Graph Meets Imaging: Knowledge Graph Approach Improves Metadata Quality Assessment in Imaging Trials

In a newly presented study by Bejaoui et al. (Charité Berlin & Universitätsmedizin Göttingen), researchers propose a knowledge graph-based method for assessing DICOM metadata quality in imaging trials. The approach was developed to address common challenges in multi-site imaging studies, where heterogeneous and incomplete metadata can limit reproducibility and data utility.

Key innovation:
DICOM metadata is transformed into a knowledge graph that models the structure and relationships of CIODs, Modules, and Attributes. This enables:

  • Customizable quality metrics (e.g., completeness, centrality)
  • Attribute prioritization based on type (e.g., mandatory, optional) or frequency per modality
  • Scalable and automated quality checks for entire datasets
  • Integration with tools like Neo4j and NetworkX for advanced graph analytics

Result:
The approach enables programmatic metadata validation in imaging trial workflows and supports reproducible assessments across modalities like CT and MRI. An open-source implementation is available, promoting community use and extension.

Conclusion:
This work shows that metadata quality is measurable and improvable with graph data science—paving the way for better-managed imaging trials and more reliable downstream analyses.

Read the full article here: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5166430