One fundamental clinical role of radiologists is to provide attending physicians with interpretations of an individual patient's radiological images essential to a treatment plan or overall patient management. Interpreting images from a newly taken radiological examination often requires reference to prior images of the same patient to establish a baseline from which to confirm a suspected pathological process or injury or to evaluate the progression of one that has been identified. Such image references are crucial to the radiologist's examination reading and when inappropriately supported can result in prolonged reading time, decreased report quality, and frustration. To address the problem of inadequate image prefetching methods used by many health care organizations, we took a knowledge-based approach and developed Image Retrieval Expert System (IRES), which incorporates relevant medical/radiological knowledge and contains image retrieval heuristics commonly shared by radiologists. This article describes the design of IRES, highlights its preliminary evaluation results, and discusses issues important for managing this and similar technologies in a health care organization.