Closed-loop artificial pancreas insulin infusion control system

Abstract

The application discloses a closed-loop artificial pancreas insulin infusion control system, which comprises a detection module, a program module and an infusion module.The detection module is used for continuously detecting a current blood glucose value G.The program module is connected with the detection module, and a mixed artificial pancreas algorithm is preset in the program module, which is used for calculating an insulin infusion amount required by a user.The mixed artificial pancreas algorithm comprises a cPID algorithm and / or a cMPC algorithm.The input of the cPID algorithm is an intermediate value of the MPC algorithm, and the input of the cMPC algorithm is an output value of the PID algorithm.The infusion module is connected with the program module, and the program module sends the insulin infusion amount to the infusion module, and the infusion module infuses insulin according to the insulin infusion amount.Through deep combination of the PID algorithm and the MPC algorithm, the advantages of the PID algorithm and the MPC algorithm are fully utilized, the infusion result is more accurate and reliable, and the precise control of the closed-loop artificial pancreas insulin infusion system is realized.

Description

[0001] Cross-reference to related applications

[0002] This application claims the benefit of and priority of the following patent application: PCT patent application filed on October 25, 2021, application number PCT / CN2021 / 126005. Technical Field

[0003] This invention relates primarily to the field of medical devices, and in particular to a closed-loop artificial pancreas insulin infusion control system. Background Technology

[0004] In a healthy person, the pancreas automatically secretes the necessary insulin / glucagon based on blood glucose levels, thus maintaining a reasonable range of blood sugar fluctuations. However, in diabetic patients, pancreatic function is abnormal, and the pancreas is unable to secrete the required insulin normally. Diabetes is a metabolic disease and is a lifelong condition. Current medical technology cannot cure diabetes; it can only control the occurrence and development of diabetes and its complications by stabilizing blood sugar levels.

[0005] Diabetic patients need to have their blood glucose levels checked before insulin injection. Current methods allow for continuous glucose monitoring, sending real-time readings to a display device for easy viewing. This method is called Continuous Glucose Monitoring (CGM). The device is attached to the skin, with its probe inserted into the subcutaneous fluid for measurement. Based on the CGM reading, the infusion device delivers the required amount of insulin subcutaneously, creating a closed-loop or semi-closed-loop artificial pancreas.

[0006] Currently, proportional-integral-derivative (PID) and model-predictive-control (MPC) algorithms are widely studied to achieve closed-loop or semi-closed-loop control of insulin infusion. However, due to the simple structure of the PID algorithm, it is not suitable for more complex scenarios, while the MPC algorithm faces the dilemma of difficulty in establishing accurate models and large computational load, which may lead to infusion prediction errors.

[0007] Therefore, there is an urgent need in the existing technology for a closed-loop artificial pancreas insulin infusion control system that incorporates a hybrid artificial pancreas algorithm. Summary of the Invention

[0008] The embodiment of the application discloses a closed-loop artificial pancreas insulin infusion control system, which is provided with a mixed artificial pancreas algorithm, the mixed artificial pancreas algorithm comprises a cPID algorithm and / or a cMPC algorithm, the input of the cPID algorithm is an intermediate value of an MPC algorithm, and the input of the cMPC algorithm is an output value of a PID algorithm. Through deep combination of the PID algorithm and the MPC algorithm, the advantages of the PID algorithm and the MPC algorithm are fully utilized, the infusion result is more accurate and reliable, and accurate control of the closed-loop artificial pancreas insulin infusion system is realized.

[0009] The application discloses a closed-loop artificial pancreas insulin infusion control system, which comprises a detection module, a program module and an infusion module. The detection module is used for continuously detecting a current blood glucose value G. The program module is connected with the detection module, and a mixed artificial pancreas algorithm is preset in the program module, which is used for calculating an insulin infusion amount required by a user. The mixed artificial pancreas algorithm comprises a cPID algorithm and / or a cMPC algorithm. The input of the cPID algorithm is an intermediate value of an MPC algorithm, and the input of the cMPC algorithm is an output value of a PID algorithm. The infusion module is connected with the program module. The program module sends the insulin infusion amount to the infusion module, and the infusion module performs insulin infusion according to the insulin infusion amount.

[0010] According to an aspect of the application, the cPID algorithm is calculated according to a blood glucose value at a current time predicted by a prediction model of the MPC algorithm, and the calculation formula is:

[0011]

[0012] Wherein:

[0013] K P is a gain coefficient of a proportional part;

[0014] K I is a gain coefficient of an integral part;

[0015] K D is a gain coefficient of a differential part;

[0016] G MPC(t) represents a blood glucose value at a current time predicted by an MPC prediction model;

[0017] G B represents a target blood glucose value;

[0018] C represents a constant;

[0019] cPID(t) represents an infusion instruction sent to an insulin infusion system.

[0020] According to an aspect of the application, the cPID algorithm is calculated according to a blood glucose value at a current time predicted by a prediction model of the MPC algorithm after risk conversion, and the calculation formula is:

[0021]

[0022] wherein:

[0023] K P is a gain coefficient of the proportional part;

[0024] K I is a gain coefficient of the integral part;

[0025] K D is a gain coefficient of the derivative part;

[0026] r MPc(t) represents a blood glucose risk after risk conversion based on a blood glucose value at the current time predicted by the MPC prediction model;

[0027] G B represents a target blood glucose value;

[0028] C represents a constant;

[0029] rcPID(t) represents an infusion instruction sent to an insulin infusion system.

[0030] According to one aspect of the present application, the blood glucose risk space conversion method includes one or more of piecewise weighted method, relative value conversion, blood glucose risk index conversion, and improved control variability grid analysis conversion.

[0031] According to one aspect of the present application, the blood glucose risk space conversion method further includes one or more of the following processing methods:

[0032] ① subtracting a component proportional to the estimated plasma insulin concentration predicted;

[0033] ② subtracting the amount of insulin that has not yet played its role in the body;

[0034] ③ compensating for blood glucose and interstitial fluid glucose concentration sensing delay using an autoregressive method.

[0035] According to one aspect of the present application, the insulin infusion amount at the current time in the prediction model of the cMPC algorithm is calculated by the rPID algorithm, and the prediction model is:

[0036] x t+1 = Ax t and BI PID(t)

[0037] G t = Cx t

[0038] wherein:

[0039] xt+1 x represents the state parameter at the next time point,

[0040] x t x represents the state parameter at the current time point,

[0041] I PID(t) I represents the insulin infusion amount at the current time point calculated by the PID algorithm;

[0042] G t G represents the blood glucose concentration at the current time point.

[0043] The parameter matrix is as follows:

[0044]

[0045]

[0046] C = [1 0 0]

[0047] b1, b2, b3, K are prior values.

[0048] According to one aspect of the present application, the insulin infusion amount at the current time point in the prediction model of the cMPC algorithm is calculated by the rPID algorithm, and the prediction model is:

[0049] x t+1 = Ax t + BI rPID(t)

[0050] G t = Cx t

[0051] Wherein:

[0052] x t+1 x represents the state parameter at the next time point,

[0053] x t x represents the state parameter at the current time point,

[0054] I rPID(t) I represents the insulin infusion amount at the current time point calculated by the rPID algorithm;

[0055] G t G represents the blood glucose concentration at the current time point.

[0056] The parameter matrix is as follows:

[0057]

[0058]

[0059] C = [1 0 0]

[0060] b1, b2, b3, K are prior values.

[0061] According to an aspect of the present application, the rPID algorithm, on the basis of the PID algorithm, converts the blood glucose asymmetric in the original physical space to the blood glucose risk approximately symmetric in the risk space.

[0062] According to an aspect of the present application, the blood glucose value in the value function of the cMPC algorithm is risk-converted, and the value function after the blood glucose risk conversion is:

[0063]

[0064] wherein,

[0065] r t+j represents the blood glucose risk value after the jth step;

[0066] I′ t+j represents the change of the insulin infusion amount after the jth step;

[0067] t represents the current time;

[0068] N and P are respectively the step numbers in the control time window and the prediction time window;

[0069] R is a weighted coefficient of the insulin component.

[0070] According to an aspect of the present application, the blood glucose risk conversion method comprises one or more of the following: the piecewise weighting method, the relative value conversion, the blood glucose risk index conversion, and the improved control variability grid analysis conversion.

