The Glossary provides short definitions of neuro-imaging concepts as well as Nilearn specific vocabulary.

If you wish to add a missing term, please create a new issue or open a Pull Request.


Analysis of variance is a collection of statistical models and their associated estimation procedures used to analyze the differences among means.


Area under the curve.


Brain Imaging Data Structure is a simple and easy to adopt way of organizing neuroimaging and behavioral data.


Blood oxygenation level dependent. This is the kind of signal measured by functional Magnetic Resonance Imaging.


Canonical independent component analysis.


Closing is, together with opening, one of the basic operations of mathematical morphology. The closing of a binary image by a structuring element is defined as the erosion of the dilation of that set.


A contrast is a linear combination of variables (parameters or statistics) whose coefficients add up to zero, allowing comparison of different treatments.


Decoding consists in predicting, from brain images, the conditions associated to trial.

Dictionary learning

Dictionary learning (or sparse coding) is a representation learning method aiming at finding a sparse representation of the input data as a linear combination of basic elements called atoms. The identification of these atoms composing the dictionary relies on a sparsity principle: maximally sparse representations of the dataset are sought for. Atoms are not required to be orthogonal.


Dilation is, with erosion one of the fundamental operations of mathematical morphology from which other operations like opening or closing are based. Dilation uses a structuring element for probing and expanding the shapes contained in the input image.


Electroencephalography is a monitoring method to record electrical activity of the brain.


Echo-Planar Imaging. This is the type of sequence used to acquire functional or diffusion MRI data.


Erosion is, with dilation, one of the fundamental operations in mathematical morphology from which other operations like opening or closing are based. Erosion uses a structuring element for probing and reducing the shapes contained in the input image.

FDR correction

False discovery rate controlling procedures are designed to control the expected proportion of “discoveries” (rejected null hypotheses) that are false (incorrect rejections of the null).


Finite impulse response. This is a type of free-form temporal filter that is used to link neural activity with hemodynamic response, when there is uncertainty on the true model.


Functional magnetic resonance imaging is based on the fact that when local neural activity increases, increases in metabolism and blood flow lead to fluctuations of the relative concentrations of oxyhaemoglobin (the red cells in the blood that carry oxygen) and deoxyhaemoglobin (the same red cells after they have delivered the oxygen). Oxyhaemoglobin and deoxyhaemoglobin have different magnetic properties (diamagnetic and paramagnetic, respectively), and they affect the local magnetic field in different ways. The signal picked up by the MRI scanner is sensitive to these modifications of the local magnetic field.


fMRIPrep is a fMRI data preprocessing pipeline designed to provide an interface robust to variations in scan acquisition protocols with minimal user input. It performs basic processing steps (coregistration, normalization, unwarping, noise component extraction, segmentation, skullstripping etc.) providing outputs, often called confounds or nuisance parameters, that can be easily submitted to a variety of group level analyses, including task-based or resting-state fMRI, graph theory measures, surface or volume-based statistics, etc.

FPR correction

False positive rate correction. This refers to the methods employed to correct false positive rates such as the Bonferroni correction which divides the significance level by the number of comparisons made.


FREM means “Fast ensembling of REgularized Models”. It uses an implicit spatial regularization through fast clustering and aggregates a high number of estimators trained on various splits of the training set, thus returning a very robust decoder at a lower computational cost than other spatially regularized methods.

functional connectivity

Functional connectivity is a measure of the similarity of the response patterns in two or more regions.

functional connectome

A functional connectome is a set of connections representing brain interactions between regions.

FWER correction

Family-wise error rate is the probability of making one or more false discoveries, or type I errors when performing multiple hypotheses tests.


General Linear Model. This is the name of the models traditionally fit to fMRI data, where one linear model is fit to each voxel time course.


Haemodynamic response function. This is a temporal filter that converts neural signals to hemodynamic signals observable with fMRI.


Independent component analysis is a computational method for separating a multivariate signal into additive subcomponents.


Magnetoencephalography is a functional neuroimaging technique for mapping brain activity by recording magnetic fields produced by electrical currents occurring naturally in the brain.


MNI stands for “Montreal Neurological Institute”. Usually, this is used to reference the MNI space/template. The current standard MNI template is the ICBM152, which is the average of 152 normal MRI scans that have been matched to the MNI305 using a 9 parameter affine transform.


Mutli-Voxel Pattern Analysis. This is the way supervised learning methods are called in the field of brain imaging.


Neurovault is a public repository of unthresholded statistical maps, parcellations, and atlases of the human brain.


Opening is, together with closing, one of the basic operations of mathematical morphology. It is defined as the dilation of the erosion of a set by a structuring element.


Act of dividing the brain into smaller regions, i.e. parcels. Parcellations can be defined by many different criteria including anatomical or functional characteristics. Parcellations can either be composed of “hard” deterministic parcels with no overlap between individual regions or “soft” probabilistic parcels with a non-zero probability of overlap.

predictive modelling

Predictive modelling uses statistics to predict outcomes.


Recursive nearest agglomeration.


Resting state fMRI is a method of functional magnetic resonance imaging that is used in brain mapping to evaluate regional interactions that occur in a resting or task-negative state, when an explicit task is not being performed.


The receiver operating characteristic curve plots the true positive rate (TPR) against the false positive rate (FPR) at various threshold settings.


Searchlight analysis consists of scanning the brain with a searchlight. That is, a ball of given radius is scanned across the brain volume and the prediction accuracy of a classifier trained on the corresponding voxels is measured.


SNR stands for “Signal to Noise Ratio” and is a measure comparing the level of a given signal to the level of the background noise.


SpaceNet is a decoder implementing spatial penalties which improve brain decoding power as well as decoder maps.


Statistical Parametric Mapping is a statistical technique for examining differences in brain activity recorded during functional neuroimaging experiments. It may alternatively refer to a software created by the Wellcome Department of Imaging Neuroscience at University College London to carry out such analyses.

supervised learning

Supervised learning is interested in predicting an output variable, or target, y, from data X. Typically, we start from labeled data (the training set). We need to know the y for each instance of X in order to train the model. Once learned, this model is then applied to new unlabeled data (the test set) to predict the labels (although we actually know them). There are essentially two possible types of problems:


In regression problems, the objective is to predict a continuous variable, such as participant age, from the data X.


In classification problems, the objective is to predict a binary variable that splits the observations into two groups, such as patients versus controls.

In neuroimaging research, supervised learning is typically used to derive an underlying cognitive process (e.g., emotional versus non-emotional theory of mind), a behavioral variable (e.g., reaction time or IQ), or diagnosis status (e.g., schizophrenia versus healthy) from brain images.


Support vector machines are a set of supervised learning methods used for classification, regression and outliers detection.


Repetition time. This is the time in seconds between the beginning of an acquisition of one volume and the beginning of acquisition of the volume following it.

Unsupervised learning

Unsupervised learning is concerned with data X without any labels. It analyzes the structure of a dataset to find coherent underlying structure, for instance using clustering, or to extract latent factors, for instance using independent components analysis (ICA).

In neuroimaging research, it is typically used to create functional and anatomical brain atlases by clustering based on connectivity or to extract the main brain networks from resting-state correlations. An important option of future research will be the identification of potential neurobiological subgroups in psychiatric and neurobiological disorders.


Voxel-Based Morphometry measures differences in local concentrations of brain tissue, through a voxel-wise comparison of multiple brain images.


A voxel represents a value on a regular grid in 3D space.

Ward clustering

Ward’s algorithm is a hierarchical clustering algorithm: it recursively merges voxels, then clusters that have similar signal (parameters, measurements or time courses).