Closed-loop BRAIN computer interface APPlication for enhanced cognition. – BrainAPP

In this study, two macaque monkeys were trained to perform complex behaviors allowing the isolation of spatial orientation of attention, while multi-electrode neural recordings were performed in the prefrontal cortex (PFC), bilaterally. In a closed-loop neurofeedback experiment, we used Brain-Machine Interface approaches that exploit machine-learning methods to quantify cortical information on single behavioral trials, with a high temporal resolution, and we provided the monkeys with real-time feedback of their attentional cortical information while performing the task.

At the same time, the same developments are being carried out in intractable implanted epileptic patients, as well as non-invasively in healthy subjects with fMRI or MEG techniques.

In addition to increased readout performances of spatial attention, real-time neurofeedback led to remarkable changes in the neural dynamics yielding mixed behavioral outcomes whihc we are currently characterizing.

Primates are able to increase their attentional abilities through a closed-loop neuro-feedback device, which allows the subject to observe his attentional cortical activity and modify it by his own will. These results have important implications for increasing the cognitive abilities of target populations.

1. Astrand E., Wardak C., Ben Hadj Hassen S., Baraduc P., Ben Hamed S. (2015) A closed-loop Brain-Machine Interface (BMI) for enhancing visuospatial attention in the non-human primate. Congrès de la Société des Neurosciences, Montpellier, 2015.
2. Ben Hamed S, Wardak C, Astrand E (2015). Real-time tagging of visual, saccadic, spatial memory and attention prefrontal representations. Congrès de la Société des Neurosciences, Montpellier, 2015
3. Di Bello F, Ben hadj Hassen S, Astrand E, Ben Hamed S (2016). Distractor suppression and distractor interference in the light of direct real-time access to the covert attentional spotlight from the frontal eye fields. Annual meeting of the Society for Neurosciences, San Diego, USA
4. Ben Hamed S, Wardak C, Astrand E (2016). Real-time tagging of visual, saccadic, spatial memory and attention prefrontal representations. Annual meeting of the Society for Neurosciences, San Diego, USA.

Organisation de symposium
Symposium retenu au sein de la société française des neurosciences, intitulé : ‘Interfaces cerveau-machines cognitives : des neurosciences fondamentales aux applications cliniques.’
Organisatrice : Suliann Ben Hamed, représentant BrainApp.
Intervenants : Elaine Astrand, représentant BrainApp, Klaus Pawelzik, Bremen, Allemagne, Nick Ramsey, Utrecht, Netherlands, Karim Jerbi, Montréal, Canada, Jérémie Mattout, Lyon, France.

At the junction between neuroscience and computer science, the field of brain machine interfaces (BCIs) has achieved, in the last decade, a remarkable growth. Its general objective is to assist, augment or repair human cortical functions.
The focus of the present project is on a subfield of BCIS, namely, invasive BCIs and neural prosthetics. Up to now, the general aim of this rapidly developing research area has been to use preserved electrophysiological nervous activities in order to counter specific dysfunctions or deficits by driving external palliative devices such as a cursor, a robotic arm or a wheelchair, thus restoring mobility and independence to patients with central or peripheral motor disabilities. One of the most representative examples of a palliative neural prosthesis is the use of the neuronal activity of the motor cortex of a tetraplegic patient to help her control a computer driven environment, thus allowing her a certain degree of assisted mobility and independence.
An important advance in the field is the demonstration that other regions than the motor cortex can be used to drive motor BCIs, such as parietal cortex, or dorso-lateral prefrontal cortex thus providing a potential substitute of motor cortex activities when these are not available, following for example an acute injury of this region. Another important finding is the fact that incorporating sensory feedback to a motor BCI improves its performance. More recently, researchers have demonstrated that motor cortex signals from one individual could be used to generated informative spinal cord stimulations leading the control of the arm of a second anesthetized individual.
By contrast, BCI research that relies on decoding higher-order cognitive processes such as attention, perception, intentions and decisions is still in its early days. A recent study shows that, in the context of a motor behavior, cognitive signals such as the expected value of a reward, i.e. the subject’s motivation, can be decoded from parietal neural activity. Another set of data demonstrates that such signals as attention engagement signals and mental calculation signals can be used to drive a cognitive BCI. Recent studies including our own work demonstrate the feasibility of real-time decoding of attentional processes. Last, our own work demonstrates that we can decode simultaneously, from cortical activity, in real-time, both the structure of the visual environment and the perception the subject has of this environment.
Here, we propose to tackle a barely scratched aspect of invasive cognitive BCIs, namely closed-loop invasive BCIs for augmented and/or restored cognition. In other words, we aim at training subjects to use the cognitive information content of their neuronal signals as a feedback to enhance this cognitive content and hence cognitive performance.
These experiments will be carried out in an animal model of human cognition as well as in surgically implanted patients. Specifically, our work will aim at four objectives:
1. Design a closed-loop invasive BCIs for augmented cognition in an animal model, i.e. provide the animals with a feedback on their cognitive content, use reinforcement learning to have them improve this cognitive content, and as an end product improve their cognitive overt performance, that is to say their behavior.
2. Design a closed-loop invasive BCIs for restored cognition, i.e. induce a cognitive loss thanks to reversible cortical inactivations and use the procedure developed in step 1 to restore cognition.
3. Design a closed-loop invasive BCIs for augmented cognition in humans. This step will be performed in pharmaco-resistant surgically implanted epileptic patients, as a first step towards translating our observations to non-invasive cognitive BCIs applications.
We believe that this work will pave the way to a new generation of intelligent BCIs with a diversity of civil, defence and clinical applications.