Content
# MedGov-AI
An agentic AI platform for clinical imaging orchestration. MedGov-AI connects a reasoning agent to medical AI tools via the [Model Context Protocol (MCP)](https://modelcontextprotocol.io), enabling autonomous multi-step clinical workflows from a natural language interface.
## What it does
- **Autonomous analysis** - Upload DICOM series or pathology images and describe what you want. The agent selects tools, runs inference, and produces a structured clinical report without manual intervention.
- **Segmentation and cell detection** - MONAI models for volumetric organ segmentation on CT/MRI; Cellpose v4 for cell detection and counting on pathology ROI images.
- **Structured report generation** - RadLex/RadReport templates for radiology; pathology-specific templates for whole-slide image analysis.
- **Workspace monitoring** - Register a directory as a workspace. When files arrive (e.g. from a PACS or scanner), the agent automatically analyses them and writes results to the configured output folder.
- **Skills** - Plain-text workflow protocols that guide the agent through domain-specific clinical sequences, improving reliability and reducing incorrect tool use.
- **Background tasks** - Inference jobs run in the background with real-time progress streamed to the UI via SSE.
- **iPath integration** - Connects to the iPath telepathology platform for whole-slide image retrieval and ROI analysis.
- **FHIR integration** - Writes clinical findings as FHIR resources (Patient, Observation) to a connected FHIR server.
## Architecture
## Quick start
**Requirements:** Python 3.12, Node.js 18+, a Gemini API key or local Ollama instance. GPU recommended for inference.
```bash
# 1. Copy and fill in your credentials
cp orchestrator/.env.example orchestrator/.env
# edit orchestrator/.env: set LLM_BACKEND, GEMINI_API_KEY, APP_ROOT
# 2. Run everything
./run.sh
```
The script sets up all virtual environments, installs dependencies, and starts the backend and frontend.
- UI: http://localhost:5173
- API: http://localhost:5001
- API docs: http://localhost:5001/docs
Open the UI and register an account — no default account is created automatically.
See [RUN.md](RUN.md) for manual setup, Docker deployment, and configuration details.
## MCP Servers
| Server | Function |
|---|---|
| `mcp-monai` | Volumetric segmentation and organ analysis via MONAI |
| `mcp-radlex` | Structured radiology report generation using RadLex templates |
| `mcp-cellpose` | Cell detection and counting on pathology ROI images |
| `mcp-ipath` | Whole-slide image retrieval from iPath telepathology platform |
| `mcp-utils` | DICOM parsing, metadata extraction, file utilities |
| `fhir-mcp-server` | FHIR resource creation via HTTP MCP (optional) |
## Documentation
- [Agent](docs/agent.md) — architecture, execution loop, modes, session, database
- [Tools](docs/tools.md) — all MCP servers and built-in tools
- [Skills](docs/skills.md) — workflow protocols and how to write your own
## Demo
[](https://youtu.be/PEgMOrTM5D4)
## Acknowledgements
This work has received support from the "Health from Portugal - Agenda Mobilizadora para a Inovação Empresarial" project, funded by Plano de Recuperação e Resiliência português under grant agreement No C644937233-00000047.
## License
TBD
