BayesBrainMap
R package BayesBrainMap implementing Bayesian brain mapping for individual functional topography and connectivity
This package contains functions implementing the BrainMap model proposed in Mejia et al. (2019) and the spatial BrainMap model proposed in proposed in Mejia et al. (2020+). (Previously, these models were named “Template ICA” and were contained in the package templateICAr.) For both models, subject-level brain networks are estimated as deviations from known population-level networks, which can be estimated using standard ICA algorithms or provided as a parcellation. Both models employ an expectation-maximization algorithm for estimation of the latent brain networks and unknown model parameters.
BrainMap consists of three steps. The main functions associated with each step are listed below.
- Prior estimation:
estimate_prior. Can export the results withexport_prior. - BrainMap model estimation (single-subject):
fit_BBM. - Identification of areas of engagement in each network (or deviation
from the prior mean):
id_engagements.
Citation
If you use BayesBrainMap please cite the following papers:
| Name | APA Citation |
|---|---|
| BrainMap | Mejia, A. F., Nebel, M. B., Wang, Y., Caffo, B. S., & Guo, Y. (2020). Template Independent Component Analysis: targeted and reliable estimation of subject-level brain networks using big data population priors. Journal of the American Statistical Association, 115(531), 1151-1177. |
| Spatial BrainMap | Mejia, A. F., Bolin, D., Yue, Y. R., Wang, J., Caffo, B. S., & Nebel, M. B. (2022). Template Independent Component Analysis with spatial priors for accurate subject-level brain network estimation and inference. Journal of Computational and Graphical Statistics, (just-accepted), 1-35. |
You can also obtain citation information from within R like so:
citation("BayesBrainMap")Installation
You can install the development version of BayesBrainMap from Github
with:
# install.packages("devtools")
devtools::install_github("mandymejia/BayesBrainMap")Important Notes on Dependencies:
To analyze or visualize CIFTI-format data, BayesBrainMap depends on
the ciftiTools package, which requires an installation of Connectome
Workbench. It can be installed from the HCP
website.
For fitting the BrainMap model with surface-based priors (with
spatial_model=TRUE in fit_BBM(), as is the default), INLA is
required. Due to a CRAN policy, INLA cannot be installed automatically.
You can obtain it by running
install.packages("INLA", repos=c(getOption("repos"), INLA="https://inla.r-inla-download.org/R/stable"), dep=TRUE).
Alternatively, dep=FALSE can be used along with manual installation of
dependencies as necessary to avoid installing all of the many INLA
dependencies, most of which are not actually required. Binaries for
alternative Linux builds can be added with the command
inla.binary.install(). Note that INLA is not required for standard
BrainMap (with spatial_model=FALSE in fit_BBM()).
Depending on the analysis, PARDISO may reduce computation time. To
obtain a free academic license forINLA-PARDISO, run inla.pardiso() in
R after running library(INLA). Provide an academic email address. Once
you obtain a license, point to it using
INLA::inla.setOption(pardiso.license = "pardiso.lic") followed by
INLA::inla.pardiso.check() to ensure that PARDISO is successfully
installed and running.