[0071] According to an aspect of the present application, the blood glucose risk conversion method further comprises one or more of the following processing methods:

[0072] ① deducting a component proportional to the estimated plasma insulin concentration;

[0073] ② deducting the amount of insulin that has not yet played its role in the body;

[0074] ③ compensating for the sensing delay of blood glucose and interstitial fluid glucose concentration by using an autoregressive method.

[0075] According to an aspect of the present application, the hybrid artificial pancreas algorithm comprises the cPID algorithm and the cMPC algorithm, the cPID algorithm calculates the first insulin infusion amount I1, the cMPC algorithm calculates the second insulin infusion amount I2, and the hybrid artificial pancreas algorithm calculates the optimized calculation of the first insulin infusion amount I1 and the second insulin infusion amount I2 to obtain the final insulin infusion amount I3.

[0076] According to one aspect of the present application, the final insulin infusion amount I3 is optimized by the average value of the first insulin infusion amount I1 and the second insulin infusion amount I2:

[0077] ① Solve the average value of the first insulin infusion amount I1 and the second insulin infusion amount I2

[0078] ② Bring the average value into the cPID algorithm and the cMPC algorithm, and adjust the algorithm parameters;

[0079] ③ Recalculate the first insulin infusion amount I1 and the second insulin infusion amount I2 based on the current blood glucose value, the cPID algorithm and the cMPC algorithm after adjusting the parameters;

[0080] ④ Perform loop calculation on steps ①-③ until I1=I2, and the final insulin infusion amount I3=I1=I2.

[0081] According to one aspect of the present application, the final insulin infusion amount I3 is optimized by the weighted value of the first insulin infusion amount I1 and the second insulin infusion amount I2:

[0082] ① Solve the weighted value of the first insulin infusion amount I1 and the second insulin infusion amount I2 Wherein α and β are the weighting coefficients of the first insulin infusion amount I1 and the second insulin infusion amount I2, respectively;

[0083] ② Bring the weighted value into the cPID algorithm and the cMPC algorithm, and adjust the algorithm parameters;

[0084] ③ Recalculate the first insulin infusion amount I1 and the second insulin infusion amount I2 based on the current blood glucose value, the cPID algorithm and the cMPC algorithm after adjusting the parameters;

[0085] ④ Perform loop calculation on steps ①-③ until I1=I2, and the final insulin infusion amount I3=I1=I2.

[0086] According to one aspect of the present application, the final insulin infusion amount I3 is obtained by comparing the first insulin infusion amount I1 and the second insulin infusion amount I2 with the statistical analysis result I4 of historical data:

[0087]

[0088] According to one aspect of the present application, two of the detection module, the program module and the infusion module are connected to each other to form an integral structure, and are respectively pasted on different positions of the skin with the third module.

[0089] According to one aspect of the present invention, the detection module, the program module, and the infusion module are connected to form an integral structure and are attached to the same location on the skin.

[0090] Compared with the prior art, the technical solution of the present invention has the following advantages:

[0091] In the closed-loop artificial pancreas insulin infusion control system disclosed in this invention, a hybrid artificial pancreas algorithm is preset. The hybrid artificial pancreas algorithm includes a cPID algorithm and / or a cMPC algorithm. The input of the cPID algorithm is the intermediate value of the MPC algorithm, and the input of the cMPC algorithm is the output value of the PID algorithm. Through the deep integration of the PID algorithm and the MPC algorithm, the advantages of the PID algorithm and the MPC algorithm are fully utilized to make the infusion results more accurate and reliable, and to achieve precise control of the closed-loop artificial pancreas insulin infusion system.

[0092] Furthermore, the cPID algorithm in the hybrid artificial pancreas algorithm further transforms the current blood glucose value predicted by the prediction model of the MPC algorithm into a blood glucose risk conversion, making the hybrid artificial pancreas algorithm more robust.

[0093] Furthermore, the cMPC algorithm in the hybrid artificial pancreas algorithm combines the current insulin infusion volume with the prediction model calculated by the PID algorithm or rPID algorithm and the input with or without risk transformation, flexibly utilizing the advantages of the PID algorithm, MPC algorithm and blood glucose risk transformation to deal with complex scenarios. This enables the artificial pancreas to provide a reliable insulin infusion volume under various conditions, thereby achieving the ideal blood glucose level at the expected time and realizing precise control of the closed-loop artificial pancreas insulin infusion system.

[0094] Furthermore, the hybrid artificial pancreas algorithm compensates for delays in insulin absorption, insulin onset, and the sensing delays in blood glucose and interstitial fluid glucose concentrations, making the output calculated by the hybrid artificial pancreas algorithm more reliable.

[0095] Furthermore, the final output of the hybrid artificial pancreas algorithm is a consistent result obtained by calculating using the cMPC and cPID algorithms, which is more feasible and reliable.

[0096] Furthermore, the final output of the composite artificial pancreas algorithm is the same result obtained by averaging or weighting the different results calculated by the cMPC algorithm and cPID algorithm. The two algorithms compensate for each other, further improving the accuracy of the output results.

[0097] Furthermore, the final output of the hybrid artificial pancreas algorithm is obtained by optimizing the calculation results of the cMPC and cPID algorithms. This optimization, combined with statistical analysis of historical control data, ensures the reliability of insulin infusion from another perspective.

[0098] Further, the detection module, the program module and the infusion module are connected to form an integral structure and are attached to the same position of the skin. The three modules are connected to form an integral structure and are attached to the same position, so that the number of devices attached to the skin of the user is reduced, and the interference of the user's activities caused by the attachment of more devices is weakened; meanwhile, the problem of poor wireless communication between separated devices is effectively solved, and the user experience is further enhanced. BRIEF DESCRIPTION OF DRAWINGS

[0099] Figure 1 A schematic diagram of module relationship of a closed-loop artificial pancreas insulin infusion control system according to an embodiment of the present application;

[0100] Figure 2 A comparison diagram of the relationship between the risk space and the blood glucose in the original physical space obtained by the piecewise weighting processing and the relative value conversion method according to an embodiment of the present application;

[0101] Figure 3 A comparison diagram of the relationship between the risk space and the blood glucose in the original physical space obtained by the BGRI and CVGA methods according to an embodiment of the present application;

[0102] Figure 4 An insulin IOB curve according to an embodiment of the present application;

[0103] Figure 5 A schematic diagram of four types of mainstream clinical optimal basal rate setting according to an embodiment of the present application. DETAILED DESCRIPTION

[0104] As described above, the PID algorithm in the prior art artificial pancreas has a simple structure and is not suitable for complex scenarios, and the MPC algorithm faces the difficulties of difficult establishment of an accurate model and large calculation amount, which may result in a predicted infusion deviation.

[0105] To solve the problem, the present application provides a closed-loop artificial pancreas insulin infusion control system, which is provided with a hybrid artificial pancreas algorithm, the hybrid artificial pancreas algorithm comprising a cPID algorithm and / or a cMPC algorithm, the input of the cPID algorithm being an intermediate value of the MPC algorithm, and the input of the cMPC algorithm being an output value of the PID algorithm. Through the deep combination of the PID algorithm and the MPC algorithm, the advantages of the PID algorithm and the MPC algorithm are fully utilized, so that the infusion result is more accurate and reliable, and the precise control of the closed-loop artificial pancreas insulin infusion system is realized.

[0106] Various exemplary embodiments of the present application will now be described in detail with reference to the accompanying drawings. It should be appreciated that the relative dimensions of the components, expressions of numbers, and numerical values set forth in these embodiments are not to be construed as limiting the scope of the present application, unless otherwise specifically stated.

[0107] Furthermore, it should be understood that the dimensions of the various components illustrated in the drawings are chosen for purposes of convenience and ease of understanding only. For example, the thickness, width, length or distance of certain elements can be exaggerated relative to other elements for clarity.

[0108] The following description of the exemplary embodiments is merely illustrative in nature and is in no way intended to limit the scope of the application, its application, or uses. Techniques, methods, and apparatus known to those of ordinary skill in the art can not be discussed in detail for the sake of brevity and / or clarity. However, it should be understood that all of these techniques, methods, and apparatus are considered to be part of the present description.

[0109] It is to be noted that like-numbers and letters on the attached drawings represent similar items and, thus, one need not concern himself with the same numbers and letters throughout the several drawings. It is also to be understood that, alternatively, the use of only distinct numbers and letters for simplicity of illustration is intended to convey the idea in the alternative that where either the same numbers or letters are intended to represent similar items in different figures of the drawings, a single number or letter is meant to be representative of those or similar items in each of the figures.

