Résumés 2023-Inferring

Hugues Berry (https://www.inrialpes.fr/Berry/) is a research director, (i.e., a research full professor, with tenure) with Inria, the French national institute for research in digital science and technology. A computational neuroscientist, his research focuses on mathematical and computer models of the spatiotemporal dynamics of biochemical reactions involved in brain cells. In recent years, he has been focusing on the modelling of the reaction networks that support synaptic plasticity, especially endocannabinoid plasticity. He has also proposed models for the metabolic and signalling networks at play in astrocyte activity and in astrocyte-neuron interactions. From 2018 to 2023, he has served as deputy scientific director of Inria, in charge of the research in digital biology and health, implementing Inria’s strategy on the application of numerical sciences (applied mathematics, computer science, artificial intelligence) to biology and health. Since 2023, he is co-heading AIstroSight, a joint research group between Inria, the Hospices Civils de Lyon (Lyon University Hospital), and Université Claude Bernard Lyon 1. The overall goal of AIstroSight is to develop innovative numerical methods for neuropharmacology, the search of new drug candidates to treat brain diseases.

That astrocytes are important for brain functions has been known for long: they provide energy support to the neurons via neurovascular coupling, they also contribute to the blood-brain barrier. Nevertheless, more recent discoveries have led to realize that their interactions with the neurons are probably more intimate. In addition to their regulation of local potassium levels, they emit fine cytoplasmic processes that interact with synapses using the same types of mechanisms as those that support information transfer between the presynaptic and postsynaptic elements. This led to the emergence of the concept of a « tripartite synapse” as a new functional unit for information transmission. However, astrocytes are not easy to study for the experimentalist. They cannot be easily characterized by electrophysiology because they lack action potentials. And a substantial part of these cells is too thin to be observed by conventional light microscopy. As a result, a number of issues are still unclear: is intracellular calcium the support of astrocyte activity? If yes, in what part of the cell and with what integration rules? Do astrocytes release chemical messages, called gliotransmitter, that directly affect the synapses in vivo? Are there different subtypes of astrocytes, functionally? Is there a correlation between the (local) shape of the cell and the (local) function it performs? What metabolites do they provide to the neurons, and under what circumstances? Of course, the answers to these questions are crucial when one wants to build mathematical models for the effects of astrocytes. In this lecture, my goal is to give an overview of these open questions, and show how they are related to the ongoing effort toward the mathematical modelling of astrocytes and their interactions with neural networks.

Up-Down synchronization of neuronal networks is a regime of collective dynamics whereby a neuronal population spontaneously switches between periods of high collective firing activity (Up state) and periods of silence (Down state). Recent experimental evidence has reported that astrocytes can control the emergence of such Up-Down regimes in neural networks, although the involved mechanisms are uncertain. To explore how astrocytes can control this phenomenon, we have studied neural network models made of three populations of cells: excitatory neurons, inhibitory neurons and astrocytes, interconnected by synaptic and gliotransmission events. Using simulation and bifurcation analysis, we show that the presence of astrocytes in these models indeed promotes the emergence of Up-Down regimes with realistic characteristics. The difference of signaling timescales between astrocytes and neurons (seconds versus milliseconds) can induce a regime where gliotransmission events released by the astrocytes alter the localization of the bifurcations points in the parameter space. In particular, the addition of astrocytes strongly enlarges the bistability region that gives rise to Up-Down synchronization. As a result, Up-Down regimes can easily be observed with astrocytes for parameter values that do not exhibit it in their absence. Taken together, this work provides a theoretical framework to test scenarios and hypotheses on the modulation of Up-Down dynamics by gliotransmission from astrocytes and some perspectives, in particular regarding epilepsy.

Dr. Joana Cabral is a distinguished researcher in the field of Theoretical and Computational Neuroscience.  With a background in Biomedical Engineering her research focuses on understanding the fundamental principles of brain function and their implications for psychiatric disorders. With a multidisciplinary approach, Joana combines advanced analytical tools and large-scale computational brain models to investigate the mechanisms supporting cognition. Joana has made significant contributions to the field, including the LEiDA algorithm, which identifies key features in whole-brain dynamics related to cognitive and behavioral conditions, and has been recognized for her achievements, including receiving the prestigious 2019 L’Oréal Award for Women in Science Portugal. Her research has provided new insights in our understanding of brain function at the macroscopic scale.

