Statistical Tools for Medical Imaging NetCDF (MINC) Files

Tools for reading, writing, analyzing, and visualizing MINC neuroimaging files with R. Supports voxel-wise linear models, mixed-effects models, correlations, ANOVA, FDR correction, TFCE, ROI-based anatomy analysis, surface-based vertex statistics, and parallel execution.

Unix only (Linux, macOS). Requires R >= 4.0. No native Windows support (WSL may work — see INSTALL).

Features
Installation
- System Requirements
- Troubleshooting
Quick Example
Dockerized Testing
Tutorials
Getting Help
Development
License

Features

Capability	Key functions
Voxel-wise linear models	`mincLm`, `mincAnova`, `mincTtest`, `mincWilcoxon`
Correlations & summaries	`mincCorrelation`, `mincMean`, `mincSummary`
Mixed-effects models	`mincLmer`, `mincLmerEstimateDF`, `mincLogLikRatio`
Surface / vertex statistics	`vertexLm`, `vertexLmer`, `vertexTFCE`
ROI / anatomy analysis	`anatLm`, `anatLmer`, `hanatLm` (hierarchical atlases)
Multiple comparisons	`mincFDR`, `mincTFCE`, `mincRandomize`
I/O	`mincGetVolume`, `mincWriteVolume`, `mincArray`
Visualization	`mincPlotSliceSeries`, `mincTriplanarSlicePlot` (2D), `rgl` (3D), `hanatView`
Parallel execution	`pMincApply`, `qMincApply`, `mincLm(parallel = ...)`

Installation

The easiest way to install is with devtools from the develop branch:

devtools::install_github(
  "Mouse-Imaging-Centre/RMINC@develop",
  upgrade_dependencies = FALSE
)

System Requirements

RMINC depends on the MINC C library. You have two options:

minc-toolkit-v2 (recommended): Download binary installers from http://bic-mni.github.io/ or build from source at https://github.com/BIC-MNI/minc-toolkit-v2. This provides the MINC library plus command-line tools.
libminc from source: If the toolkit is not found, RMINC’s configure will attempt to build libminc for you. This requires CMake (> 2.6), git, and the HDF5 development headers.

Debian / Ubuntu:

sudo apt-get install libhdf5-dev cmake git lsof autoconf automake libtool

macOS (Homebrew):

brew install hdf5 cmake autoconf automake libtool

macOS note: RMINC’s compiled backend needs a Fortran compiler. Apple’s default clang does not include one — install gcc via Homebrew (brew install gcc libssh2 libgit2 libomp zlib) or gfortran directly, and point R at it via ~/.R/Makevars. See the INSTALL file for the exact FLIBS/LDFLAGS setup and detailed instructions.

Troubleshooting

If configure cannot find MINC, set the install prefix via an environment variable:

export MINC_TOOLKIT=/opt/minc/1.9.18

Or pass the path at install time from R:

devtools::install_github(
  "Mouse-Imaging-Centre/RMINC@develop",
  args = "--configure-args='--with-build-path=/opt/minc/1.9.18'",
  upgrade_dependencies = FALSE
)

Quick Example

Fit a voxel-wise linear model across a set of MINC volumes. The response column in data holds MINC file paths; the model is fit independently at every voxel:

library(RMINC)

# `gf` is a data frame whose `volumes` column holds MINC file paths
fit <- mincLm(volumes ~ group, data = gf)

# Output columns: "F-statistic", "R-squared", beta coefficients,
# t-statistics, "logLik". Write the whole-model F-statistic (column 1):
mincWriteVolume(fit, output.filename = "group_Fstat.mnc", column = 1)

Function-level documentation is available in R via ?mincLm, ?mincLmer, ?vertexLm, ?anatLm, etc.

Dockerized Testing

Build the test image:

docker build -t rminc-test .

Run checks and tests (mount the source directory):

docker run -v $(pwd):/RMINC rminc-test              # check + tests (default)
docker run -v $(pwd):/RMINC rminc-test check          # R CMD check --as-cran only
docker run -v $(pwd):/RMINC rminc-test test            # tests only
docker run -v $(pwd):/RMINC rminc-test test mincLm     # filtered tests
docker run -v $(pwd):/RMINC rminc-test shell           # interactive bash shell

Tutorials

Voxel-wise Statistics (somewhat out of date)
How RMINC Parallel Works
RMINC 2D Visualization
Visualizing 3D Objects with RMINC
Analyzing Anatomy with Hierarchical Atlases

Getting Help

Function documentation is available through R’s ? interface (e.g. ?mincLm). The MICePUB wiki has additional background and usage notes. Report bugs or request features at the issues page.

Development

The develop branch is where active development happens; the master branch may lag behind. Contributions via pull requests to develop are welcome.

The C/C++ backend is built with Rcpp and autoconf — src/Makevars is generated from src/Makevars.in, and RcppExports.{R,cpp} are auto-generated by Rcpp. Do not edit these files directly.

After editing R source containing roxygen comments, regenerate documentation and the NAMESPACE with:

roxygen2::roxygenise()

License

RMINC is released under the BSD 3-Clause License.

RMINC