[0110] Figure 1 A schematic diagram of the modules of the closed-loop artificial pancreas insulin infusion control system of the embodiments of the present application.

[0111] The closed-loop artificial pancreas insulin infusion control system disclosed in the embodiments of the present application mainly comprises a detection module 100, a program module 101, and an infusion module 102.

[0112] The detection module 100 is used to continuously detect the current blood glucose value of the user. Generally, the detection module 100 is a continuous glucose monitor (CGM), which can detect the current blood glucose value of the user in real time and monitor the change in blood glucose, and send the current blood glucose value to the program module 101.

[0113] The program module 101 is used to control the operation of the detection module 100 and the infusion module 102. Therefore, the program module 101 is connected to the detection module 100 and the infusion module 102, respectively. Here, the connection includes conventional electrical connection or wireless connection.

[0114] The infusion module 102 contains the mechanical structure necessary for insulin infusion, and is controlled by the program module 101. According to the current insulin infusion amount data issued by the program module 101, the infusion module 102 infuses the current required insulin into the user's body. At the same time, the infusion state of the infusion module 102 can also be fed back to the program module 101 in real time.

[0115] The embodiment of the present application does not limit the specific positions and connection relationship of the detection module 100, the program module 101 and the infusion module 102, as long as the above-mentioned function conditions can be met.

[0116] As in one embodiment of the present application, the three are electrically connected to each other to form an integral structure. Therefore, the three are pasted at the same position on the user's skin. The three modules are connected to form an integral structure and are pasted at the same position, so that the number of devices pasted on the user's skin will be reduced, thereby weakening the interference of the pasted devices on the user's activities; at the same time, the problem of the reliability of wireless communication between separated devices is effectively solved, further enhancing the user experience.

[0117] As in another embodiment of the present application, the program module 101 and the infusion module 102 are connected to each other to form an integral structure, and the detection module 100 is separately arranged in another structure. At this time, the detection module 100 and the program module 101 transmit wireless signals to each other to realize the connection with each other. Therefore, the program module 101 and the infusion module 102 are pasted at a certain position on the user's skin, and the detection module 100 is pasted at another position on the user's skin.

[0118] As in still another embodiment of the present application, the program module 101 and the detection module 100 are connected to each other to form the same device, and the infusion module 102 is separately arranged in another structure. The infusion module 102 and the program module 101 transmit wireless signals to each other to realize the connection with each other. Therefore, the program module 101 and the detection module 100 can be pasted at a certain position on the user's skin, and the infusion module 102 can be pasted at another position on the user's skin.

[0119] As in still another embodiment of the present application, the three are arranged in different structures respectively. Therefore, the three are pasted at different positions on the user's skin respectively. At this time, the program module 101 transmits wireless signals to the detection module 100 and the infusion module 102 respectively to realize the connection with each other.

[0120] It should be noted that the program module 101 of the embodiment of the present application also has the functions of storage, recording and access to the database, and therefore the program module 101 can be reused. In this way, not only the user's physical condition data can be stored, but also the production cost and the use cost of the user can be saved. As described above, when the detection module 100 or the infusion module 102 reaches the end of life, the program module 101 can be separated from the detection module 100, the infusion module 102 or both the detection module 100 and the infusion module 102.

[0121] Generally, the service life of the detection module 100, the program module 101 and the infusion module 102 is different. Therefore, when the three are electrically connected to form one device, the three can also be separated from each other in pairs. For example, when one module reaches the end of its service life, the user can replace only the module, and keep the other two modules for continued use.

[0122] It should be noted that the program module 101 of the embodiment of the application can also include a plurality of sub-modules. According to the functions of the sub-modules, different sub-modules can be arranged in different structures, which are not specifically limited here, as long as the control conditions of the program module 101 can be met.

[0123] Specifically, the program module 101 is preconfigured with an rPID (risk-proportion-integral-derivative) algorithm for converting blood glucose that is asymmetric in the original physical space to blood glucose risk that is approximately symmetric in the risk space. The rPID algorithm is obtained by converting a classical PID (proportion-integral-derivative) algorithm, and the specific processing manner will be described in detail below. According to the corresponding infusion instruction calculated by the rPID algorithm, the program module 101 controls the infusion module 102 to infuse insulin.

[0124] The classical PID algorithm can be expressed by the following formula:

[0125]

[0126] Wherein:

[0127] K P is a gain coefficient of the proportional part;

[0128] K I is a gain coefficient of the integral part;

[0129] K D is a gain coefficient of the derivative part;

[0130] G represents the current blood glucose value;

[0131] G B represents the target blood glucose value;

[0132] C represents a constant;

[0133] PID(t) represents the infusion instruction sent to the insulin infusion system.

[0134] Considering the actual distribution characteristics of the glucose concentration of diabetic patients, such as the normal blood glucose range of 80-140 mg / dL, which can also be relaxed to 70-180 mg / dL, the general low blood glucose can reach 20-40 mg / dL, and the high blood glucose can reach 400-600 mg / dL.

[0135] The distribution of high / low blood glucose has significant asymmetry in the original physical space, and the risk of high blood glucose and the risk of low blood glucose corresponding to the same deviation of blood glucose from the normal range in clinical practice are obviously different. For example, a decrease of 70 mg / dL from 120 mg / dL to 50 mg / dL is considered as severe hypoglycemia with high clinical risk, and emergency measures such as supplement of carbohydrates need to be taken; while an increase of 70 mg / dL from 120 mg / dL to 190 mg / dL just exceeds the normal range, and the degree of high blood glucose is not serious for diabetic patients, and the blood glucose is often reached in daily situation, and basically no treatment measures need to be taken.

[0136] In view of the asymmetric characteristics of the clinical risk of glucose concentration, the asymmetric blood glucose in the original physical space is converted to the blood glucose risk in the risk space which is approximately symmetric, so that the PID algorithm is more robust.

[0137] Correspondingly, the rPID algorithm formula is converted as follows:

[0138]

[0139] Among them:

[0140] rPID(t) represents the infusion instruction sent to the insulin infusion system after the risk conversion;

[0141] r represents the blood glucose risk;

[0142] The meanings of other symbols are as described above.

[0143] In order to maintain the stability of the PID integral, combined with the physiological effect of insulin in reducing blood glucose, in an embodiment of the present application, the input parameter of the PID, the blood glucose deviation Ge=G-G B is processed, such as the segmented weighted processing on Ge=G-G B , as follows:

[0144]

[0145] In another embodiment of the present application, the relative value is used to convert the deviation greater than the target blood glucose G B , as follows:

[0146]

[0147] Figure 2 The comparison chart of the blood glucose risk space obtained by the segmented weighted processing and the relative value conversion and the blood glucose in the original physical space.

[0148] In the original PID algorithm, the blood glucose risk (i.e. Ge) on both sides of the target blood glucose value presents a serious asymmetry consistent with the original physical space. After conversion to the blood glucose risk space, the blood glucose risk on both sides of the target blood glucose value is approximately symmetrical, so that the integral term can remain stable, making the rPID algorithm more robust.

[0149] In another embodiment of the present application, there is a fixed zero risk point when converting risk, and data deviating from both sides of the zero risk point is processed. The original parameter corresponding to the risk point greater than zero is positive when converted to the risk space, and the original parameter corresponding to the risk point less than zero is negative when converted to the risk space. Specifically, the classic blood glucose risk index (BGRI) method can be used for reference. This method is based on clinical practice and considers that the clinical risk of hypoglycemia of 20 mg / dL and hyperglycemia of 600 mg / dL is equivalent, and the blood glucose in the range of 20-600 mg / dL is processed as a whole by logarithmization. The blood glucose value corresponding to the zero risk point of this method is set to the target blood glucose value G B . Its risk space conversion formula is as follows:

[0150]

[0151] Wherein:

[0152] r(G) = 10 * f(G) 2

[0153] The conversion function f(G) is as follows:

[0154] f(G) = 1.509 * [(ln(G)) 1.084 - 5.381]

[0155] In the classic blood glucose risk index method, the blood glucose value corresponding to the zero risk point of this method is 112 mg / dL. In other embodiments of the present application, the zero risk point blood glucose value can also be adjusted in combination with the risk and data trend of clinical practice, which is not specifically limited here. The risk space of the blood glucose value greater than the zero risk point is fitted, and the specific fitting method is not specifically limited.

