Marie-Constance Corsi, an explorer of cerebral intention

Marie-Constance Corsi’s exploration of the mysteries of the human brain has been far from a linear journey. Her academic career, marked by forays into physics and telecommunications, bears witness to her attraction for multidisciplinarity. It was at Télécom Bretagne, now IMT Atlantique, that she discovered the power of interdisciplinary collaboration: “Interacting with people from different backgrounds is incredibly enriching”, she confides. “It enables you to approach problems from new angles and find innovative solutions.” 

Driven by her desire to apply her knowledge to the biomedical field, Marie-Constance Corsi’s engineering studies culminated in an Information and Communication Technologies for Health (ICT & Health) course run by Institut Mines Télécom, the University of Montpellier and Mines d'Alès, in which engineering students work alongside medical and paramedical professionals.

Her Master's degree in neuropsychology and clinical neuroscience, obtained at the same time as her PhD, complements her scientific background and gives her a unique perspective on how the brain works. This dual skill set, combining technical expertise with an understanding of cognitive mechanisms, has proved invaluable in her work on brain-computer interfaces (BCIs), which translate brain activity into commands capable of controlling a computer, a prosthesis or any other automated system.

Her thesis was defended in 2015 at Grenoble Alpes University and focused on the development of optically-pumped helium-4 magnetometers. These extremely sensitive sensors are capable of measuring the magnetic fields generated by the heart and brain. “The aim was to replace bulky, expensive and dangerous cryogenic systems with more compact, higher-performance devices that operate at room temperature”, explains Marie-Constance. A major technological challenge, requiring the reduction of background noise and an improvement in sensor sensitivity to obtain more accurate and reliable measurements. This work has paved the way for new applications in the study of brain function, since, at the end of Marie-Constance’s thesis, her supervisors founded a start-up company: Mag4Health. 

Today, her efforts are focused on the effectiveness of brain-computer interfaces, particularly for patients who struggle to control these devices. “One of the main obstacles to the use of BCIs is that some patients struggle to master them”, she stresses. Therefore, the aim is to decipher the subject's “intention” by taking account of their specific needs in order to provide them with the most relevant information. 

To this end, Marie-Constance explores innovative neurophysiological markers, such as a neurophysiological manifestation called “desynchronisation”, which is linked to motor imagination. This phenomenon is characterised by a reduction in brain signal power when someone imagines or performs a movement. This marker is traditionally used in BCI, but has one drawback: it only provides information on activity in a specific brain area, without taking account of interactions between the different areas. After several projects to identify biomarkers of BCI performance and develop new Riemannian classification approaches based on traditional functional connectivity metrics, Marie-Constance started studying “neuronal avalanches”: cascades of brain activity that could further improve the design of BCIs...

What exactly is a neuronal avalanche? It is a phenomenon of rapid, aperiodic propagation of neuronal activity throughout the brain. These avalanches are defined as cascades of activity in neural networks, whose size distribution can be approximated by a power law. And what is interesting is that the manner in which these “avalanches” propagate changes according to the activity being performed, such as whether the subject is at rest or imagining a particular movement.

Video showing the real-time propagation of neuronal avalanches from magnetoencephalography (MEG) recordings. Credit: P. Sorrentino.

In an article published in iScience (Measuring neuronal avalanches to inform brain-computer interfaces, M.C. Corsi et al) in 2024, Marie-Constance and Pierpaolo Sorrentino, from the Institut des Neurosciences des Systèmes (INS) in Marseille, studied the role of these neuronal avalanches as possible markers for improving the design of BCIs. Their aim was to track the probability of avalanche propagation between two brain regions and to construct an avalanche transition matrix (ATM), a type of “map” that indicates, for each pair of regions (X,Y), the probability of region Y being active at time t+1, given that region X was active at time t. The team then compared these transition probabilities for the resting and motor imagery conditions in order to identify significant differences and assess whether these properties could enable the decoding of tasks.

Comparing these matrices enabled Marie-Constance and her colleagues to identify significant differences in the manner in which avalanches propagate. Even more importantly, the properties of these neural avalanches enable tasks to be decoded more accurately than with conventional methods, and with less inter-individual variability. These results suggest that their properties could not only help to improve the interpretation of users’ intentions but also open up new opportunities to improve the performance and adaptation of BCIs to suit each individual. These aspects are currently being studied by Camilla Mannino, a doctoral student in the team. 

Within the NERV team, Marie-Constance works on a daily basis with computer scientists, neurologists and specialists from other disciplines. “These constant interactions are key to furthering research,” she says. “They enable us to compare ideas, break down barriers between disciplines, and ensure that the tools developed meet the actual needs of patients, carers and healthcare staff.”

HappyFeat – software produced as the result of collaboration between different disciplines – is designed to facilitate the use of BCIs in clinical settings. It provides doctors with a decision-making tool for personalising patient training. In practical terms, HappyFeat, whose lead developer is Arthur Desbois, takes the raw data recorded by a BCI system and extracts the relevant information from it, such as the brain signal power or the way in which different areas of the brain communicate with each other. The software then suggests several methods for controlling the machine. Once these characteristics have been selected, it trains a machine learning algorithm, called a “classifier”, which learns to recognise the patterns of brain activity associated with a particular intention, such as moving one’s right hand.

What reading matter would she recommend to anyone wishing to go into neuroscience? My Stroke of Insight (2008), by Dr Jill Bolte Taylor, the testimony of a neuroscientist who survived a stroke, which provides a unique insight into how the brain works. And for anyone interested in several fields of research, Marie-Constance suggests Chaos: Making a New Science (1989) by James Gleick: “An interesting book, an ode to multidisciplinarity. It highlights the difficulties of compartmentalisation between scientific disciplines, even though certain fields rely on common mathematical models”, she explains. 

“Cultivate your curiosity! You need to reach out to others and seize the opportunity to broaden your horizons by taking part in conferences and workshops. Research is above all a human adventure. This passion, perseverance and sharing enable us to push back the boundaries of knowledge”, concludes Marie-Constance.

After her telecommunications engineering studies, Mary-Constance Corsi was keen to apply what she had learned to health issues. She studied for her PhD in Biomedical Instrumentation on the latest generation of magnetoencephalography (MEG) sensors while simultaneously completing a Master’s in Clinical Neuroscience. In 2016, she started as a postdoctoral researcher in the ARAMIS team at Inria, with a focus on proposing new computational methods to improve brain-computer interfaces (BCI). She then joined the NERV joint project team as a researcher. She is currently focusing on developing tools to reduce the proportion of BCI users who are unable to control their device even after several training sessions, and on developing diagnostic tools for neurological pathologies.