FSL (FMRIB Software Library): Neuroimaging Guide

The FMRIB Software Library (FSL) is a comprehensive software library for analyzing functional magnetic resonance imaging (fMRI), structural MRI, and diffusion-weighted brain imaging data. Developed by researchers at the University of Oxford, FSL provides command-line programs and graphical interfaces for advanced neuroimaging workflows across cognitive neuroscience and translational neurology. In contemporary neuroscience research, FSL serves as an industry-standard framework alongside platforms such as SPM and AFNI for processing volumetric neuroimaging data.

What It Is

FSL is a suite of image analysis and statistical tools designed to process 3D and 4D neuroimaging acquisitions. Originally created by the Analysis Group at the Oxford Centre for Functional MRI of the Brain (FMRIB)—now part of the Wellcome Centre for Integrative Neuroimaging (WIN)—the software operates on standard NIfTI-1 and NIfTI-2 file formats (compressed .nii.gz files). The platform handles structural tissue segmentation, linear and non-linear spatial normalization, voxel-wise diffusion tensor modeling, and model-based statistical analysis of functional blood-oxygen-level-dependent (BOLD) signals.

FSL bridges raw scan reconstruction with reproducible statistical inference. It is integrated into high-throughput scientific pipelines, cloud compute environments, and collaborative open-science repositories like those utilizing OpenNeuro and BIDS data standards. Researchers apply FSL to study anatomical variations in healthy brains, measure pathological atrophy, and map functional activation patterns across task and resting states.

Who Runs It and What It Covers

FSL is developed and maintained by the Analysis Group at the Wellcome Centre for Integrative Neuroimaging (WIN), University of Oxford. Academic and non-commercial educational users access FSL under a dedicated non-commercial research license. Commercial applications, clinical software integrations, or proprietary ventures require a commercial license issued through Oxford University Innovation (OUI).

The library covers three primary neuroimaging modalities:

Structural MRI (T1-weighted, T2-weighted): Includes the Brain Extraction Tool (BET) for skull-stripping, FMRIB’s Automated Segmentation Tool (FAST) for bias field correction and tissue classification, and FIRST for subcortical brain structure segmentation. Longitudinal atrophy and cross-sectional brain volume estimation are handled via SIENA and SIENAX.
Functional MRI (fMRI): Encompasses FMRI Expert Analysis Tool (FEAT) for model-based General Linear Model (GLM) analysis with FILM prewhitening, as well as MELODIC for model-free exploratory analysis via Probabilistic Independent Component Analysis (PICA).
Diffusion MRI (DWI/DTI): Supported by the FMRIB Diffusion Toolbox (FDT), Tract-Based Spatial Statistics (TBSS) for skeletonized white matter group analysis, and ProbtrackX for probabilistic tractography. Susceptibility and eddy current distortions are corrected using TOPUP and EDDY.
Visualization and Utilities: Visual inspection is provided by FSLeyes, the Python-based image viewer bundled with modern FSL distributions, replacing legacy tools like FSLView.

How to Use FSL: Step-by-Step Analysis Workflow

Researchers execute FSL through both interactive graphical interfaces and shell scripts. The following six-step tutorial outlines a standard end-to-end analytical pipeline for processing structural and task-based functional MRI data:

FSL Download and Installation: Download the official FSL installation script from the Oxford FMRIB portal. Install the software via conda environments on Linux, macOS (native Apple Silicon and Intel support), or Windows via the Windows Subsystem for Linux (WSL). Set environment variables ($FSLDIR and path configurations) to enable command-line tools.
Data Conversion to NIfTI: Convert raw scanner DICOM files to 4D NIfTI format using dcm2niix, adhering to the Brain Imaging Data Structure (BIDS) naming conventions. Inspect data dimensions and voxel orientation.
Brain Extraction (FSL BET): Run the Brain Extraction Tool on the structural T1 image using bet input_t1.nii.gz output_brain.nii.gz -f 0.5 -g 0 to remove non-brain tissue (skull, scalp, neck) and generate an isolated brain mask.
Tissue Segmentation (FSL FAST): Segment the brain-extracted volume into grey matter, white matter, and cerebrospinal fluid (CSF) using fast -t 1 -n 3 -H 0.1 -I 4 -l 20.0 output_brain.nii.gz. FAST simultaneously corrects spatial intensity variations caused by RF coil inhomogeneities.
Spatial Registration (FLIRT and FNIRT): Align functional images to high-resolution structural scans using FLIRT (affine linear registration) and normalize structural scans to standard stereotaxic space (MNI152) using FNIRT (non-linear registration).
Model-Based fMRI Analysis (FSL FEAT): Configure a first-level analysis in the FEAT interface or generate a design script (design.fsf). FEAT performs motion correction via MCFLIRT, high-pass temporal filtering, prewhitening via FILM, and GLM contrast fitting across experimental conditions.
Interactive Quality Control (FSLeyes): Launch fsleyes to verify brain extraction quality, confirm registration overlays between functional and MNI template volumes, and inspect statistical z-statistic maps.

