dicom2rdf

dicom2rdf is a data pipeline that converts millions of DICOM SR documents to RDF Turtle, makes its metadata searchable with SPARQL, and is powered by QLever, a high performance graph database.

Unlike other approaches that only translate the bare hierarchical structure of a DICOM SR document to RDF triples, dicom2rdf applies a second processing step to transform this raw structure via a set of SPARQL CONSTRUCT queries. The resulting schema yields simpler SPARQL queries that are both highly explorable and run faster.

Example: Chronological SSDE by Acquisition Protocol

Prerequisites

• Docker with Docker Compose V2, or
• Podman with podman-compose

Usage

We use docker in the following examples, but podman works just as well.

1. Clone this repository and cd into it:

   git clone https://github.com/ad-freiburg/dicom2rdf.git ~/dicom2rdf && cd ~/dicom2rdf

2. Create empty bind mounts:

   mkdir -p data/{ttl-raw,ttl-semantic,index-raw,index-semantic}

3. Create an .env file from the example and adjust as needed:

   cp .env.example .env

4. Start the pipeline:

   docker compose up --build

   1. In the future, you may also start QLever and QLever UI without a full pipeline run:

      docker compose -f compose.yml -f compose.isolated.yml up qlever qlever-ui

5. Wait for the "ready" message that displays the URLs of the QLever and QLever UI instances.

6. Open the QLever UI, start typing a subject, and explore the available predicates: