The aim of a Brain Computer Interface (BCI) is to translate EEG features from scalp or invasive recording into a command for a specilized device (most often using bio-feedback). BCI for communication purpose is not a recent idea, there is indeed no need of muscular activity to use it that is why the BCI stands for a real alternative for restoration of control and communication for severely disabled people. One can notice that even if BCI systems are an attractive field of study for computer sciences laboratories, we will still hardly find a usable BCI system among the users (typicaly Locked-In Syndrom person). We have identified some raesons to that: 1. lack of robustness : a. of the measurement : stability of the EEG signal quality durnig the time, bad artefacts processing b. of the application : signal processing for a given function, no adaptation of signal processing techniques with time, no aplication that fit with BCI needs) 2. lack of ergonomics : complexe and time consuming setup of amplification systems and softwares, need of techniciens for setup and during the use, very bad "acceptability" from patients who do not want to look like "cyber-man", and at lst but not least very unconfortable Human Machine Interface looking like old MS-DOS console As a consequence of what, despite a growing number of publications in that field of study, there is still no BCI system for daily use that fits to user's needs. The aim of our project can be defined as follow : 1. gather main actors of BCI signal processing pipe and propose true key innovations to raise robustness needed to transfert BCI technology from bench to bedside. 2. work close to patients in order to propose a system which will truly fit to their needs. At the end of the project, we will get, after a 36-month-long experimental developpement, a functionnal prototype of virtual keyboard for communication called RoBIK, commanded through a BCI system which will be robust and avaible for a daily use by patients.

Dysfunctions of the central nervous system are a major economic and social issue. Neural prostheses and brain-computer interfaces offer promising perspectives to restore motor functions and communication capabilities in patients suffering from severe paralysis. These approaches require the implantation of arrays of microelectrodes offering the possibility to record brain activity with stability on the long term. However, to date, the fabrication of brain implants housing a large number of microelectrodes and offering a stable connection with the neural tissue on the long term remains impaired by two major limitations. The first one stems from the electrode material itself when the size of the electrodes becomes small. Noble metals such as Platinum or Iridium have been used for decades to make macroscopic electrodes, which are now used in routine for neural recording and stimulation in several clinical applications such as cochlear implants, deep brain stimulation for Parkinson disease, and also the pre-surgical functional evaluation of epilepsy. However, thanks to the development of microfabrication technologies, the past decades have seen the development of new types of implants housing tens or even hundreds of microelectrodes on the micrometer scale. Yet, when the size of the electrodes diminishes, two problems arise: The intrinsic (thermal) noise level of the electrodes increase, and their safe charge injection limits decrease, which prevents delivering sufficient currents to activate neural networks without inducing lesions due to electrochemical reactions at the electrode/tissue interface. In this context, the first goal of the NeuroMeddle project will be to consider new types of materials based on the electrodeposition of pure or doped PEDOT/PSS to develop electrodes with improved performance and stable on the long term. A second main problem of existing brain implants (for instance like the Utah array) is the instability over days and even the loss of neural signals along time after a few weeks or months. This is especially the case for action potential signals, either of single or multiple units. This instability is mostly due to the combined effect of the movements of the brain and the inadequacy of the rigidity of implant materials versus the soft properties of the brain tissue. For this reason, an important line of research worldwide is the development of flexible implants matching the geometry and the mechanical properties of the brain, while still compatible with intracortical recordings. In this quest, an important open challenge remains to find strategies to insert flexible microelectrodes so that they meddle into the brain to create intimate and stable connections with individual neurons on the long term. Hence, the second goal of the NeuroMeddle project will be to develop implants offering such conformational stability, based on transient rigidification of flexible electrodes using biodegradable embedding materials (PEG, PLA, Chitosan, Silk fibroin) for the time of implantation. We will particularly focus on silk fibroin, which offers high rigidity and is not yet used in Europe for this type of application, while well mastered by one of the NeuroMeddle partner. The new electrode materials based on PEDOT/PSS, as well as biodegradable materials will first be tested in vivo in the rat. In a second step, we will consider another model closer to the human brain in order to face similar problems of stability of implants. We will use a paradigm for cortical recordings underlying vocalization in the awake mini pig, which allows to test the stability of long term recordings of unit and multiunit signals using the new conformational implants developed in the project.