古谷 栄光

<title>Abstract</title><sec> <title>Background</title> Nondepolarizing neuromuscular blocking drugs (NDNBs) are clinically used to produce muscle relaxation during general anesthesia. To better understand clinical properties of NDNBs, comparative in vitro pharmacologic studies have been performed. In these studies, a receptor binding model, which relies on the assumption that the inhibition, i.e., the effect of an NDNB, is proportional to the receptor occupancy by the drug, has been effectively used to describe obtained experimental data. However, it has not been studied in literature under which conditions the above assumption can be justified nor the assumption still holds in vivo. The purpose of this study is to explore the in vivo relationship between the inhibition and the receptor occupancy by an NDNB and to draw implications on how in vitro experimental results can be used to discuss the in vivo properties of NDNBs. </sec><sec> <title>Methods</title> An ordinary differential equation model is employed to simulate physiologic processes of the activation of receptors by acetylcholine (ACh) as well as inhibition by an NDNB. With this model, the degree of inhibition is quantified by the fractional amount of receptors that are not activated by ACh due to the presence of an NDNB. The results are visualized by plotting the fractional amounts of the activated receptors as a function of the receptor occupancy. </sec><sec> <title>Results</title> Numerical investigations reflecting in vivo conditions show that the degree of inhibition is not proportional to the receptor occupancy, i.e., there is a nonlinear relationship between the inhibition and the receptor occupancy. However, under a setting of high concentration of ACh reflecting a typical situation of in vitro experiments, the relationship between the inhibition and the receptor occupancy becomes linear, suggesting the validity of the receptor binding model. Also, it is found that the extent of nonlinearity depends on the selectivity of NDNBs for the two binding sites of the receptors. </sec><sec> <title>Conclusions</title> While the receptor binding model may be effective for estimating affinity of an NDNB through in vitro experiments, these models do not directly describe in vivo properties of NDNBs, because the nonlinearity between the inhibition and the receptor occupancy causes the modulation of the resultant concentration-effect relationships of NDNBs. </sec>

In critically ill patients suffering from hyperglycemia, it has been recently shown that mortality and morbidity can be reduced by keeping blood glucose within the range of 80-110 mg/dL. However, maintaining glycemia within such range is difficult due to the time variability in insulin sensitivity in critically ill patients. In this paper, we propose a novel glycometabolism model of critically ill patients with an insulin sensitivity parameter and develop a nonlinear model predictive glycemic control system with online identification of insulin sensitivity at one-hour intervals. Simulation results show that our system keeps 70% of BG measurements within the range of 80-110 mg/dL without any severe hypoglycemic incidents, which indicates the effectiveness and safety of our system.

<p>An increasing number of closed-loop blood glucose (BG) control algorithms have been developed in recent years with the 'artificial pancreas' as the ultimate goal, although tight postprandial BG control remains an elusive goal. In this report, the authors propose a novel semi closed-loop BG control algorithm with meal announcement, which involves computation of the optimal continuous subcutaneous insulin infusion for a specific meal 60 min prior to mealtime. It utilizes a mathematical model of glucose-insulin metabolism to predict the impact of carbohydrates on postprandial BG levels based on carbohydrate intake and glycemic index (GI) value. The optimal pre-meal insulin is infused until mealtime, after which the control algorithm switches to model predictive control (MPC) to stabilize postprandial glycemia at the target value of 100 mg/dL (5.55 mmol/L). In silico results for four representative foods with GI values spanning a wide range show that in the case of exact patient-model match with precise information of carbohydrate composition and mealtime, postprandial BG levels can be maintained between 86–134 mg/dL (4.78–7.44 mmol/L) and 86–152 mg/dL (4.78–8.44 mmol/L) for 50 g and 100 g of carbohydrates, respectively. With consideration of intra-patient variability and meal-related uncertainties regarding the estimated carbohydrate amount and start of meal consumption, the BG control range is 75–159 mg/dL (4.17–8.83 mmol/L) with no critical hypoglycemic episodes.</p>

An accurate mathematical representation of the impact of carbohydrates on postprandial glucose-insulin metabolism in type 1 diabetes (T1D) is essential for the development of model-based diabetes-related technologies. In this study, the physiological representation of a previous model of gut absorption from carbohydrates is enhanced by including a maximum rate of exogenous glucose appearance as observed in clinical studies in the literature. Simulation results for the same four representative carbohydrate-rich foods with varying glycemic index values as in our previous study, in addition to a sugary beverage, show that the postprandial glycemic excursion relates more closely to clinical data of postprandial glycemic excursion in patients with T1D.

