A model-based approach to the diagnosis of the cardiac arrhythmias

Abstract

The cardiac arrhythmias are abnormalities of the heart beat as observed on the electrocardiogram. They are interesting to workers in medical knowledge-based systems because their full interpretation requires anatomical and temporal reasoning using causal models. This paper presents an architecture for a knowledge-based cardiac arrhythmia monitor in which model-based hypotheses are built dynamically to explain the current portion of a continuous electrocardiographic signal. Based on the hypothesize-and-test paradigm, the control algorithm selects members of a hierarchy of electrophysiologic models and adapts them to the current rhythm to form the current hypotheses. Expectations are generated for each of the current hypotheses and compared with the incoming signal. Mismatched hypotheses are discarded except in special cases relating to artifact or changes in underlying rhythm type. A prototype has been implemented to test the basic concepts of this architecture. © 1992.