Deep-MI/FastSurfer

PyTorch implementation of FastSurferCNN

FastSurfer – Quick‑look

What it does – FastSurfer is a deep‑learning‑powered neuro‑imaging pipeline that replaces the classic FreeSurfer workflow. It takes a high‑resolution T1‑weighted MRI (and optionally a T2) and produces:

  • Whole‑brain tissue and cortical‑parcellation (95 classes) in a few minutes on a GPU.
  • Sub‑structures such as the corpus callosum, cerebellum, and hypothalamus via dedicated networks.
  • Surface reconstruction and cortical‑thickness maps in ~20‑40 min, comparable to FreeSurfer but far faster.

Why it matters – Traditional FreeSurfer can take many hours per subject and struggles with sub‑millimetre data. FastSurfer’s VIN‑style networks are resolution‑independent, so they run quickly even on 0.7 mm isotropic scans, enabling large‑scale studies and higher‑resolution analyses.

How to get it – Pre‑built Docker, Apptainer/Singularity, or macOS packages are provided; a native install is also possible. All containers bundle the required software, but you must supply a free FreeSurfer license for the surface stage.

Typical workflow

  1. run_fastsurfer.sh with flags like --t1 <path> --sd <subjects_dir> --sid <id> --fs_license <file>.
  2. Choose modules: --seg_only for fast segmentation, --surf_only for surface reconstruction, or run the full pipeline.
  3. Optional extensions: FastSurfer‑LIT adds lesion inpainting when a mask is given.

Outputs – Segmentation volumes (asegdkt, cerebnet, hypvinn), corpus callosum masks, surface files (.surf, .thickness), and CSV/TSV statistics ready for downstream FreeSurfer tools.

Hardware – GPU‑accelerated mode is default; CPU‑only works but is much slower. Recommended: 6‑core CPU, ≥16 GB RAM, NVIDIA ≥2016 with ≥12 GB VRAM (or Apple‑silicon for macOS).

Citations – If you publish, cite the series of NeuroImage papers listed in the README (FastSurfer, FastSurferVINN, CerebNet, FastSurfer‑HypVINN, FastSurfer‑CC, FastSurfer‑LIT).

Related

  • Project
    freesurfer/freesurfer

    FreeSurfer is an open-source software suite for processing, analyzing, and visualizing human brain MRI data from structural and functional studies.

  • Project
    iot-salzburg/gpu-jupyter

    GPU‑Jupyter provides a Docker‑based, GPU‑enabled JupyterLab environment pre‑installed with Python, PyTorch, TensorFlow, Julia, and R. It offers versioned images, easy customization, and Docker‑Compose support, enabling reproducible deep‑learning experiments on any machine with an NVIDIA GPU.

  • Project
    hao-ai-lab/FastVideo

    FastVideo is an open‑source framework for training, fine‑tuning, and real‑time inference of video diffusion models. It offers sparse‑attention distillation, scalable multi‑GPU training, and cross‑platform support (NVIDIA GPUs, Apple Silicon via MLX, ARM DGX Spark). The library includes a CLI, Python API, and the Dreamverse app for live video editing. Install via `uv` and start generating videos with a few lines of Python.

  • Project
    FAST-Imaging/FAST

    An open-source framework for high-performance medical image processing, neural network inference, and real-time visualization using CPUs and GPUs.

  • Project
    perseval-BLR/NeuralScreen

    NeuralScreen is a Windows‑only overlay that runs NVIDIA’s DLSS 5 neural renderer on the whole desktop (or a single window) in real time, optionally adding DLSS‑G frame generation. It works with RTX GPUs, offers hot‑key control, per‑window mode, recording, and a configurable UI.