Brain activity exhibits chaotic signals comparable with the ones observed in networks of delay-coupled oscillators. On one side, transient brain rhythms are detected in EEG/MEG signals. On the other, fMRI reveals slow and spatiotemporally organized signal fluctuations. Using mathematical models of coupled oscillators, I will show how the network system can engage in a critical regime where intermittent cluster synchronization can generate signals sharing qualitative and quantitative features with human brain activity.

Spontaneous fluctuations in functional magnetic resonance imaging (fMRI) signals correlate across distant brain areas, forming functional networks that appear disrupted in numerous psychiatric and neurological disorders, pointing to a key role in brain function. However, the generative mechanism of fMRI signal correlations are not fully understood. In this talk, I will give an overview of the evidence gathered applying Leading Eigenvector Dynamics Analysis (LEiDA) on different neuroimaging datasets and the insights it provides to understand brain function. I will further describe recent insights on the origin of LEiDA patterns obtained from experiments in rodent. Overall, despite the convincing evidence, it is crucial to obtain a better mechanistic understanding of  functional newokrs and their organizing principles, in order to design informed strategies to rebalance long-range interactions between brain areas that appear disrupted in disease.

Maxime DESCOTEAUX, PhD is Professor in Computer Science since 2009 at the Science Faculty of Sherbrooke University and Scientific Advisor at Imeka solutions (www.imeka.ca). In 2021, he became a member of the college of the Royal Society of Canada. He is the founder and director of the Sherbrooke Connectivity Imaging Laboratory (SCIL) (http://scil.usherbroeoke.ca/), where there are 10-15 graduate and medical students, and postdocs. He holds the Research Chair in NeuroInformatics and his research focuses on brain connectivity from state-of-the-art diffusion MRI and multi-modality acquisition, reconstruction, tractography, processing and visualization. The aim of the SCIL is to better understand structural connectivity, develop novel tractography algorithms, validate them and use them for human brain mapping and connectomics applications. Pr Descoteaux holds the USherbrooke Institutional Research Chair in NeuroInformatics. He has been cited more than 14500 times, his h/i-10index are 59/174 and has 175+ journal publications (google scholar).

This talk will focus on the 160,000 km of human brain white matter, its integrity and how non-invasive diffusion MRI can be used to develop novel biomarkers sensitive to demyelination, axonal disruption and neuroinflammation. I will introduce the basics of the technology and show how to compute quantitative track-specific diffusion MRI markers that can be used for white matter assessment. I will present methods that combine cortical surface meshes with tractography reconstruction. Mapping diffusion MRI tractography to the cortical surface facilitates the integration of white matter features onto gray matter, especially for connectivity analysis, which can serve as the computational backbone for electrophysiological and functional human brain mapping.

In this talk, I will cover what tractography and tractometry can do well and what are some of the important current tractography limitations such as the length, position, shape and gyral biases, and will present some solutions that start addressing these limitations using machine learning, anatomical priors and advanced acquisition methods. The lack of ground truth on long‐range connectivity of the human brain makes it hard to quantitatively evaluate results. A key challenge for future tractography algorithms will be to control for false positives, while identifying the full extent of existing fiber bundles.

