Whole-brain linear effective connectivity (WBLEC) estimation

These Python notebooks reproduce some figures in the following preprint using the libraries pyMOU and NetDynFlow: https://www.biorxiv.org/content/10.1101/531830v2

The notebook 1_MOUEC_Estimation.ipynb should be executed first to tune the model to the fMRI data. The other notebooks can be used for classification and interpretation of the model (using the flow for network analysis).

The data files are:

• BOLD time series in ts_emp.npy
• structural connectivity in SC_anat.npy
• ROI labels in ROI_labels.npy

####Notebook 1_MOUEC_Estimation.ipynb

This notebook calculates the functional connectivity and the model-based effective connectivity for each session (or run) and subject from the BOLD time series. The model is a multivariate Ornstein-Uhlenbeck (MOU) process, whose estimation procedure is implemented in the pyMOU library. The model estimates and other measures are stored in the form of arrays in the model_param_movie folder.

####Notebooks 2a_ClassificationTasks.ipynb and 2b_ClassificationSubjects.ipynb

These notebooks compare the performances of the several type of connectivity measures (including functional and effective connectivity) in identifying cognitive tasks and subjects. They rely on the scikit.learn library.

####Notebook 3a_Flow.ipynb

This notebook uses the NetDynFlow library to calculate the flow, which is network-oriented analysis of the MOU model fitted to the BOLD data. The flow corresponds to the input response of the network to perturbation (or stimulation of given regions). The flow captures network effects that arise from the recurrent connectivity, i.e. also taking into account indirect paths between all pairs of regions.

####Notebook 3b_Communities.ipynb

This notebook detects communities based on the flow, namely brain regions are grouped together if they exchange strong flow in the network. It also compares the community structure between rest and movie.