Managing Provenance for Medical Datasets - An Example Case for Documenting the Workflow for Image Processing

In this paper, we present a novel data repository architecture that is capable of handling the complex image processing workflows and its associated provenance for clinical image data. This novel system has unique and outstanding properties versus existing systems. Among the most relevant features are a flexible and intuitively usable data and metadata management that includes the use of a graph-based provenance management strategy based on a standard provenance model. Annotation is supported to allow for flexible text descriptors as being widespread found for clinical data when structured templates are not yet available. The architecture presented here is based on a modern database and management concepts and allows to overcome the limitations of current systems namely limited provenance support, lacking flexibility, and extensibility to novel requests. To demonstrate the practical applicability of our architecture, we consider a use case of automated image data processing workflow for identifying vascular lesions in the lower extremities, and describe the provenance graph generated for this workflow. Although presented for image data, the proposed concept applies to more general context of arbitrary clinical data and could serve as an additional service to existing clinical IT systems.