HP Labs Technical Reports

Learning Qualitative Models from Physiological Signals

Hau, David T.; Coiera, Enrico W.

HPL-95-31

Keyword(s):

Abstract: Physiological models represent a useful form of knowledge, but are both difficult and time consuming to contruct by hand. Futher, most physiological systems are incompletely understood. This article addresses these two issues with a system that learns qualitative models from physiological systems. We describe the Genmodel learning system in detail, including the front-end processing and segmenting that transforms a signal into a set of qualitative states. Next we report results of experiments on data obtained from six patients during cardiac bypass surgery. Useful models were obtained, representing both normal physiology and pathophysiology paticular to the patient being monitored. Model variations across time and across different levels of temporal abtraction and fault tolerance are examined. Implications for the design of intelligent monitoring systems and smart alarms are explored.

Back to Index