Embedded Files
Functional harmonics reveal multi-dimensional basis functions underlying cortical organization
Katharina Glomb, Morten L. Kringelbach, Gustavo Deco, Patric Hagmann, Joel Pearson, Selen Atasoy
Cell Reports 2021; doi: https://doi.org/10.1016/j.celrep.2021.109554
In this paper, we looked at the harmonic modes of the dense functional connectome of the HCP (functional connectivity from resting state fMRI, averaged over ~800 subjects). We found that many of the functional harmonics, as we called them, very much look like known functional networks, and that they capture organizational features of the cortex on different scales, ranging from retinotopy - within single brain regions - to principal cortical gradients reported previously. When we used functional harmonics to characterize task activation maps, we found that they could explain these maps in a specific and compact manner: a small and specific set of functional harmonics put together would reproduce the task activation map.
Recent work from my time at the Hagmann lab in Lausanne:
Connectome spectral analysis to track EEG task dynamics on a subsecond scale
Katharina Glomb, Joan Rue Queralt, David Pascucci, Michaël Defferrard, Sebastien Tourbier, Margherita Carboni, Maria Rubega, Serge Vulliemoz, Gijs Plomp, Patric Hagmann
NeuroImage 2020; doi: https://doi.org/10.1016/j.neuroimage.2020.117137
We were looking for a way to track networks in our source-reconstructed EEG data, recorded during a visual task - and we found a solution using harmonic modes of the structural connectivity. In other words, using harmonic modes extracted from an average white matter connectome, we get a set of building blocks that can be used equally for all subjects. We expressed our EEG data with the help of these building blocks and we found that it captures the stages of stimulus processing well. Using harmonic modes means that the signal is reduced in dimensionality, as only a relatively low number of building blocks is necessary.
Using structural connectivity to augment community structure in EEG functional connectivity
Katharina Glomb, Emeline Mullier, Margherita Carboni, Maria Rubega, Giannarita Iannotti, Sebastien Tourbier, Martin Seeber, Serge Vulliemoz, Patric Hagmann
Network Neuroscience 2020; doi: https://doi.org/10.1162/netn_a_00147
This paper was sort of our first dive into source-reconstructed EEG data. We wanted to combine structural and functional connectivity, but first we needed to show that the white matter connectivity explains EEG connectivity beyond their common dependence on Euclidean distance. And it does, at least in the resting state EEG data that we used here. We then smoothed the EEG signal in graph space, meaning that we augmented the signal in each brain region with the signal from the brain regions it is most closely connected to anatomically. We found that this way, the community structure in the functional graph becomes more similar to the resting state networks found in fMRI, suggesting that this method can be used to reduce noise in source-reconstructed EEG data.
Proud to be a co-author on this work lead by PhD candidate Joan Rué Queralt:
The connectome spectrum as a canonical basis for a sparse representation of fast brain activity
Joan Rué-Queralt, Katharina Glomb, David Pascucci, Sébastien Tourbier, Margherita Carboni, Serge Vulliémoz, Gijs Plomp, Patric Hagmann
NeuroImage 2021; doi: https://doi.org/10.1016/j.neuroimage.2021.118611
This paper dives into the temporal dynamics of the EEG signal in graph space. More specifically, Joan shows that the compactness of the signal - i.e., how many harmonic modes are necessary to capture the EEG signal - varies over the course of a visual task. As it turns out, this is tied to integration and segregation of brain regions as they process the stimulus up to the point where participants make a decision.
Review that resulted from a workshop held at the conference of the Organization for computational neuroscience in 2019:
Computational models in Electroencephalography
Katharina Glomb, Joana Cabral, Anna Cattani, Alberto Mazzoni, Ashish Raj, Benedetta Franceschiello
Brain Topography 2021; doi: https://doi.org/10.1007/s10548-021-00828-2
In this review, which is written in a bit of a tutorial-y kind of way, we ask the questions: Why is source-reconstruced EEG not more commonly used in more clinically-oriented research? And what has to happen in order for this to change? We argue that one big advantage of looking at EEG on the source level is that it offers the opportunity of in silico experiments. What is missing is a good understanding of which models of EEG activity are around, and how to decide which one to use. We look at models on different spatial scales - micro-, meso- and macro-scale - and introduce some common ones (without claiming the list to be nearly exhaustive). We show some recent examples of where they are already being used in research in order to illustrate their potential and point out future directions of the field.
Work from my PhD in the Deco lab:
Stereotypical modulations in dynamic functional connectivity explained by changes in BOLD variance
Katharina Glomb, Adrián Ponce-Alvarez, Matthieu Gilson, Petra Ritter, Gustavo Deco
NeuroImage 2018; doi: https://doi.org/10.1016/j.neuroimage.2017.12.074
and
Resting state networks in empirical and simulated dynamic functional connectivity
Katharina Glomb, Adrián Ponce-Alvarez, Matthieu Gilson, Petra Ritter, Gustavo Deco
NeuroImage 2017; doi: https://doi.org/10.1016/j.neuroimage.2017.07.065
The goal with these two papers, and of my PhD, was to figure out whether BOLD resting state networks really activate in sequence and whether that constitutes different brain states in the sense of non-stationary dynamics; back then, it seemed to me like this was an assumption of the field, but as a beginner-PhD student, I wasn't convinced by the evidence ;) In my first paper, I applied tensor decomposition to both, empirical and simulated data and tried to show that resting state networks are present in the simulated data, and not just the average functional connectivity.
Page updated
Google Sites
Report abuse