[0156] In another embodiment of the present application, the zero risk point blood glucose value defined by the original CVGA method is 110 mg / dL, and the following pairs of blood glucose values with equal risk are assumed (90 mg / dL, 180 mg / dL; 70 mg / dL, 300 mg / dL; 50 mg / dL, 400 mg / dL). In the embodiment of the present application, the real risk and data trend characteristics in clinical practice are considered, and the pair of equal risk (70 mg / dL, 300 mg / dL) is modified to (70 mg / dL, 250 mg / dL), and the zero risk point blood glucose value is set as the target blood glucose value G B . The polynomial model fitting is performed, and the following risk functions for the two sides of the zero risk point are obtained:

[0157]

[0158] The maximum value is limited:

[0159] |r|=min(|r|,n)

[0160] The value range of the maximum value n is 0-80 mg / dL, and the value of n is preferably 60 mg / dL.

[0161] In other embodiments of the present application, the zero risk point blood glucose value and the pair of equal risk data can also be adjusted according to the real risk and data trend in clinical practice, which is not specifically limited herein. The equal risk point is fitted again, and the specific fitting method is not specifically limited; and the specific value for limiting the maximum value is also not specifically limited.

[0162] Figure 3 The comparison chart of the blood glucose risk converted to the risk space by the BGRI and CVGA methods and the blood glucose in the original physical space.

[0163] Similar to the processing of Zone-MPC, the blood glucose risk converted by the BGRI and CVGA methods is relatively flat in the normal range of blood glucose, especially in the range of 80-140 mg / dL. Unlike Zone-MPC, which is completely 0 in this range and loses the ability to further optimize, the risk of rPID is flat in this range, but still has stable and slow adjustment ability, which can further adjust the blood glucose to the target value and achieve more accurate blood glucose control.

[0164] In another embodiment of the present application, uniform processing can be applied to the data deviating from both sides of the zero risk point, as in the foregoing embodiment, the data deviating from both sides of the zero risk point can be processed by BGRI or CVGA method; or different processing can be applied, such as combining BGRI and CVGA method, in which case the same zero risk point blood glucose value, such as target blood glucose value G B , can be used. When the blood glucose value is less than the target blood glucose value G B , the BGRI method is used, and when the blood glucose value is greater than the target blood glucose value G B , the CVGA method is used, in which case:

[0165] r = -r(G), if G≤G B

[0166] Wherein:

[0167] r(G) = 10*f(G) 2

[0168] The conversion function f(G) is as follows:

[0169] f(G) = 1.509*[(ln(G)) 1.084 -5.381]

[0170] r = -4.8265*10 4 -4*G 2 +0.45563*G-44.855, if G>G B .

[0171] Similarly, the CVGA method can be used when the blood glucose value is less than the target blood glucose value G B , and the BGRI method can be used when the blood glucose value is greater than the target blood glucose value G B , in which case:

[0172] r = r(G), if G>G B ,

[0173] Wherein:

[0174] r(G) = 10*f(G) 2

[0175] The conversion function f(G) is as follows:

[0176] f(G) = 1.509*[(ln(G)) 1.084 -5.381]

[0177] r = G-G B , if G≤G B .

[0178] Meanwhile, the maximum value can also be limited:

[0179] |r| = min(|r|, n)

[0180] wherein the maximum value n is defined in the range of 0-80 mg / dL, preferably n is 60 mg / dL.

[0181] In other embodiments of the present application, the blood glucose value at the zero risk point can also be set as the target blood glucose value G B For data less than or equal to the target blood glucose value G B , the BGRI method is used, while for data greater than the target blood glucose value G B , the processing method of deviation amount is used, such as piecewise weighted processing or relative value processing.

[0182] When piecewise weighted processing is used, at this time:

[0183] r = -r(G), if G≤G B

[0184] wherein:

[0185] r(G) = 10*f(G) 2

[0186] Under the conversion function f(G):

[0187] f(G) = 1.509*[(ln(G)) 1.084 -5.381]

[0188]

[0189] When relative value processing is used:

[0190] r = -r(G), if G≤G B ,

[0191] wherein:

[0192] r(G) = 10*f(G) 2

[0193] Under the fitted symmetric conversion function f(G):

[0194] f(G) = 1.509*[(ln(G)) 1.084 -5.381]

[0195] r = 100*(G-G B ) / G, if G>G B

[0196] When the blood glucose values corresponding to the zero risk points are all the target blood glucose value G BWhen the data is less than or equal to the target blood glucose value G B , the processing functions of the piecewise weighted processing, the relative value processing and the CVGA method are consistent, so when the piecewise weighted processing or the relative value processing is adopted for the data less than or equal to the target blood glucose value G B , and the BGRI method is adopted for the data greater than the zero risk point blood glucose value, the processing results are equivalent to the aforementioned CVGA method for the blood glucose value less than or equal to the target blood glucose value G B , and the BGRI method for the blood glucose value greater than the target blood glucose value G B , and the calculation formula is not repeated.

[0197] It should be noted that in each embodiment of the present application, the target blood glucose value G B is 80-140 mg / dL, and preferably, the target blood glucose value G B is 110-120 mg / dL.

[0198] The above processing methods can make the rPID algorithm convert the asymmetric blood glucose in the original physical space to the approximately symmetric blood glucose risk in the risk space, so as to not only retain the simple and robust characteristics of the PID algorithm, but also have the targeted and clinically valuable blood glucose risk control function, and realize the precise control of the closed-loop artificial pancreas insulin infusion system.

[0199] In the closed-loop artificial pancreas control system, there are three delay effects: insulin absorption delay (about 20 minutes from subcutaneous to blood circulation tissue, and about 100 minutes to reach the liver), insulin onset delay (about 30-100 minutes), and sensing delay of interstitial fluid glucose concentration and blood glucose (about 5-15 minutes). Any attempt to accelerate the responsiveness of the closed-loop system can lead to unstable system behavior and system oscillation. In order to compensate for the insulin absorption delay in the closed-loop artificial pancreas control system, in an embodiment of the present application, an insulin feedback compensation mechanism is introduced. The amount of insulin that has not been absorbed in the body is deducted from the output, and a component proportional to the estimated plasma insulin concentration (the actual human insulin secretion also uses the insulin concentration in the blood as a negative feedback signal). The formula is as follows:

[0200]

[0201] Wherein:

[0202] PID(t) represents the infusion instruction sent to the insulin infusion system;

[0203] PID c (t) represents the infusion instruction with compensation sent to the insulin infusion system;

[0204] γ represents the compensation coefficient of the estimated plasma insulin concentration to the algorithm output, the coefficient becomes larger, the algorithm becomes relatively conservative, and the coefficient becomes smaller, the algorithm becomes relatively aggressive, therefore, in the embodiment of the application, the range of γ is 0.4-0.6, preferably, γ is 0.5.

[0205] represents the estimation of the plasma insulin concentration, which can be obtained by various conventional prediction algorithms, such as being directly calculated from the infused insulin according to the pharmacokinetic curve of the insulin, or using a conventional autoregressive method:

[0206]

[0207] wherein:

[0208] represents the estimation of the plasma insulin concentration at the current time;

[0209] PID c (n-1) represents the output with compensation at the last time;

[0210] represents the estimation of the plasma insulin concentration at the last time;

[0211] represents the estimation of the plasma insulin concentration at the time before the last time;

[0212] K0 represents the coefficient of the output part with compensation at the last time;

[0213] K1 represents the coefficient of the estimation part of the plasma insulin concentration at the last time;

[0214] K2 represents the coefficient of the estimation part of the plasma insulin concentration at the time before the last time;

[0215] wherein, the initial value is The time interval between each time can be selected according to actual needs.

[0216] Correspondingly, the compensation output formula after the risk conversion by the foregoing method is as follows:

[0217]

[0218] wherein:

[0219] rPIDc(t) represents the infusion instruction with compensation sent to the insulin infusion system after the risk conversion;

[0220] rPID(t) represents the infusion instruction sent to the insulin infusion system after the risk conversion;

[0221] The meanings of other characters are as described above.

[0222] In order to compensate for the delay of insulin action in the closed-loop artificial pancreas control system, in one embodiment of the present application, insulin on board (IOB) is introduced, which is the amount of insulin that has not yet acted in the body, and the IOB is deducted from the output of insulin, preventing the accumulation of insulin infusion, excessive amount, and the risk of postprandial hypoglycemia.

