---

The Problem

FreeSurfer's recon-all produces cortical statistics for its built-in atlases (Desikan-Killiany, Destrieux, DKT), but extracting measures with any other atlas requires manually chaining together multiple commands:

mri_surf2surf → mris_anatomical_stats → parse output
mri_vol2vol   → mri_segstats         → parse output

fsatlas automates this entire workflow into a single command — for any atlas, for any number of subjects.

---

Features

• 22 built-in atlases — Schaefer 2018 (100–1000 parcels × 7/17 networks), Tian 2020 subcortical (Scales I–IV), HCP-MMP1, DKT, Desikan, Destrieux. Auto-downloads on first use.
• Custom atlas support — Point at any .annot (surface) or .nii.gz (volumetric, MNI space) file.
• Surface pipeline — Transfers annotations from fsaverage → subject via mri_surf2surf, extracts 9 cortical measures via mris_anatomical_stats.
• Volumetric pipeline — Registers MNI-space NIfTI → native space via mri_vol2vol + talairach.xfm, extracts 7 subcortical measures via mri_segstats.
• Tidy TSV output — Long-format output: subject_id | atlas | hemisphere | region | measure | value.
• Batch processing — Process all subjects in $SUBJECTS_DIR or a specified list.
• Failure resilience — Pipeline continues on per-subject errors; all failures logged to a separate TSV.
• Docker image — Run without a local FreeSurfer installation.

---

Requirements

Requirement	Version
Python	≥ 3.10
FreeSurfer	8.x
FREESURFER_HOME	must be set
SUBJECTS_DIR	must be set

Subjects must have completed recon-all.

---

Installation

pip install fsatlas

From source:

git clone https://github.com/GalKepler/fsatlas.git
cd fsatlas
pip install -e ".[dev]"

---

Quick Start

List available atlases

fsatlas list-atlases

Extract Schaefer 100-parcel cortical stats for all subjects

fsatlas extract --atlas schaefer100-7 --output-dir ./results

Target specific subjects

fsatlas extract --atlas schaefer100-7 -s sub-01 -s sub-02 -o ./results

Load subjects from a file

fsatlas extract --atlas schaefer400-17 --subjects-file subjects.txt -o ./results

Extract subcortical stats with the Tian atlas

fsatlas extract --atlas tian-s2 -o ./results

Use a custom surface atlas

# Provide one hemisphere; the other is auto-detected
fsatlas extract --atlas /path/to/lh.myatlas.annot -o ./results

Use a custom volumetric atlas (MNI space)

fsatlas extract --atlas /path/to/subcortical_atlas.nii.gz -o ./results

Pre-download an atlas

fsatlas download schaefer400-7

---

Built-in Atlas Catalog

ID	Family	Type	Parcels	Space
schaefer100-7	Schaefer 2018	Surface	100	fsaverage
schaefer200-7	Schaefer 2018	Surface	200	fsaverage
schaefer300-7	Schaefer 2018	Surface	300	fsaverage
schaefer400-7	Schaefer 2018	Surface	400	fsaverage
schaefer500-7	Schaefer 2018	Surface	500	fsaverage
schaefer600-7	Schaefer 2018	Surface	600	fsaverage
schaefer700-7	Schaefer 2018	Surface	700	fsaverage
schaefer800-7	Schaefer 2018	Surface	800	fsaverage
schaefer900-7	Schaefer 2018	Surface	900	fsaverage
schaefer1000-7	Schaefer 2018	Surface	1000	fsaverage
schaefer100-17	Schaefer 2018	Surface	100	fsaverage
schaefer200-17	Schaefer 2018	Surface	200	fsaverage
schaefer300-17	Schaefer 2018	Surface	300	fsaverage
schaefer400-17	Schaefer 2018	Surface	400	fsaverage
tian-s1	Tian 2020	Volumetric	16	MNI152NLin6Asym
tian-s2	Tian 2020	Volumetric	32	MNI152NLin6Asym
tian-s3	Tian 2020	Volumetric	50	MNI152NLin6Asym
tian-s4	Tian 2020	Volumetric	54	MNI152NLin6Asym
hcp-mmp	HCP-MMP 1.0	Surface	360	fsaverage
desikan	FreeSurfer	Surface	68	built-in
destrieux	FreeSurfer	Surface	148	built-in
dkt	FreeSurfer	Surface	62	built-in

---

Output Format

Results are written as long-format (tidy) TSV files to the specified output directory.

Cortical ({atlas}_cortical.tsv):

subject_id	atlas	hemisphere	region	measure	value
sub-01	schaefer100-7	lh	7Networks_LH_Vis_1	thickness_mean_mm	2.341
sub-01	schaefer100-7	lh	7Networks_LH_Vis_1	surface_area_mm2	843.0
sub-01	schaefer100-7	lh	7Networks_LH_Vis_1	gray_matter_volume_mm3	2110.0

Cortical measures: thickness_mean_mm, surface_area_mm2, gray_matter_volume_mm3, mean_curvature, gauss_curvature, fold_index, curvature_index, integrated_rect_curvature, eTIV

Subcortical ({atlas}_subcortical.tsv):

subject_id	atlas	hemisphere	region	measure	value
sub-01	tian-s2	lh	CAU-lh	volume_mm3	1248.0
sub-01	tian-s2	lh	CAU-lh	intensity_mean	72.34

Subcortical measures: volume_mm3, intensity_mean, intensity_std, intensity_min, intensity_max, intensity_range, voxel_count

Failures ({atlas}_failures.tsv) — subjects that could not be processed, with error messages.

---

Docker

docker build -t fsatlas .

docker run --rm \
    -v /path/to/SUBJECTS_DIR:/subjects \
    -v /path/to/license.txt:/opt/freesurfer/license.txt \
    -e SUBJECTS_DIR=/subjects \
    fsatlas extract --atlas schaefer100-7 -o /subjects/results

The Docker image is based on freesurfer/freesurfer:8.0.0 and includes a self-contained Python environment at /opt/fsatlas-venv.

---

Architecture

src/fsatlas/
├── cli/
│   └── main.py           # Click CLI: extract, list-atlases, download
├── atlases/
│   ├── catalog.yaml      # Built-in atlas definitions (13 atlases)
│   └── registry.py       # Atlas loading, downloading, custom atlas support
└── core/
    ├── environment.py    # FreeSurfer detection, subject discovery, path resolution
    ├── pipeline.py       # Orchestrator: validate → transfer → extract → aggregate
    ├── transfer.py       # mri_surf2surf / mri_vol2vol wrappers (600s timeout)
    └── extract.py        # mris_anatomical_stats / mri_segstats + long-format parsing

Data flow:

CLI
 └─ resolve atlas + discover subjects
     └─ Pipeline (per subject)
         ├─ validate subject directory structure
         ├─ transfer atlas → subject space (cached)
         │   ├─ surface: mri_surf2surf (fsaverage → subject)
         │   └─ volumetric: mri_vol2vol (MNI → native via talairach.xfm)
         └─ extract stats
             ├─ cortical: mris_anatomical_stats → 9 measures (long format)
             └─ subcortical: mri_segstats → 7 measures (long format)
 └─ concatenate → {atlas}_cortical.tsv / {atlas}_subcortical.tsv

---

Development

pip install -e ".[dev]"
pytest                 # run tests
ruff check src/        # lint
ruff format src/       # format
mypy                   # type check

---

Citation

If you use fsatlas in your research, please cite the relevant atlas papers (shown via fsatlas list-atlases) and:

fsatlas: Automated atlas-based morphometry extraction for FreeSurfer. https://github.com/GalKepler/fsatlas

---

License

MIT © Gal Kepler