Relational Databases versus Search Engines: A Performance Comparison for Storing and Querying DICOM Metadata

The Digital Imaging and Communications in Medicine (DICOM) standard adopts files as individual, self-contained repositories for the storage of a mixed of alphanumerical and binary content regarding radiological images. Usually, groups of DICOM files are hierarchically organized in studies and series, physically disposed into filesystem directory trees. Despite its simplicity in storing content, ordinary filesystems do not provide index capabilities allowing searches by content – restricting access by directory names and file names. To surpass such limitation, Picture Archiving and Communication Systems (PACSs) often adopt Relational Database Management Systems (RDBMSs) as metadata repositories, benefiting from its general-purposed index structures. An alternative approach, not quite explored, considers the adoption of search engines as metadata catalogs, aiming to minimize the search time by exploring the engine’s index optimizations. In order to evaluate the performance on managing DICOM metadata, this work compares relational database instances to a search engine in terms of storage space, storage time, and query time. Results show that, in the best case, the search engine is slightly slower in storing content; however, it requires 69% less disk space for the same dataset. For queries, in turn, the search engine performs up to 8.3 times faster in retrieving groups of tags.