Neuroscience

Brain-computer interfaces: a multidisciplinary project team seeking to help people with paralysis

Date:
Changed on 26/01/2024
NERV is a brand new joint project team combining networks, computer science, neuroscience and mathematics with the aim of perfecting brain-computer interfaces in order to help people with paralysis. We caught up with researchers Fabrizio de Vico Fallani and Marie-Constance Corsi to chat about their ambitions to drive neuroengineering forward and improve the performance of algorithms.
231228-image-chapo-NERV

Developing more reliable brain-computer interfaces

NERV is a new project team that was launched back on 1st October. A joint undertaking involving Inria, Sorbonne University, Inserm and the CNRS, located at the Paris Brain Institute (ICM), it was set up with the goal of improving the performance of brain computer interfaces (BCI), which enable people to control computers, artificial limbs and other automated systems using their thoughts.

“We currently have algorithms capable of detecting cortical signatures corresponding to individual tasks, such as imagining forming a fist, explique Fabrizio de Vico Fallani, head of the NERV team. But the level of precision is only 70 to 80%. When you're controlling a wheelchair and you want to cross the road, you can’t have any errors, which is why we’re seeking to get as close as possible to 100%.” Marie-Constance Corsi, researcher with the NERV team, adds: “15 to 30% of users are unable to control the tool, even after a lot of training. This can be down to the machine, or the individual. What we want to do is to identify markers which factor in the unique nature of people's brains.” 

A new team would be needed to tackle such a vast project. This was made possible in large part through the ERC Consolidator Grant awarded to Fabrizio de Vico Fallani, providing two million euros’ worth of funding for a period of seven years. The Nerv project team has set up shop at the Institut du Cerveau (ICM), an Inria partner laboratory, and is very much in keeping with the joint research strategy the two institutes launched in 2021 aimed at developing computational neuroscience, neuro-engineering and data science.

One team representing more than six different disciplines

Of course, interfaces and interdisciplinarity go hand-in-hand. The NERV team work with computer scientists, mathematicians, physicists and engineers, as well as neurologists, psychologists and physiotherapists, with a vibrant mix of researchers, PhD students, postdoctoral researchers and clinicians.

 

Photo extérieure de l'entrée de l'institut du cerveau.

The Institut du Cerveau (formerly the Institut du Cerveau et de la Moelle, Brain and Spinal Cord Institute) is a research centre which brings together patients, doctors and researchers, the goal being to enable the rapid development of treatments for nervous system injuries which can then be used on patients.
 

“The ICM provides a unique environment within Pitié-Salpêtrière Hospital, enabling us to interact with clinicians and to gather data on both healthy people and patients”, says Marie-Constance Corsi. NERV will seek to combine data analysis, mathematical modelling and experiments. “It’s a somewhat unusual environment for an Inria researcher as we’re surrounded by neuroscience”, says Fabrizio de Vico Fallani. “But 360° discussions are hugely beneficial: it’s like mathematicians are becoming neurologists, and vice versa.” 

 

Supplying algorithms with better data

Seeking to improve BCIs from both a machine and a human perspective, the team will spend part of their time working on the data to be integrated into the algorithm and the rest of their time on the neuronal mechanisms specific to each individual. “With regard to the former, the performance levels of algorithms are not yet high enough as a result of the characteristics they look for”, says Fabrizio de Vico Fallani. “Instead of observing each region of the brain individually, it might make more sense to look to see how they activate together, viewing the brain as a network.” 

 

Capture d'écran du logiciel interactif Vizaj qui visualise des réseaux spatiaux en 3D.

Screenshot of Vizaj, a free online graphic interface for the visualisation of spatial data in 3D (doi: 10.1371/journal.pone.0282181). NERV developed Vizaj in 2022 for spatial networks for which visualisation is more complex than for standard, non-spatial graphs. Long-term Vizaj will be integrated into Happyfeat as an interactive visualisation tool for use in BCIs (Happyfeat, a Python open-source software program, will be available in 2024).

 

The team will conduct an in-depth analysis of brain interactions using electroencephalography (EEG), magnetoencephalography (MEG) and new sensors which Marie-Constance Corsi has been working on since her PhD. The aim is to enhance the data available to algorithms, thereby improving their ability to understand people’s intentions.

 

Photo d'une femme assise devant trois écrans d'ordinateurs, vue de dos. Elle porte un casque d’électroencéphalographie

Electrical signals emitted by the brain are recorded using an electroencephalography (EEG) headset, before then being processed by software and interpreted into commands (brain-computer interface).  As a result it is possible to move a cursor around a screen simply by imagining a movement, which produces brain activity similar to that produced during actual movement. © Inria / Photo C. Morel.
 

