MedVision

Provides AI-powered medical image analysis via MCP for chest X-rays using DenseNet, RAD-DINO embeddings, and FAISS.

python

0

GitHub Stars

python

Language

4 months ago

First Indexed

2 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "u9401066-medvision-mcp": {
      "command": "uv",
      "args": [
        "run",
        "--directory",
        "/path/to/medvision-mcp",
        "python",
        "-m",
        "src.medvision_mcp.server"
      ]
    }
  }
}

MedVision MCP provides AI-powered medical image analysis tools accessible through the Model Context Protocol. It enables you to analyze chest X-rays using a Visual RAG workflow that combines DenseNet classification, RAD-DINO embeddings, and FAISS similarity search for fast, context-aware analyses.

How to use

You will interact with the MedVision MCP server through an MCP client that connects via standard input/output (stdio) or via a remote HTTP endpoint. The server offers tools to classify chest X-rays, search for similar historical cases, and perform full Visual RAG analysis. Use the client to request analysis, pass an image path or data, and receive structured results that your agent can use to make medical inferences.

How to install

Prerequisites: ensure you have Python installed and access to the UV toolchain used to run MCP servers.

# Clone the project
git clone https://github.com/u9401066/medvision-mcp.git
cd medvision-mcp

# Install development dependencies using UV
uv sync

# Optional: run tests or type checks as you develop
uv run pytest
uv run pyright

# Start the MCP server (example command used in configuration guidance)
uv run --directory /path/to/medvision-mcp python -m src.medvision_mcp.server
```} ,{

Additional notes

Configuration and usage details focus on practical steps you can take to run and interact with the MedVision MCP server. You will configure the MCP client to connect to the MedVision server, start the server process, and issue requests that return classification results, region-specific analyses, and similar-case findings.

Gradio-based UI is available for interaction with ROIs and analysis results if you choose to run the UI example provided in the project. The UI exposes tabs for full analysis, quick classification, canvas ROI drawing, index building, and index loading.

Available tools

analyze_xray

Performs full Visual RAG analysis on a chest X-ray, combining classification results with a similarity search against a curated case index.

classify_xray

Executes DenseNet-121 based classification to detect 18 pathologies in chest X-rays.

search_similar_cases

Searches the FAISS index to find historically similar chest X-ray cases for contextually relevant comparisons.

build_rag_index

Builds a FAISS index from a directory of chest X-ray images to enable fast similarity searches.

load_rag_index

Loads a pre-built FAISS index for immediate use in similarity queries.

get_engine_status

Checks whether the MCP model components are loaded and ready to serve requests.