Brain-Computer Interfaces (BCIs) that rely on motor imagery currently allow subjects to control quad-copters, robotic arms, and computer cursors. Recent advancements have been made possible because of breakthroughs in fields such as electrical engineering, computer science, and neuroscience. Currently, most real-time BCIs use hand-crafted feature extractors, feature selectors, and classification algorithms. In this work, we explore the different classification algorithms currently used in electroencephalographic (EEG) signal classification and assess their performance on intracranial EEG (iEEG) data. We first discuss the motor imagery task employed using iEEG signals and find features that clearly distinguish between different classes. Second, we compare the different state-of-the-art classifiers used in EEG BCIs in terms of their error rate, computational requirements, and feature interpret-ability. Next, we show the effectiveness of these classifiers in iEEG BCIs and last, show that our new classification algorithm that is designed to use spatial, spectral, and temporal information reaches performance comparable to other state-of-the-art classifiers while also allowing increased feature interpret-ability. / Master of Science / Brain-Computer Interfaces (BCIs) as the name suggests allows individuals to interact with computers using electrical activity captured from different regions of the brain. These devices have been shown to allows subjects to control a number of devices such as quad-copters, robotic arms, and computer cursors. Applications such as these obtain electrical signals from the brain using electrodes either placed non-invasively on the scalp (also known as an electroencephalographic signal, EEG) or invasively on the surface of the brain (Electrocorticographic signal, ECoG). Before a participant can effectively communicate with the computer, the computer is calibrated to recognize different signals by collecting data from the subject and learning to distinguish them using a classification algorithm. In this work, we were interested in analyzing the effectiveness of using signals obtained from deep brain structures by using electrodes place invasively (also known as intracranial EEG, iEEG). We collected iEEG data during a two hand movement task and manually analyzed the data to find regions of the brain that are most effective in allowing us to distinguish signals during movements. We later showed that this task could be automated by using classification algorithms that are borrowed from electroencephalographic (EEG) signal experiments.

Intracranial electroencephalography (iEEG) provides superior diagnostic and research benefits over non-invasive EEG in terms of spatial resolution and the level of electrophysiological detail. Post-operative Computed Tomography (CT) scans provide the precision in electrode localization required for clinical purposes; however, to use this data for basic sleep research the challenge lies in identifying the precise locations of the implanted electrodes’ recording sites in terms of neuroanatomical regions as well as reliable scoring of their sleep data without the aid of facial electrodes. While existing methods can be combined to determine their exact locations in three-dimensional space, they fail to identify the functionally relevant gray matter areas that lie closest to them, especially if the points lie in the white matter. We introduce an iterative sphere inflation algorithm in conjunction with a unified pipeline to detect the exact as well as nearest regions of interest for these recording sites. Next, for sleep scoring purposes, we establish differences observed in alpha band activity between wakefulness and rapid eye movement (REM) sleep in frontal and temporal regions of iEEG patients. Lastly, we implement an automated sleep scoring method relying on the variations in alpha and delta bands power during sleep which can be applied to large sets of iEEG data recorded without accompanying electrooculogram (EOG) and electromyogram (EMG) electrodes available across labs for use in studies pertaining to neural dynamics during sleep. / M.S. / Patients with epilepsy (a neurological disorder characterized by seizures) who do not respond to medication often undergo invasive monitoring of their brains’ electrical activity using intracranial electroencephalography (iEEG). iEEG requires a surgery in which electrodes are inserted directly into the patient’s brain for better measurements. While they are monitored, these patients offer a unique opportunity for research studies that investigate the role of sleep in various learning, memory mechanisms and other health-related areas. This is because the direct contact of the electrodes with the brain tissue provides far superior quality and resolution of brain activity data in comparison to non-invasive cap-based EEG that healthy subjects wear over their scalp. However, in order to derive meaningful conclusions from these invasive recordings, we must first know the exact areas of the brain from which each site records the electrical data. We must then be able to identify which stage of sleep the patient is in at any given point in time, to be able to successfully correlate specific sleep stage-related activity with our research objectives; these patients often lack the facial electrodes used for standard sleep scoring procedures. To solve the first problem, we present an electrode localization method along with an algorithm to determine which neighboring regions contribute most to a given site’s recorded data. For the second problem, we first establish a difference in the behavior of alpha waves in the brain between wakefulness and rapid eye movement (REM) sleep. Lastly, we present an automated method to classify sleep data into different stages based on the variation in alpha waves and delta waves found during sleep.

