Post-Doctoral Research Visit F/M Campagne Post-doctorant : Higher-order approaches in brain-computer interfaces

Brain-computer interfaces (BCIs) enable interaction with the external environment by decoding individuals’ mental intentions. BCIs find applications in addressing basic neuroscience questions and in practical use cases, from exoskeleton control to neurofeedback rehabilitation. In general, BCI usability relies on the ability to characterize brain functioning and correctly classify users’ mental states. Despite significant progress, BCI performance remains inconsistent due in part to overly simplistic brain function decoding descriptors. In this context, network models have arisen as potent tools for investigating new brain organizational dynamics and augmenting BCI classification accuracy. While powerful, network models are limited to describing pairwise interactions, whereas real-world dynamics often involve higher-order interactions (HOIs) involving groups of three or more units. Recent approaches have started to infer HOIs from temporal brain signals, like functional magnetic resonance imaging, showing potential in distinguishing patients’ consciousness states and linking HOIs to ageing. Yet, to date, little is known about the HOI effects on BCI in healthy and clinical populations.

[1] Wolpaw et al, 2002, Clin Neurophysiol 113: 767–91. [2] Gonzalez-Astudillo et al, 2021, J Neural Eng 18: 011001 [3] Battiston et al. Nat. Physics 17.10 (2021): 1093-1098. [4] Gatica et al. Brain connectivity 11.9 (2021): 734-744 [5] Luppi et al. Nat. Neuroscience 25.6 (2022): 771-782.

The project aims to characterize the importance of HOIs in brain functional data of both healthy individuals and stroke patients in BCI. Using recent higher-order frameworks, the postdoctoral researcher will need to analyze HOIs from brain signals measured by Magneto/Electroencephalography (M/EEG). By investigating the potential of these HOIs to enhance BCI accuracy among healthy subjects, the project aims to uncover the underlying learning mechanisms of these systems and improve our ability to decode various mental states. Ultimately, the project seeks to integrate HOI statistics derived from M/EEG data of stroke patients engaged in BCI tasks into predictive models for better prediction of their post-stroke rehabilitation outcomes.

• PhD, preferentially in Applied Mathematics or Physics;

• A strong background in computational modeling, information theory, complex systems;

• Have good programming skills in at least one of the following languages: Python, Julia, Matlab, C;

• Knowledge of computational neuroscience or neuroimaging is a plus;

• Fluency in English both spoken and written.