Applying partially observable Markov decision processes to anesthesia control: A simulated study.

To have a better controller, this research applies Partially Observable Markov Decision Process (POMDP) framework to achieve better drug delivery policy, even when there is incomplete information about a patients' current states during a general anesthesia. In this dissertation, a robust POMDP controller model is introduced for better closed-loop anesthesia control. This model is solved using a state-of-the-art POMDP planner to compute for propofol rates to administer to a patient. The new controller is tested on 1000 simulated patients, and performance metrics are analyzed and compared against previously published results.