SELF-LEARNING FUZZY-LOGIC CONTROL OF NEUROMUSCULAR BLOCK

We have assessed the performance of a ''self-learning'' fuzzy logic co ntroller to administer atracurium to a required depth of neuromuscular block. We studied 20 ASA I and II patients undergoing surgery anticip ated to last longer than 30 min. A Datex Relaxograph was used to measu re the degree of neuromuscular block, and control to a T1 twitch heigh t set point of 10% of baseline neuromuscular function was selected. Th e controller commenced with a blank rule-base and instructed a Graseby 3400 infusion pump to administer an atracurium infusion to maintain t his level of block. The system achieved stable control of neuromuscula r block with a mean T1 error of -0.52% (SD 0.55%) accommodating a rang e in mean atracurium infusion rate of 0.25-0.44 mg kg(-1) h(-1). These results compare favourably with the more computationally intensive an d unwieldy adaptive control strategies for atracurium infusion used pr eviously. There was less variation in infusion rates than in our previ ously studied fixed rules fuzzy controller.