Abstract
It is a commonplace observation that despite the availability of several very good expert systems in various domains of medicine, very few of them are in routine clinical use. We believe that this is partly because current expert systems do not integrate painlessly into a physician's routine work. We have used a commercial voice-activated radiology reporting system to double as voice-input to a prototype microcomputer-based expert system, IMAGE/ICON. This allows a radiologist to dictate a case-report as usual and as a by-product, receive diagnostic case-advice from IMAGE/ICON. This advice is presented both as written text as well as a group of images that are relevant to the dictated case in one of several ways. The voice input into IMAGE/ICON is carried out by taking advantage of trigger words used in voice-reporting, so no parsing of free text is necessary. The combined text-and-image format of the generated advice is designed to aid both analytical thought as well as the pattern recognition facility so critical in visually oriented disciplines such as radiology.