Branesta is a tool for computing brain network stability, a biomarker for brain aging.
- Free software: MIT license
- Documentation: https://branesta.readthedocs.io.
Please cite our article:
Mujica-Parodi, Lilianne R., et al. "Diet modulates brain network stability, a biomarker for brain aging, in young adults." Proceedings of the National Academy of Sciences 117.11 (2020): 6170-6177. link: https://www.pnas.org/content/117/11/6170
Brain network stability measures the extent of temporal reorganization that takes place in brain networks. Brain networks describe inter-regional communication across the brain. Lower network stability (represented by higher values) is related to weaker persistence of brain networks. The terms Network Stability and Network INstability are used interchangibly and they refer to the exact same metric.
The procedure of computing brain network stability is as follows: fMRI time-series that were previously parcelled into ROIs are first binned into time windows (=snapshots) of N timepoints without overlaps (N = window length). Next, pairwise correlations among all ROIs are computed separately for each time window. For the whole brain, brain network stability (scalar) is quantified by taking the l2 norm of the element-wise differences of correlation matrices corresponding to two different snapshots. τ is the number of steps separating two snapshots from which a given value of brain network stability is calculated from. For instance, if τ=1, two consecutive snapshots snapshots are used (e.g. #4 and #5). If τ=16, then 16 snapshots are separating the two snapshots (e.g. #3 and #19). Given a window length of 30 timepoints, if the time-series have a length of 720 timepoints, then there will be 24 snapshots (720/30=24). At τ=1, there are 23 instability values, whereas at τ=20, 4 different instability values would be computed.
For functional networks (labeled as "subnetworks" in our program), the procedure is analog to the above. The only difference is that once correlations are computed for each time window, element-wise differences are calculated only across those ROIs that spatially overlap with the functional network. In order to facilitate comparison of network instability among networks consisting of different number of nodes, network stability is normalized with the number of edges in the correlation matrix.
- computes network stability from parcelled time-series
- performs computations at every τ
- computes for subnetworks (optional)
- allows user-defined time window length
- easy to install (pip)
- command line tool
This package was developed within the Laboratory for Computational Neurodiagnostics (LCNeuro) at Stony Brook University, New York.
This package was created with Cookiecutter and the audreyr/cookiecutter-pypackage project template.