Data-inferred multiscale brain models for psychiatric disorders
Poster No:
1455
Submission Type:
Abstract Submission
Authors:
Martin Breyton1, Spase Petkoski2, Viktor Sip3, Marmaduke Woddman3, Romain Guilhaumou4, Viktor Jirsa5
Institutions:
1Aix-Marseille Université, MARSEILLE, France, 2Université Aix-Marseille, Marseille, Marseille, 3Aix-Marseille Université, Marseille, France, 4AP-HM, Marseille, France, 5Aix-Marseille University, Marseille, Bouches du Rhône
First Author:
Co-Author(s):
Introduction:
Whole brain models are slowly emerging as promising tools in medicine (Wang, H. E. 2024). They integrate data of various modality into one mechanistic framework to generate personalized predictions. A network of neural mass models is constructed, constrained by an individual connectome, and parameters involved in mechanisms at lower scales are inferred from large-scale brain recordings (EEG, fMRI). But the range of parameters is limited by the choice of neural mass model. Phenomenological models offer an abstract representation of neural activity for which the link between mathematical parameters and biological constraints is arbitrary. For mean-field models, the link with neuronal properties is more obvious, but some dimensions are often lost during mathematical derivation. Future clinical applications of virtual brain twins heavily depend on their ability to incorporate highly detailed lower scale mechanisms. This is particularly true in Psychiatry where most treatments are pharmacological and involve complex interactions between receptor activations and physiological alterations. Here, we validate a method to upscale detailed microscopic dynamics to the whole brain and infer parameters that are not available in the ground truth model.
Methods:
We generated (Fig. 1A) a dataset of spiking activity of a heterogenous QIF network (Izhikevich, E. M. 2006) of 10000 neurons, systematically varying excitability (η) and the probability of connections between neurons (p). For this network, an exact mean-field model was previously derived by Montbrió et al. (MPR model, Montbrió, E. 2015) describing the average firing rate (r) and membrane potential (v) of the population. The MPR model assumes an all-to-all connectivity between neurons (p=1) and an infinite population size. An MLP was trained on the QIF data to estimate the dynamics of (r,v) as a function of η and p and the results were compared to the ground-truth MPR model. Brain network models were constructed with The Virtual Brain (Sanz-Leon, P. 2015), using the structural connectivity of a schizophrenic patient whose brain was parcellated according to the Scheafer (Schaefer, A. 2018) atlas (100 regions, 7 Networks). The fluidity of brain activity is quantified by the variance of the upper triangular part of the Functional Connectivity Dynamics matrix (Hansen, E. C. A. 2015). Inference was performed using Simulation Based Inference (Tejero-Cantero, A. 2020) on synthetic BOLD time series by sampling random values of p and η for each functional network.
Results:
We found that the reconstructed dynamics from QIF data captured all the core features of the ground-truth MPR model and that microscale parameters can be successfully retrieved from brain imaging synthetic data. We show in Fig 1.B, the phase plane estimates by the MLP (when p=1) for different values of η and find that the topology of the nullclines and of the flow of the MPR model are well retrieved. When building a whole brain network with the estimated flow (TVB-QIF) and comparing it with a whole brain model using the MPR model (TVB-MPR), we find dynamical properties to also be the same (Fig 1.C). We find that the relationship between excitability and the global coupling is preserved in the whole brain network. Lastly, we inferred the global coupling and both excitability and connectivity between neurons for all 7 functional networks from synthetic BOLD data. We find that the posterior distributions converge to the ground truth (Fig 1. D, only global coupling and one network are shown for clarity).
Conclusions:
Inferring microscopic dynamics using machine learning techniques can successfully capture and reproduce results obtained mathematically in mean-field models. New data generated by highly detailed simulators, or from in vitro manipulations of brain tissue (e.g. organoids) can now be used to estimate microscopic effects such as synaptic density or pharmacological interventions from brain imaging recordings.
Modeling and Analysis Methods:
fMRI Connectivity and Network Modeling 1
Methods Development 2
Keywords:
FUNCTIONAL MRI
Machine Learning
Modeling
Psychiatric Disorders
Statistical Methods
Systems
1|2Indicates the priority used for review
Abstract Information
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
For human MRI, what field strength scanner do you use?
Which processing packages did you use for your study?
Provide references using APA citation style.
Izhikevich, E. M. (2006). Dynamical Systems in Neuroscience : The Geometry of Excitability and Bursting. The MIT Press.
Montbrió, E. (2015). Macroscopic Description for Networks of Spiking Neurons. Physical Review X, 5(2), 021028.
Sanz-Leon, P. (2015). Mathematical framework for large-scale brain network modeling in The Virtual Brain. NeuroImage, 111, 385‑430.
Schaefer, A. (2018). Local-Global Parcellation of the Human Cerebral Cortex from Intrinsic Functional Connectivity MRI. Cerebral Cortex, 28(9), 3095‑3114.
Tejero-Cantero, A. (2020). sbi : A toolkit for simulation-based inference. Journal of Open Source Software, 5(52), 2505.
Wang, H. E. (2024). Virtual brain twins : From basic neuroscience to clinical use. National Science Review, 11(5).