Lessons learned from closed loops in engineering: towards a multivariable approach regulating depth of anaesthesia

Event-based MPC for propofol administration in anesthesia

BACKGROUND AND OBJECTIVE

The automatic control of anesthesia is a demanding task mostly due to the presence of nonlinearities, intra- and inter-patient variability and specific clinical requirements to be meet. The traditional approach to achieve the desired depth of hypnosis level is based on knowledge and experience of the anesthesiologist. In contrast to a typical automatic control system, their actions are based on events that are related to the effect of the administrated drug. Thus, it is interesting to build a control system that will be able to mimic the behavior of the human way of actuation, simultaneously keeping the advantages of an automatic system.

METHODS

In this work, an event-based model predictive control system is proposed and analyzed. The nonlinear patient model is used to form the predictor structure and its linear part is exploited to design the predictive controller, resulting in an individualized approach. In such a scenario, the BIS is the controlled variable and the propofol infusion rate is the control variable. The event generator governs the computation of control action applying a dead-band sampling technique. The proposed control architecture has been tested in simulation considering process noise and unmeasurable disturbances. The evaluation has been made for a set of patients using nonlinear pharmacokinetic/pharmacodynamic models allowing realistic tests scenarios, including inter- and intra-patient variability. Results For the considered patients dataset the number of control signal changes has been reduced of about 55% when compared to the classical control system approach and the drug usage has been reduced of about 2%. At the same time the control performance expressed by the integrated absolute error has been degraded of about 11%.

CONCLUSIONS

The event-based MPC control system meets all the clinical requirements. The robustness analysis also demonstrates that the event-based architecture is able to satisfy the specifications in the presence of significant process noise and modelling errors related to inter- and intra-patient variability, providing a balanced solution between complexity and performance.

Pain Detection with Bioimpedance Methodology from 3-Dimensional Exploration of Nociception in a Postoperative Observational Trial

Although the measurement of dielectric properties of the skin is a long-known tool for assessing the changes caused by nociception, the frequency modulated response has not been considered yet. However, for a rigorous characterization of the biological tissue during noxious stimulation, the bioimpedance needs to be analyzed over time as well as over frequency. The 3-dimensional analysis of nociception, including bioimpedance, time, and frequency changes, is provided by ANSPEC-PRO device. The objective of this observational trial is the validation of the new pain monitor, named as ANSPEC-PRO. After ethics committee approval and informed consent, 26 patients were monitored during the postoperative recovery period: 13 patients with the in-house developed prototype ANSPEC-PRO and 13 with the commercial device MEDSTORM. At every 7 min, the pain intensity was measured using the index of Anspec-pro or Medstorm and the 0-10 numeric rating scale (NRS), pre-surgery for 14 min and post-anesthesia for 140 min. Non-significant differences were reported for specificity-sensitivity analysis between ANSPEC-PRO (AUC = 0.49) and MEDSTORM (AUC = 0.52) measured indexes. A statistically significant positive linear relationship was observed between Anspec-pro index and NRS (r2 = 0.15, p < 0.01). Hence, we have obtained a validation of the prototype Anspec-pro which performs equally well as the commercial device under similar conditions.

Credibility Evidence for Computational Patient Models Used in the Development of Physiological Closed-Loop Controlled Devices for Critical Care Medicine

Physiological closed-loop controlled medical devices automatically adjust therapy delivered to a patient to adjust a measured physiological variable. In critical care scenarios, these types of devices could automate, for example, fluid resuscitation, drug delivery, mechanical ventilation, and/or anesthesia and sedation. Evidence from simulations using computational models of physiological systems can play a crucial role in the development of physiological closed-loop controlled devices; but the utility of this evidence will depend on the credibility of the computational model used. Computational models of physiological systems can be complex with numerous non-linearities, time-varying properties, and unknown parameters, which leads to challenges in model assessment. Given the wide range of potential uses of computational patient models in the design and evaluation of physiological closed-loop controlled systems, and the varying risks associated with the diverse uses, the specific model as well as the necessary evidence to make a model credible for a use case may vary. In this review, we examine the various uses of computational patient models in the design and evaluation of critical care physiological closed-loop controlled systems (e.g., hemodynamic stability, mechanical ventilation, anesthetic delivery) as well as the types of evidence (e.g., verification, validation, and uncertainty quantification activities) presented to support the model for that use. We then examine and discuss how a credibility assessment framework (American Society of Mechanical Engineers Verification and Validation Subcommittee, V&V 40 Verification and Validation in Computational Modeling of Medical Devices) for medical devices can be applied to computational patient models used to test physiological closed-loop controlled systems.

