SynthSeg 2.0 Model Weights

SynthSeg 2.0 pre-trained model weights (~355 MB total).

These files are needed for the pure-Python SynthSeg fallback on systems without FreeSurfer.
The original Dropbox/SharePoint download links are dead (see BBillot/SynthSeg#116).

Files:

• synthseg_robust_2.0.h5 (205 MB) — robust segmentation (slower, more accurate)
• synthseg_2.0.h5 (51 MB) — fast segmentation
• synthseg_parc_2.0.h5 (51 MB) — cortical parcellation
• synthseg_qc_2.0.h5 (48 MB) — QC scores

Source: FreeSurfer 7.4.1 (/Applications/freesurfer/7.4.1/models/)
License: Apache 2.0 (same as SynthSeg)

Stable v2.0.0

Fully automated mode using 4 neural networks trained on identifying veins and arteries, as well as the implementation of FastSurfer Segmentation.

stable v1.1.1

Addons have been included as optional downloads, and bug fixes made to ensure stability and flexibility.

stable v1.1.0 (addons)

In this release an addon framework has been added such that they can be dynamically loaded into the addon folder and loaded from the menu.

An addon, "boundary", is available for download (see here) which allows for the segmentation and analysis of the grey/white matter boundary of an axial t2 slice. It is quite specific, and requires an axial T1 and T2 reconstruction from 3D, as well as a DCE and attached T2 axial image. To run the segmentation, fsl_anat is required.

stable v1.0.0

This is the first stable version of the program. It contains no extra addons or features that the subsequent releases will, however this is a minimal version that can be used to modify, extend or use as a lightweight neuroimaging solution without too many bells and whistles.