Christophe Grova is Associate Professor affiliated to the Department of Physics of Concordia University and a research member of PERFORM center since July 2014, while remaining adjunct Professor affiliated to Biomedical Engineering Dpt and Neurology and Neurosurgery Dpt at McGill Faculty of Medicine. He is also affiliated to the epilepsy group of the Montreal Neurological Institute (MNI), the McConnell Brain Imaging Center of the MNI and a member of Physnum team at Centre de Recherches Mathématiques. He received his Engineering and Master degrees in biomedical engineering at the University of Technology of Compiègne (France) in 1998, followed by a Ph.D. in SPECT/MRI registration at University of Rennes (France). From 2003 to 2008, his postdoctoral studies at the MNI were focussed on EEG source imaging of epileptic discharges and the correspondence with EEG/fMRI results, while acting as part time research associate for the set-up of the MEG centre of Université de Montreal (2006-2008). Dr Grova has been assistant Professor at McGill from July 2008 to July 2014. Since 2008, he is the director of the “Multimodal Functional Imaging Laboratory” (MultiFunkIm) which is now located on both McGill and Concordia campus. His areas of expertise are EEG/MEG source localization, multimodal data fusion involving EEG/MEG, fMRI and fNIRS, for application in epilepsy and sleep research. C. Grova is the scientific lead of the physiology platform at PERFORM promoting neuroimaging in realistic lifestyle environments using wearable technology (EEG, fNIRS). His team is also handling the development and validation of two software packages: MEM in Brainstorm for EEG/MEG source localization and NIRSTORM for fNIRS data analysis.

Localizing along the cortical surface the underlying generators of scalp electrophysiology measured with Electroencephalography (EEG) or Magnetoencephalography (MEG) consists in solving a challenging ill-posed inverse problem. The first part if the course will consist of an overview of the methods dedicated to solve this problem. With an excellent temporal resolution, EEG/MEG offer unique data to investigate communications/interactions within and between resting state networks, by localizing ongoing resting state activity. We will review the different metric that have been proposed to characterize brain networks in different frequency bands, either from EEG/MEG source imaging data or from in situ recordings using intracranial EEG data.

Accurate delineation of the epileptogenic zone (EZ) during presurgical workup of focal drug-resistant epilepsy patients can be challenging. Stereo-electroencephalography (SEEG) recordings, considered as the gold-standard for the localization of the EZ, might be the step towards mapping the seizure-onset zone (SOZ) and determining surgical candidacy. However, a successful investigation requires a strong pre-implantation hypothesis on the localization of the EZ, which can be derived from non-invasive investigations such as EEG/magnetoencephalography (MEG) and EEG/functional MRI (EEG/fMRI). In the present study, we propose a quantitative validation of the ability of EEG/MEG source imaging of transient epileptic discharges and simultaneous EEG/fMRI responses to similar discharges to localize the epileptogenic zone and epileptogenic networks. To do so, we compared EEG/MEG source imaging and EEG/fMRI responses to the Gold Standard provided by in situ SEEG measurements. We quantified concordance between all data, as spatial overlap and distances in the SEEG channel space. Using this objective quantification approach, we demonstrated that EEG/MEG source imaging was more sensitive to the main peak of epileptic discharges, whereas EEG/fMRI was more sensitive to the seizure onset zone, localizing regions where a large increase oxygen consumption can be expected at the time of epileptic discharges (Abdallah et al Neurology 2022). During the second part of the talk, I will present how brain network features estimating when analyzing resting state activity using EEG/MEG or EEG/fMRI could provide relevant information on the organization of underlying brain networks and how epilepsy specific reorganizations of brain networks might become useful to predict postsurgical outcome.

Viktor Jirsa is Director of the Inserm Institut de Neurosciences des Systèmes at Aix-Marseille-Université in Marseille, France. Dr. Jirsa received his PhD in 1996 in Theoretical Physics and Applied Mathematics and has since then contributed to the field of Theoretical Neuroscience, in particular through the development of large-scale brain network models based on realistic connectivity. His work has been foundational for network science in brain medicine and the use of personalized virtual brain models in epilepsy. He is Scientific Director of the clinical trial EPINOV, evaluating the use of virtual brain technology in epilepsy surgery. Dr. Jirsa serves as Chief Science Officer of the European digital neuroscience infrastructure EBRAINS (https://ebrains.eu) and lead investigator in the Human Brain Project (HBP) (https://www.humanbrainproject.eu/). Dr. Jirsa has been awarded several international prizes for his research including the first HBP Innovation prize (2021) and Grand Prix de Recherche en Provence (2018) and has published more than 160 scientific articles.

