AFNI is a functional imaging analysis package. It is funded by the NIMH, based in Bethesda, Maryland, and directed by Robert Cox. Like FSL, it is written in C, and it’s very common to use shell scripting of AFNI command line utilities to automate analyses. Users often describe liking AFNI’s scriptability, and image visualization. It uses the GPL license.


Berkeley software distribution license. The BSD license is permissive, in that it allows you to modify and use the code without requiring that you use the same license. It allows you to distribute closed-source binaries.


Contrast that is blood oxygen level dependent. When a brain area becomes active, blood flow increases to that area. It turns out that, with the blood flow increase, there is a change in the relative concentrations of oxygenated and deoxygenated hemoglobin. Oxy- and deoxy- hemoglobin have different magnetic properties. This in turn leads to a change in MRI signal that can be detected by collecting suitably sensitive MRI images at regular short intervals during the blood flow change. See the Wikipedia FMRI article for more detail.


BrainVISA is a sister project to NIPY. It also uses Python, and provides a carefully designed framework and automatic GUI for defining imaging processing workflows. It has tools to integrate command line and other utilities into these workflows. Its particular strength is anatomical image processing but it also supports FMRI and other imaging modalities. BrainVISA is based in NeuroSpin, outside Paris.


Diffusion tensor imaging. DTI is rather poorly named, because it is a model of the diffusion signal, and an analysis method, rather than an imaging method. The simplest and most common diffusion tensor model assumes that diffusion direction and velocity at every voxel can be modeled by a single tensor - that is, by an ellipse of regular shape, fully described by the length and orientation of its three orthogonal axes. This model can easily fail in fairly common situations, such as white-matter fiber track crossings.


Diffusion-weighted imaging. DWI is the general term for MRI imaging designed to image diffusion processes. Sometimes researchers use DTI to have the same meaning, but DTI is a common DWI signal model and analysis method.


The most widely-used open-source package for analyzing electro-physiological data. EEGlab is written in matlab and uses a GPL license.


Functional magnetic resonance imaging. It refers to MRI image acquisitions and analysis designed to look at brain function rather than structure. Most people use FMRI to refer to BOLD imaging in particular. See the Wikipedia FMRI article for more detail.


FSL is the FMRIB software library, written by the FMRIB analysis group, and directed by Steve Smith. Like AFNI, it is a large collection of C / C++ command line utilities that can be scripted with a custom GUI / batch system, or using shell scripting. Its particular strength is analysis of DWI data, and ICA functional data analysis, although it has strong tools for the standard SPM approach to FMRI. It is free for academic use, and open-source, but not free for commercial use.


The GPL is the GNU general public license. It is one of the most commonly-used open-source software licenses. The distinctive feature of the GPL license is that it requires that any code derived from GPL code also uses a GPL license. It also requires that any code that is statically or dynamically linked to GPL code has a GPL-compatible license. See: Wikipedia GPL and


Independent component analysis is a multivariate technique related to PCA, to estimate independent components of signal from multiple sensors. In functional imaging, this usually means detecting underlying spatial and temporal components within the brain, where the brain voxels can be considered to be different sensors of the signal. See the Wikipedia ICA page.


The lesser GNU public license. LGPL differs from the GPL in that you can link to LGPL code from non-LGPL code without having to adopt a GPL-compatible license. However, if you modify the code (create a “derivative work”), that modification has to be released under the LGPL. See Wikipedia LGPL for more discussion.


matlab began as a high-level programming language for working with matrices. Over time it has expanded to become a fairly general-purpose language. See also: Wikipedia MATLAB. It has good numerical algorithms, 2D graphics, and documentation. There are several large neuroscience software projects written in MATLAB, including SPM software, and EEGlab.


Principal component analysis is a multivariate technique to determine orthogonal components across multiple sources (or sensors). See ICA and the Wikipedia PCA page.


Positron emission tomography is a method of detecting the spatial distributions of certain radio-labeled compounds - usually in the brain. The scanner detectors pick up the spatial distribution of emitted radiation from within the body. From this pattern, it is possible to reconstruct the distribution of radiactivity in the body, using techniques such as filtered back projection. PET was the first mainstream technique used for detecting regional changes in blood-flow as an index of which brain areas were active when the subject is doing various tasks, or at rest. These studies nearly all used water activation PET. See the Wikipedia PET entry.


SPM (statistical parametric mapping) refers either to the SPM approach to analysis or the SPM software package.

SPM approach

Statistical parametric mapping is a way of analyzing data, that involves creating an image (the map) containing statistics, and then doing tests on this statistic image. For example, we often create a t statistic image where each voxel contains a t statistic value for the time-series from that voxel. The SPM software package implements this approach - as do several others, including FSL and AFNI.

SPM software

SPM (statistical parametric mapping) is the name of the matlab based package written by John Ashburner, Karl Friston and others at the Functional Imaging Laboratory in London. More people use the SPM package to analyze FMRI and PET data than any other. It has good lab and community support, and the matlab source code is available under the GPL license.


Quoting from the Voxbo webpage - “VoxBo is a software package for the processing, analysis, and display of data from functional neuroimaging experiments”. Like SPM, FSL and AFNI, VoxBo provides algorithms for a full FMRI analysis, including statistics. It also provides software for lesion-symptom analysis, and has a parallel scripting engine. VoxBo has a GPL license. Dan Kimberg leads development.


Voxels are volumetric pixels - that is, they are values in a regular grid in three dimensional space - see the Wikipedia voxel entry.

water activation PET

A PET technique to detect regional changes in blood flow. Before each scan, we inject the subject with radio-labeled water. The radio-labeled water reaches the arterial blood, and then distributes (to some extent) in the brain. The concentration of radioactive water increases in brain areas with higher blood flow. Thus, the image of estimated counts in the brain has an intensity that is influenced by blood flow. This use has been almost completely replaced by the less invasive BOLD FMRI technique.