by Prof. Viktor Jirsa, Aix-Marseille Université
When: 24 September 2026 at 3:00 pm
Where: Aula Magna, Istituto Teologico S. Antonio Dottore (Via San Massimo 25, Padova)
Abstract: Over the past two decades, neuroimaging has transformed how we describe the brain, yielding ever-richer maps of structure, function, and connectivity. The same workflow generalizes to healthy aging, neurodegenerationngly face are mechanistic and individual: why does this brain generate these seizures, and what will this intervention do?
Virtual brain twins address this gap by augmenting empirical brain maps with generative mechanism. Epistemically, a digital brain twin is neither a simulation nor a statistical model alone, but a coupling between three elements: (i) the individual brain, characterized through multimodal imaging; (ii) a generative mechanistic brain model, expressing network activity on the subject’s connectome; and (iii) a bidirectional inference loop linking the two, inverting the model against individual data to estimate latent parameters with quantified uncertainty. This triad of individual brain data, mechanism, and inversion distinguishes a twin from a map or a free-running simulation, and grants it clinical actionability.
We illustrate the workflow along the examples of clinical applications to epilepsy and schizophrenia. The connectome is reconstructed from DTI and co-registered with anatomical MRI and CT; region-specific neural mass models are placed on this connectome; and pathophysiology is implemented as epileptogenicity of a brain region for epilepsy and pertubations of dopamine and serotonin gated neuromodulation for schizophrenia. The parameters are at the interface of data and model and are estimated by Bayesian inversion using Hamiltonian Monte Carlo. The inference yields full posteriors and inference diagnostics. Both examples are undergoing prospective randomized clinical trialing, in which the epilepsy application (Virtual Epileptic Patient (VEP) has just finished clinical evaluation (EPINOV, 2019 – 2023, 356 patients with drug resistant epilepsy undergoing neurosurgery).
The same workflow generalizes to healthy aging, neurodegeneration, and other disorders, and is deployed through EBRAINS, the European digital research infrastructure for neuroscience.