Diffusion Imaging in Python

DIPY, short for Diffusion Imaging in Python, is an open-source Python library. Its main emphasis is diffusion MRI of the brain. The library also covers normalizing images across space, processing signals, applying machine learning, statistics and visual display of medical scans, along with specialized methods for computational anatomy, including diffusion, perfusion and structural imaging. The project describes itself as a user-friendly library that keeps growing, and it is written in Python.

Diffusion models for single-shell and multi-shell data

The library provides a wide range of diffusion models. For single-shell data it includes methods such as DTI, CSA, SFM, SDT, Q-Ball and CSD. For multi-shell data it includes GQI, DTI, DKI, SHORE, MAPMRI and MSMT-CSD, and the project says there are many more. These models let researchers estimate the structure of tissue from the way water diffuses through it.

Registration, tractography and tract analysis

DIPY offers affine registration, diffeomorphic 2D and 3D registration, and registration based on streamlines, which aligns the fiber pathways of different scans. Its tractography options include probabilistic tracking, deterministic tracking and PFT tracking, which follow white matter pathways. The project also supports analysis approaches such as BUAN, AFQ and K-fold cross-validation, which help researchers compare pathways across people.

Noise reduction, preprocessing and brain extraction

The preprocessing tools include Patch2Self, non-local means, Gibbs unringing, MP-PCA and L-PCA for reducing noise and artifacts. The library also offers brain extraction, estimation of the signal-to-noise ratio and the reslicing of datasets, so that data can be prepared before analysis.

Visualization and command-line workflows

The library visualizes orientation distribution functions and provides interactive tractogram viewing, with an advanced user interface and shaders. Every algorithm can be run from the command line, and the project says users can also create their own algorithms, which makes the library suited to workflows that need to be repeated across many scans.

Published methods and recent research

The project's own methods have appeared in peer-reviewed journals. These include BUAN, published in Scientific Reports in 2020, EVAC+, published in Communications Medicine in 2024, and BundleWarp, published in Medical Image Analysis in 2026. The project's site presents these as examples of methods built on the library and developed through research.

Support, consulting and training

The project is open source, and it notes that partners can provide professional support and services. These cover consulting, custom development and training for organizations that need help with their imaging work. The project lists its sponsors, which include funders that have supported its development, and it asks users to cite the original 2014 paper when they publish work that relies on it.