Computer-Aided Diagnosis with an Application to Endocrinology
by W. Schild, B. Lunenfeld, B. Gavish
An interactive system for computer-assisted medical diagnosis is described. Medical experts are provided with a framework in which to translate the knowledge on which their decisions are based. The technique is based on simulating the physician's own diagnostic process. The approach taken allows consistency and completeness checks to be made, thereby achieving a high degree of reliability. Information related to a given specialty is described in terms of disorder patterns, clinical facts, and logical relationships among the clinical data. Laboratory results are processed by a subsystem which uses physician-supplied rules to establish validity and to interpret the test data with respect to their diagnostic significance. The system incorporates a dynamically generated questionnaire which provides efficient gathering of anamnestic, observational, and other clinical parameters. The physician is prompted for relevant patient data by an algorithm which assures that an almost minimal amount of information is requested. Simulation facilities allow the user to examine clustering phenomena among disorders, and an option is included that traces the logic of any decision taken to exclude candidates for a final diagnosis.