Within the realm of medical decision making, pathological diagnosis is noteworthy in that practitioners depend to a large extent on visual experience in the interpretation of histologial and cytological preparations. Since standard computer-based decision systems support only textual interactions, they are inadequate for image-intensive diagnosis. For this reason we are proposing a different sort of consultant system with the ability to manipulate both text and images. The name of the system is EIDETIC. Its design is novel in that the textual decision system, akin to a conventional expert system, has the ability to retrieve and display digitized micrographs from a large image data base. The images called by the program are intended to provide an appropriate, annotated and complete visual experience for the pathologist at the point of decision. The goal of the project is to develop a device which can measurably improve the quality of diagnosis in the domain of neuropathology. In addition to compressing visual experience around the point of decision, other new diagnostic techniques will be engendered. Facilities will be provided which incorporate image processing into the diagnostic routine, and which delineate appropriate use of immuno- and cytochemical stains. Means are being developed to digitize microscopic images, create overlays, and store images with associated textual descriptors. The consultant system will define the current diagnostic context and call appropriate images by searching these textual descriptors. EIDETIC is designed to segregate domain specific from domain independent components. This has been done to facilitate the extension of these methodologies to other image-intensive disciplines. These might include areas of radiology, ophthalmology, endoscopy, and dermatology. The system is intended as a first generation tool to explore the manner in which visual interpretation and diagnosis can be assisted by newly evolved digital technologies.
