A Graph-Based Representation for Magnetic Resonance Imaging Metadata Leveraging Neo4j

The Digital Imaging and Communications in Medicine (DICOM) standard, whose adoption has enabled interoperability across diagnostic devices and streamlined radio-logical workflows, serves as the foundation for managing and exchanging medical imaging data. At the core of DICOM are the metadata, which provide additional information describing image and patient characteristics and offer a crucial context for analysis. Despite its strengths in storage and reporting, DICOM's file-centric architecture hinders large-scale metadata extraction, aggregation, and semantic querying, which limits the effectiveness of analytical workflows and artificial intelligence (AI) applications. To address these limitations, this paper proposes a graph-based representation of DICOM metadata using Neo4j, a scalable graph-based database management system that stores and queries data as a graph. Neo4j enables flexible metadata integration, semantic exploration across studies, and provides efficient support for AI-driven downstream analysis. An examination of DICOM files from publicly available Magnetic Resonance Imaging (MRI) images led to the identification of a subset of the most relevant and frequent metadata, which we used to define our graph-based data schema. By showing the results of three queries performed on the defined schema, we demonstrate how Neo4j-hased solutions can be effective in investigating bioimage metadata, thereby improving the findability of images with specific characteristics within large datasets.