Conformance testing of DICOM image objects

The DICOM Standard contains Information Object Definitions (IODs) for images generated by each modality type. In order to test if a specific image conforms to its specification it is necessary to formalize the IODs, their mandatory and optional aspects, including the associated natural language conditions. The approach described in this paper is to define a IOD language capable of representing the attribute collections and conditional expressions found in DICOM IODs. The syntax of the TOD language retains as far as possible the familiar tabular form of the DICOM documents allowing easy translation. The set of condition primitives contained in the IOD language is based on an analysis of the types of conditions found in DICOM IODs, The testing of a specific image follows a two level process. First, the description of each IOD defined using the IOD language is compiled into an efficient representation. Second, an image instance is compared against the compiled IOD representation by an image parser and a detailed description of any errors is generated, Formal IOD descriptions have been constructed for Computed Tomography (CT), Magnetic Resonance (MR) and Secondary Capture (SC) DICOM IODs and used to evaluate several sets of test images generated by a range of manufacturers during the period 1993 to 1996. All images of the same type, produced using the same software version, normally contain the same kind of errors. We found that overall 92% of CT studies, 91% of MR studies and 64% of SC studies contained at least one violation of their corresponding IOD. Most errors are caused by only a small number of common situations.