A model-based expert system for interpretation of hemodynamic data from ICU patients
Ruilin Zhao, Roger G. Mark, Peter Szolovits · 2002
With the proliferation of modern monitoring and laboratory procedures, physicians in intensive care areas may face "information overload", in dealing with very large, complex and ever-changing quantities of clinical data, which often lacks efficient organization. This research analyzes the medical knowledge required for formulating decision models in the domain of hemodynamics. Based on such analysis, a knowledge based expert system to track a patient's hemodynamic state has been developed and evaluated in a laboratory setting. The initial phase of the work utilizes a cardiovascular simulator to generate "pseudo-ICU" waveforms as input to the expert system in order to guide the development of the matrix of rules and search strategies. A number of pathological simulations have been successfully analyzed by this model-based expert system, including examples of hypertension, left ventricular failure, hypovolemia, pulmonary hypertension, etc. We conclude that our approach is practical, and provides a mechanism for transforming and reducing real-time physiologic data into pathophysiologic hypotheses relevant to the management of patients.