Coord2Region

Coord2Region maps brain coordinates (or atlas region names) to anatomical labels, nearby studies, LLM summaries, and optional AI-generated images. It combines NiMARE, Nilearn, and MNE under a single CLI/Python API and ships with a companion web interface for configuration authoring.

Why Coord2Region?

• Atlas + studies in one stop. Fetch atlases, convert MNI ↔ Talairach, and query datasets such as Neurosynth, NeuroQuery, and NiMARE without wiring them up yourself.
• Optional AI enrichments. Provide API keys once (OpenAI, Gemini, Hugging Face, etc.) and the same workflow can emit human-friendly summaries or illustrative images.
• Reproducible outputs. Every command can emit YAML, JSON, and CSV artefacts so collaborators can re-run the exact pipeline.
• Browser builder. The React/Vite builder mirrors the CLI schema so first-time users can generate configs and commands without installing Python up front.

Quick Start

1. Install the package (Python 3.10-3.14):

   python -m venv .venv
   source .venv/bin/activate
   pip install coord2region
   

2. Configure credentials and defaults. Run the helper once to create a private config/coord2region-config.yaml. It covers atlas fetch directories as well as AI provider API keys (all optional).

   python scripts/configure_coord2region.py
   

   Prefer environment variables? Set OPENAI_API_KEY, GEMINI_API_KEY, HUGGINGFACE_API_KEY, etc. instead of generating the YAML.

3. Run a CLI recipe.

   # Atlas labels only
   coord2region coords-to-atlas 30 -22 50 --atlas harvard-oxford
   
   # Labels + studies + LLM summary (requires API key)
   coord2region coords-to-summary 30 -22 50 --atlas harvard-oxford --model gemini-2.0-flash
   
   # Region name workflow
   coord2region region-to-insights "Left Amygdala" --atlas harvard-oxford
   

4. Explore the builder. Visit the Config Builder to generate YAML/CLI commands interactively. Import/export configs to stay in sync with local runs.

5. Jump into Python (optional).

   from coord2region import AtlasFetcher, AtlasMapper, AIModelInterface, generate_summary
   
   atlas = AtlasFetcher().fetch_atlas("harvard-oxford")
   mapper = AtlasMapper("harvard-oxford", atlas["vol"], atlas["hdr"], atlas["labels"])
   print(mapper.mni_to_region_name([30, -22, 50]))
   
   ai = AIModelInterface(huggingface_api_key="YOUR_KEY")
   studies = []  # populate via coord2region.coord2study helpers
   print(generate_summary(ai, studies, [30, -22, 50]))
   

CLI recipes at a glance

All commands emit YAML/JSON/CSV outputs under coord2region-output/ by default. Use --result-dir to customise the export path.

Web interface

The web interface mirrors the CLI schema and lives at babasanfour.github.io/Coord2Region. It provides:

• Guided forms for inputs (coordinates or region names), atlas selection, study radius, summaries, and image options.
• Live YAML + CLI previews you can copy or download.
• Presets to learn common workflows (single peak lookup, region → coords, multi-peak insights).
• Import/export so you can iterate on a config in the browser and run the CLI locally.

Builder (inputs & atlas)	Builder (outputs & providers)	Runner

To preview or hack on the web stack locally, follow web-interface/README.md (Vite dev server + Jekyll shell + Playwright tests).

Further reading

• Documentation – user guide, pipeline walkthrough, API reference, tutorials.
• Examples gallery
• Web interface overview
• License · Contributing · Code of Conduct · Security Policy
• Preprint

API workflow

A compact overview of the Coord2Region pipeline: shows how inputs (coordinates or region names) are mapped to atlas labels, linked to study results, optionally enriched by AI summaries/images, and exported as reproducible artifacts.