Volumetric data rendering with Neuroglancer

Introduction

Main Features:

• Voxel-based rendering: Efficiently visualize large 3D datasets by streaming the data directly into the browser.
• Support for multiple data formats: Neuroglancer works with a variety of formats like ZARR and OME-ZARR for volumetric images, as well as annotation formats.
• Interactive, shareable views: Customize and share views by copying the URL directly from the Neuroglancer interface.

Available datasets

Popular datasets using Neuroglancer

• MICrONS Explorer: A large-scale dataset from the MICrONS Project, providing high-resolution volumetric reconstructions of a mouse brain.
• FlyEM Hemibrain: This dataset offers a detailed 3D reconstruction of the Drosophila melanogaster brain at nanometer resolution.
• “H01” Dataset: 1.4 Petabyte for one cubic millimeter of the human brain, released by the Harvard University and the Connectomics at Google team.

Dataset requirements

Supported Data Types

• Neuroglancer precomputed format
• N5
• Zarr v2/v3 (including OME-ZARR)
• Python in-memory volumes (with automatic mesh generation)
• BOSS , DVID, Render, Single NIfTI files, Deep Zoom images

Data preparation

Volumetric images

Conversion Steps:

• Step 1: Convert the TIFF slices into ZARR format, which supports chunked, compressed storage.
• Step 2: Convert the ZARR files into OME-ZARR format, which includes metadata for biological imaging data and supports better integration with visualization tools like Neuroglancer.

Notebook: TIFF slices to multiscale OME-ZARR

Streaming data locally

With Python:

• Step 1: Use Python’s built-in HTTP server to serve the data directory:

  python3 -m http.server
  

• Step 2: Open the Neuroglancer demo page and enter the local URL of your data (e.g., zarr://http://localhost:8000/my-dataset.ome.zarr).

Sharing views

Example of JSON in URL:

• Copy the URL from your browser, and you’ll notice that it contains encoded parameters for layers, views, and more.
• The JSON configuration embedded in the URL can be customized to adjust the view programmatically.

{
  "dimensions": {...},
  "position": [
    3744.089599609375,
    4562.4326171875,
    837.0703125
  ],
  "layers": [
    {
      "type": "image",
      "source": "zarr://https://hifis-storage.desy.de:2443/Helmholtz/HIP/collaborations/2405_MDC_Treier/public/G5111-S4/647nm_cFOS",
      "tab": "rendering",
      "shaderControls": {
        "normalized": {
          "range": [
            0,
            500
          ]
        }
      },
      "crossSectionRenderScale": 0.47742080195520836,
      "name": "cFOS"
    }
    {
      "type": "annotation",
      "source": "precomputed://https://hifis-storage.desy.de:2443/Helmholtz/HIP/collaborations/2405_MDC_Treier/public/G5111-S4/cells",
      "tab": "source",
      "name": "Cells"
    }
  ],
  "selectedLayer": {
    "visible": true,
    "layer": "cFOS"
  }
}

Programmatically generate URL

Link to the full notebook showing how to generate a Neuroglancer URL programmatically:

Data hosting for Neuroglancer

• Local Hosting: Serve data from your own machine or institution’s servers.
• AWS S3 or Google Cloud Storage: Store large datasets and stream them to Neuroglancer from the cloud.
• Scientific Data Services: e.g. BioImage Archive
• Do you know other compatible hosting places? Let us know!