The VoxBo suite of tools for brain image analysis supports a variety of formats for 1D (vector), 2D (matrix), 3D (volume) and 4D (volume time series) data.

File format support in VoxBo is modular - most of the tools can read and write any supported file format transparently, with output file formats usually determined by the file extension.

The i/o library supports several special access methods. For 4D files, filename:mask is used to access the inclusion mask of a 4D file. The mask is treated as a 3D volume. Similarly, filename:n accesses the nth volume of the 4D series in filename, where n is an integer and the first volume is numbered 0.

When writing, VoxBo honors filename tags that can be used to specify the fileformat or byteorder. Tags are specified in square brackets at the end of the filename, and multiple tags can be separated within the brackets by commas. For example, foo[imgdir] can be used to specify the imgdir file format (useful because imgdir format doesn't have a standard extension). foo.nii[little] forces little-endian storage for file formats that allow it.

VoxBo native file formats

VoxBo's native file formats, all supported for read and write, include .ref (1D), .mat (2D), .cub (3D), and .tes (4D). Although these formats are not widely supported in external software, they offer some useful features, including user-editable text headers. gzip-compressed cub and tes files are supported transparently.


VoxBo supports 3D and 4D NIfTI files (compressed or not) for read and write. VoxBo does not currently make intelligent use of the qform/sform orientation information.


VoxBo supports SPM-style Analyze(TM) files, including 3D and 4D files, as well as 4D time series composed of directories containing 3D files, for read/write. All individual files are stored in .img/.hdr pairs.


VoxBo supports DICOM (as well as some DICOM-like files in the older ACR/NEMA format) for reading only. VoxBo supports a corner of the DICOM standard frequently encountered in brain imaging research, including mosaic format volumes. Files containing single slice data are read as 3D volumes. Directories containing multiple files are generally parsed as 3D or 4D data, as appropriate, as long as no extraneous files are in the directory.

Text Matrices and Vectors

Plain text files containing just numbers (or comment lines beginning with #, %, or ;), can be parsed as matrix or vector data.



Pre-built VoxBo binaries are available for Linux, OSX, and Cygwin, and via the NeuroDebian project ( for Debian derivatives.


VoxBo is provided with no warranty whatsoever.


For an historical roster of the VoxBo development team, visit

RELATED TO voxbo-fileformats…

For detailed help on any VoxBo command-line program, run it with no arguments. For graphical programs, use the -h flag. For more help with VoxBo, for information about the mailing list, or to report bugs, visit the web site at

For general information about VoxBo, see voxbo(7).