Kaplan R S
National Library of Medicine, National Institutes of Health, Rochester, Minnesota.
Proc Annu Symp Comput Appl Med Care. 1991:639-43.
An expert system is designed which uses the AI/Rheum knowledge base as the basis for a directed workup system for rheumatological disorders. A Turbo Pascal prototype is demonstrated which, unlike AI/Rheum, permits entry by the clinician of a patient's chief complaint(s) and subsequently develops a dynamic differential diagnosis for this limited set of findings. The system restricts its line of questioning to those pertinent in order to rule in or rule out items on this dynamic differential diagnosis, and, unlike AI/Rheum, it provides immediate notification to the user when a critical mass of information has been entered in order to meet a diagnosis at one of the AI/Rheum criteria table's three levels of diagnosis (possible, probable, definite). Thus the system allows for a more rapid, focused decision making approach than does AI/Consult, while it follows the trend established by QMR in that it abandons the Greek Oracle problem solving approach and instead adopts a physician-assisted hypothesis investigation approach. The system is currently a prototype, without AI/Rheum's page oriented and mouse driven interface, and thus the total user interaction time may be longer than with AI/Rheum even though fewer user interactions (responses) are required. Plans for future development include optimization of the knowledge base to allow for more efficient, problem oriented questioning and modification of the knowledge base compiler in order to dynamically rule out diseases through the addition of three new levels of diagnosis in the knowledge base criteria tables.
设计了一个专家系统，它将人工智能/风湿病知识库用作风湿病紊乱定向检查系统的基础。展示了一个Turbo Pascal原型，与人工智能/风湿病不同，它允许临床医生输入患者的主要症状，随后针对这组有限的症状得出动态鉴别诊断。该系统将其提问范围限制在相关问题上，以便对动态鉴别诊断中的项目进行确诊或排除，并且与人工智能/风湿病不同，当输入了足够多的信息以符合人工智能/风湿病标准表三个诊断级别（可能、很可能、肯定）之一的诊断时，它会立即向用户发出通知。因此，与人工智能咨询相比，该系统允许采用更快速、更有针对性的决策方法，同时它遵循了QMR确立的趋势，即放弃希腊神谕式的问题解决方法，转而采用医生辅助的假设调查方法。该系统目前是一个原型，没有人工智能/风湿病的页面导向和鼠标驱动界面，因此即使所需的用户交互（响应）较少，总用户交互时间可能比使用人工智能/风湿病时更长。未来的开发计划包括优化知识库，以实现更高效、以问题为导向的提问，并修改知识库编译器，以便通过在知识库标准表中添加三个新的诊断级别来动态排除疾病。