In this study we introduce an extension of a previously developed model of glucose-insulin metabolism in type 1 diabetes (T1D) from carbohydrates that includes the effect of dietary fat on postprandial glycemia. We include two compartments that represent plasma triglyceride and non-esterified fatty acid (NEFA) concentration, in addition to a mathematical representation of delayed gastric emptying and insulin resistance, which are the most well-known effects of dietary fat metabolism. Simulation results show that postprandial glucose as well as lipid levels in our model approximates clinical data from T1D patients.

The authors propose a mathematical model of glucose-insulin metabolism in type 1 diabetes based on Bergman and Shimoda insulin models, which are adjusted to represent diabetic state and improve the accuracy of subcutaneous insulin absorption, respectively. The authors also propose a model of digestion and absorption from carbohydrates based on the glycemic index (GI) of foods and carbohydrate bioavailability concepts that provide a glucose-equivalent representation of the impact of carbohydrates on blood glucose levels. Comparison with clinical data demonstrates that the proposed model is able to represent postprandial blood glucose excursion for carbohydrates with varying GI values.

我々は，適切な鎮静度維持や麻酔科医の負担軽減を目的として，鎮静度指標aepEXに基づく鎮静度制御システムの研究を行ってきた．適切な鎮静度維持のためには鎮静状態と覚醒状態を正確に判別できる必要があり，従来研究で麻酔薬に対する応答のモデル(薬物動態モデル)として時変モデルを用いることにより適切な制御が実現できることを示したが，本研究では既存の薬物動態モデルについて時変性を考慮した場合に最適となるモデルを検討した．

医療上の目的から患者の状態制御の研究が盛んに行われている．これらの研究において制御技術が非常に大きな役割を果たしうると考えられるが，必ずしも臨床的に十分な信頼性を持つシステムが開発されるに至っていない．本講演では，手術中の麻酔制御や1型糖尿病患者の血糖値制御などを例にとり，研究の現状と臨床における状況を述べ，問題点を示したうえで，今後望まれる方向について述べる．

Treatment of type 1 diabetes consists of subcutaneous insulin administration to maintain blood glucose levels within normal range. In particular, a large prandial insulin dose is necessary at mealtime depending on the carbohydrate amount and type ingested. Nonetheless, such prandial insulin bolus is often miscalculated, which results in postprandial hyper/hypoglycemia. In this study we develop a feedback-feedforward blood glucose control that comprises an optimal preprandial insulin infusion followed by a MPC-based postprandial blood glucose control to maintain normoglycemia. The former includes an optimization problem using a mathematical model of glucose-insulin metabolism in type 1 diabetes previously developed and considers insulin infusion 60 min prior to food ingestion. Simulation results of the proposed prandial blood glucose control algorithm demonstrate the ability of the proposed blood glucose control algorithm to reduce postprandial hyperglycemia and prevent hypoglycemia in type 1 diabetic patients.

Treatment of type 1 diabetes consists of maintaining postprandial normoglycemia using the correct prandial insulin dose according to food intake. Nonetheless, it is hardly achieved in practice, which results in several diabetes-related complications. In this study we present a feedforward plus feedback blood glucose control system that considers the glycemic index of foods. It consists of a preprandial insulin bolus whose optimal bolus dose and timing are stated as a minimization problem, which is followed by a postprandial closed-loop control based on model predictive control. Simulation results show that, for a representative carbohydrate intake of 50 g, the present control system is able to maintain postprandial glycemia below 140 mg/dL while preventing postprandial hypoglycemia as well.

