Farmer Terry G, Edgar Thomas F, Peppas Nicholas A
Department of Chemical, The University of Texas at Austin, 1 University Station C0400, Austin, TX 78712-0231, USA.
Ind Eng Chem Res. 2009 Mar 24;48(9):4402-4414. doi: 10.1021/ie800871t.
The effectiveness of closed-loop insulin infusion algorithms is assessed for three different mathematical models describing insulin and glucose dynamics within a Type I diabetes patient. Simulations are performed to assess the effectiveness of proportional plus integral plus derivative (PID) control, feedforward control, and a physiologically-based control system with respect to maintaining normal glucose levels during a meal and during exercise. Control effectiveness is assessed by comparing the simulated response to a simulation of a healthy patient during both a meal and exercise and establishing maximum and minimum glucose levels and insulin infusion levels, as well as maximum duration of hyperglycemia. Controller effectiveness is assessed within the minimal model, the Sorensen model, and the Hovorka model. Results showed that no type of control was able to maintain normal conditions when simulations were performed using the minimal model. For both the Sorensen model and the Hovorka model, proportional control was sufficient to maintain normal glucose levels. Given published clinical data showing the ineffectiveness of PID control in patients, the work demonstrates that controller success based on simulation results can be misleading, and that future work should focus on addressing the model discrepancies.
针对描述I型糖尿病患者体内胰岛素和葡萄糖动态变化的三种不同数学模型，评估了闭环胰岛素输注算法的有效性。进行模拟以评估比例加积分加微分（PID）控制、前馈控制以及基于生理学的控制系统在进餐和运动期间维持正常血糖水平方面的有效性。通过将模拟响应与健康患者在进餐和运动期间的模拟进行比较，并确定最大和最小血糖水平、胰岛素输注水平以及高血糖的最长持续时间，来评估控制效果。在最小模型、索伦森模型和霍沃卡模型中评估控制器的有效性。结果表明，当使用最小模型进行模拟时，没有一种控制类型能够维持正常状态。对于索伦森模型和霍沃卡模型，比例控制足以维持正常血糖水平。鉴于已发表的临床数据表明PID控制在患者中无效，这项工作表明基于模拟结果的控制器成功可能会产生误导，并且未来的工作应专注于解决模型差异。
