Control aspects of clinical support systems

Abstract

Standard control theory approaches alone are not sufficient for the design of clinical control systems. Additional medical expertise has to be incorporated in order to achieve high quality and robust performance. The main problem to be overcome is the fact that no real time identifiability of the patient’s characteristics can be guaranteed, due to the low signal-to-noise ratio of the feedback signals in combination with unknown and possibly rapid changes in patient characteristics. Solutions to this problem are suggested by the manual control methods of expert clinicians. One solution is to adjust the controller’s parameters through monitoring its control performance. Another solution, in multi-input-multi-output controllers, is to allow variation of only one drug at a time. In addition, sometimes other goals are more important than keeping feedback signals at setpoint values. These goals need to be part of the clinical support system as well. The success of adding clinician-derived heuristics to standard control systems is illustrated by a number of single and multiple drug controllers, whose goal it is to support the clinician in the stabilization of the systemic arterial and/or the pulmonary artery pressure during and after cardiac surgery.