Neural fields are spatially continuous vector fields of neural activity described by differential-integral equations. When placed in a large scale context using long distance connectivity, time delays via signal propagation are introduced, which cannot be described using spatial differentials. The resulting set of functional differential equations is the natural mathematical basis for full brain networks used nowadays in digital brain twin modeling. High resolution in space at mm2 is the minimum to obtain viable digital twins. I will unpack the mathematics linked to this field, present typical bifurcations, resonances, flows and manifolds in state space at rest and disease, and discuss principles of emergence in networks of this nature.

Over the past decade we have demonstrated that the fusion of subject-specific structural information of the human brain with mathematical dynamic models allows building biologically realistic brain network models, which have a predictive value, beyond the explanatory power of each approach independently. The network nodes hold neural population models, which are derived using mean field techniques from statistical physics expressing ensemble activity via collective variables. Our hybrid approach fuses data-driven with forward-modeling-based techniques and has been successfully applied to explain healthy brain function and clinical translation including aging, stroke and epilepsy. Here we illustrate the workflow along the example of epilepsy: we reconstruct personalized connectivity matrices of human epileptic patients using Diffusion Tensor weighted Imaging (DTI). Subsets of brain regions generating seizures in patients with refractory partial epilepsy are referred to as the epileptogenic zone (EZ). During a seizure, paroxysmal activity is not restricted to the EZ, but may recruit other healthy brain regions and propagate activity through large brain networks. The identification of the EZ is crucial for the success of neurosurgery and presents one of the historically difficult questions in clinical neuroscience. The application of latest techniques in Bayesian inference and model inversion, in particular Hamiltonian Monte Carlo, allows the estimation of the EZ, including estimates of confidence and diagnostics of performance of the inference. The example of epilepsy nicely underwrites the predictive value of personalized large-scale brain network models. The workflow of end-to-end modeling is an integral part of the European neuroinformatics platform EBRAINS and enables neuroscientists worldwide to build and estimate personalized virtual brains.

Randy McIntosh is the director of the Institute for Neuroscience and Neurotechnology, at Simon Fraser University.  His research program involves computational modeling and brain imaging to explore changes in cognition across the lifespan and changes in the face of brain damage or disease. The program builds on an international collaboration that delivered TheVirtualBrain (thevirtualbrain.org) and integrates research efforts globally to accelerate research and translation. The goals are 1) to integrate the modeling platform into the standard workflow for clinical decision support, and 2) develop a cloud-based system where anyone can create brain models for research, clinical use, or education.

TheVirtualBrain (TVB) was introduced to the neuroscience community a decade ago. It was the first platform for the creation of large-scale simulations of human brain networks, and has continued to evolve as a community project, extending to multiple basic and clinical applications, as well as extensions to models of rodent and macaque brains. My talk will review some key milestones and cover the good and not-so-good decisions we made in building this platform.

A hidden repertoire is a functional configuration in the brain that supports behaviour but is seldom used. As a complex system, the brain can show a broad range of configurations for the same function. This “many-to-one” property imparts our brain with resilience during normal operations but also in the face of adverse events, such as damage or disease. I will cover the evidence for these repertoires and cover strategies for investigation, and the implication for the I will also relate the existence of such repertoires to variations in the qualia of our experience.

À venir

Neurodegenerative disorders may take decades to develop before clinical symptoms are detectable. Predicting and understanding the individual course of a progressive disorder is of fundamental importance for accurate diagnosis and therapeutic intervention. In practice, there are numerous challenges associated, including high inter-subject variability and the lack of robust descriptors. A growing number of studies have attempted to overcome this gap by using Big-Data and sophisticated statistical and computational models. In this lecture, we will cover main principles of multi-modal data-driven brain models focused on tracking and understanding disease progression. Multiple available computational tools will be presented.

We will present our work on integrating several biological scales and data modalities (from multilevel molecular omics to multimodal neuroimaging and clinical data) via empirical and mechanistic brain models. These models, based on dynamic system analysis and machine learning, allow to track neurodegenerative progression and heterogeneity while clarifying the underlying multilevel biological mechanisms. They also focus on providing individually-tailored predictions of therapeutic needs.

