The NiConnect project: understanding the brain, even in a resting state
© Inria / CEA, Neurospin
NiConnect is a Future Investments project that just been approved a few weeks ago. Gaël Varoquaux, the project's young promoter and member of the Parietal team, explains the aims of this research programme in bioinformatics.
Can you present the area of expertise concerned by NiConnect?
The project concerns questions in the fields of neuroscience and medicine. Our basic tool is therefore a standard MRI for examining brain activity. The first step consists of recreating a "film" using different snapshots of the brain provided by the MRI in functional imaging mode. This means we can see the oxygen consumed by each part of the brain. We can also determine which activated areas consume more, how long each zone is active, the intensity, etc. Many experiments in cognitive sciences are based on a stimulus (thinking about a particular action or object) and an analysis of the subject's brain in reaction to it. As a result of these experiments, it is possible to determine a functional architecture of the brain.
What is unique about NiConnect is that it analyses the brain in a resting state while it executes "background tasks”, according to Andreas Kleinschmidt, an Inserm researcher and member of the project team. In the absence of stimuli, there is still an architecture that manages the brain. Fortunately, when we shift into sleep mode, the brain does not just switch off. However, there have not been many studies of how the areas that govern the normal execution of these activities are activated. Our role is to offer computing and statistical tools to extract these resting state architectures and analyse their variability in a quantitative way. This is important research in applied mathematics and the NiConnect project represents the result of many years of work on the subject.
Working on the brain in its resting state allows us to help
people who otherwise would not easily have access to an MRI
We can question the interest of analysing the way the brain works in its resting state, especially because the subjects, since they do not receive instruction, are free to think of many things, making analyses even more difficult. However, it is possible to perform our analyses on sedated subjects, for example, on a patient with Parkinson's disease who would otherwise have difficulty controlling his or her movements and could not have an MRI. Our goal is to help people who would not generally have access to this type of test. In the long term, we hope to extend our work to people in an unconscious state and perhaps improve prognoses for coma patients, for example.
Can you tell us in more detail about the different research themes of NiConnect?
The project is organised around several stages. First, we need to produce a software programme that clinical researchers can use to analyse MRI images. Above all, we need to consolidate the available basic research and create a robust tool based on all of this knowledge. This means we need to recruit three engineers over the next four years. After that, when we have capitalised on existing research, we want to develop new models and algorithms for modelling the resting state. For this part, we are working with the CATI (Centre Analyse et Traitement d’Images ) and the CEA, which is a data processing centre for the French Alzheimer Plan, as well as a small scientific computing firm called Logilab.
During a second stage, we are going to automatically analyse a significant database of standard cerebral images in order to calibrate our models. More specifically, we want to automatically analyse these images in partnership with the Inserm functional imaging laboratory at Pitié-Salpêtrière Hospital in Paris and try to detect anomalies, but only those resulting from pathologies. In an image, we should be able to distinguish a real neurological problem from a scanner malfunction. By correcting the algorithm that reacts poorly in the second case, the software will gradually allow practitioners to focus exclusively on real problems.
Cerebral areas and their interactions understood using MRI of the resting brain, © Inria
Finally, the third theme is our participation in an on-going clinical trial dedicated to Parkinson's disease conducted at Henri Mondor Hospital with the neurosurgeon Stéphane Palfi. This is a test for a new treatment, which has been validated with several patients. Today we have reached a more important stage, with a trial involving around twenty patients. It involves a large number of different measurements, including cerebral images. There is a lot of "background noise" in the data, that is to say it is hard to analyse because of all the interference, which is a real statistical hurdle. Our goal in this particular case is to offer doctors more reliable information so they can make the right decisions.
The first stage allows us to create the tool, the second to calibrate it, and the third to test it in a real situation. This is a comprehensive project where the different stages follow each other, but with periods that overlap to allow for feedback, adjustments, and review of previous points. This is very important because we are communicating with people with very different cultures, and we need to understand each other in order to achieve a result that works. Another fundamental point is that all the partners know each other already. We are all from the Paris area so we can meet on a regular basis and find a way forward.
Being in charge of such a project right after being recruited as a research scientist is rather uncommon, isn't it?
Yes, and I am still amazed! I have had a rather atypical career path, as I started out in physics. My first post-doctoral work in Florence was on quantum physics! Then I wanted to tackle more open-ended problems and work on complex systems like the brain. I took some time to think about the next step in my career and left for the United States, where I worked in a private scientific software firm, which happened to make applications for neuroscience. When I returned to France, I contacted the Parietal team in order to do another post-doc project. I then did a third one to perfect my mastery of the field, once again at Neurospin, but this time in an Inserm team. And now I've been hired on in Bertrand Thirion's team.
This winding path has given me special skills , particularly modelling in physics and conducting experiments on real problems.
The fact I am the youngest member of the team
has not stopped me from leading the project,
because my background made me the best person to coordinate NiConnect.
But I am still amazed because I was just recruited in October 2011, and the deadline for submitting the project was the beginning of November! Work had started before that of course, but it was still a lot of responsibility to take on so soon. Being the youngest member of the team hasn't stopped me from taking the lead, because this really is my area of expertise. That does not mean there aren't more senior members, who are also working on the project, it's just that my particular career path made me the best person to coordinate it.
We knew we would be selected because we are addressing an important problem and we had formed a strong partnership. But nothing was certain, because the project is rather atypical, since traditionally research in statistics and algorithms consist of developing algorithms and models, and we are taking this much further. In our project, we are saying we need to go beyond this to have an impact, that is to say we need to go out in the field and make a real effort in software engineering so that we can make it work with a large mass of data. So we are delighted the project was accepted, because this means we have the means to consolidate and apply real data. This is just the beginning. The project was only approved a few weeks ago, and the team will be finally up and running when we have recruited the engineers and post-doc researchers we need.
These articles could interest you:
Find out more
Gaël Varoquaux , research scientist in Parietal research-team