Journée GDR-STIC santé, BCI @ Paris: 20 Mai, INRIA Place d'Italie
Compte rendu de la journée à télécharger ici.
BCI ds la littérature...
Programme de la journée
Matin
- 10:00 -> 10:50 : Bertrand Thirion (INRIA/Neurospin), Decoding brain states from MRI/MEG data: Some common challenges with BCI Download the slides
- Abstract : Decoding brain states from neuroimaging (functional MRI, MEG/EEG) data is a recent approach in neuroimaging data analysis that has been adopted as a sensitive characterization of subtle signal biases that are related to the performance of various perceptual, cognitive or motor tasks. While having many differences with brain-computer interfaces based e.g. on EEG data due to the reduced temporal resolution, the absence of real-time constraints and of feedback, the decoding paradigm shares with BCI the necessity to reduce the data while retaining the information of interest, and is more targeted toward providing a meaningful and interpretable model of standard brain functional networks, as in the standard brain mapping perspective. We will review the state of the art in decoding approaches, describe some recent developments performed at Parietal and finally discuss possible convergence with the BCI domain (real-time analysis and feedback).
- 10:50 -> 11h20 : Pause
- 11:20 -> 11h45 : Louis Mayaud (Oxford/Hopital de Garches CICIT), Robust Brain Computer Interface Keyboard (RoBIK) How to bring BCI from bench to bedside ? Locks and key success factors Download the slides
- 11:45 -> 12:10 : Théodore Papadopoulo (INRIA Sophia Antipolis) Co-Adapt : new directions for the design of BCI systems by using co-adaptation Download the slides
- Abstract : The partners of this projects are the INSERM U821 laboratory of Bron, the "laboratoire de Neurologie de la cognition" UMR6155 CNRS of Marseille, The INRIA Lille Sequel project-team and the "laboratoire d'Analyse Topologie et Probabilités" UMR6632/CNRS of Aix en Provence.Brain Computer Interfaces (BCI) provide a direct communication channel from the brain to a computer, bypassing traditional interfaces such as keyboard or mouse, and also providing a feedback to the user, through a sensory modality (visual, auditory or haptic). A target application of BCI is to restore mobility or autonomy to severely disabled patients, but more generally BCI opens up many new opportunities for better understanding the brain at work, for enhancing Human Computer Interaction, and for developing new therapies for mental illnesses. In BCI, new modes of perception and interaction come into play, and a new user must learn to operate a BCI, as an infant learns to explore his/her sensorymotor system. Central to BCI operation are the notions of feedback and of reward, which we believe should hold a more central position in BCI research. The goal of this project is to study the co-adaptation between a user and a BCI system in the course of training and operation. The quality of the interface will be judged according to several criteria (reliability, learning curve, error correction, bit rate). BCI will be considered under a joint perspective: the user's and the system's. From the user's brain activity, features must be extracted, and translated into commands to drive the BCI system. Feature extraction from data, and classification issues, are very active research topics in BCI. However, additional markers may also be extracted to modulate the system's behavior. It is for instance possible to monitor the brain's reaction to the BCI outcome, compared to the user's expectations. This type of information we refer to as meta-data because it is not directly related to the command, and it may be qualitative rather than quantitative. To our knowledge, there is so far no BCI system that integrates such meta-data from the user's brain. From the point of view of the system, it is important to devise adaptive learning strategies, because the brain activity is not stable in time. How to adapt the features in the course of BCI operation is a difficult and important topic of research. A Machine Learning method known as Reinforcement Learning (RL) may prove very relevant to address the above questions. Indeed, it is an adaptive learning method that explicitly incorporates a reward signal, which may be qualitative (hence allowing meta-data integration). The aim of CO-ADAPT is to propose new directions for BCI design, by modeling explicitly the co-adaptation taking place between the user and the system.
