FastSurfer is a fast, accurate, and extensively validated pipeline for fully automated processing of structural human brain MRI — with FreeSurfer-conform outputs.
Whole-brain segmentation in under a minute · full surface reconstruction in about an hour · free & open source.
Everything you expect from a full surface-based morphometry pipeline — at a fraction of the runtime.
Whole-brain segmentation in under a minute on a GPU; the full surface pipeline in about an hour — instead of many hours.
Extensively validated for accuracy, generalizability, reliability and sensitivity across many public datasets.
Produces outputs compatible with FreeSurfer, so it drops into your existing analysis workflows.
Built for big-data cohorts and time-critical clinical use. Free and open source, with an active community.
Run the segmentation pipeline on a single T1-weighted scan. Choose the option that fits your setup: Apptainer for HPC clusters, Docker for workstations and servers, or the native package for macOS. See the documentation for GPU setup and all options.
Full documentation →On macOS? Use the native package installer — see the installation guide →
FastSurfer combines an advanced neural-network segmentation (FastSurferCNN / VINN) with a fast surface pipeline (recon-surf) for volumes, cortical surfaces, thickness and parcellations.
Integrated module for fast, reliable and detailed cerebellum sub-segmentation.
Integrated module for automated sub-segmentation of the hypothalamus and adjacent structures.
Automated morphometry of the corpus callosum and its sub-regions, derived from the whole-brain segmentation.
Lesion inpainting module enabling robust surface reconstruction in brains with tumors or cavities.
Explore a real FastSurfer whole-brain segmentation interactively in your browser with NeuroCade. No installation, no account required.
Open in NeuroCade ↗Web-based viewer · full-brain segmentation · runs in any modern browser
FastSurfer and its modules are backed by peer-reviewed publications. Using FastSurfer in your work? Please cite the relevant papers.
Get up and running with the documentation, or explore the source on GitHub.