[0223] Figure 4 is an insulin IOB curve according to an embodiment of the present application.

[0224] According to the IOB curve shown in FIG. 1, the cumulative residual amount of previously infused insulin can be calculated, and the specific curve can be selected according to the actual insulin action time of the user. Figure 4

[0225] PID'(t) = PID(t) - IOB(t)

[0226] wherein:

[0227] PID'(t) represents the infusion instruction sent to the insulin infusion system after deducting the IOB;

[0228] PID(t) represents the infusion instruction sent to the insulin infusion system;

[0229] IOB(t) represents the amount of insulin that has not yet acted in the body at time t.

[0230] Correspondingly, the output formula for deducting the amount of insulin that has not yet acted in the body after risk conversion by the foregoing method is as follows:

[0231] rPID'(t) = rPID(t) - IOB(t)

[0232] wherein:

[0233] rPID'(t) represents the infusion instruction sent to the insulin infusion system after deducting the amount of insulin that has not yet acted in the body after risk conversion;

[0234] rPID(t) represents the infusion instruction sent to the insulin infusion system after risk conversion;

[0235] The meanings of other characters are as described above.

[0236] In order to obtain a more ideal control effect, the calculation of IOB is processed as follows, IOB m , IOB o respectively correspond to the IOB of meal insulin and other insulin except meal. The formula is as follows:

[0237] IOB(t) = IOB m，t + IOB o，t ​

[0238] wherein:

[0239]

[0240] wherein:

[0241] IOB m，t represents the amount of meal insulin that has not yet acted in the body at time t;

[0242] IOB o，t represents the amount of non-meal insulin that has not yet acted in the body at time t;

[0243] D i (i = 2-8) represents the respective coefficients of the IOB curve corresponding to the insulin action time of i, respectively;

[0244] I m，t represents the amount of meal insulin;

[0245] I 0，t represents the amount of non-meal insulin;

[0246] IOB(t) represents the amount of insulin that has not yet acted in the body at time t.

[0247] The meal insulin and non-meal insulin are distinguished in the processing of IOB, which can make the insulin be cleared faster when eating and blood glucose is too high, can obtain greater insulin output, and blood glucose regulation is faster. When close to the target, a longer insulin action time curve is used to make the insulin be cleared slower, and blood glucose regulation is more conservative and stable.

[0248] When PID'(t) > 0 or rPID'(t) > 0, the final amount of insulin infused is PID'(t) or rPID'(t);

[0249] When PID'(t) < 0 or rPID'(t) < 0, the final amount of insulin infused is 0.

[0250] In order to compensate for the sensing delay of tissue fluid glucose concentration and blood glucose in the closed-loop artificial pancreas control system, in an embodiment of the present application, an autoregressive method is used for compensation, and the formula is as follows:

[0251]

[0252] wherein,

[0253] G SC (n) represents the current time interstitial fluid glucose concentration, that is, the measurement value of the sensing system;

[0254] represents the estimated concentration of blood glucose at the last time point;

[0255] G SC (n-1) and G SC (n-2) represent the interstitial fluid glucose concentration at the last time point and the time point before the last time point respectively;

[0256] K0 represents the coefficient of the estimated concentration of blood glucose at the last time point;

[0257] K1 and K2 represent the coefficients of the interstitial fluid glucose concentration at the last time point and the time point before the last time point respectively.

[0258] wherein, at the initial time point,

[0259] By estimating the blood glucose concentration through the interstitial fluid glucose concentration, the sensing delay of the interstitial fluid glucose concentration and the blood glucose is compensated, the PID algorithm is more accurate, and accordingly, the rPID algorithm can more accurately calculate the actual demand of the human body for insulin.

[0260] In the embodiment of the present application, the insulin absorption delay, the insulin onset delay, the sensing delay of the interstitial fluid glucose concentration and the blood glucose can be partially compensated or fully compensated, preferably, all the delay factors are considered for full compensation, so that the rPID algorithm is more accurate.

[0261] In another embodiment of the present application, the program module 101 is pre-provided with an rMPC (risk-model-predictive-control) algorithm for converting the blood glucose asymmetric in the original physical space to the blood glucose risk approximately symmetric in the risk space, the rMPC algorithm is obtained by conversion processing on the basis of the classical MPC (model-predictive-control) algorithm, and the corresponding infusion instruction calculated according to the rMPC algorithm is used by the program module 101 to control the infusion module 102 to infuse insulin.

[0262] The classical MPC algorithm is composed of three elements, a prediction model, a value function and a constraint condition. The prediction model of the classical MPC is as follows:

[0263] x t+1 =Ax t +BI t

[0264] G t =Cx t

[0265] wherein:

[0266] x t+1 represents the state parameter at the next time point,

[0267] xt a state parameter representing a current time point,

[0268] I t an insulin infusion amount representing a current time point;

[0269] G t a blood glucose concentration representing a current time point.

[0270] The parameter matrix is as follows:

[0271]

[0272]

[0273] C=[1 0 0]

[0274] b1, b2, b3, K are prior values.

[0275] The value function of the MPC is composed of the square sum of the deviation of the output G (blood glucose value) and the square sum of the change of the input I (insulin amount). The MPC needs to obtain the minimum solution of the value function.

[0276]

[0277] wherein:

[0278] I′ t+j represents the change of the insulin infusion amount after the jth step;

[0279] represents the difference between the predicted blood glucose concentration and the target blood glucose value after the jth step;

[0280] t represents a current time point;

[0281] N and P are respectively the step number in the control time window and the prediction time window;

[0282] R is the weighted coefficient of the insulin component.

[0283] The insulin infusion amount of the jth step is I t +I′ t+j .

[0284] In the embodiment of the present application, the control time window T c = 30 min, the prediction time window T p = 60 min, and the weighted coefficient R of the insulin amount is 11000. It should be noted that although the control time window adopted in the calculation is 30 min, only the first step calculation result of the insulin output is adopted in the actual operation, and after the operation, the minimum solution of the above value function is recalculated according to the latest blood glucose value obtained.

[0285] In the embodiment of the present application, the infusion time step j in the control time window is controlled n , j n The value range of j is 0-30 min, preferably 2 min. The number of steps N = T c / j n The range of j is 0 to N.

[0286] In other embodiments of the present application, the control time window, the prediction time window and the weighted coefficient of the insulin amount can also be selected as other values, which are not specifically limited here.

[0287] As mentioned earlier, due to the significant asymmetry of the distribution of high / low blood glucose (original physical space), the same degree of deviation of blood glucose from the normal range corresponds to significantly different high blood glucose risk and low blood glucose risk in clinical practice. In view of the asymmetric characteristics of the clinical risk of glucose concentration, the asymmetric blood glucose in the original physical space is converted to the approximately symmetric blood glucose risk in the risk space, so that the MPC algorithm is more accurate and flexible. The value function of the rMPC algorithm after the risk conversion is as follows:

[0288]

[0289] Wherein,

[0290] r t+j represents the blood glucose risk value after the jth step;

[0291] I′ t+j represents the change of insulin infusion amount after the jth step.

[0292] The deviation of the blood glucose value is converted into the corresponding blood glucose risk, and the specific conversion method is the same as that in the aforementioned rPID algorithm, such as segmented weighted processing and relative value processing; it also includes setting a fixed zero risk point in the risk space, and the blood glucose concentration of the zero risk point can be set as the target blood glucose value. The data deviating from both sides of the zero risk point is processed, such as using the BGRI and improved CVGA method; it also includes using different methods to process the data deviating from both sides of the target blood glucose value.

[0293] Specifically, when using segmented weighted processing:

[0294]

[0295] When using relative value processing:

[0296]

[0297] When using the classical blood glucose risk index method:

[0298]

[0299] wherein:

[0300] r(G t+j ) = 10*f(G t+j ) 2

[0301] The conversion function f(G t+j ) is as follows:

[0302] f(G t+j ) = 1.509*[(ln(G t+j )) 1.084 - 5.381]

[0303] When the control variability grid analysis method is used:

[0304]

[0305] The maximum value is also limited:

[0306] |r t+j | = min(|r t+j |, n)

[0307] wherein the value of the maximum value n is 0-80 mg / dL, preferably the value of n is 60 mg / d.