Optimal adaptive control of drug dosing using integral reinforcement learning

In this paper, a reinforcement learning (RL)-based optimal adaptive control approach is proposed for the continuous infusion of a sedative drug to maintain a required level of sedation. To illustrate the proposed method, we use the common anesthetic drug propofol used in intensive care units (ICUs). The proposed online integral reinforcement learning (IRL) algorithm is designed to provide optimal drug dosing for a given performance measure that iteratively updates the control solution with respect to the pharmacology of the patient while guaranteeing convergence to the optimal solution. Numerical results are presented using 10 simulated patients that demonstrate the efficacy of the proposed IRL-based controller.

A comparison of propofol-to-BIS post-operative intensive care sedation by means of target controlled infusion, Bayesian-based and predictive control methods: an observational, open-label pilot study

Purpose

We evaluated the feasibility and robustness of three methods for propofol-to-bispectral index (BIS) post-operative intensive care sedation, a manually-adapted target controlled infusion protocol (HUMAN), a computer-controlled predictive control strategy (EPSAC) and a computer-controlled Bayesian rule-based optimized control strategy (BAYES).

Methods

Thirty-six patients undergoing short lasting sedation following cardiac surgery were included to receive propofol to maintain a BIS between 40 and 60. Robustness of control for all groups was analysed using prediction error and spectrographic analysis.

Results

Although similar time courses of measured BIS were obtained in all groups, a higher median propofol effect-site concentration (CePROP) was required in the HUMAN group compared to the BAYES and EPSAC groups. The time course analysis of the remifentanil effect-site concentration (CeREMI) revealed a significant increase in CeREMI in the EPSAC group compared to BAYES and HUMAN during the case. Although similar bias and divergence in control was found in all groups, larger control inaccuracy was observed in HUMAN versus EPSAC and BAYES. Spectrographic analysis of the system behavior shows that BAYES covers the largest spectrum of frequencies, followed by EPSAC and HUMAN.

Conclusions

Both computer-based control systems are feasible to be used during ICU sedation with overall tighter control than HUMAN and even with lower required CePROP. EPSAC control required higher CeREMI than BAYES or HUMAN to maintain stable control.

Clinical trial number: NCT00735631.

Electronic supplementary material

The online version of this article (10.1007/s10877-018-0208-2) contains supplementary material, which is available to authorized users.

Fractional-Order Closed-Loop Model Reference Adaptive Control for Anesthesia

The design of a fractional-order closed-loop model reference adaptive control (FOCMRAC) for anesthesia based on a fractional-order model (FOM) is proposed in the paper. This proposed model gets around many difficulties, namely, unknown parameters, lack of state measurement, inter and intra-patient variability, and variable time-delay, encountered in controller designs based on the PK/PD model commonly used for control of anesthesia, and allows to design a simple adaptive controller based on the Lyapunov analysis. Simulations illustrate the effectiveness and robustness of the proposed control.

Adaptive pharmacokinetic and pharmacodynamic modelling to predict propofol effect using BIS-guided anesthesia

BACKGROUND AND OBJECTIVE

Propofol is widely used for hypnosis induction and maintenance of general anesthesia. Its effect can be assessed using the bispectral index (BIS). Many automatic infusion systems are based in pharmacokinetics (PK) and pharmacodynamics (PD) models to predict the response of the patient to the drug. However, all these models do not take into account intra and inter-patient variability. An adjusted intraoperative drug administration allows faster recovery and provides post-operative side-effect mitigation

METHODS

BIS evolution and surgery-recorded propofol infusion data of a group of 60 adult patients (30 males/30 females) with ASA I/II physical status were used to test a real time PK/PD compartmental model. This new algorithm tunes three model parameters (ce50, γ and ke0), minimizing a performance function online.

RESULTS

The error in the BIS signal predicted by the real time PK/PD model was smaller than the error measured with fixed parameter equations. This model shows that ce50, γ and ke0 change with time and patients, given a mean (95% confidence interval) of 3.89 (3.52-4.26)mg/l, 4.63 (4.13-5.13) and 0.36 (0.31-0.4)min(-1), respectively.

CONCLUSIONS

The real time PK/PD model proposed provides a closer description of the patient real state at each sample time. This allows for greater control of the drug infusion, and thus the quantity of drug administered can be titrated to achieve the desired effect for the desired duration, and reduce unnecessary waste or post-operative effects.