The present thesis investigated memory load and memory decay in auditory working memory. Alpha power as a marker for memory load served as the primary indicator for load and decay fluctuations hypothetically reflecting functional inhibition of irrelevant information. Memory load was induced by presenting auditory signals (syllables and pure-tone sequences) in noise because speech-in-noise has been shown before to increase memory load. The aim of the thesis was to assess with magnetoencephalography whether a-priori temporal expectations for the onset-time of a to-be-remembered stimulus reduces memory load. It was reported previously that top-down modulations such as spatial expectations reduce memory load and improve memory performance. However, this effect has neither been investigated with temporal expectations nor in the auditory domain. The present thesis showed that temporal expectations for a syllable in noise reduced memory load. Reduced alpha power during stimulus maintenance as well as improved performance indicated the decrease in memory load. Alpha power effects emerged from the right cingulo-opercular network, presumably reflecting a reduced need for functional inhibition. Critically, symbolic cues induced temporal expectations. This effect could not be replicated for clear speech. However, more implicit temporal expectations based on the passage of time elicited a similar decrease in alpha power for clear speech reflecting reduced memory load. Memory decay was assessed with variable delay phases in an auditory sensory memory task with pure-tone sequences. Similarly to memory performance, alpha power decreased with longer delay phases. Critically, temporal expectations counteracted memory decay and led to more sustained performance as well as alpha power across different delay phases. These alpha-power effects were localized to frontal and parietal attention networks as well as primary auditory and visual sensory areas. This implies the involvement of different brain regions relevant for encoding and maintenance in auditory memory and questions a parsimonious functional inhibition explanation. A correlation of alpha power and behavioral performance underpinned the importance of alpha power for auditory working memory. Altogether, the results of the present thesis provide evidence for a beneficial effect of a-priori temporal expectations for an auditory signal on working memory. Moreover, alpha dynamics were shown to be a distinct marker for the neural efficiency of managing working memory limitations.

La magnétocardiographie (MCG) et la magnétoencéphalographie (MEG) sont deux techniques d'imagerie non-invasives mesurant respectivement les champs magnétiques cardiaques et cérébraux. Les dispositifs actuels utilisent des capteurs supraconducteurs de haute performance mais nécessitant un dispositif de refroidissement cryogénique, engendrant de fortes contraintes tant techniques que financières. Les magnétomètres à pompage optique (OPM) tendent à constituer une réelle alternative. Parmi eux figurent ceux développés au CEA-LETI, basés sur l'utilisation de l'hélium 4. Cette thèse a pour objectif de développer des magnétomètres vectoriels à 4He (fonctionnant à température ambiante) dédiés aux applications MCG et MEG.Après une optimisation des paramètres-clés d'un prototype non-miniaturisé préexistant, une sensibilité inférieure à 100 fT/sqrt(Hz) a pu être obtenue suivant deux axes. Afin de respecter les besoins spécifiques de la MCG et de la MEG une étape de miniaturisation a dû être menée et une architecture gradient-métrique a été mise en place. Parallèlement, des tests précliniques menés à Clinatec nous ont permis de concevoir un nouveau conditionnement du prototype, ainsi qu'un système réduisant les perturbations magnétiques. Une analyse des principales sources de bruit a révélé que les deux principaux contributeurs au bruit sont le laser et le système de décharge HF. Nous avons ainsi envisagé plusieurs pistes d'amélioration du niveau de bruit dont une nouvelle technique de détection. Le prototype issu de ces travaux comporte une pièce élémentaire (la cellule) d'un centimètre de côté, et présente une sensibilité intrinsèque de 350 fT/sqrt(Hz).Le dispositif a ensuite été testé avec succès dans le cadre de mesures MCG sur un sujet sain, précédées de tests sur fantôme ayant permis de prouver l'opérabilité de nos capteurs dans un environnement clinique. Par ailleurs, la reproductibilité des résultats ainsi que la possibilité de réduire à 30 s le temps d'acquisition des données ont pu être démontrées. Une optimisation spécifique de la partie optique du prototype a permis d'obtenir une sensibilité de l'ordre de 210 fT/sqrt(Hz) entre 3 et 300 Hz, compatible avec l'application MEG. Après des tests menés avec succès sur fantôme, trois séries d'essais ont été réalisées sur un sujet sain. Nous avons pu ainsi détecter des potentiels évoqués auditifs, visuels ainsi qu'une modulation de l'activité cérébrale spontanée sous l'effet de l'ouverture des paupières. L'ensemble des résultats obtenus constitue les premières preuves de concept cliniques du dispositif en MCG et MEG. / Magnetocardiography (MCG) and magnetoencephalography (MEG) are non-invasive techniques consisting in measuring respectively cardiac and brain magnetic fields. Despite their performance, the sensors currently used need a cryogenic cooling system which engenders technical and financial constraints. New cryogenic-free sensors have recently emerged: the OPMs (Optically-Pumped Magnetometers). Among them, vector 4He magnetometers developed by CEA-LETI which work at room-temperature. This thesis is focused on the development of 4He magnetometers dedicated to MCG and MEG.After having optimized the key-parameters of a first non-miniaturized prototype, a sensitivity inferior to 100 fT/sqrt(Hz) has been obtained along two axes. In order to meet biomedical constraints, a miniaturization of the device has been processed. In parallel, preclinical tests were carried out which have enabled us to design a gradiometer mode, a new packaging, and a magnetically isolated system. A noise analysis revealed that laser and HF discharge were the main sources of disturbance, and lead us to consider improvements such as a new detection mode. Eventually, a sensor, with a 1cm-sided cell, and an intrinsic sensitivity of 350 fT/√Hz has been developed.Then, device has been successfully tested in the frame of the MCG application from a healthy subject, preceded by a simulation study with a phantom which enables us to demonstrate its operability within a clinical environment. Moreover, we have proved the reproducibility of the measurements and the possibility to detect the main features of the cardiac cycle within a 30 s recording time. A specific optimization of the optical part has enabled us to obtain a 210 fT/sqrt(Hz) sensitivity between 3 and 300 Hz, suitable for the MEG application. After having tested our device with a phantom, three MEG experiments were performed with a healthy subject: auditory evoked field, visual evoked field and spontaneous activities have been detected. The obtained results form the first clinical proof of concept of the device for MCG and MEG applications.

