Assessing the relational database model for optimization of content discovery services in medical imaging repositories

Medical imaging has been an essential contributor to high-quality medical decisions. In the past few years, the production of medical imaging data has grown impressively, thanks to the increasing number of imaging centers and higher resolution modalities. Keeping high availability and acceptable performance in this scenario raises new challenges related to storage, discovery and distribution of imaging data. Nowadays Picture Archiving and Communication System (PACS) must optimize these processes to the limit to cope with Big Data usage scenarios. In this regard, this work explores novel technologies to improve the performance of query and retrieve services in medical imaging context, ensuring always the compatibility with Digital Imaging and Communications in Medicine (DICOM) standard. The focus is the optimization of querying services. Namely, we conducted several controlled experiments to determine the best database model to support these services. More precisely, we studied the performance of a traditional PACS archive, based on a relational database, against a more recent NoSQL database. We used large datasets with 7 million medical images that represent accurately a year of medical practice. The result of this work is a set of guidelines for the correct usage of analyzed databases in big data medical imaging scenarios, including the advantages and limitations of each model.