🧠 DFDP2 – DICOM to RDF Processing and Visualization Demo

This project is a self-contained web application that demonstrates a complete pipeline for:

• Processing DICOM files
• Extracting key metadata
• Mapping it to semantic ontologies (ROO, SNOMED CT, FOAF, etc.)
• Generating an in-memory RDF knowledge graph
• Providing a web-based interface for dataset discovery, SPARQL querying, and graph visualization

Built with Python, the app uses:

• FastAPI for the web server
• pydicom for handling DICOM files
• rdflib for RDF generation and SPARQL querying
• D3.js for frontend graph visualization

---

🚀 Features

Feature	Description
🏥 DICOM Processing	Extracting Metadata from DICOM files picked from TCIA
📑 Metadata Extraction	Extracts Patient ID, Study Date, Modality, Accession Number etc
📚 Semantic Mapping	Maps values to ROO, SNOMED CT, and FOAF ontologies
🔗 RDF Generation	Builds triples and populates an in-memory knowledge graph
🔍 SPARQL Endpoint	Supports SPARQL 1.1 queries via a web form
📂 Metadata Catalog	Web interface styled after FAIR Data Platforms (Health DCAT-AP)
🕸️ Knowledge Graph Visualization	In-browser graph using force-directed layout

---

📦 Installation

✅ Prerequisites

• Python 3.7+

• pip (Python package manager)

  🐙 GitHub Codespaces (Recommended)

# 1. Open in Codespaces (click badge above)

### 🔧 Install Dependencies


```bash
pip install "fastapi[all]" uvicorn  pydicom rdflib requests "python-multipart" Jinja2

▶️ Running the Application

#Load the Dicom images Hugging Face Repository and convert them to DICOM files or download from Cancer Imaging Archive which are now currently stored in dicom_files folder
python fetch_dicom.py # Generates dicom_metadata.json
python map_dicom_complete.py # Generates dicom_mapped_with_catalog.ttl
# then start the FAST API App.
uvicorn main:app --reload

In case you get an error like this : ERROR: [Errno 98] Address already in use Fix it using these commands :

lsof -i :8000 #uvicorn 18407 codespace    3u  IPv4 257296      0t0  TCP localhost:8000 (LISTEN)
kill -9 XXXX # displayed nos next to uvicorn, so 18407

Output : [1]+ Killed uvicorn main:app --reload

3. Visit: http://127.0.0.1:8000

You’ll see logs like:

INFO:     Uvicorn running on http://127.0.0.1:8000 (Press CTRL+C to quit)
INFO:     Started server process [xxxxx]
INFO:     Application startup complete.

🌐 Web Interface Overview

Page	URL	Description
🏠 Home	/	Upload DICOMs, view summary
📚 Catalog	/catalog	View/search processed DICOM datasets
🌐 SPARQL	/ (form)	Query the RDF graph using SPARQL
🧬 Visualize	/visualize	Interactive graph of datasets and relationships

⚙️ How It Works

Run Scripts to

• Extracting Metadata from DICOM files picked from TCIA and stored in dicom_files/
• Metadata like PatientID, StudyDate, and Modality are extracted
• Map to SNOMED CT and ROO terms
• RDF triples are generated

🧪 SPARQL Query Example

Query all DICOM datasets where the modality is CT:

#Query 1
PREFIX dcat: <http://www.w3.org/ns/dcat#>
PREFIX dcterms: <http://purl.org/dc/terms/>
PREFIX dicom: <http://dicom.nema.org/resources/ontology/DCM#>

SELECT ?file ?patientId ?studyDate ?modality ?accessionNumber
WHERE {
  ?dataset a dcat:Dataset ;
           dcat:distribution ?file .
  
  ?file dicom:PatientID ?patientId ;
        dicom:SeriesDate ?studyDate ;
        dicom:Modality ?modality ;
        dicom:AccessionNumber ?accessionNumber .
}
ORDER BY ?studyDate
LIMIT 10

#Query 2 : 
PREFIX dicom: <http://dicom.nema.org/resources/ontology/DCM#>

SELECT DISTINCT ?patientId
WHERE {
  ?subject dicom:Manufacturer "GE MEDICAL SYSTEMS" .
  ?subject dicom:PatientID ?patientId .
}

#Query 3
PREFIX dicom: <http://dicom.nema.org/resources/ontology/DCM#>

SELECT DISTINCT ?manufacturer ?modelName
WHERE {
  ?file dicom:Manufacturer ?manufacturer .
  ?file dicom:ManufacturerModelName ?modelName .
}
ORDER BY ?manufacturer ?modelName
#Query 4
PREFIX dicom: <http://dicom.nema.org/resources/ontology/DCM#>
PREFIX roo: <http://www.cancerdata.org/roo/>

SELECT DISTINCT ?patientId ?bodyPart ?age ?sex ?reasonForStudy
WHERE {
  ?subject dicom:PatientID ?patientId .
  ?subject dicom:BodyPartExamined ?bodyPart .

  OPTIONAL { ?subject roo:hasAge ?age . }
  OPTIONAL { ?subject roo:hasSex ?sex . }
  OPTIONAL { ?subject roo:hasReasonForStudy ?reasonForStudy . }
}
#Query 5
PREFIX dicom: <http://dicom.nema.org/resources/ontology/DCM#>
PREFIX dcat: <http://www.w3.org/ns/dcat#>

SELECT ?patientId (COUNT(?file) AS ?numberOfScans)
WHERE {
  ?dataset a dcat:Dataset ;
           dcat:distribution ?file .
  ?file dicom:PatientID ?patientId .
}
GROUP BY ?patientId
ORDER BY ?patientId

📚 Ontologies Used

Prefix	URI
ROO	https://www.cancerdata.org/roo-information
SNOMED	https://bioportal.bioontology.org/ontologies/SNOMEDCT
DCAT	http://www.w3.org/ns/dcat#
FOAF	http://xmlns.com/foaf/0.1/
dicom	http://dicom.nema.org/resources/ontology/DCM
DCTERMS	http://purl.org/dc/terms/

---

📁 Directory Structure

.
├── main.py             # FastAPI application
├── templates/          # HTML templates (auto-generated)
├── static/             # Static assets (CSS/JS)
├── dicom_files/        # DICOMs downloaded via Script

---

📈 Visualization

At /visualize, you'll find a D3.js-based graph of the RDF data:

• Nodes are color-coded by type (e.g., Patient, Modality)
• Drag nodes to explore relationships
• Hover over nodes and edges to view URIs and labels

---

📌 To-Do / Ideas for Future

• Persistent RDF store (e.g., Blazegraph, Fuseki)
• Support for real-world DICOM tags and vocabularies
• Authentication for upload and SPARQL features
• Multi-user catalog and permission system

---

Screentshots for reference

---

📄 License

MIT License
Free to use, modify, and distribute with proper attribution.