No description available.

Le sommeil est défini comme un état comportemental de repos où nous perdons conscience de notre environnement et notre réactivité aux stimuli extérieurs est drastiquement réduite. Pourtant, lorsque nous dormons, l’appel par notre prénom ou à la sonnerie du réveil peuvent encore nous réveiller, suggérant qu’un certain degré de traitement des stimuli reste possible. Dans ce travail, nous soulevons la question de la profondeur du traitement de l’information extérieure pendant le sommeil. Nous avons enregistré simultanément l’activité cérébrale de sujets sains adultes en électro- et magnéto-encéphalographie (EEG et MEG) en réponse à des stimulations auditives, avant, pendant, et après une courte période de sommeil. Afin de tester la profondeur du traitement de l’information à travers la hiérarchie corticale, nous nous sommes concentrés sur les capacités de codage prédictif hiérarchique, qui permettent au cerveau d’anticiper les évènements futurs à partir d’une connaissance passée. Les prédictions sont faites à de nombreuses si ce n’est toutes les étapes de la hiérarchie corticale. Tester les différents niveaux de prédiction nous permet donc d’évaluer précisément à quel niveau l’intégration de l’information est interrompue. Nous avons d’abord testé les capacités du cerveau à détecter la nouveauté auditive. Nous avons présenté aux sujets des séquences de sons comprenant des régularités temporelles à courte (locale) ou à longue (globale) échelle de temps, et analysé les réponses cérébrales à des sons violant ces régularités. Les réponses cérébrales à ces violations locales ou globales se traduisent respectivement en EEG à l’éveil par l’émergence de deux signaux d’erreur de prédiction : la négativité de mismatch (MMN) et la P300. Notre analyse révèle que la MMN et la P300 disparaissent toutes deux dans le sommeil avec la perte des activations des aires associatives préfrontales et pariétales. Au cours de l’endormissement, la MMN diminue progressivement, tandis que la P300 disparait brutalement avec la perte de conscience des stimuli. Ce comportement tout-ou-rien renforce l’hypothèse que la P300 est un marqueur de la conscience. Malgré tout, nous avons montré que le cerveau détecte toujours les nouveaux sons et peut s’y habituer, mais seulement dans un contexte limité d’adaptation sensorielle de bas niveau. Après avoir démontré la perte des capacités de codage prédictif dans le sommeil dans le cadre de régularités statistiques arbitraires et nouvellement acquises, dans une deuxième série d’expériences nous avons testé la capacité du cerveau endormi à établir des prédictions sur les sons à venir dans le cadre de connaissances sémantiques connues déjà stockées en mémoire à long terme. Nous avons présenté à des sujets endormis des opérations arithmétiques simples, comme “deux plus deux égal neuf”, et nous avons enregistré les réponses cérébrales aux résultats corrects et aux résultats faux. Nous avons découvert que le cerveau était toujours capable de détecter les violations arithmétiques dans le sommeil, avec des activations en partie similaires à celles de l’éveil. Nous suggérons que, bien que le sommeil prévienne tout calcul explicite, il y a conservation des signaux d’erreur de prédiction pour les opérations arithmétiques simple déjà mémorisées. Ce travail clarifie à quel niveau l’intégration de l’information auditive est interrompue pendant le sommeil, et quelles fonctions cognitives persistent ou s’altèrent. La persistance de l’adaptation sensorielle et des capacités de prédiction à partir de connaissances déjà mémorisées sont probablement responsables de la réactivité résiduelle qui peut être observée pendant le sommeil alors que les sujets sont inconscients. Finalement, ces résultats aident aussi à mieux comprendre pourquoi un stimulus donné sera traité ou non dans le sommeil. (...) / Sleep can be defined as a behavioral state of rest in which consciousness of external stimuli vanishes and responsiveness to the environment is drastically reduced. When we sleep, however, we may still react and wake up to our name or to the alarm clock, suggesting that some processing of external stimuli remains. We address in the present work the question of how deeply external information is processed during sleep. We recorded brain activity in adult human subjects simultaneously in electro and magnetoencephalography (EEG and MEG) in response to auditory stimulation, before, during and after a short period of sleep. In order to test information integration through the brain hierarchy, we focused on hierarchical predictive coding capabilities, which enable the brain to anticipate the future from previous knowledge. Predictions occur at many if not all steps of the cortical hierarchy. Testing different levels of predictions enables us to assess the steps at which information integration is disrupted during sleep. We first tested the capacity of the sleeping brain to detect auditory novelty. We analyzed brain responses to violations of local and global temporal regularities, which are respectively reflected in EEG during wakefulness by two successive prediction error signals, the mismatch negativity (MMN) and the P300. Our analysis revealed that both the MMN and the P300 vanish during sleep, along with the loss of activations in prefrontal and parietal associative areas. The MMN gradually decreased in the descent to sleep, whereas the P300 vanished abruptly with the loss of awareness during N1 sleep. This all-or-none behavior strongly reinforces the hypothesis that the P300 is a marker of consciousness. Even so, we showed that sounds still activate sensory cortices, and that the brain remains able to detect new sounds and to habituate to them, but only in the limited context of sensory adaptation. Having demonstrated the disruption of predictive coding for arbitrary and newly acquired statistical regularities, in a second set of experiments we tested the capacity of the sleeping brain to develop predictions of future auditory stimuli for over-learned semantic knowledge stored in long-term memory. We presented sleeping subjects with simple arithmetic facts such as “two plus two is nine” and recorded brain responses to correct or incorrect results. We discovered that the sleeping brain was still able to detect arithmetic violations, with activations in part similar to wakefulness. We suggest that, although sleep disrupts explicit arithmetic computations, there is a preservation of prediction error signals for arithmetic facts stored in long-term memory. The present work clarifies the steps at which auditory information integration is disrupted during sleep, and which cognitive functions remain or vanish. The preservation of low-level sensory adaptation and of predictions from long term memory may account for the residual responsiveness that can be observed during sleep, while subjects are unconscious. Finally, these results also help to better understand why a given stimulus may or may not be processed during sleep. The depth of information integration is function of the ongoing spontaneous oscillations of the sleeping brain, but also of the nature of the stimulus, i.e. its salience, its knowledge, and its relevance.