Over the course of a typical night, sleep is comprised of multiple different stages that involve changes in brainwave patterns. Intracranial EEG (iEEG) is an invasive brain recording technique used in hospital settings in epileptic patients to determine the focus of their seizure activity. The intracranial data recorded allows one to directly observe the neural activity of deep brain structures such as the hippocampus and to detect single unit activity and local field potentials, thus providing a level of physiological detail normally available only in animal studies. In this thesis we employ intracranial data to advance our understanding of sleep neural dynamics in humans, and to this end its focus is in two areas : (1) developing a way of sleep scoring iEEG data and (2) investigating the neural dynamics of a particular waveform found during sleep, the sleep spindle, and its role in memory consolidation. Typically, iEEG recordings do not include electrooculogram or electromyogram recordings, which are normally needed for sleep scoring—especially for scoring rapid-eye movement (REM) sleep. We identified differences in alpha power between wake and REM sleep to develop a methodological way to reliably differentiate between wake and REM sleep states. We also wanted to investigate the neural dynamics involved with a particular brainwave seen during sleep, the sleep spindle, which is thought to be important for sleep-mediated memory consolidation. Historically, sleep spindles were thought to occur synchronously across the cortex, but recent findings using iEEG have identified that sleep spindles can also be local. We utilized intracranial EEG to confirm previous findings that iEEG can identify local sleep spindles. In addition to identifying local sleep spindles, we aimed to investigate the potential role that sleep spindles have on learning and memory using standard targeted memory reactivation paradigms for iii both procedural and declarative memories. We found that local sleep spindles occurred at a specific time following auditory stimulation for both procedural and declarative memories. This work has opened up the use of iEEG recordings to investigations of REM sleep dynamics and laid the groundwork for examining the role of local sleep spindles in memory consolidation. / Master of Science / During a night of sleep, our brain goes through different stages that exhibit changes in brainwave patterns. Intracranial EEG (iEEG) is an invasive brain recording technique used in hospital settings in epileptic patients to determine the focus of their seizure activity; this particular brain recording technique allows one to observe the brain activity of deep brain structures. By using iEEG data, we aimed to (1) develop a way of sleep scoring iEEG data and (2) investigate the neural dynamics of a particular waveform found during sleep, the sleep spindle, and its role in memory consolidation.  Electrooculograms (EOG) are used to record the electrical activity of eye movements, and electromyograms (EMG) are used to measure muscle activity. Both of these recording techniques, in addition to EEG, are needed for sleep scoring, especially rapid eye movement (REM) sleep. However, typical iEEG recordings do not have EOGs and EMGs applied to the patient. Using iEEG data, we were able to identify differences in a specific brainwave, the alpha rhythm, between wakeful brain activity and REM sleep brain activity. Furthermore, we were able to use this difference to reliably score REM sleep in iEEG data without the need for EOGs and EMGs.  We also wanted to investigate the brainwave changes in a particular waveform, the sleep spindle, that has been thought to be important for sleep-mediated memory consolidation. Previous research using typical EEG recordings showed that sleep spindles occur synchronously across the cortex, but recent findings using iEEG have identified that sleep spindles can also occur asynchronously across the cortex. We replicated previous research showing that these local sleep spindles are identifiable using iEEG recordings. In addition to identifying local sleep spindles, we investigated the potential role that sleep spindles have on learning and memory. To do so, we used standard targeted memory reactivation paradigms for two types of memory: declarative and procedural memory. We found that local sleep spindles occurred at a specific time following auditory stimulation for both procedural and declarative memories.  This work has opened up the use of iEEG recordings to investigations of REM sleep dynamics and laid the groundwork for examining the role of local sleep spindles in memory consolidation.

Our brain controls the processes of the body including movement. In this thesis, we try to understand how the information about hand movement is encoded into the brain's electrical activity and how this activity can be used to predict the velocity and absolute velocity of hand movements. Using a well-established deep neural network architecture for EEG decoding - the Deep4Net - we predict hand movement velocity and absolute velocity from intracranial EEG signals. While reaching the expected performance level, we determine the influence of different frequency bands on the network's prediction. We find that modulations in the high-gamma frequency band are less informative than expected based on previous studies. We also identify two architectural modifications which lead to higher performances. 1. the removal of max-pooling layers in the architecture leads to significantly higher correlations. 2. the non-uniform receptive field of the network is a potential drawback making the network biased towards less relevant information. 1