Strengths and Limits

Evaluating FSL within an institutional neuroimaging lab involves balancing analytical capabilities against software requirements:

Strengths

Methodological Rigor: FSL incorporates robust statistical algorithms, including FILM prewhitening for temporal autocorrelation in fMRI and FLAME for mixed-effects multi-level group statistics.
High-Performance Automation: All GUI actions generate reproducible shell scripts and parameter files (.fsf), allowing seamless execution on High-Performance Computing (HPC) clusters and containerized pipelines.
Specialized Diffusion Tools: Tools like TOPUP, EDDY, and TBSS remain gold standards for correcting and quantifying diffusion MRI datasets.
Modern Visualization: FSLeyes provides dynamic 3D rendering, customizable colour look-up tables, atlas integration, and Python scripting support.

Limits

OS Dependencies: Native Windows execution is not supported; Windows researchers must run FSL through WSL2 or virtualized containers.
Commercial Licensing Restrictions: While free for academic research, commercial deployment requires formal licensing agreements through Oxford University Innovation.
GUI Customization Constraints: Advanced batch workflows demand shell scripting proficiency, which presents a learning curve for researchers accustomed to graphical environments.

Alternatives

Neuroimaging researchers select analysis packages based on study modalities, programming ecosystems, and institutional infrastructure. The table below compares FSL with other major neuroimaging platforms:

Platform	Primary Modalities	Runtime Environment	Licensing Model	Scripting Interface
FSL	fMRI, Structural MRI, Diffusion (DTI/DWI)	Linux, macOS, Windows (WSL)	Free academic; commercial via OUI	Bash, Python / Conda, C++
SPM12	fMRI, PET, Structural VBM, EEG/MEG	MATLAB / Standalone runtime	GNU General Public License (GPL)	MATLAB scripts, batch GUI
AFNI	fMRI, Surface analysis, Real-time MRI	Linux, macOS (Unix-native)	Open-source (GPL)	C, Tcsh / Bash, Python
FreeSurfer	Cortical surface reconstruction, subcortical parcellation	Linux, macOS	Open-source (FreeSurfer License)	Shell scripts, C / C++

Investigators examining cellular and systems-level neurophysiology across neuroscience and neurobiology often combine FSL with cortical surface mapping tools to analyze multi-modal structural and functional connectivity.

Frequently Asked Questions

Is FSL free for academic research?

Yes. FSL is freely accessible for non-commercial academic research and educational use. For commercial applications or revenue-generating activities, institutions must obtain a commercial license through Oxford University Innovation.

What is the difference between BET and FAST in FSL?

BET (Brain Extraction Tool) removes skull, scalp, and non-brain tissue to isolate the cerebral volume. FAST (FMRIB’s Automated Segmentation Tool) operates on the brain-extracted image to segment neural tissue into grey matter, white matter, and cerebrospinal fluid while correcting RF coil bias field inhomogeneities.

How does FSL handle fMRI statistical analysis?

FSL performs fMRI analysis primarily through FEAT. FEAT applies a General Linear Model (GLM) using FILM (FMRIB’s Improved Linear Model), which prewhitens timeseries data to account for temporal autocorrelation. Multi-subject group statistics are modeled using FLAME (FMRIB’s Local Analysis of Mixed Effects) or non-parametric permutations via Randomise.

Can FSL run on Windows?

FSL does not offer a native Windows binary. Windows users can install FSL through the Windows Subsystem for Linux (WSL2), running an Ubuntu distribution, or via containerized environments such as Docker and Singularity.

Sources

University of Oxford, Wellcome Centre for Integrative Neuroimaging (WIN) / FMRIB Analysis Group. FSL Documentation: Overview and Installation. https://fsl.fmrib.ox.ac.uk/fsl/docs/ (2026-11-08).
University of Oxford, WIN / FMRIB Analysis Group. BET (Brain Extraction Tool) User Guide and Reference. https://fsl.fmrib.ox.ac.uk/fsl/docs/structural/bet.html (2026-11-08).
University of Oxford, WIN / FMRIB Analysis Group. FAST (FMRIB’s Automated Segmentation Tool) User Guide. https://fsl.fmrib.ox.ac.uk/fsl/docs/structural/fast.html (2026-11-08).
University of Oxford, WIN / FMRIB Analysis Group. FEAT (FMRI Expert Analysis Tool) Overview and Design Rules. https://fsl.fmrib.ox.ac.uk/fsl/docs/task_fmri/feat/index.html (2026-11-08).
University of Oxford, WIN / FMRIB Analysis Group. FSLeyes Image Viewer Documentation. https://fsl.fmrib.ox.ac.uk/fsl/docs/utilities/fsleyes.html (2026-11-08).
University of Oxford, Oxford University Innovation. FSL License Agreement and Commercial Inquiries. https://fsl.fmrib.ox.ac.uk/fsl/docs/license.html (2026-11-08).