Automatic Regulation of Anesthesia via Ultra-Local Model Control

As a part of the BMS2021 Benchmark Challenge, this paper deals with the design and testing of a closed-loop anesthesia delivery regulation system by exploiting the open-source Matlab-based patient simulator. Because of system inherent complexity together with intra-and inter-patient parameters variability and partially unknown disturbances, traditional model-based approaches may suffer. To overcome these limitations, we opt for a data-driven approach using real-time ultra-local models coupled with the corresponding so-called intelligent controllers. In this way, one maintains the hemodynamic variables while regulating the levels of hypnosis, analgesia, and neuromuscular blockade in anesthesia by automatic delivery of drugs. The performance of the proposed approach has been evaluated in silico by considering a representative dataset composed of 24 patients, the presence of disturbances mimicking both surgical stimulations and actions of “anesthesiologist in the loop”, including also noise effects and time-varying system delays.