This paper proposes a novel hypnosis control method using Auditory Evoked Potential Index (aepEX) as a hypnosis index. In order to avoid side effects of an anesthetic drug, it is desirable to reduce the amount of an anesthetic drug during surgery. For this purpose many studies of hypnosis control systems have been done. Most of them use Bispectral Index (BIS), another hypnosis index, but it has problems of dependence on anesthetic drugs and nonsmooth change near some particular values. On the other hand, aepEX has an ability of clear distinction between patient consciousness and unconsciousness and independence of anesthetic drugs. The control method proposed in this paper consists of two elements: estimating the minimum effect-site concentration for maintaining appropriate hypnosis and adjusting infusion rate of an anesthetic drug, propofol, using model predictive control. The minimum effect-site concentration is estimated utilizing the property of aepEX pharmacodynamics. The infusion rate of propofol is adjusted so that effect-site concentration of propofol may be kept near and always above the minimum effect-site concentration. Simulation results of hypnosis control using the proposed method show that the minimum concentration can be estimated appropriately and that the proposed control method can maintain hypnosis adequately and reduce the total infusion amount of propofol.

In this paper, we propose an extension to our previously developed mathematical model of glucose-insulin metabolism in type 1 diabetes (T1D). It comprises a carbohydrate metabolism subsystem with a solid foundation of the relative impact of carbohydrates on blood glucose levels; an exogenous insulin subsystem that considers the effect of subcutaneous and intravenous rapid-acting insulin; and a glucose-insulin metabolism subsystem in T1D for postprandial and postabsorptive states. Simulations of blood glucose excursion in our model during both states demonstrate the reliability and accuracy of our model.

We present some improvements to our previous model of postprandial glucose excursion in diabetics under rapid-acting insulin therapy. In the glucose sub-model, we correct Glycemic Index values from the glucose relative function and as a result, adjust absorption rate parameters for carbohydrates. In the insulin sub-model, we modify Shimoda insulin model parameters to obtain a more realistic representation of the effect of rapid-acting insulin on glycemic levels. Additionally, we derive a more intelligible expression of the relative exogenous insulin function. Results from computer simulations show that our enhanced model is now accurate throughout the GI range of carbohydrates.

In ambulatory surgery, anesthetic drugs must be administered at a suitable rate to prevent adverse reactions after discharge from the hospital. To realize more appropriate anesthesia, we have developed a hypnosis control system, which administers propofol as an anesthetic drug to regulate the bispectral index (BIS), an electroencephalography (EEG)-derived index reflecting the hypnosis of a patient. This system consists of three functions: 1) a feedback controller using a model-predictive control method, which can adequately accommodate the effects of time delays; 2) a parameter estimation function of individual differences; and 3) a risk control function for preventing undesirable states such as drug overinfusion or intraoperative arousal. With the approval of the ethics committee of our institute, 79 clinical trials took place since July 2002. The results show that our system can reduce the total amount of propofol infusion and maintain the BIS more accurately than anesthesiologist's manual adjustment.

プロポフォール投与時のBispectral Index (BIS) 変化のモデルには, 薬物動態モデルと薬力学モデルを縦続接続したモデルが広く用いられている. この薬物動態モデルのパラメータは, 少人数に対する間欠的な血中濃度の測定より得られたものであり, BIS変化のモデルとして最良のものである保証はない. また従来のモデルでは, BIS変化に含まれる時間遅れを適切に考慮していない. そこで時間遅れを考慮したモデルについて, 投与速度から予測したBISと測定BISの誤差を最小とするパラメータを同定した. その結果, 従来のものよりも望ましいと考えられる値を得た. また時間遅れの適切な考慮の必要性が示唆された.

In this paper, we study indices based on electroencephalogram (EEG) for estimating effect of electroconvulsive therapy (ECT) on patients with depression. ECT is an effective treatment for psychiatric disorders such as depression, schizophrenia, and applied to patients who fail to respond to drug therapy. Since the parameters of ECT such as intensity, duration, and timing of electrical simulation are determined from the experience of medical doctors, they are not necessarily appropriate. In order to find the best parameters for each patient an objective index of the efficacy of ECT has to be established. For this purpose we examine correlation between the recovery of some clinical scales and the changes of feature quantities obtained from EEG. The result suggests that the change of the increasing speed of Spectral Entropy (SE) after cerebral seizure may be used as an index of the efficacy of ECT. © 2007 SICE.

