Skip to content

Repository for Sanchez-Romero, R., Ito, T., Mill, R. D., Hanson, S. J., & Cole, M. W. (2023). "Causally informed activity flow models provide mechanistic insight into network-generated cognitive activations". NeuroImage, 120300. https://doi.org/10.1016/j.neuroimage.2023.120300

Notifications You must be signed in to change notification settings

ColeLab/DirectedActflow_release

Repository files navigation

Directed Activity Flow Modelling

Repository for code and results from
Sanchez-Romero, R., Ito, T., Mill, R. D., Hanson, S. J., & Cole, M. W. (2023). Causally informed activity flow models provide mechanistic insight into network-generated cognitive activations. NeuroImage, 120300.

Use the repository Discussions for questions or email ruben.saro.at.rutgers.edu.

We present an approach to build generative directed activity flow models using fMRI functional connectivity (FC) inferred with the PC causal search algorithm, and task-related activations.

Alt text

We also compared commonly used FC methods ordered in a continuum in terms of the amount of statistical conditional independence information and causal principles they used.

Alt text

PC algorithm for directed functional connectivity

Importantly, we provide a Python wrapper to run the version of the PC algorithm used in this paper (PCalgWrapper.py), where we remove some orientation rules that assume no-cycles in the true connectivity. This accounts for the knowledge that brain networks contain cyclic patterns. In practice, we prefer to leave connections unoriented (that may or may not be a cycle) instead of incorrectly inferring the abscence of a cycle. Full description and pseudocode of the PC algorithm are in Sanchez-Romero et al..
We highly recommend to check PCalgorithm_run_example.ipynb for an example run of the PC algorithm detailing aspects of the wrapper. See also the Tetrad project website for the original implementation of PC and other causal search algorithms.

Dependencies

Our Tetrad-PC implementation requires Java. See instructions here

Pseudo-empirical simulations data and results

We include a Python Jupyter notebook to reproduce our pseudo-empirical simulations analysis and results: simulations_results_release.ipynb. This simulation strategy was introduced in Sanchez-Romero and Cole (2021) and accompanying repository github.com/ColeLab/CombinedFC.

Dependencies

Our simulations require igraph. You can download it using "pip install python-igraph" or "conda install -c conda-forge python-igraph"

Alt text

(Figure 2 from Sanchez-Romero et al.).

Empirical data and results

We include a Python Jupyter notebook to reproduce fMRI empirical analysis and results using Human Connectome Project (HCP) data: empirical_results_release.ipynb

Alt text

(Figure 3. See Sanchez-Romero et al., for the rest of the results figures.)

About

Repository for Sanchez-Romero, R., Ito, T., Mill, R. D., Hanson, S. J., & Cole, M. W. (2023). "Causally informed activity flow models provide mechanistic insight into network-generated cognitive activations". NeuroImage, 120300. https://doi.org/10.1016/j.neuroimage.2023.120300

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published