Reinforcement Q-learning for Closed-loop Hypnosis Depth Control in Anesthesia

The management of anesthesia is one of the most critical challenges since millions of subjects undergo surgeries every day. For instance, providing an inaccurate drug dose to the patient may entail adverse effects and postoperative complications. In this context, the use of computer-controlled drug dosing systems nowadays available provides significant ad-vantages to effectively regulate anesthesia. This paper presents the control of the depth of hypnosis in anesthesia through a Machine Learning approach, i.e. a Q-learning, that enables an agent to learn in an interactive environment by trial and error using feedback from its own actions and experiences. The design and performance evaluation of the proposed solution are done in silico by exploiting an Open-source Patient Simulator, that can describe the main relevant system characteristics.