À venir

The aim of this lecture is to provide familiarities with important concepts and computational methods from Algebraic Topology and Network Science that have been greatly used to model and study a variety of phenomena in different areas of science. Specifically, I will give a short introduction to homology of simplicial complexes, Hodge-Laplacian, and multilayer networks.

Great efforts have been made these last decades to understand how individual neurons and/or brain regions organize and connects to each other’s in space and time to support brain activities. In the network science parlance, the current standard paradigm used to model brain activity patterns and the various neuronal interactions is based mainly on the concepts of structural connectivity and functional connectivity networks. The former network depicts intrinsic anatomical (though non-stationary) wiring structures in the brain, while the latter is dynamically brought about by context-dependant variations of neuronal connections that are bounded within physiological constraints as well. In particular, the extent to which the brain anatomy contributes to its functional dynamics has been of major interest for scientists. This talk uses a novel mathematical framework we have developed to model and analyse the topology and spectral properties of the interconnectivities between the structural and functional brain networks.

À venir

The lecture will introduce how the Ising model’s dynamics could be associated with local metabolic activity, as captured through 18F-fluorodeoxyglucose positron emission tomography (FDG-PET). This integration aims to further refine the Ising model’s representation of functional dynamics in various brain states or conditions.

During the upcoming Montreal workshop on Brain Modeling and Simulation, I will discuss the application of the generalized Ising model in understanding the structure-function dynamics of the human brain. One of the features I’ll touch upon is the integration of the Ising model with the Integrated Information Theory (IIT) developed by Giulio Tononi. This combination offers a method to potentially quantify levels of “consciousness” and better grasp the interplay between brain structural and functional aspects.
The lecture will also introduce how the Ising model’s dynamics could be associated with local metabolic activity, as captured through 18F-fluorodeoxyglucose positron emission tomography (FDG-PET). This integration aims to further refine the Ising model’s representation of functional dynamics in various brain states or conditions.
While I will briefly discuss the conceptual idea of the Ising model transitioning to a « sleeping » state through magnetic field coupling, the main focus will be on the core applications of the Ising model. This includes its capability to provide insights into “consciousness” and the potential correlations between brain structure and function.
By expanding the generalized Ising model in these ways, the model becomes a valuable tool for understanding brain structure-function relationships and for investigating the influence of metabolic activity on brain dynamics. I look forward to sharing these insights during the workshop in Montreal.

À Christine Tardif is an Assistant Professor of Biomedical Engineering, and Neurology & Neurosurgery at McGill University since 2017.  She is also co-director of the MRI Unit at the McConnell Brain Imaging Centre, and co-director of the Quebec Bio-Imaging Network. Dr. Tardif’s lab develops novel MRI techniques to generate high-resolution and quantitative MR images of the brain in-vivo, and relates them to microstructural features of the tissue. She has a particular interest in myelin plasticity across the lifespan in healthy subject and neurological disorders. The lab has a translational approach, working on both small animal (7 Tesla) and human (3 and 7 Tesla) MRI systems.

Conventional “weighted” magnetic resonance imaging (MRI) is commonly used in the clinic and research to visualize brain anatomy based on tissue contrast. These images are sensitive to many aspects of brain tissue, as well as the acquisition parameters selected and the imperfections of the imaging system. Quantitative MRI goes beyond conventional MRI by providing a quantitative measure (with a physical unit) of the physico-chemical parameters of the tissue that are related to its microstructure and composition. In this lecture, I will review quantitative MRI techniques to map the T1, T2 and T2* relaxation times of the brain and magnetization transfer saturation. I will also discuss the interpretation of these MR parameters in the context of brain imaging.

Magnetic resonance imaging is widely used to study the relationship between the structural connectivity and functional connectivity of the brain in vivo. Yet the microstructure of the white matter tracts that form the structural connectome are rarely considered. During this talk, I will present how quantitative MRI can be combined with diffusion-weighted imaging to map the microstructure of individual white matter tracts. This includes the use of multi-contrast encoding during image acquisition as well as image processing techniques to disentangle the MR properties of the complex network of crossing white matter tracts in the brain. These novel imaging techniques can be used investigate how tract microstructure modulates brain network function.