[0308] When the BGRI method is used when the blood glucose value is less than the target blood glucose value G B and the CVGA method is used when the blood glucose value is greater than the target blood glucose value G B :

[0309] r t+j = -r(G t+j ), if G t+j ≤ G B

[0310] wherein:

[0311] r(G t+j ) = 10*f(G t+j ) 2

[0312] The conversion function f(G t+j ) is as follows:

[0313] f(G t+j ) = 1.509*[(ln(G t+j )) 1.084 - 5.381]

[0314] r t+j = -4.8265*10 4 - 4*G t+j2 +0.45563*G t+j -44.855, if G t+j >G B

[0315] When the CVGA method is used when the blood glucose value is less than the target blood glucose value G B , and the BGRI method is used when the blood glucose value is greater than the target blood glucose value G B :

[0316] r t+j =r(G t+j ), if G t+j >G B

[0317] wherein:

[0318] r(G t+j )=10*f(G t+j ) 2

[0319] The conversion function f(G t+j ) is as follows:

[0320] f(G t+j )=1.509*[(ln(G t+j )) 1.084 -5.381]

[0321] r t+j =G t+j -G B , if G t+j ≤G B .

[0322] The maximum value can also be limited:

[0323] |r t+j | = min(|r t+j |, n)

[0324] wherein the value of the maximum value n is in the range of 0-80 mg / dL, and preferably, the value of n is 60 mg / dL.

[0325] When the BGRI method is used when the blood glucose value is less than the target blood glucose value G B , and the piecewise weighted method is used when the blood glucose value is greater than the target blood glucose value G B :

[0326] r t+j =-r(G t+j ), if G t+j ≤G B

[0327] Wherein:

[0328] r(G t+j ) = 10*f(G t+j ) 2

[0329] The conversion function f(G t+j ) is as follows:

[0330] f(G t+j ) = 1.509*[(ln(G t+j )) 1.084 -5.381]

[0331]

[0332] When the blood glucose value is less than the target blood glucose value G B , the BGRI method is adopted, and when the blood glucose value is greater than the target blood glucose value G B , the relative value conversion is adopted:

[0333] r t+j = -r(G t+j ), if G t+j ≤ G B

[0334] Wherein:

[0335] r(G t+j ) = 10*f(G t+j ) 2

[0336] The conversion function f(G t+j ) is as follows:

[0337] f(G t+j ) = 1.509*[(ln(G t+j )) 1.084 -5.381]

[0338]

[0339] When the data less than or equal to the target blood glucose value G B is processed by the segmented weighted processing or the relative value processing, and the data greater than the zero risk point blood glucose value is processed by the BGRI method, the processing result is equivalent to the aforementioned CVGA method when the blood glucose value is less than or equal to the target blood glucose value G B , and the BGRI method when the blood glucose value is greater than the target blood glucose value G B , and the calculation formula is not repeated.

[0340] It should be noted that in the above various conversion formulas:

[0341] r t+jGj is the blood glucose risk value at the jth step;

[0342] G t+j Gj is the blood glucose value detected at the jth step.

[0343] G is the target blood glucose value B G is 80-140 mg / dL, preferably, G is 110-120 mg / dL. B G is 110-120 mg / dL.

[0344] The beneficial effects after risk conversion and the relationship between blood glucose and blood glucose risk are consistent with those in the rPID algorithm, which are not repeated here.

[0345] Similarly, in order to compensate for the insulin absorption delay, an insulin feedback compensation mechanism can also be used for compensation; in order to compensate for the delay in the effect of insulin, IOB compensation can also be used; the sensing delay of interstitial fluid glucose concentration and blood glucose concentration can also be compensated by autoregressive compensation, and the specific compensation methods are consistent with those in the rPID algorithm, specifically:

[0346] For insulin absorption delay, the compensation formula is as follows:

[0347]

[0348] Wherein:

[0349] I t+j Ij represents the infusion instruction sent to the insulin infusion system at the jth step;

[0350] rI c(t+j) rIj represents the infusion instruction sent to the insulin infusion system at the jth step after risk conversion;

[0351] γ represents the compensation coefficient of the estimated plasma insulin concentration to the algorithm output, and the increase of the coefficient will lead to the relative conservatism of the algorithm, and the decrease of the coefficient will lead to the relative aggressiveness, therefore, in the embodiment of the present application, the range of γ is 0.4-0.6, preferably, γ is 0.5. Ij represents the estimation of plasma insulin concentration at the jth step.

[0352] For insulin effect delay, the compensation formula is as follows:

[0353] rI′ t+j rI′j = rIj t+j -IOB(t+j)

[0354] Wherein:

[0355] rI′ t+j rI′j represents the infusion instruction sent to the insulin infusion system at the jth step after risk conversion and deduction of IOB;

[0356] rI t+jThis indicates the infusion instruction sent to the insulin infusion system at step j after risk conversion;

[0357] IOB(t+j) represents the amount of insulin in the body that has not yet taken effect at time t+j.

[0358] Similarly, IOB(t+j) can be distinguished between being in a meal and not being in a meal, in which case:

[0359] IOB(t+j)=IOB m，t+j +IOB o，t+j

[0360] in:

[0361]

[0362] in:

[0363] IOB m，t+j This represents the amount of insulin consumed after a meal that has not yet taken effect in the body at time t+j.

[0364] IoB o，t+j This represents the amount of non-meal insulin that has not yet taken effect in the body at time t+j;

[0365] D i (i = 2 - 8) represent the corresponding coefficients of the IOB curves for insulin action time i;

[0366] I m，t+j This represents the insulin level at time t+j after the meal.

[0367] I 0，t+j This represents the non-meal insulin amount at time t+j;

[0368] IOB(t+j) represents the amount of insulin in the body at time t+j that has not yet taken effect.

[0369] When rI′ t+j When >0, the final infused insulin volume is rI′. t+j ;

[0370] When rI′ t+j When <0, the final amount of insulin infused is 0.

[0371] For the sensing delay of tissue fluid glucose concentration and blood glucose concentration, autoregressive compensation can also be used, as shown in the following formula:

[0372]

[0373] in,

[0374] G SC(t+j) represents the interstitial fluid glucose concentration at t+j, i.e. the measurement value of the sensing system;

[0375] G (t+j-1) represents the estimated concentration of blood glucose at t+j-1;

[0376] G (t+j-1) represents the estimated concentration of blood glucose at t+j-1; SC G (t+j-1) represents the estimated concentration of blood glucose at t+j-1; SC G (t+j-2) represents the interstitial fluid glucose concentration at t+j-1 and t+j-2, respectively;

[0377] K0 represents the coefficient of the estimated concentration part of blood glucose at t+j-1;

[0378] K1 and K2 represent the coefficients of the interstitial fluid glucose concentration at t+j-1 and t+j-2, respectively.

[0379] wherein, at the initial time,

[0380] The beneficial effects of various compensation methods are consistent with those in the rPID algorithm, which will not be repeated here.

[0381] It should be noted that in the rMPC algorithm, it is preferred to compensate for the delay in the effect of insulin and the sensing delay of the interstitial fluid glucose concentration and the blood glucose concentration.

[0382] In another embodiment of the present application, the program module 101 is pre-provided with a composite artificial pancreas algorithm, the composite artificial pancreas algorithm comprises a first algorithm and a second algorithm, when the detection module 100 detects the current blood glucose value and sends the current blood glucose value to the program module 101, the first algorithm calculates a first insulin infusion amount I1, the second algorithm calculates a second insulin infusion amount I2, the composite artificial pancreas algorithm optimizes and calculates the first insulin infusion amount I1 and the second insulin infusion amount I2 to obtain a final insulin infusion amount I3, and sends the final insulin infusion amount I3 to the infusion module 102, and the infusion module 102 infuses insulin according to the final infusion amount I3.

[0383] The first algorithm and the second algorithm are one of a classic PID algorithm, a classic MPC algorithm, an rMPC algorithm or an rPID algorithm. The rMPC algorithm or the rPID algorithm is an algorithm for converting blood glucose that is asymmetric in the original physical space to blood glucose risk that is approximately symmetric in the risk space. The conversion method of blood glucose risk in the rMPC algorithm and the rPID algorithm is as described above.

