The practice of radiology is a non-linear process where image processing and routing decisions for best practice are often done by humans on the fly. Commercial PACS and image processing workstations assume a fixed workflow that is enshrined in their database schemas, and may only account for 4-5 stages in the lifetime of a diagnostic imaging study. DEWEY uses DICOM (with proposed expansion to HL7) information contained in the study itself to automatically select the workflow(s) that should be pursued, including routing to one or more post-acquisition image processing engines before submission to the Radiologist for interpretation. At its core, DEWEY listens for medical images in DICOM format to be sent to a server, constructs one or more workflow processes, and performs the tasks specified in the workflow on the DICOM images at the study, series, or image level (such as automatic segmentation, classification, filtering, windowing, and others). This allows for intelligent processing of medical images directly from the modality without time consuming manual intervention. Additionally, DEWEY has the capability to process DICOM images via external and decoupled image analytics. Structurally, DEWEY is based around DICOM receivers, a workflow engine for classifying and mapping image objects to workflows, a message queue for analytic engines to subscribe to, consume from and return results to, and a noSQL database for persistent storage of analytic results.
DEWEY is applicable where any PACS system would benefit from an improved, intelligent, and automated workflow. For a study where repeated images are processed with the same algorithms, and displayed in the same manner, DEWEY would provide for substantial time and cost savings, while also providing for increased precision.