Self-hosted research workspace

Neuroimaging workflows.
One Docker install.

Inspect MRI data, organize cases, run FastSurfer, and explore results with a viewer-aware assistant.

Install NeuroCade Explore sample case GitHub
NeuroCade sample case workspace showing MRI views, layers, and assistant panel
One workspace

From source image to
surface reconstruction

Keep case organization, interactive review, processing, and results together in a browser-based research environment.

Cases

Research data workspace

Create workspaces and cases, then import NIfTI volumes, DICOM ZIP archives, or existing FreeSurfer outputs.

Data and outputs persist in a host directory you configure.
Viewer

Volumes, labels, and surfaces

Review MRI volumes and segmentation overlays alongside cortical surfaces, with synchronized navigation and label lookup.

Load FastSurfer outputs directly into the active case.
Processing

FastSurfer built in

Launch FastSurfer cortical reconstruction and segmentation from the interface and follow long-running job status.

Current configured analysis runtime: FastSurfer.
Integrated processing

FastSurfer,
ready in the workspace

Run structural MRI segmentation and surface reconstruction from the active case, then review generated volumes, labels, statistics, and surfaces in the same interface.

FastSurfer logo
FastSurfer Structural image analysis

Whole-brain segmentation

Corpus callosum morphometry

Cerebellum subsegmentation

Hypothalamus subsegmentation

Lesion-robust processing

Cortical surface reconstruction

Viewer-aware assistant

Work with the
active case in context

The optional assistant can inspect case files, control the viewer, read segmentation statistics, and launch configured workflows.

Connected to the active case

Ask about the volume, segmentation, surface, or label currently under review.

Actions, not just answers

Run configured tools, query measured statistics, and load generated results back into the viewer.

Your choice of provider

Configure OpenAI-compatible, Anthropic, Google, or Ollama models—or install without an LLM.

Assistant
Report the left and right hippocampal volumes from the FastSurfer segmentation.
NeuroCade I read the measured segmentation statistics for the active case.
Left hippocampus3,182 mm³
Right hippocampus3,409 mm³
Focus the viewer on the left hippocampus.
NeuroCadeThe crosshair is now centered on the left hippocampus and its segmentation label is highlighted.
Ask about the active case…
How it runs

A reproducible stack,
without the setup maze

The web interface, API, and background jobs ship together. Processing tools run through a dedicated container runtime.

01 / Browser

NeuroCade UI

Open the workspace from a local machine or an internally deployed server.

02 / Docker

Single app container

The built web UI, API, job worker, and Apptainer runtime ship in one published image.

03 / Storage

Mounted host data

Cases, outputs, tool images, and the SQLite database persist outside the app container.

04 / Apptainer

Neuroimage Containers

Run neuroimaging packages, such as FastSurfer, against the active case data. A large catalogue of tools is provided by NeuroDesk.

Know your data boundary. Image processing and case storage run on the host you configure. If you enable an external AI provider, assistant requests are sent according to that provider’s configuration and policies. Choose Ollama for a host-local model, or install with --llm-provider no-llm.
Quick install

Docker is the only
runtime dependency

The installer creates the local configuration and data directories, pulls the published image, downloads the prepared sample case, and starts NeuroCade.

The default local profile binds to your machine and opens at http://localhost:8000. No host Python environment, Node.js installation, or Docker Compose setup is required.

TerminalLOCAL INSTALL
bash <(curl -fsSL https://raw.githubusercontent.com/Deep-MI/NeuroCade/main/scripts/install.sh) --mode local
PrerequisiteDocker Engine or Docker Desktop
Applicationhttp://localhost:8000
AI setupCloud, Ollama, or no LLM
Read the installation guide ↗ · Versioned releases ↗

Local, without an assistant

Install the viewer, case workspace, and processing runtime without configuring an LLM provider.

bash <(curl -fsSL https://raw.githubusercontent.com/Deep-MI/NeuroCade/main/scripts/install.sh) --mode local --llm-provider no-llm

Local assistant with Ollama

Connect NeuroCade to an Ollama model running on the Docker host.

bash <(curl -fsSL https://raw.githubusercontent.com/Deep-MI/NeuroCade/main/scripts/install.sh) --mode local --llm-provider ollama

Internal server

Bind for internal access and configure the authentication values requested by the installer.

bash <(curl -fsSL https://raw.githubusercontent.com/Deep-MI/NeuroCade/main/scripts/install.sh) --mode internal

Reproducible release

Pin an exact published GHCR image tag instead of following the latest channel.

bash <(curl -fsSL https://raw.githubusercontent.com/Deep-MI/NeuroCade/main/scripts/install.sh) --mode local --image ghcr.io/deep-mi/neurocade:<tag>

Ready to explore NeuroCade?

Start with the read-only sample case, install the research preview locally, or join the project on GitHub.

Explore sample case GitHub issues Contact