The NERV team will then turn their attention to researching biomarkers for performance and learning, the goal being to be able to adapt the training strategy and programme to individual patients.

Verbatim

We showed in a 2020 study that when someone is learning how to use a BCI, their brain modifies the way in which the different regions of the brain communicate with each other. So if we train the algorithm using the starting characteristics, it won't work: you need data which tracks changes to performance levels.

Auteur

Marie-Constance Corsi

Poste

Researcher with NERV

A direct application for patients recovering from stroke

Long-term, enhanced BCIs could have a significant impact on the lives of people with paralysis, including those recovering from stroke.

 

Une personne, avec des électrodes sur la tête, est assise devant une plateforme BCI comprenant un bras motorisé.

NERV is responsible for coordinating the development of the BCI platform at the Institut du Cerveau in Paris. This platform is for designing and experimenting with innovative prototypes in areas such as multimodal control (Tobii Pro Glasses), robot end effectors (Pollen Robotic 7 dof), and augmented reality (Hololens 2 microsoft).
 
Verbatim

The brain can be highly flexible. For around a year after a stroke tasks which were previously carried out by the motor areas of the brain can be handled by other brain regions. But for most patients this isn't enough, and that’s where BCIs can be really useful in promoting brain rehabilitation.

Auteur

Fabrizio de Vico Fallani

Poste

Head of the NERV team

What the researchers want to do is to identify the brain signals emitted when a patient imagines completing a particular movement, before then using the algorithm to get a muscle prosthetic to carry out the movement. In conjunction with transcranial magnetic stimulation this closed circuit will make it possible to reopen the “plasticity window” and complete functional rehabilitation, thereby increasing the chances of recovery. This experiment is set to be launched on 50 patients in 2024.

Potential applications in other fields

NERV is also focused on the technological development of BCIs, and has already developed Happyfeat, a software program designed to make life easier for clinicians looking to use them. “This software can be used to visualise, personalise and save certain settings in order to ensure the reproducibility of experiments”, explains Marie-Constance Corsi. An open-source version of Happyfeat is scheduled for release in 2024 for use by researchers, doctors and manufacturers of medical devices. 

 

Infographie représentant les différentes caractéristiques du logiciel HappyFeat : les extractions de données, la visualisation, le tri ou encore la configuration du BCI.

HappyFeat is an open-source software program written in Python which is designed to simplify the use of BCI pipelines in clinical environments, in addition to helping researchers to introduce network and graph-based approaches into BCIs through the use of characteristics based on functional connectivity. An open-source version of Happyfeat is scheduled for release in 2024 for use by researchers, doctors and manufacturers of medical devices.

 

But the potential applications of the research carried out by NERV are not limited to neuroscience. The algorithms they have developed could be used in other disciplines involving the modelling of complex systems and networks, from energy and communications to the environment, public health and genetics. “We’re working on collaborations in some of these fields, but we haven’t lost sight of our overarching goal of improving BCIs”, concludes Fabrizio de Vico Fallani.

Titre

“This collaboration will ensure that patients don’t have to wait long to benefit from breakthroughs in research”

Verbatim

Working with NERV assists with the formulation of basic research questions with their roots in clinical issues, while making it easier to adapt technical solutions to patient needs. Our objective is to provide patients recovering from chronic stroke with more effective and personalised rehabilitation strategies, thereby improving their quality of life and their functional independence.

Auteur

Camile Bousfiha

Poste

Neurologist with NERV

Fabrizio de Vico Fallani - brief bio

Portrait de Fabrizio de Vico Fallani
“It was Mario Chavez, a CNRS researcher with Nerv, who brought me to the ICM as a postdoctoral researcher in 2013. Then in 2014 I secured a Starting Research Position within ARAMIS, a joint project team at the Inria Paris centre. While continuing to work at the ICM, I became a contracted researcher at Inria in 2017. My research was focused on both the theoretical side of things - modelling the brain as a network - as well as the more experimental, seeking to improve BCIs. Setting NERV was the logical progression after the previous ten years.”

 

Marie-Constance Corsi - brief bio

Portrait de Marie-Constance Corsi
“After studying telecommunications engineering I was keen to apply what I had learned to the field of health. I studied for a PhD in Biomedical Instrumentation on the latest generation of MEG sensors, alongside a Master’s in Clinical Neuroscience. In 2016 I joined the ARAMIS project team at Inria as a postdoctoral researcher where I worked on new computational methods aimed at improving BCIs, which is how the collaboration with Fabrizio came about.”

 

Find out more