[0384] When I1=I2, I3=I1=I2;

[0385] When I1≠I2, the arithmetic mean of I1 and I2 can be substituted into the first algorithm and the second algorithm to re-optimize the algorithm parameters, and the insulin infusion amount required at the current time is calculated again by the first algorithm and the second algorithm after the parameter optimization, if I1 and I2 are still not the same, the arithmetic mean of I1 and I2 is taken again to repeat the above process until I1 and I2 are the same, that is:

[0386] ①Solving the average value of the first insulin infusion amount I1 and the second insulin infusion amount I2

[0387] ②The average value is substituted into the first algorithm and the second algorithm, respectively, to adjust the algorithm parameters;

[0388] ③The first insulin infusion amount I1 and the second insulin infusion amount I2 are recalculated based on the current blood glucose value, the first algorithm and the second algorithm after adjusting the parameters;

[0389] ④The steps ①-③ are calculated in a loop until I1=I2, and the final insulin infusion amount I3=I1=I2.

[0390] At this time, when the first algorithm or the second algorithm is a PID or rPID algorithm, the algorithm parameters are K P , and K D =T D / K P , T D can be taken as 60min-90min, K I =T I *K P , T I can be taken as 150min-450min. When the first algorithm or the second algorithm is a MPC or rPMC algorithm, the algorithm parameters are K.

[0391] When I1≠I2, I1 and I2 can also be weighted, and the calculated values after the weighted processing are substituted into the first algorithm and the second algorithm to re-optimize the algorithm parameters, and the insulin infusion amount required at the current time is calculated again by the first algorithm and the second algorithm after the parameter optimization, if I1 and I2 are still not the same, I1 and I2 are weighted again, the weighting coefficient is adjusted, and the above process is repeated until I1 and I2 are the same, that is:

[0392] ①Solving the weighted value of the first insulin infusion amount I1 and the second insulin infusion amount I2 Wherein α and β are the weighting coefficients of the first insulin infusion amount I1 and the second insulin infusion amount I2, respectively;

[0393] ②The weighted value The algorithm parameters are adjusted and brought into the rMPC algorithm and the rPID algorithm;

[0394] The first insulin infusion amount I1 and the second insulin infusion amount I2 are recalculated based on the current blood glucose value, the adjusted rMPC algorithm and the rPID algorithm;

[0395] The steps 1-3 are cyclically calculated until I1=I2, and the final insulin infusion amount I3=I1=I2.

[0396] Similarly, when the first algorithm or the second algorithm is a PID or rPID algorithm, the algorithm parameters are K P , and K D =T D / K P , T D may be 60min-90min, and K I =T I *K P , T I may be 150min-450min. When the first algorithm or the second algorithm is a MPC or rPMC algorithm, the algorithm parameters are K.

[0397] In the embodiment of the present application, the values of α and β can be adjusted according to the sizes of the first insulin infusion amount I1 and the second insulin infusion amount I2, when I1≥I2, α≤β; when I1≤I2, α≥β; preferably, α+β=1. In other embodiments of the present application, the values of α and β can also be in other ranges, which are not limited here.

[0398] When the calculation results of the two are the same, that is, I3=I1=I2, it can be considered that the insulin infusion amount at the current time can make the blood glucose value reach the ideal level. Through the processing of the above-mentioned method, each algorithm is mutually referenced, preferably, the rMPC algorithm and the rPID algorithm are mutually referenced, further improving the accuracy of the output result, making the result more feasible and reliable.

[0399] In another embodiment of the present application, the program module 101 further has a memory for storing information of the user's historical body state, blood glucose value and insulin infusion amount, etc. Statistical analysis can be performed based on the information in the memory to obtain a statistical analysis result I4. When I1≠I2, I1, I2 and I4 are compared respectively, and the final insulin infusion amount I3 is calculated. One of I1 and I2 which is closer to the statistical analysis result I4 is selected as the calculation result of the final composite artificial pancreas algorithm, that is, the final insulin infusion amount I3. The program module 101 sends the final insulin infusion amount I3 to the infusion module 102 for infusion; that is:

[0400]

[0401] The reliability of the insulin infusion amount is ensured by comparison with historical data.

[0402] In another embodiment of the present application, when both I1 and I2 are inconsistent and the difference is large, the blood glucose risk space conversion manner and / or the compensation manner for the delay effect in the rMPC algorithm and / or the rPID algorithm can also be adjusted to be similar, and then the output result of the compound artificial pancreas algorithm is finally determined by the above-mentioned arithmetic mean, weighted processing, or comparison with statistical analysis results.

[0403] In another embodiment of the present application, the closed-loop artificial pancreas control system further comprises a meal recognition module and a motion recognition module. The meal recognition commonly used to identify whether the user is having a meal or exercising can be based on the blood glucose change rate and judged by a specific threshold. The blood glucose change rate can be calculated from two time points before and after or obtained by linear regression of multiple time points in a period of time. Specifically, when the change rate of two time points before and after is used for calculation, the calculation formula is:

[0404] dG t / dt=(G t -G t-1 ) / Δt

[0405] wherein:

[0406] G t represents the blood glucose value at the current time point;

[0407] G t-1 represents the blood glucose value at the previous time point;

[0408] Δt represents the time interval between the current time point and the previous time point.

[0409] When the change rate calculation formula of three time points is used, the calculation formula is:

[0410] dG t / dt=(3G t -4G t-1 +G t-2 ) / 2Δt

[0411] wherein:

[0412] G t represents the blood glucose value at the current time point;

[0413] G t-1 represents the blood glucose value at the previous time point;

[0414] G t-2 represents the blood glucose value at the time point two hours before;

[0415] Δt represents the time interval between the current time and the last time.

[0416] Before calculating the blood glucose rate of change, the original continuous glucose data can also be filtered or smoothed. The threshold can be set to 1.8mg / mL-3mg / mL, or personalized.

[0417] Similar to meal recognition, exercise recognition can also be based on the blood glucose rate of change and determined by a specific threshold value, since exercise will cause a rapid decrease in blood glucose. The calculation of the blood glucose rate of change can also be as described above, and the threshold value can be personalized. In order to determine the occurrence of exercise more quickly, the closed-loop artificial pancreas insulin infusion control system also includes an exercise sensor (not shown). The exercise sensor is used to automatically detect the physical activity of the user, and the program module 101 can receive the physical activity information. The exercise sensor can automatically and accurately sense the physical activity state of the user and send the activity state parameter to the program module 101, thereby improving the output reliability of the composite artificial pancreas algorithm in the exercise scenario.

[0418] The exercise sensor can be provided in the detection module 100, the program module 101, or the infusion module 102. Preferably, in the embodiment of the present application, the exercise sensor is provided in the program module 101.

[0419] It should be noted that the number of exercise sensors and the positions of the multiple exercise sensors are not limited in the embodiment of the present application, as long as the exercise sensor can sense the activity state of the user.

[0420] The exercise sensor includes a three-axis acceleration sensor or a gyroscope. The three-axis acceleration sensor or the gyroscope can more accurately sense the activity intensity, activity mode, or body posture. Preferably, in the embodiment of the present application, the exercise sensor is a combination of a three-axis acceleration sensor and a gyroscope.

[0421] It should be noted that the blood glucose risk conversion methods used by the rMPC algorithm and the rPID algorithm in the calculation process can be the same or different, the compensation methods for the delay effect can be the same or different, and the calculation process can also be adjusted according to the actual situation.

[0422] In another embodiment of the present application, the program module 101 is provided with a hybrid artificial pancreas algorithm, and the hybrid artificial pancreas algorithm includes a cPID algorithm and / or a cMPC algorithm, wherein the input of the cPID algorithm is the intermediate value of the MPC algorithm, and the input of the cMPC algorithm is the output value of the PID algorithm.

[0423] Specifically, the cPID algorithm is calculated according to the blood glucose value at the current time predicted by the MPC prediction model, that is:

[0424]

[0425] wherein:

[0426] K P is a gain coefficient of the proportional part;

[0427] K I is a gain coefficient of the integral part;

[0428] K D is a gain coefficient of the derivative part;

[0429] G MPC(t) represents a blood glucose value at the current time predicted by the MPC prediction model;

[0430] G B represents a target blood glucose value;

[0431] C represents a constant;

[0432] cPID(t) represents an infusion instruction sent to the insulin infusion system.

[0433] Similarly, the cPID algorithm can also be converted in the risk conversion manner as described above, further improving the robustness of the hybrid artificial pancreas algorithm. That is:

[0434]

[0435] wherein:

[0436] K P is a gain coefficient of the proportional part;

[0437] K I is a gain coefficient of the integral part;

[0438] K D is a gain coefficient of the derivative part;

[0439] r MPC(t) represents a blood glucose risk after risk conversion based on the blood glucose value at the current time predicted by the MPC prediction model;

[0440] G B represents a target blood glucose value;

[0441] C represents a constant;

[0442] rcPID(t) represents an infusion instruction sent to the insulin infusion system.

