GSoC 2021 project idea 1.2: Visualizing high-order interactions in neuroimaging data and beyond

The functioning of complex systems — such as the human brain, the climate, and many more — is characterised by emergent phenomena, which is a consequence of non-linear interactions between various of its constituent units. Crucially, the resulting dynamics observed in such systems is qualitatively different from the sum of the dynamics of each part. A promising way to deepen our understanding of these systems is to leverage existing datasets of these complex systems; however, the analysis of these data requires specific analytic tools to make sense of their highly non-trivial interplay. Several tools and frameworks have been developed to look at different statistical dependencies that can be observed in these fascinating multivariate datasets. Among these, information theory offers a powerful and versatile framework; notably it allows to detect and quantify high-order interactions: namely, the characteristic of the joint informational role of a group of variables. However, metrics of high-order interactions are challenging to visualize, and the lack of adequate visualization paradigms is hindering their potential impact in numerous research fields. While visualization is easy and well explored for local measures and pairwise connectivity ones, it becomes trickier when it comes to higher order measures. In fact, a lot of the appeal and understandability of these metrics data analyses (in general, and in particular in neuroimaging) relies on their visual representation, with or without an anatomical overlay.

The goal of this project is to conceptualize and implement strategies and tools for the visualization of high-order interactions, both for general datasets, and also specifically tailored to neuroimaging data analysis — mainly M/EEG and fMRI. The main deliverable of the project will be a toolbox that integrates the developed approaches. (A parallel project will focus on the implementation of these high-order metrics in a python-based toolbox.) Any open source tool (starting with python and javascript) is welcome.

Experience in statistics and neuroimaging data analysis is a plus, but not a requirement.

Lead mentor: @f.rosas Fernando Rosas, Imperial College London; [GitHub]
Co-mentor: @Daniele_Marinazzo, Ghent University,[GitHub]

Skills: neuroimaging, statistics, visualization, Python, Javascript, MEG, EEG

@f.rosas @Daniele_Marinazzo I’m a junior undergraduate in biomedical engineering and highly interested in this project owing to my previous experience working with multi-modal imaging facilities. To get started with this, is there any github link or documentation for this project. Or could we have a discussion over a call regarding the details

Dear Debabrata
thanks a lot for your interest!
For this project there’s not a repo yet, it’s something to start from scratch.
This twin project is about the implementation of the measures whose results will need to be visualized.

Hypergraphs visualization is an idea one could start with.

All the best

Dear @f.rosas @Daniele_Marinazzo,
I’m a PhD student in neuroscience & philosophy at the University of Milan working on methods for quantifying complexity and higher-order interactions (e.g. integrated information) applied to brain stimulation recordings. I’m very interested in this project – besides resonating with my current work, I’m keen on data visualization and (more recently) the ΦID framework. Yet, I’m not familiar with the GSoC format, could you give me some coordinates on how to collaborate with the project? From what I’ve understood I should send a proposal according to the guidelines here:

In the section detailing the project I want to work on, should I then refer to this project (1.2) and for example detail a feature I could help developing? It also mentions I can share my draft with INCF so they can comment on it before the application deadline. Thanks in advance.

1 Like

Dear Renzo

glad to hear you’re interested.
Yes, you should follow the instructions on the GSoC page (and check elegibility).
You can either share your proposal with us on a google doc, or edit it directly on the platform, this year is possible to save a draft and to have the promoters see it directly from there (this is what is meant by “shared with INCF”). In this case please notify us, since we don’t get an automated message.
At this link you can find all the info and FAQs compiled by INCF.
Looking forward to your proposal!