Installation

AutoHS can be installed as a BIDS App using Docker, Apptainer/Singularity, or directly from source on HPC systems. For a first run after install, see Quick Start.

Architecture

The published BIDS App image (autohs/autohs) is an orchestrator: it discovers BIDS T1w scans, invokes segmentation (FreeSurfer or FastSurfer in a separate container or SIF), then runs AI-compute (hippocampal volumes, asymmetry index, reports). Segmentation backends are not embedded in the orchestrator image — you provide Docker images or Apptainer SIF paths at runtime (see below).

Docker Container

Pull the published image (after maintainers configure Docker Hub):

docker pull autohs/autohs:latest

Or build locally:

docker build -f docker/Dockerfile.bidsapp -t autohs/autohs:latest .

Run examples: Quick Start.

Apptainer / Singularity Container

Download a release artifact or build from Docker:

apptainer build autohs.sif docker://autohs/autohs:latest

Or download autohs_<version>.sif from GitHub Releases.

Set segmentation SIF paths when running outside the bundled Docker stack:

export FREESURFER_SIF=/path/to/freesurfer_7.4.1.sif
export FASTSURFER_SIF=/path/to/fastsurfer_latest.sif
export AUTOHS_RUNTIME=apptainer

These variables are required for Apptainer runs (no site-specific defaults are shipped in the repository).

Python / Source Install

For development or Apptainer-only HPC nodes:

git clone https://github.com/phindagijimana/AutoHS.git
cd AutoHS
./AutoHS install
pip install -r requirements-bids.txt

External Dependencies

Dependency

Required for

Notes

T1w BIDS dataset

Always

At least one *_T1w.nii*

FreeSurfer

Default segmentation

license.txt required

FastSurfer

--fastsurfer

No FreeSurfer license needed

Docker

Default runtime when available

Or Apptainer on HPC

bids-validator

Optional

Use --skip-bids-validator to skip

FreeSurfer license

Place license.txt in the repo root or pass --fs-license-file. FreeSurfer is not required when using --fastsurfer.

Documentation is hosted at https://autohs.readthedocs.io/en/latest/.