ARTIFICIAL INTELLIGENCE AND CLINICAL DECISION MAKING

点击翻译按钮获取中文摘要

This proposal is to develop an integrated program for the representation of complex knowledge and reasoning processes in medicine, using the techniques of artificial intelligence (AI) and decision analysis (DA). The project will address problems in both diagnostic and therapeutic reasoning in two medical domains: coronary disease, and fluid and electrolyte disturbances. The planned large-scale AI system will incorporate a common knowledge representation formalism and alternative reasoning programs to do data interpretation, hypothesis generation, hypothesis testing and revision, diagnostic query planning, therapeutic planning and automatically-generated explanations. DA reasoning will be integrated into the system, especially to improve the test and treatment planning components. A uniform knowledge base will be developed in the two identified medical domains, utilizing knowledge acquired from the analysis of protocols, expert opinion, case-based learning, hand-coded knowledge from the literature, and strictly limited attempts to automatically acquire structured knowledge from English free-text. A panel of specific cases will be developed that can be used to validate additions to the knowledge base and changes in the reasoning components of the system. A series of protocol analyses will be done, involving both diagnostic and therapeutic decision making in clinical settings where uncertainty and risk are predominant factors. Specific attention will focus on the degree to which experienced clinicians rely on case-based reasoning and means will be developed to represent, index and utilize such case-based experience in the reasoning program. A new generation of DA tools will be built to aid the construction of DA models for use in actual clinical settings. These tools will utilize the same knowledge bases as the AI models, and will share with them the ability to look in varying depth of detail at problems and to generate explanations of their workings. A set of template analyses dealing with rapidly progressive glomerulonephritis, renal transplantation, use of anticoagulant therapy and the timing of surgery for valvular heart disease, and a knowledge base of risk and prognosis estimates in these domains will also be constructed. An AI based system for quantitative and qualitative reasoning about patient preferences will augment a system to help health professionals and patients to establish individual patient utility structures.