Comprendre comment différentes régions du cerveau interagissent afin d’exécuter une tâche, est un défi très complexe. La magnéto- et l’électroencéphalographie (MEEG) sont deux techniques non-invasive d’imagerie fonctionnelle utilisées pour mesurer avec une bonne résolution temporelle l’activité cérébrale. Estimer cette activité à partir des mesures MEEG est un problème mal posé. Il est donc crucial de le régulariser pour obtenir une solution unique. Il a été montré que l’homogénéité structurelle des régions corticales reflète leur homogénéité fonctionnelle. Un des buts principaux de ce travail est d’utiliser cette information structurelle pour définir des a priori permettant de contraindre de manière plus anatomique ce problème inverse de reconstruction de sources. L’imagerie par résonance magnétique de diffusion (IRMd) est, à ce jour, la seule technique non-invasive qui fournisse des informations sur l’organisation structurelle de la matière blanche. Cela justifie son utilisation pour contraindre notre problème inverse. Nous utilisons l’information fournie par l’IRMd de deux manière différentes pour reconstruire les activations du cerveau : (1) via une méthode spatiale qui utilise une parcellisation du cerveau pour contraindre l’activité des sources. Ces parcelles sont obtenues par un algorithme qui permet d’obtenir un ensemble optimal de régions structurellement homogènes pour une mesure de similarité donnée sur tout le cerveau. (2) dans une approche spatio-temporelle qui utilise les connexions anatomiques, calculées à partir des données d’IRMd, pour contraindre la dynamique des sources. Ces méthodes sont appliquée à des données synthétiques et réelles. / Understanding how brain regions interact to perform a given task is a very challenging task. Electroencephalography (EEG) and Magnetoencephalography (MEG) are two non-invasive functional imaging modalities used to record brain activity with high temporal resolution. As estimating brain activity from these measurements is an ill-posed problem. We thus must set a prior on the sources to obtain a unique solution. It has been shown in previous studies that structural homogeneity of brain regions reflect their functional homogeneity. One of the main goals of this work is to use this structural information to define priors to constrain more anatomically the MEG/EEG source reconstruction problem. This structural information is obtained using diffusion magnetic resonance imaging (dMRI), which is, as of today, the unique non-invasive structural imaging modality that provides an insight on the structural organization of white matter. This makes its use to constrain the EEG/MEG inverse problem justified. In our work, dMRI information is used to reconstruct brain activation in two ways: (1) In a spatial method which uses brain parcels to constrain the sources activity. These parcels are obtained by our whole brain parcellation algorithm which computes cortical regions with the most structural homogeneity with respect to a similarity measure. (2) In a spatio-temporal method that makes use of the anatomical connections computed from dMRI to constrain the sources’ dynamics. These different methods are validated using synthetic and real data.

The present thesis investigated memory load and memory decay in auditory working memory. Alpha power as a marker for memory load served as the primary indicator for load and decay fluctuations hypothetically reflecting functional inhibition of irrelevant information. Memory load was induced by presenting auditory signals (syllables and pure-tone sequences) in noise because speech-in-noise has been shown before to increase memory load. The aim of the thesis was to assess with magnetoencephalography whether a-priori temporal expectations for the onset-time of a to-be-remembered stimulus reduces memory load. It was reported previously that top-down modulations such as spatial expectations reduce memory load and improve memory performance. However, this effect has neither been investigated with temporal expectations nor in the auditory domain. The present thesis showed that temporal expectations for a syllable in noise reduced memory load. Reduced alpha power during stimulus maintenance as well as improved performance indicated the decrease in memory load. Alpha power effects emerged from the right cingulo-opercular network, presumably reflecting a reduced need for functional inhibition. Critically, symbolic cues induced temporal expectations. This effect could not be replicated for clear speech. However, more implicit temporal expectations based on the passage of time elicited a similar decrease in alpha power for clear speech reflecting reduced memory load. Memory decay was assessed with variable delay phases in an auditory sensory memory task with pure-tone sequences. Similarly to memory performance, alpha power decreased with longer delay phases. Critically, temporal expectations counteracted memory decay and led to more sustained performance as well as alpha power across different delay phases. These alpha-power effects were localized to frontal and parietal attention networks as well as primary auditory and visual sensory areas. This implies the involvement of different brain regions relevant for encoding and maintenance in auditory memory and questions a parsimonious functional inhibition explanation. A correlation of alpha power and behavioral performance underpinned the importance of alpha power for auditory working memory. Altogether, the results of the present thesis provide evidence for a beneficial effect of a-priori temporal expectations for an auditory signal on working memory. Moreover, alpha dynamics were shown to be a distinct marker for the neural efficiency of managing working memory limitations.

