches based on averaged evoked responses, should reflect the learning involved. We used a general linear model (GLM) and Bayesian Model Reduction (BMR) to assess the respective contributions of experimental manipulations and learning mechanisms under probabilistic assumptions. (ii) To validate and expand on previous findings regarding the effect of changes in predictability using simultaneous EEG-MEG recordings. Our trial-by-trial analysis revealed only a few stimulus-responsive sensors but the measured effects appear to be consistent over subjects in both time and space. In time, they occur at the typical latency of the MMN (between 100 and 250 ms post-stimulus). In space, we found a dissociation between time-independent effects in more anterior temporal locations and time-dependent (learning) effects in more posterior locations. However, we could not observe any clear and reliable effect of our manipulation of predictability modulation onto the above learning process. Overall, these findings clearly demonstrate the potential of trial-to-trial modeling to unravel perceptual learning processes and their neurophysiological counterparts.

The development of reliable assistive devices for patients that suffer from motor impairments following central nervous system lesions remains a major challenge in the field of non-invasive Brain-Computer Interfaces (BCIs). These approaches are predominated by electroencephalography and rely on advanced signal processing and machine learning methods to extract neural correlates of motor activity. However, despite tremendous and still ongoing efforts, their value as effective clinical tools remains limited. We advocate that a rather overlooked research avenue lies in efforts to question neurophysiological markers traditionally targeted in non-invasive motor BCIs. We propose an alternative approach grounded by recent fundamental advances in non-invasive neurophysiology, specifically subject-specific feature extraction of sensorimotor bursts of activity recorded via (possibly magnetoencephalography-optimized) electroencephalography. This path holds promise in overcoming a significant proportion of existing limitations, and could foster the wider adoption of online BCIs in rehabilitation protocols.

We here turn the general and theoretical question of the complementarity of EEG and MEG for source reconstruction, into a practical empirical one. Precisely, we address the challenge of evaluating multimodal data fusion on real data. For this purpose, we build on the flexibility of Parametric Empirical Bayes, namely for EEG-MEG data fusion, group level inference and formal hypothesis testing.The proposed approach follows a two-step procedure by first using unimodal or multimodal inference to derive a cortical solution at the group level; and second by using this solution as a prior model for single subject level inference based on either unimodal or multimodal data. Interestingly, for inference based on the same data (EEG, MEG or both), one can then formally compare, as alternative hypotheses, the relative plausibility of the two unimodal and the multimodal group priors. Using auditory data, we show that this approach enables to draw important conclusions, namely on (i) the superiority of multimodal inference, (ii) the greater spatial sensitivity of MEG compared to EEG, (iii) the ability of EEG data alone to source reconstruct temporal lobe activity, (iv) the usefulness of EEG to improve MEG based source reconstruction.Importantly, we largely reproduce those findings over two different experimental conditions. We here focused on Mismatch Negativity (MMN) responses for which generators have been extensively investigated with little homogeneity in the reported results. Our multimodal inference at the group level revealed spatio-temporal activity within the supratemporal plane with a precision which, to our knowledge, has never been achieved before with non-invasive recordings.

Determining the anatomical source of brain activity non-invasively measured from EEG or MEG sensors is challenging. In order to simplify the source localization problem, many techniques introduce the assumption that current sources lie on the cortical surface. Another common assumption is that this current flow is orthogonal to the cortical surface, thereby approximating the orientation of cortical columns. However, it is not clear which cortical surface to use to define the current source locations, and normal vectors computed from a single cortical surface may not be the best approximation to the orientation of cortical columns. We compared three different surface location priors and five different approaches for estimating dipole vector orientation, both in simulations and visual and motor evoked MEG responses. We show that models with source locations on the white matter surface and using methods based on establishing correspondences between white matter and pial cortical surfaces dramatically outperform models with source locations on the pial or combined pial/white surfaces and which use methods based on the geometry of a single cortical surface in fitting evoked visual and motor responses. These methods can be easily implemented and adopted in most M/EEG analysis pipelines, with the potential to significantly improve source localization of evoked responses.

Brain-computer interfaces (BCIs) mostly rely on electrophysiological brain signals. Methodological and technical progress has largely solved the challenge of processing these signals online. The main issue that remains, however, is the identification of a reliable mapping between electrophysiological measures and relevant states of mind. This is why BCIs are highly dependent upon advances in cognitive neuroscience and neuroimaging research. Recently, psychological theories became more biologically plausible, leading to more realistic generative models of psychophysiological observations. Such complex interpretations of empirical data call for efficient and robust computational approaches that can deal with statistical model comparison, such as approximate Bayesian inference schemes. Importantly, the latter enable the optimization of a model selection error rate with respect to experimental control variables, yielding maximally powerful designs. In this paper, we use a Bayesian decision theoretic approach to cast model comparison in an online adaptive design optimization procedure. We show how to maximize design efficiency for individual healthy subjects or patients. Using simulated data, we demonstrate the face- and construct-validity of this approach and illustrate its extension to electrophysiology and multiple hypothesis testing based on recent psychophysiological models of perception. Finally, we discuss its implications for basic neuroscience and BCI itself.

We present a brain-computer interface (BCI) version of the famous “Connect Four”. Target selection is based on brain event-related responses measured with nine EEG sensors. Two players compete against each other using their brain activity only. Importantly, we turned the general difficulty of producing a reliable BCI command into an advantage, by extending the game play and rules, in a way that adds fun to the game and might well prove to trigger up motivation in future studies. The principle of this new BCI is directly inspired from our own implementation of the classical P300 Speller (Maby et al. 2010, Perrin et al. 2011). We here establish a proof of principle that the same electrophysiological markers can be used to design an efficient two-player game. Experimental evaluation on two competing healthy subjects yielded an average accuracy of 82%, which is in line with our previous results on many participants and demonstrates that the BCI “Connect Four” can effectively be controlled. Interestingly, the duration of the game is not significantly affected by the usual slowness of BCI commands. This suggests that this kind of BCI games could be of interest to healthy players as well as to disabled people who cannot play with classical games.