An adaptive control structure to mitigate lack of feasibility in systems with poor identifiability. Case Study on General Anesthesia

Poor identifiability in medical systems is due to ethical and regulatory limitations in free decisions regarding design of excitation signals and application of persistent excitation to human body areas of interest. Furthermore, drug-dose effect relationship is o particular challenge due to inter and intra patient variability. Complexity of interacting biological self-and extra- regulatory networks of vital processes in humans add more to the uncertainty limiting the feasibility of obtaining a digital twin of the patient for better clinical outcome. Epistemic and aleatory uncertainty is clustered around the problem of identifiability. This is furthermore strengthen by the lack of suitable instrumentation to measure the necessary information, rather making available inferences and surrogate infometrics. In this paper we propose a theoretical framework build on a commonly encountered problem of model mismatch in closed loop control of anesthesia. The solution uses ratio control of two drug co-administration in an online model estimation with adaptive setpoint strategy to achieve general anesthesia. Our simulations indicate that the proposed solution helps in calibrating the dose-effect response model to actual response of the patient in online setting and uses setpoint adaptation to guide the induction of anesthesia.