The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

MSc Thesis #94129

Title:Supporting Choice in Medical Diagnosis
Date: 1994
Abstract:The nature of medical diagnosis suggests the division between two kinds of medical knowledge: knowledge about medical concepts such as diseases, symptoms and relationships between them (domain knowledge) and knowledge about the reasoning techniques used with the domain knowledge in order to achieve a diagnosis (task knowledge). The connection between these two kinds of knowledge is established when a decision about which course to pursue next has to be made. However, diagnosticians carry out diagnosis in different ways depending on their experience and the data under consideration at each moment of the diagnosis. All these aspects are considered for the development of the current system, which has been built through several stages. Firstly, the problem of Knowledge Acquisition was considered in order to get the source of medical expertise needed for the problem, distinguishing between domain knowledge (knowledge about diseases and symptoms) and task knowledge (diagnostic tasks). Secondly, a model of that knowledge was built, taking the example of TOMKAT, a medical knowledge acquisition toolkit which models both domain and task knowledge. The task model is built in such a way that allows the diagnostician to execute diagnostic subtasks in different ways, such as considering first the most dangerous disease, the most common disease, and so on. Finally, according to the nature of medical knowledge, we take an object-oriented approach for implementing this medical diagnostic system. This system successfully fixes a connection between domain and task knowledge for medical diagnosis, thus establishing a further step in the application of Artificial Intelligence to medicine.

[Search These Pages] [DAI Home Page] [Comment]