AIM: A new information model and standard for annotation and mark-up of nuclear medicine images

Learning Objectives A standard for communication between image analysis and reporting software can improve reporting efficiency and workflow and minimize transcription errors that can occur when quantitative results are dictated rather than transferred electronically. The use of this standard also facilitates the adoption of structured reporting with its numerous benefits including data mining and research. Using well-defined common data elements and controlled terminologies, image annotations can be queried, retrieved and computed upon.

Image annotations such as quantitative measurements and mark-ups are typically captured on clinical imaging workstations in proprietary formats while human image interpretations are captured in clinical or research reports. This makes image annotations difficult to query, retrieve and analyze. The NCI’s Cancer Biomedical Informatics Grid (caBIG™) Imaging Workspace has developed a standards-based model for Annotation and Image Markup (AIM). AIM functionality has been incorporated into multiple imaging workstations. The AIM programming library creates, validates and transforms AIM annotations between AIM XML and DICOM SR. We linked core annotation functionality to the AIM software development kit such that annotations and measurements made on displayed images are captured in an AIM data service. Controlled terms from RadLex can be presented in a structured report template or annotations can be entered ad hoc. We have modified an open source imaging workstation to capture image annotation and markup in the AIM format using controlled terminologies. The resulting annotation can then be serialized as XML or DICOM SR. This then enables the creation of large collections of AIM annotations that can be queried not only to find images of similar content but to correlate pixel interpretation with other biomedical information content. This project demonstrates how a clinical imaging workstation can be easily modified to support AIM