Multivariable Anesthesia Control Using Reinforcement Learning

The anesthesia community has recently witnessed numerous advances in monitoring of the anesthetic state. This development has spurred a renewed interest in the automation of the clinical anesthesia. The absence of a precise model in biomedical field has motivated the authors to propose a novel approach capable of handling the uncertainties present in the model and moreover, to design a new approach which is subjective and can be adopted to patient specific requirements. In this regard a study has been conducted in this paper, in order to fit the intelligence of fuzzy controllers based on reinforcement learning in this clinical application. This study suggests that the chosen approach fits the scenario of controlling two anesthetic drugs; Atracurium and Isoflurane, to achieve a surgical anesthetic state, in terms of muscle relaxation (paralysis) and blood pressure (mean arterial pressure (MAP)) as monitoring indices.