Knowledge elicitation for fuzzy logic control of drug infusion

In the work described in the paper the control of neuro-muscular blockade via relaxant drug administration to patients undergoing surgery is considered. Adequate levels of blockade are required to assist the surgeon perform an operation whilst minimising the risks of tissue damage. Relaxant drugs are conventionally administered via bolus injections determined by anaesthetists according to their clinical judgement. It has been found that, like many biomedical systems, the patient dynamics are nonlinear, time-varying and exhibit large inter-patient variability. Under such circumstances, the methodology of fuzzy logic control, introduced by Zadeh, may be suitable for such an application provided that a suitable rule-base can be elicited. Two approaches are presented, one of which employed a training simulator, and the other used logged data taken from an anaesthetist performing manual control of drug infusion