The Power of Abstract Temporal Diagnosis in Medical Domains

Johann Gamper and Wolfgang Nejdl

Abstract

Two observations motivate this paper: first, the model-based diagnosis paradigm has been recognized to provide powerful techniques for automated medical diagnosis, and second, time is inherent in medical domains, where the capability to deal with various kinds of temporal information is necessary for a sophisticated diagnosis system in medicine. In this paper we describe a model-based framework for diagnosing time-varying systems by using the diagnosis of hepatitis~B as an example. Based on a logical framework extended by qualitative and quantitative temporal constraints we show how to describe complex temporal behavior of systems. We introduce the concept of abstract observations as an abstraction from observations at time points into assumptions over time intervals. This leads to a more intuitive representation and makes the diagnosis system independent of the number of actual observations and the granularity of time. Additionally, the concept of abstract temporal diagnoses captures in a natural way indefinite temporal knowledge we often have about diagnoses, in particular in medical domains.

Keywords: Diagnosis, Model-based Reasoning, Temporal Reasoning

The full report is available as a postscript file .