1 ABSTRACT The presented dissertation deals with prognostic factors influencing a favorable postoperative outcome in patients undergoing surgical treatment of epilepsy and the possibilities of improving the methods used in the localization of epileptogenic lesions. This work is based on the results of four published studies. The first study evaluated the factors influencing the long-term outcomes of epilepsy surgery in MRI-negative (nonlesional) extratemporal lobe epilepsy (nETLE). The aim of the study was to evaluate the benefit of non-invasive diagnostic tests and their relationship with a favorable surgical outcome in a group nETLE patients. Univariate analysis showed that localized interictal epileptiform discharges (IEDs) on the scalp EEG were associated with a favorable surgical outcome. Diagnostic difficulty in this group of patients is highlighted by the fact that, although 9 of 24 patients undergoing surgery had a favorable outcome, and only nine of 85 patients with nETLE achieved such a favorable outcome. The second work evaluated the benefit of SPECT (Single Photon Emission Tomography) statistical processing over traditional subtraction methods in patients with MRI-negative temporal lobe epilepsy (nTLE) and MRI-negative extratemporal epilepsy (nETLE). 49 consecutive patients who underwent...

Comment notre cerveau traite l’information de la récompense, et comment un tel traitement est influence par des paramètres tels que la probabilité et le risque sont devenues des questions cruciales des neurosciences cognitives. De plus, des recherches récentes suggèrent un effet modulateur d’un certain nombre d’hormones sur le cerveau et sur le comportement, et également qu’un dysfonctionnement du système de récompense pourrait expliquer des comportements addictifs tels que le jeu pathologique. Durant cette thèse, nous avons eu recours à de l’EEG stéréotaxique (SEEG) et à une combinaison d’Imagerie à Résonnance Magnétique fonctionnelle (IRMf) et d’endocrinologie pour réaliser trois études s’intéressant au traitement de la récompense chez des sujets sains, chez des patients souffrant d’épilepsie chez qui des macroélectrodes ont été implantées, et chez des joueurs pathologiques. Ensemble, nos études améliorent la compréhension de nouveaux aspects du traitement de la récompense chez les sujets sains, chez les patients épileptiques, et chez les joueurs pathologiques / How our brain processes reward information and how such processing is influenced by parameters such as reward probability and risk have become key questions in cognitive neuroscience. In addition, recent researches suggest a modulatory effect of a number of hormones on brain and behavior and a dysfunction of the reward system in a number of behavioral addictions, such as gambling disorder. This Ph.D. used intracranial EEG (iEEG) and combined Functional Magnetic Resonance Imaging (fMRI) and endocrinology to perform four studies investigating reward processing in healthy subjects, patients with epilepsy implanted with depth electrodes and individuals with gambling disorder. Together, our series of studies advance our understanding of new aspects concerning reward processing in healthy subjects, patients with epilepsy and individuals with gambling disorder