Various actual systems include time delays due to measurement and/or computational delays, and transmission and transport lags. In this paper, the authors propose a state predictor for a certain class of multivariable systems including multiple output delays. The predictor consists of full-order observers, each estimates a past state from a delayed output, and finite interval integrators, which compensate the effect of the delays using state transition equations. The error of the state prediction converges to zero at the arbitrary rate adjusted by choosing a finite number of poles of the full-order observers. The output matrix used for observer design is not affected by the delays, whereas that of a conventional observer greatly depends on the length of the delays. Numerical examples of an integral process and an unstable process demonstrate that the errors of numerical computation in the proposed predictor are smaller than those in the conventional observer.

代表的な手術時の鎮静度指標として脳波を利用したBispectral IndexおよびEntropyがあげられる。これらの算出には短時間フーリエ変換が用いられているが、利用する脳波の時間幅は、Bispectral Indexについては一定であり、Entropyについても解析周波数帯域に応じて段階的に変化させているだけで、いずれも時間分解能が悪い。本研究では、より良い鎮静度指標を作ることを目的として、時間分解能の良いウェーブレット変換を用い、短時間フーリエ変換を用いた場合と比較検討する。

入出力に未知の複数むだ時間を含む多変数系の同定法は，未だ確立されていない．これに対し我々は，部分空間法に基づいてインパルス応答行列を推定し，これをむだ時間推定に用いて，さらに集中定数部を部分空間法で同定する方法を提案している．本発表では，この方法を外生入力を受ける確率システムの実現問題に適用し，有色雑音が加わる入出力むだ時間系についても，入出力データに基づき良好な状態空間実現を得る方法を示す．

近年、生体機能の複雑なメカニズムをシミュレーションすることにより、病気の原因解明などに役立てようという研究が行われている。このような研究対象の一つに血糖値調節機構がある。我々は昨年、門脈コンパートメントを含む血糖値調節機構のモデルを提案し、膵全摘犬のデータに基づいてパラメータの同定を行なった。本研究では、健常な犬に対するグルコースおよびインスリン投与時の血糖値・インスリン血中濃度データも利用して、さらに正確なパラメータを求めることを試みた。

筆者らが提案するスライディングモード制御を応用した制御対象の同定法は，スライディングモード手法の特徴を利用し，複数回の試行により，同定対象とモデルとの出力誤差が0となるようなパラメータを同定するものである．本稿では産業機械に多い油圧シリンダの位置決め制御を行う場合を想定し，「積分+二次遅れ」系に対して同定アルゴリズムを構成し，シミュレーションにより本手法の有効性を示す．

We have been engaged in developments of computer control systems to regulate various physiological states of patients under or after surgical operation. We started the research in 1991 in cooperation with surgeons and anesthesiologists. We first tried hypotensive control of the blood pressure of patients under surgical operation. The system was clinically applied to 34 cases since 1995 with the approval of the Ethics Committee of Kyoto University Hospital. As a result, we succeeded in decreasing the blood transfusion and the operation time roughly to the half and to the two thirds, respectively, in average. Based on this experience, we continued the research and developed the hypnosis control system that is under clinical application at the Day Surgery Unit of Kyoto University Hospital at present. We are now working for improving this system as well as trying new applications such as blood sugar control of patients after surgical operation. In this paper, our experience is introduced from the view point of complex-system control engineering. (c) 2005 Elsevier Ltd. All rights reserved.

Various practical systems include time delays due to measurement and computational delays, and transmission and transport lags. In this paper, the authors propose a novel state predictor for a certain class of multivariable systems including multiple output delays. The predictor consists of full-order observers estimating past state from each delayed output and finite interval integrators compensating the effect of the delays using state transition equations. State prediction error converges to zero at an arbitrary rate, which can be determined by choosing a finite number of poles of the full-order observers. In this predictor, the distance to instability of the state transition matrix is not affected by the delays. This means that large delays have no influence on the numerical stability, whereas that of a conventional observer highly depends on the delays. Numerical examples for an integral process and an unstable process demonstrate the effectiveness of the proposed predictor.

