Skip to content

Latest commit

 

History

History
36 lines (27 loc) · 2.66 KB

README.md

File metadata and controls

36 lines (27 loc) · 2.66 KB

event_detection

Perform event detection, clustering, and FC prediction as in Betzel et al (2021): Individualized event structure drives individual differences in whole-brain functional connectivity.

hypothesis for how high amplitude co-fluctuations might contribute to FC

dependencies and requirements

This script uses MATLAB. It was tested using MATLAB 2020a on macOS Mojave 10.14.6. The example script uses the Generalized Louvain function for performing community detection. It can be downloaded here.

installation and usage

Download and unzip the directory, open MATLAB, and then navigate to wherever you downloaded the files. You can run the script from within that directory.

what is here

  • Example parcellated resting fMRI BOLD time series from here.
  • System labels for each brain region.
  • Analysis script.
  • Some helper functions.

what does the script do?

  1. Reads in the parcel time series data, generates edge time series, and calculates the root sum squared (RSS) amplitude at each frame.
  2. Uses a circular-shift null model to generate null time series and repeats the procedure from Step 1, resulting in a null distribution of RSS values.
  3. Identifies statistically significant frames (observed RSS > than that of null) and partitions time series into segments of temporally contiguous supra-threshold frames.
  4. Identifies peak RSS within each segment and extracts pattern of activity (parcel time series) and edge time series (co-fluctuation) at that instant.
  5. Clusters those co-fluctuation patterns using modularity maximization and estimates cluster (community) centroids.
  6. Generated a predicted FC matrix given the detected communities and the frequencies with which different clusters appear.
  7. Makes a couple figures.

if you want to apply this method to your data

  • Simply replace the "ts.mat" variable with a [time x node] matrix of neural recordings.

Runtime for entire script depends on size of neural recordings variable, but for the data included here, it takes approximately 100 s.

If you use this code, please cite: R Betzel, S Cutts, S Greenwell, O Sporns (2021). Individualized event structure drives individual differences in whole-brain functional connectivity. bioRxiv link to paper

For the Generalized Louvain algorithm, please also cite: Jutla, I. S., Jeub, L. G., & Mucha, P. J. (2011). A generalized Louvain method for community detection implemented in MATLAB. URL http://netwiki. amath. unc. edu/GenLouvain.