Building models that interpret neuroimaging data

[maedoc]

Title

Building models that interpret neuroimaging data

Leaders

Marmaduke Woodman

Collaborators

it's early days

Brainhack Global 2023 Event

Brainhack Marseille

Project Description

• What are you doing, for whom, and why?

We are writing a new implementation of whole brain models oriented towards recent machine learning algorithms. This implementation is for students & post docs who will come up with tomorrow's theory of brain (dys)function and need better tools for doing so.

• What makes your project special and exciting?

Our project is 🦄🦄🦄 because we are building fine-grained models of neural dynamics in entire cohorts where current whole brain models only maps coarse-grained statistics.

• How to get started?

The package is being developed at https://github.com/ins-amu/vbjax and includes neural mass and field models, forward models for MEG/fMRI and some data fitting examples.

For the brain hack we will use data from HCP MEG with Brainstorm based preprocessing; scripts at https://github.com/maedoc/friedchicken, but this is not so much the focus of the project.

• Where to find key resources?

More resources on the background of the modeling is available at https://thevirtualbrain.org and https://www.ebrains.eu/tools/the-virtual-brain.

Link to project repository/sources

https://github.com/ins-amu/vbjax

Goals for Brainhack Global

• discuss use cases with potential users, even those unfamiliar with modeling
• help new users install and run examples
• extend existing set of models
• write new examples for data users have already prepared
• test deep neural network for more flexible time series modeling

Good first issues

• Generate Sphinx docs ins-amu/vbjax#11
• Add tests for partial observed node dynamics ins-amu/vbjax#29
• Add colored noise option ins-amu/vbjax#7

Communication channels

nothing for the moment

Skills

• Brainstorming use cases, data features & models 50%
• Python coding & debugging 50%

Onboarding documentation

https://github.com/ins-amu/vbjax#readme

What will participants learn?

• how to run a whole-brain simulation
• how to use Jax & NumPyro, potentiall w/ GPUs
• how to optimize a model to fit data
• how to do Bayesian MCMC to find parameters consistent with data

Data to use

We are using mostly data from HCP https://db.humanconnectome.org,

• preprocessed connectomes and BOLD from EBRAINS (https://search.kg.ebrains.eu)
• MEG w/ Brainstorm based preprocessing (https://github.com/maedoc/friedchicken)

Number of collaborators

2

Credit to collaborators

https://github.com/ins-amu/vbjax/graphs/contributors

Image

https://avatars.githubusercontent.com/u/6401490?s=48&v=4

Type

method_development, other

Development status

2_releases_existing

Topic

bayesian_approaches, causality, connectome, deep_learning, machine_learning, neural_networks, reproducible_scientific_methods, statistical_modelling, systems_neuroscience

Tools

Brainstorm, Jupyter, MNE, other

Programming language

containerization, documentation, Matlab, Python

Modalities

MEG, other

Git skills

0_no_git_skills, 1_commit_push, 2_branches_PRs

Anything else?

No response

Things to do after the project is submitted and ready to review.

• Add a comment below the main post of your issue saying: Hi @brainhackorg/project-monitors my project is ready!
• Twitter-sized summary of your project pitch.

[maedoc]

Hi @brainhackorg/project-monitors my project is ready!

[arnaudletroter]

Hi Marmaduke,
Thanks for submitting your project to BHM 2023.
Your project is now visible online !
Arnaud for The BHM organization team.