RESUMO
In this paper we present data collection and methods for the selection of a model class with the goal of automated neuromuscular blockade (NMB). Neuromuscular response was measured in the presence of rocuronium in rabbits (N=5) and humans (N=14). An average response was formed and used to determine optimal ARX and Laguerre representations for a wide range of orders and parameters. A 6th order Laguerre model was selected based on its accuracy and simplicity. Models were identified for each subject. For each group, variation was measured by comparison to the average response. The standard deviation of the average impulse response static gain was 45.4 and 45.8% of the mean for the rabbit and human models, respectively. The range of static gain was 121 and 159% of the mean for the rabbit and human datasets. Frequency domain analysis showed differences in gain of 12 and 15dB, and phase of 45 and 75° for the rabbit and human models respectively. With this knowledge, design and development of appropriate controllers for NMB will proceed.