User guide

Download for offline viewing

Download the user guide and examples.

• Introduction
  • What is nilearn: MVPA, decoding, predictive models, functional connectivity
    • Why is machine learning relevant to NeuroImaging? A few examples!
  • Installing nilearn
  • Python for NeuroImaging, a quick start
    • Your first steps with nilearn
    • Scientific computing with Python
      • Basic numerics
      • Scikit-learn: machine learning in Python
    • Finding help
• Decoding and MVPA: predicting from brain images
  • An introduction to decoding
    • Loading and preparing the data
      • The Haxby 2001 experiment
      • Loading the data into nilearn
    • Performing a simple decoding analysis
      • A few definitions
      • A first estimator
      • Decoding made easy
      • Measuring prediction performance
        • Cross-validation
        • Choosing a good cross-validation strategy
        • Choice of the prediction accuracy measure
        • Prediction accuracy at chance using simple strategies
      • Visualizing the decoder’s weights
    • Decoding without a mask: Anova-SVM
      • Dimension reduction with feature selection
      • Visualizing the results
  • Choosing the right predictive model for neuroimaging
    • Predictions: regression, classification and multi-class
      • Regression
      • Classification: two classes or multi-class
    • Different linear models
    • Setting estimator parameters
    • Bagging several models
    • References
  • FREM: fast ensembling of regularized models for robust decoding
    • FREM pipeline
    • Empirical comparisons
      • Decoding performance increase on Haxby dataset
      • Spatial regularization of decoding maps on mixed gambles study
    • References
  • SpaceNet: decoding with spatial structure for better maps
    • The SpaceNet decoder
    • Related example
    • References
  • Searchlight : finding voxels containing information
    • Principle of the Searchlight
    • Preparing the data
      • Masking
    • Setting up the searchlight
      • Classifier
      • Score function
      • Cross validation
      • Sphere radius
    • Visualization
      • Searchlight
      • Comparing to massively univariate analysis: F_score or SPM
    • References
  • Running scikit-learn functions for more control on the analysis
    • Performing decoding with scikit-learn
      • Using scikit-learn estimators
      • Cross-validation with scikit-learn
      • Measuring the chance level
    • Going further with scikit-learn
      • Decoding without a mask: Anova-SVM using scikit-learn
      • Using any other model in the pipeline
    • Setting estimator parameters
• Functional connectivity and resting state
  • Extracting times series to build a functional connectome
    • Time-series from a brain parcellation or “MaxProb” atlas
      • Brain parcellations
      • Extracting signals on a parcellation
    • Time-series from a probabilistic atlas
      • Probabilistic atlases
      • Extracting signals from a probabilistic atlas
    • A functional connectome: a graph of interactions
    • A functional connectome: extracting coordinates of regions
  • Connectome extraction: inverse covariance for direct connections
    • Sparse inverse covariance for functional connectomes
    • Sparse inverse covariance on multiple subjects
    • Comparing the different approaches on simulated data
    • Linking total and direct interactions at the group level
  • Extracting functional brain networks: ICA and related
    • Multi-subject ICA: CanICA
      • Objective
      • Fitting CanICA model with nilearn
      • Visualizing results
      • Interpreting such components
    • An alternative to ICA: Dictionary learning
  • Region Extraction for better brain parcellations
    • Fetching movie-watching based functional datasets
    • Brain maps using Dictionary learning
    • Visualization of Dictionary learning maps
    • Region Extraction with Dictionary learning maps
    • Visualization of Region Extraction results
    • Computing functional connectivity matrices
    • Visualization of functional connectivity matrices
    • Validating results
  • Clustering to parcellate the brain in regions
    • Data loading: movie-watching data
    • Applying clustering
    • Using and visualizing the resulting parcellation
      • Visualizing the parcellation
      • Compressed representation
• Plotting brain images
  • Different plotting functions
  • Different display modes
  • Available Colormaps
  • Adding overlays, edges, contours, contour fillings, markers, scale bar
  • Displaying or saving to an image file
  • Surface plotting
  • Interactive plots
    • 3D Plots of statistical maps or atlases on the cortical surface
      • view_img_on_surf: Surface plot using a 3D statistical map
      • view_surf: Surface plot using a surface map and a cortical mesh
      • plot_surf_stat_map: Surface plot using a surface map and a cortical mesh
    • 3D Plots of connectomes
    • 3D Plots of markers
    • Interactive visualization of statistical map slices
• Analyzing fMRI using GLMs
  • An introduction to GLMs in fMRI statistical analysis
    • A primer on BOLD-fMRI data analysis
      • What is fMRI ?
      • fMRI data modelling
      • fMRI statistical analysis
      • Multiple Comparisons
  • First level models
    • HRF models
    • Design matrix: event-based and time series-based
      • Event-based
      • Time series-based
    • Fitting a first level model
      • Computing contrasts
    • Extracting predicted time series and residuals
    • Surface-based analysis
  • Second level models
    • Fitting a second level model
    • Thresholding statistical maps
    • Multiple comparisons correction
    • Voxel based morphometry
• Manipulation brain volumes with nilearn
  • Input and output: neuroimaging data representation
    • Inputing data: file names or image objects
      • File names and objects, 3D and 4D images
      • File name matching: “globbing” and user path expansion
    • Fetching open datasets from Internet
    • Understanding neuroimaging data
      • Nifti and Analyze data
      • Niimg-like objects
      • Text files: phenotype or behavior
  • Manipulating images: resampling, smoothing, masking, ROIs…
    • Functions for data preparation and image transformation
    • Resampling images
      • Resampling one image to match another one
      • Resampling to a specific target affine, shape, or resolution
    • Accessing individual volumes in 4D images
    • Computing and applying spatial masks
      • Extracting a brain mask
      • Masking data: from 4D Nifti images to 2D data arrays
    • Image operations: creating a ROI mask manually
  • From neuroimaging volumes to data matrices: the masker objects
    • The concept of “masker” objects
    • NiftiMasker: applying a mask to load time-series
      • Custom data loading: loading only the first 100 time points
      • Controlling how the mask is computed from the data
        • Visualizing the computed mask
        • Different masking strategies
        • Extra mask parameters: opening, cutoff…
      • Common data preparation steps: smoothing, filtering, resampling
        • Smoothing
        • Temporal Filtering and confound removal
        • Resampling: resizing and changing resolutions of images
      • Inverse transform: unmasking data
    • Extraction of signals from regions: NiftiLabelsMasker, NiftiMapsMasker
      • Regions definition
      • NiftiLabelsMasker Usage
      • NiftiMapsMasker Usage
    • Extraction of signals from seeds: NiftiSpheresMasker
• Advanced usage: manual pipelines and scaling up
  • Building your own neuroimaging machine-learning pipeline
    • Data loading and preprocessing
      • Downloading the data
      • Loading non image data: experiment description
      • Masking the data: from 4D image to 2D array
        • Applying a mask
        • Automatically computing a mask
    • Applying a scikit-learn machine learning method
    • Unmasking (inverse_transform)
    • Visualizing results
    • Going further
  • Downloading statistical maps from the Neurovault repository
    • Specific images or collections
    • Selection filters
    • Output
    • Neurosynth annotations
    • Examples using Neurovault
    • References