With MEG, the tiny magnetic fields produced by neuronal currents within the brain can be measured completely non-invasively. But the signals are very small (~100 fT) and often obscured by spontaneous brain activity and external noise. So, a recurrent issue in MEG data analysis is the identification and elimination of this unwanted interference within the recordings. Various strategies exist to meet this purpose. In this thesis, two of these strategies are scrutinized in detail. The first is the commonly used procedure of averaging over trials which is a successfully applied data reduction method in many neurocognitive studies. However, the brain does not always respond identically to repeated stimuli, so averaging can eliminate valuable information. Alternative approaches aiming at single trial analysis are difficult to realize and many of them focus on temporal patterns. Here, a compromise involving random subaveraging of trials and repeated source localization is presented. A simulation study with numerous examples demonstrates the applicability of the new method. As a result, inferences about the generators of single trials can be drawn which allows deeper insight into neuronal processes of the human brain. The second technique examined in this thesis is a preprocessing tool termed Signal Space Separation (SSS). It is widely used for preprocessing of MEG data, including noise reduction by suppression of external interference, as well as movement correction. Here, the mathematical principles of the SSS series expansion and the rules for its application are investigated. The most important mathematical precondition is a source-free sensor space. Using three data sets, the influence of a violation of this convergence criterion on source localization accuracy is demonstrated. The analysis reveals that the SSS method works reliably, even when the convergence criterion is not fully obeyed. This leads to utilizing the SSS method for the transformation of MEG data to virtual sensors on the scalp surface. Having MEG data directly on the individual scalp surface would alleviate sensor space analysis across subjects and comparability with EEG. A comparison study of the transformation results obtained with SSS and those produced by inverse and subsequent forward computation is performed. It shows strong dependence on the relative position of sources and sensors. In addition, the latter approach yields superior results for the intended purpose of data transformation.

Current theories of auditory attention are largely based upon studies examining either the presentation of a single auditory stimulus or requiring the identification and labeling of stimuli presented sequentially. Whether or not these theories apply in more complex ecologically-valid environments where multiple sound sources are simultaneously active is still unknown. This study examined the pattern of neuromagnetic responses elicited when participants had to perform a search in an auditory language-based `scene` for a stimulus matching an imperative target held in working memory. The analysis of source waveforms revealed left lateralized patterns of activity that distinguished target present from target absent trials. Similar source waveform amplitudes were found when the target was presented in the left or right hemispace. The results suggest that auditory search for speech sounds engage a left lateralized process in the superior temporal gyrus.

Current theories of auditory attention are largely based upon studies examining either the presentation of a single auditory stimulus or requiring the identification and labeling of stimuli presented sequentially. Whether or not these theories apply in more complex ecologically-valid environments where multiple sound sources are simultaneously active is still unknown. This study examined the pattern of neuromagnetic responses elicited when participants had to perform a search in an auditory language-based `scene` for a stimulus matching an imperative target held in working memory. The analysis of source waveforms revealed left lateralized patterns of activity that distinguished target present from target absent trials. Similar source waveform amplitudes were found when the target was presented in the left or right hemispace. The results suggest that auditory search for speech sounds engage a left lateralized process in the superior temporal gyrus.

We examined the spatiotemporal dynamics underlying the processing of a false belief task using magnetoencephalography (MEG). Twenty adults performed a false belief task adapted for MEG. Regions of interest were selected based on source analyses on the contrast between false and true belief, and MEG source time-course reconstructions were generated and analyzed to determine the temporal architecture of neural activations specific to false belief reasoning. We found frontal, temporal and parietal regions to activate during false belief processing, confirming prior findings. We also extend previous findings by adding information about the temporal profile of neural activity during theory of mind processing, an area lacking in the literature. We found that increased frontal activity began at 100 ms bilaterally, followed by parietal regions from 200 to 330 ms and temporal regions at 350 ms, at which point frontal activity became lateralized to the right hemisphere.