Repas (libre)
Voici une selection de restaurants à moins de 1km du lieu de la réunion : restaurants proches
Apres-midi
- 13:45 -> 14:25 : Fabien Lotte (I2R Singapour), Towards more practical Brain-Computer Interfaces for future use outside laboratories Download the slides
- Abstract : Brain-Computer Interfaces (BCI) are communication systems that enable users to send commands to computers by using only their brain activity, this activity being measured and processed by the system (typically using ElectroEncephaloGraphy or EEG). Recently, BCI research has grown rapidly and several prototypes of BCI have been developed, mainly in the medical field, as assistive devices for disabled users, but also for a more general audience, with applications such as video games or virtual reality. Despite these promising results, current BCI systems are only prototypes that are limited to a laboratory use. Indeed, BCI suffer from several limitations which restraint their use in practical applications. Among these limitations, we can mention the non-robustness of BCI to noise and user’s motions (which severely pollute EEG signals), as well as their long calibration times. Indeed, for each new BCI user, a specific BCI must be calibrated, which requires acquiring many examples of EEG signals from this user. This presentation describes our recent works aiming at solving such limitations. More precisely, it describes various EEG signal processing and classification methods to design robust BCI systems that can be trained with a very small amount of data. These methods are generally based on regularization methods that make use of a-priori we have on BCI design. They notably exploit useful information contained in the EEG signals of previous users (different from the target user) or spatial a-priori about the EEG signals measured by neighbouring electrodes.
- 14:25 -> 14:50 : Pause
- 14:50 -> 15:15 : Marco Congedo (CNRS/GIPSA-lab), BCI and Brain Research in Grenoble: l’Equipe ViBS (Vision and Brain Signal Processing) du GIPSA-lab
- Abstract : At GIPSA-lab (Grenoble Images Parole Signal Automatique) in Grenoble we have formed a new team named ViBS (Vision and Brain Signal Processing), which makes of BCI (Brain-Computer Interface) a fondamental subject of investigation. In this talk current BCI-related activities of the team supported by ANR projects Open-ViBE2, RoBiK and Gaze&EEG, will be presented.
- 15:15 -> 15:40 : Florian Yger et Rémi Flamary (INSA-Rouen), Méthodes vaste marge pour la classification de signaux BCI Download the slides by Rémi Download the corresponding Cap 2010 paper by Rémi Download the slides by Florian Download the corresponding Cap 2010 paper by Florian
- 15:40 -> 16:05 : Michel Besserve (Max Planck Institute for Biological Cybernetics, Tubingen) Interactions causales entre bandes de fréquences dans les signaux électrophysiologies: un aperçu de la dynamique des traitements neuronaux
- Abstract : Bien que les signaux électriques fonctionnels (EEG,LFP) soient actuellement utilisés dans les interfaces cerveau machines, beaucoup de questions se posent encore sur les informations qu'ils contiennent a propos de l'activité sous-jacente des réseaux neuronaux. J'illustrerai ceci en mettant en évidence des interactions non-linéaires entre différents aspects des signaux extra-cellulaires enregistrés dans le cortex visuel chez le singe. En utilisant des méthodes issues de la théorie de l'information et des méthodes a noyau, je mettrai en évidence le rôle de la bande gamma qui influence les activités basses fréquences et très hautes fréquences des réseau pendant le traitement de stimuli visuels. Ces résultats montrent la richesse et la complexité des signaux cérébraux qui est probablement liée a l'organisation en micro-circuits des assemblées neurones corticaux sous-jacents. Le prise en compte de ces aspects est a mon avis fondamentale pour les développements futurs des ICM.
- 16:05 -> 16:15 : Mini-Pause
- 16:15 -> 16:40 : Bertrand Rivet (GIPSA-lab), Réduction de capteurs pour les BCI de type P300 speller
- 16:40 -> 17:05 : Alexandre Barachant (CEA Grenoble) : Mise en oeuvre d’un brain switch EEG Download the slides
Cette journée est soutenue par le GDR-STIC santé, INRIA-Jalons, l'équipe TAO du LRI, INRIA Italie (place d'), le CNRS et l'AFIA
Inscriptions et Infos pratiques
Inscriptions
La participation à la journée est gratuite, néanmoins le nombre de participants est limité à une quarantaine. L'inscription est donc obligatoire :- Email à cedric.gouy-pailler à lri.fr (Sujet du message : Inscription BCI meeting)
Missions pour les membres du GDR-STIC Santé
- Inscription sur le site du GDR GDR-STIC Santé (les membres du GDR-STIC Santé peuvent demander ici un ordre de mission pour la journée et avoir leurs frais pris en charge)
- Contacter Martine Trani pour la prise en charge de votre mission (mtrani@ibisc.fr). Vous devez remplir un ordre de mission avant la mission en joignant votre RIB ainsi qu'un état de frais au retour de la mission avec tous les papiers justificatifs. Télécharger les documents pour la mission.