We propose to develop automated technique to facilitate the creation and maintenance of large electronic knowledge bases of medical treatment plans. These techniques for knowledge-base construction ultimately will aid clinical experts in the design and validation of new treatment protocols. The research will foster dissemination of protocol guidelines in a format that can be incorporated within computer-based systems that physicians will use to obtain patient-specific treatment advice. We will complete development of a computer system, PROTEGE, that will allow users to create protocol-entry tools that are each custom-tailored to a particular medical specialty. These PROTEGE-generated tools will permit expert physicians to describe treatment protocols using simple, graphical notations; the graphical specifications will automatically be converted to an internal format that a special version of the ONCOCIN program will interpret to render patient-specific therapy advice. In the course of our work, we will address issues related both to computer-assisted development and dissemination of clinical protocols, and to the general problem of acquiring and representing medical knowledge for use within computer-based advice systems. Our research plan comprises a four-step process: 1. We will implement PROTEGE on a general-purpose computer, developing a new user-interface-management system that will facilitate generation of more flexible protocol-entry tools. 2. We will develop new models of problem solving that will allow PROTEGE users to represent more complex therapy-planning actions. We will postulate new terms and relationships that can be used to describe medical- therapy actions in a domain-independent fashion, and test the effectiveness of these terms and relationships in the specification of new classes of treatment protocols. 3. We will evaluate PROTEGE by creating tools with which physicians will enter new protocols for the therapy of AIDS and of AIDS-related infections. 4. We will develop methods for computer-aided design of new clinical protocols, allowing developers to define libraries of knowledge concerning expected efficacy and expected toxicity of therapeutic agents. We will evaluate these design methods in the area of clinical trials for HIV- related diseases.
Showing the most recent 10 out of 15 publications
