Usage
dicom-mcp is a server based on the Model Context Protocol (MCP), designed to enable AI assistants to interact with DICOM servers (such as PACS, VNA, etc.). It enables AI assistants to query, read, and move medical imaging data, thereby supporting more complex medical imaging-related tasks.
Main Capabilities
This server provides four categories of tools:
Query Metadata
query_patients: Search for patients by criteria such as name, ID, or date of birth.query_studies: Find studies using patient ID, date, modality, description, accession number, or Study UID.query_series: Locate series within a specific study using modality, series number/description, or Series UID.query_instances: Find a single instance (image/object) within a series using instance number or SOP Instance UID.
Read DICOM Reports (PDF)
extract_pdf_text_from_dicom: Retrieve a specific DICOM instance containing an encapsulated PDF and extract its text content.
Send DICOM Images
move_series: Use C-MOVE to send a specific DICOM series to another configured DICOM node.move_study: Use C-MOVE to send an entire DICOM study to another configured DICOM node.
Utility Tools
list_dicom_nodes: Display the currently active DICOM node and list all configured nodes.switch_dicom_node: Change the active DICOM node for subsequent operations.verify_connection: Use C-ECHO to test DICOM network connectivity with the currently active node.get_attribute_presets: List the available detail levels (minimal, standard, extended) for metadata query results.
These tools can be chained together to answer complex questions.
How to Use
Installation: Can be installed using uv or pip, or by cloning the repository.
Configuration: dicom-mcp requires a YAML configuration file (such as config.yaml) that defines DICOM nodes and the calling AE title. The configuration can be adjusted, or kept as default for compatibility with the example ORTHANC server.
MCP Integration: Add the server to the client configuration (such as claude_desktop_config.json).
Optional Example ORTHANC Server: If no DICOM server is available, a local ORTHANC server can be run using Docker. Clone the repository and install the test dependencies:pip install -e ".[dev]", the UI is located at http://localhost:8042。
Testing: Testing requires a running Orthanc DICOM server, Docker can be used, run tests using pytest, stop the Orthanc container. Use MCP Inspector to debug server communication.
Applicable Scenarios
- Enable AI assistants to query and retrieve medical imaging metadata.
- Extract clinical report text from DICOM-encapsulated PDFs.
- Send imaging series or studies to an AI endpoint for image segmentation, classification, and other processing.
- Work with locally deployed open-weight LLMs to achieve complete data privacy.
Limitations
DICOM-MCP is not intended for clinical use and should not be connected to real hospital databases or databases containing sensitive patient data. Otherwise, it may lead to loss of patient data and leakage of patient data to the internet.
Technical Details
- Built on pynetdicom.
- Uses PyPDF2 for PDF text extraction.
- Follows the Model Context Protocol (MCP).
Information Not Provided
The official materials do not provide specific pricing, charging models, a list of supported platforms, performance metrics, or compatibility details with other systems.
