ABSTRACT
Essential to the conduct of research in the eld of computeraided diagnosis (CAD) is a collection of clinical images that capture the range of the disease state under investigation. Such an image collection, known as a database, preferably requires metadata, supplied by a domain expert radiologist, on the nature, extent, and/or location of the abnormality. Herein lies a major impediment to investigators wishing to contribute to CAD research: the collection of image databases is a laborious and expensive task. Limited capabilities of the hospital information system or radiology information system may complicate the initial search for relevant clinical images. Limited institutional informatics infrastructure and restricted access to the clinical picture archiving and communications system that stores clinical images may physically hinder access to required images. Issues of patient condentiality, codied in the Health Insurance Portability and Accountability Act (Department of Health and Human Services 2002), mandate strict anonymization schemes. e time required for radiologists to review, assess, and annotate images as required by the research may be quite onerous and infringes on their clinical responsibilities. Finally, some institutions not aliated with a medical center may not even have access to
clinical data of any kind. Consequently, for those investigators able to collect an image database from their own institutional resources, the number of accumulated cases is often smaller than desired, and the rigor of the associated “truth” is often less than preferred.
