Home



Wellcome to my page!

image

Currently, I am a Senior Research Fellow at INS (UMR1106).

My work aims to infer the dynamics of personalized virtual brain models using a Bayesian framework to enhance diagnostics, interventions, therapies, and decision-making in brain-related medicine and digital health.

Hire me for consulting services and tutorials!


What is inference and how to make inference?

☞ See here in Video.
☞ See here in Keynote.
☞ See here in Powerpoint.

I develop open-source statistical, machine learning and AI algorithms for probabilistic inference on complex systems such as the brain. Bridging this with clinical translation, I have built Virtual Brain Twins to study brain disorders, leveraging principles from statistical physics, dynamical systems, and Bayesian inference (Monte Carlo sampling, simulation-based inference, deep neural density estimators, generative AI.


I perform probabilistic inference for:

  • Cohort studies, focusing on longitudinal trajectories, which do not necessarily require a control group.

  • Case-Control studies, disregarding heterogeneity and instead focusing on the average patient (a potential predictor for rare outcomes). This approach, however, is noninformative for constructing models in brain diseases such as neurodegenerative and psychiatric disorders.

  • Normative studies, an emerging approach for quantifying and describing how individuals deviate from expected patterns learned from a healthy population, using Generalized Additive Model for Location, Scale and Shape (GAMLSS) models.

  • Personalized Causal studies in precision medicine, focusing on the individualized anatomical data, and mechanistic models, inverted by the state-of-the-art probabilistic inference methods, such as adaptive Hamiltonian Monte Carlo and Normalizing Flows.

My tools for identifying spatiotemporal patterns emerging from high-dimensional nonlinear complex systems are automatic and flexible, employing SATO inference algorithms;
(1) Markov chain Monte Carlo (MCMC) algorithms in probabilistic programming languages,
(2) Simulation-based inference (SBI) using deep neural density estimators.


I have worked with spiking, mass, mean-field, and whole-brain models to fit neurophysiological data, including SEEG, EEG, MEG, and fMRI. These models aim to reveal causal mechanisms in neurological disorders such as epilepsy, Alzheimer's, Parkinson's, multiple sclerosis, and alcohol use disorder, as well as in resting-state, general anesthesia and healthy aging.


I have collaborated closely with scientist, engineers and medical doctors in hospitals (e.g., La Timon Marseille) and industrial companies (e.g., SATT Sud-Est). This has led to to multiple publications, patents (e.g., BVEP), and software development currently used in clinical trials, such as (EPINOV; the first clinical trial in epilepsy leveraging virtual brain models, involving data from 400 patients across 12 centers in France). I have extensively used the softwares: TVB, Stan, PyMC3, NumPyro, scikit-learn, SBI.