In this work, we propose an approach that allows to explore the potential pathophysiological mechanisms (at neuronal population level) of ictogenesis by combining clinical intracranial electroencephalographic (iEEG) recordings with a neural mass model. IEEG recordings from temporal lobe epilepsy (TLE) patients around seizure onset were investigated. Physiologically meaningful parameters (average synaptic gains of the excitatory, slow and fast inhibitory population, Ae, B and G) were identified during interictal to ictal transition. We analyzed the temporal evolution of four ratios, i.e. Ae/G, Ae/B, Ae/(B + G), and B/G. The excitation/inhibition ratio increased around seizure onset and decreased before seizure offset, suggesting the disturbance and restoration of balance between excitation and inhibition around seizure onset and before seizure offset, respectively. Moreover, the slow inhibition may have an earlier effect on the breakdown of excitation/inhibition balance. Results confirm the decrease in excitation/inhibition ratio upon seizure termination in human temporal lobe epilepsy, as revealed by optogenetic approaches both in vivo in animal models and in vitro. We further explored the distribution of the average synaptic gains in parameter space and their temporal evolution, i.e. the path through the model parameter space, in TLE patients. Results showed that the synaptic gain values located roughly on a plane before seizure onset, dispersed during ictal and returned when the seizure terminated. Cluster analysis was performed on seizure paths and demonstrated consistency in synaptic gain evolution across different seizures from individual patients. Furthermore, two patient groups were identified, each one corresponding to a specific synaptic gain evolution in the parameter space during a seizure. Results were validated by a bootstrapping approach based on comparison with random paths. The differences in the path revealed variations in EEG dynamics for patients despite showing an identical seizure onset pattern. Our approach may have the potential to classify the epileptic patients into subgroups based on different mechanisms revealed by subtle changes in synaptic gains and further enable more robust decisions regarding treatment strategy. The increase of excitation/inhibition ratios, i.e. Ae/G, Ae/B and Ae/(B+G), around seizure onset makes them potential cues for seizure detection. We explored the feasibility of a model based seizure detection algorithm. A simple thresholding method was employed. We evaluated the algorithm against the manual scoring of a human expert on iEEG samples from patients suffering from different types of epilepsy. Results suggest that Ae/(B+G), i.e. excitation/(slow + fast inhibition) ratio, allowed the best performance and that the algorithm best suited TLE patients. Leave-one-out cross-validation showed that the algorithm achieved 94.74% sensitivity for TLE patients. The median false positive rate was 0.16 per hour, and median detection delay was -1.0 s. Of interest, the values of the threshold determined by leave-one-out cross-validation for TLE patients were quite constant, suggesting a general excitation/inhibition balance baseline in background iEEG among TLE patients. Such a model-based seizure detection approach is of clinical interest and could also achieve good performance for other types of epilepsy provided that more appropriate model, i.e. better describe epileptic EEG waveforms for other types of epilepsy, is implemented. Altogether, this thesis contributes to the field of epilepsy research from two perspectives. Scientifically, it gives new insights into the mechanisms underlying interictal to ictal transition, and facilitates better understanding of epileptic seizures. Clinically, it provides a tool for reviewing EEG data in a more efficient and objective manner and offers an opportunity for on-demand therapeutic devices. / Doctorat en Sciences de l'ingénieur et technologie / info:eu-repo/semantics/nonPublished

Les mécanismes cérébraux engagés dans la prise de décision sont loin d’être compris. Ils peuvent cependant être décomposés en plusieurs étapes : il s’agit premièrement d’assigner une « valeur » aux options considérées, c’est-à-dire une quantification subjective du désir d’obtenir chacune d’entre elles. Ensuite, il faut les comparer afin d’être capable de sélectionner celle qui a la plus grande valeur. L’assignation d’une valeur à un objet semble être effectuée par un réseau cérébral qui recoupe le réseau de la récompense identifié chez l’animal et il a été logiquement nommé le « système cérébral des valeurs ». Le travail réalisé dans cette thèse s’intéresse à la notion de valeur et aux moyens d’y avoir accès, aux propriétés du réseau cérébral d’évaluation et à son implication dans le processus de décision. La première étude a montré que les moyens utilisés pour mesurer les valeurs pouvaient être considérés comme équivalents. La deuxième étude, réalisée sur des données d’intra-électroencéphalographie humaine, a permis d’étudier la dynamique neurale du réseau cérébral d’évaluation, mais aussi d’étudier ses propriétés. La dernière expérience, faite en IRMf propose une solution générale sur l’implémentation neurale du processus de décision et révèle des mécanismes sources de biais dans le comportement jusqu’ici inexplorés. Les résultats de ces études considérés dans leur ensemble mettent en lumière certains mécanismes cognitifs de la prise de décision en explorant les propriétés neurales d’assignation de valeurs mais également en proposant un nouveau cadre d’implémentation de la décision elle-même. / Neural processes engaged in decision-making remain unclear. A decomposition of these processes might help us to understand the involved mechanisms. Indeed, first we need to assign what we will call a ‘subjective value’ to each option – i.e. the quantification of how much we like each of these options. Then, we need to compare those values to finally being able to select one of them. Assigning a value seems to be the function of an interesting brain network which overlaps the reward circuitry identified in animal studies – and which is called the Brain Valuation System (BVS). In the first study of this PhD thesis, we investigated and compared three behavioral ways to have an access to these ‘subjective values’. We found that subjective values were relatively robust to the way they were elicited. In the second study, we investigated the specific properties of the Brain Valuation System established through fMRI in humans in a large dataset of intra-EEG recordings in epileptic patients. Finally, in the last study we investigated how this brain network was involved during a binary choice in fMRI. Altogether, our findings shed light on the distinct cognitive mechanisms underlying value-based decision-making i) by exploring the neural properties of value assignment and ii) by proposing a general solution to the neural implementation of the comparison between option values. We believe this demonstration points to hidden default policies as sources of bias in choices.