For the purpose of maintaining hypnosis level of patients during general anesthesia appropriately, we have developed a hypnosis control system with a function of online identification of individual parameters. As the control strategy, model predictive control, which can take the dead time included in the bispectral index (BIS) response to propofol infusion into account rigorously, is used. The online identification function obtains pharmacodynamic parameters and the dead time for each individual patient from the BIS response for the first few minutes of anesthesia induction. With the approval by the Ethics Committee of Kyoto University Hospital, the system was applied clinically, and its effectiveness in maintaining hypnosis level accurately and reducing the amount of propofol infusion was confirmed.

We developed an automatic blood pressure control system to maintain the blood pressure of patients at a substantially low level during a surgical operation. The developed system discharges two functions, continuous feedback control of the mean arterial pressure (MAP) by a state-predictive servo controller and risk control based on the inference by fuzzy-like logics and rules using measured data. Twenty-eight clinical applications were made beginning in November 1995, and the effects of the automatic blood pressure control on the operation time and on bleeding were assessed affirmatively by means of Wilcoxon testing. This paper essentially reports the engineering details of the control system.

We have been engaged in developments of computer control systems to regulate various physiological states of patients under or after surgical operation. We started this sort of research in 1991 in cooperation with surgeons and anesthesiologists. We first tried hypotensive control of the blood pressure of patients under surgical operation. The system was clinically applied to 31 cases since 1095 with the approval of the Ethics Committee of Kyoto University Hospital As a result, we succeeded in decreasing the blood transfusion and the operation time, roughly speaking, to the half in average. Based on this experience, we continued the research and developed the hypnosis control system that is under clinical application at the day surgery unit of Kyoto University Hospital at present. We are now working for improving this system as well as trying new applications such as blood sugar control of patients after surgical operation. In this talk, our experience will be introduced from the view point of complex-system control engineering.

A newly improved method for controlling blood glucose was compared with the standard model predictive controller under continuous glucose infusion. Continuous intravenous glucose infusion at rates of 50 or 100 mg/kg/hour was conducted on pancreatectomized dogs. An improved blood glucose control method using a combination of the proportional controller in the initial stage and the model predictive controller in the later stage was compared with the simple model predictive controller. The parameters of the controller were determined by identifying individual responses to the infused insulin during the first 60 minutes. The parameters of the proportional controller were changed at 60, 90, and 120 minutes to reflect the response to the infused insulin. The simple model predictive controller was able to reach the target level in the usual manner under the low infusion rate of glucose. However, under glucose infusion rates of 100 mg/kg/hour and more, it was difficult to reach the target level within 8 hours. In contrast, the improved system could reach the target level within 5 to 8 hours even under continuous glucose challenge. Addition of the modified proportional controller to the model predictive controller can stabilize the blood glucose control even under continuous glucose infusion.

In ambulatory surgery, anesthetic drug must be administered at an appropriate rate to prevent side effects after discharge from hospital. In this study, we have been developing a sedation control system using a model predictive controller, which uses propofol as the anesthetic drug and the bispectral index (BIS) as the index of sedation, and has an identification function of the individual response and a risk control function for preventing drug over-infusion and intra-operative arousal. After obtaining the approval from the Ethics Committee of Kyoto University Hospital, 80 clinical trials were made from July 2002. The results show that our system can reduce propofol infusion dose, and can maintain the BIS more accurately than manual administration.

A method has been proposed for designing an LQI servosystem in which reference characteristics and disturbance characteristics can be adjusted separately. The paper extends the method to the case in which the plant has a cascaded pure delay. It is proved that the resulting optimal compensator is the same as that for the plant with the delay removed, except for the existence of a state-prediction mechanism to cope with the delay, and that all the fundamental desirable features are inherited from the delay-free case. From the results of the paper, a complete solution to the optimal-design problem of state-predictive two-degree-of-freedom LQI servosystems for step references and step disturbances is obtained.