[0443] The insulin infusion amount at the current time in the prediction model of the cMPC algorithm is calculated by the PID algorithm, that is, the prediction model of the cMPC algorithm is:

[0444] x t+1 = Ax t + BI PID(t)

[0445] G t = Cx t

[0446] wherein:

[0447] x t+1 denotes the state parameter at the next time instant,

[0448] x t denotes the state parameter at the current time instant,

[0449] I PID(t) denotes the insulin infusion at the current time instant calculated by the PID algorithm;

[0450] G t denotes the blood glucose concentration at the current time instant.

[0451] The parameter matrix is as follows:

[0452]

[0453]

[0454] C = [1 0 0]

[0455] b1, b2, b3, K are prior values.

[0456] Similarly, the insulin infusion at the current time instant in the prediction model of the cMPC algorithm can also be calculated by the rPID algorithm, and the specific blood glucose risk conversion mode is as described above. That is, the cMPC model is:

[0457] x t+1 = Ax t + BI rPID(t)

[0458] G t = Cx t

[0459] wherein:

[0460] x t+1 denotes the state parameter at the next time instant,

[0461] x t denotes the state parameter at the current time instant,

[0462] I rPID(t)This represents the insulin infusion volume at the current moment, calculated using the rPID algorithm.

[0463] G t This indicates the blood glucose concentration at the current moment.

[0464] The parameter matrix is ​​as follows:

[0465]

[0466]

[0467] C = [1 0 0]

[0468] b1, b2, b3, and K are prior values.

[0469] The value function of the cMPC algorithm can be composed of the sum of squares of the deviations in the output G (blood glucose level) and the sum of squares of the changes in the input I (insulin level). MPC aims to find the minimum solution of the value function.

[0470]

[0471] in:

[0472] I' t+j This indicates the change in insulin infusion volume after step j;

[0473] This represents the difference between the predicted blood glucose concentration and the target blood glucose value after step j;

[0474] t represents the current time;

[0475] N and P are the number of steps within the control time window and the prediction time window, respectively;

[0476] R is the weighting coefficient for the insulin component.

[0477] The insulin infusion volume in step j is I. t +I′ t+j .

[0478] Similarly, the output G (blood glucose level) in the value function of the cMPC algorithm can also undergo risk transformation. The transformed value function is as described above:

[0479]

[0480] in,

[0481] r t+j This represents the blood glucose risk value after step j;

[0482] I' t+j This indicates the change in insulin infusion volume after step j;

[0483] t represents the current time;

[0484] N, P are respectively the step number in the control time window and the prediction time window;

[0485] R is the weighted coefficient of the insulin component.

[0486] In the embodiment of the present application, the cMPC algorithm is a combination of the prediction model calculated by the PID algorithm or rPID at the current time and the value function with or without risk conversion. The advantages of the PID algorithm, MPC algorithm and blood glucose risk conversion are flexibly utilized to face complex situations, so that the artificial pancreas can provide reliable insulin infusion in various situations, so that the blood glucose reaches the ideal level at the expected time, and the precise control of the closed-loop artificial pancreas insulin infusion system is realized.

[0487] In the PID algorithm and MPC algorithm of each stage described above, the risk conversion method is as described above and will not be repeated here. The conversion methods can be the same or different. Similarly, the three delay effects can also be compensated according to the method described above.

[0488] In an embodiment of the present application, the hybrid artificial pancreas algorithm only includes cPID algorithm or cMPC algorithm.

[0489] In another embodiment of the present application, the hybrid artificial pancreas algorithm includes cPID algorithm and cMPC algorithm, one of which is used to calculate the required insulin of the user, and the other is used as a backup.

[0490] In another embodiment of the present application, the hybrid artificial pancreas algorithm includes cPID algorithm and cMPC algorithm, the cPID algorithm is used to calculate the first insulin infusion amount I1, and the cMPC algorithm is used to calculate the second insulin infusion amount I2. The hybrid artificial pancreas algorithm further optimizes the first insulin infusion amount I1 and the second insulin infusion amount I2 to obtain the final insulin infusion amount I3. The specific optimization method is as described above, that is:

[0491] When I1=I2, I3=I1=I2;

[0492] When I1≠I2, the arithmetic mean or the value after weighted processing of the two is substituted into the algorithm to recalculate the current insulin infusion amount I1 and I2. If the data is still not the same, the above process is repeated until I3=I1=I2, that is:

[0493] ①Solving the average value of the first insulin infusion amount I1 and the second insulin infusion amount I2

[0494] ②Substituting the average value Substitute these parameters into the cPID and cMPC algorithms and adjust the algorithm parameters.

[0495] ③ Based on the current blood glucose level, the cPID algorithm and cMPC algorithm with adjusted parameters recalculate the first insulin infusion volume I1 and the second insulin infusion volume I2;

[0496] ④ Repeat steps ① to ③ until I1 = I2, and finally the insulin infusion volume I3 = I1 = I2.

[0497] or:

[0498] ① Calculate the weighted value of the first insulin infusion volume I1 and the second insulin infusion volume I2. Where α and β are the weighting coefficients of the first insulin infusion volume I1 and the second insulin infusion volume I2, respectively;

[0499] ②Weighted values Substitute these parameters into the cPID and cMPC algorithms and adjust the algorithm parameters.

[0500] ③ Based on the current blood glucose level, the cPID algorithm and cMPC algorithm with adjusted parameters recalculate the first insulin infusion volume I1 and the second insulin infusion volume I2;

[0501] ④ Repeat steps ① to ③ until I1 = I2, and the final insulin infusion volume I3 = I1 = I2.

[0502] When the two differ, they can be statistically analyzed against historical information such as the user's past physical condition, blood glucose levels, and insulin infusion volume to obtain the current statistical analysis result I4. The one that is closer to the statistical analysis result I4 between I1 and I2 is selected as the final insulin infusion volume I3, i.e.:

[0503]

[0504] The beneficial effects of optimizing the first insulin infusion volume I1 and the second insulin infusion volume I2 described above are as previously stated and will not be repeated here.

[0505] In summary, this invention discloses a closed-loop artificial pancreas insulin infusion control system, which is pre-programmed with a composite artificial pancreas algorithm, including a cPID algorithm and a cMPC algorithm. The input of the cPID algorithm is the intermediate value of the MPC algorithm, and the input of the cMPC algorithm is the output value of the PID algorithm. By deeply combining the PID algorithm and the MPC algorithm, the advantages of the PID algorithm and the MPC algorithm are fully utilized to achieve precise control of the closed-loop artificial pancreas insulin infusion system.

[0506] While certain specific embodiments of the application have been described in detail herein for the purposes of exemplification, numerous other variations and modifications will be apparent to persons skilled in the art. Any and all such variations and modifications are within the scope of this application as defined in the appended claims.

Claims

1. A closed-loop artificial pancreas insulin infusion control system, characterized in that, include: The detection module is used to continuously detect the current blood glucose value G; A program module connected to the detection module has a preset hybrid artificial pancreas algorithm for calculating the amount of insulin to be infused by the user. The hybrid artificial pancreas algorithm includes a cPID algorithm and a cMPC algorithm. The input of the cPID algorithm is the intermediate value of the MPC algorithm, and the input of the cMPC algorithm is the output value of the PID algorithm. and An infusion module is connected to the program module. The program module sends the insulin infusion volume to the infusion module, and the infusion module infuses insulin according to the insulin infusion volume. Wherein, the blood glucose value in the value function of the cMPC algorithm undergoes risk transformation, and the value function after blood glucose risk transformation is: in, r t+j This represents the blood glucose risk value after step j; I' t+j This indicates the change in insulin infusion volume after step j; t represents the current time; N and P are the number of steps within the control time window and the prediction time window, respectively; R is the weighting coefficient for the insulin component; The blood glucose risk conversion method is an improved conversion method for controlled variability grid analysis. When using an improved method for controlling volatile grid analysis transformation: At the same time, its maximum value was also limited: |r t+j |=min(|r t+j |,n) The maximum value of n is limited to a range of 0–80 mg / dL; Among them G B Indicates the target blood glucose level; G t+j This represents the blood glucose value at step j.

2. The closed-loop artificial pancreas insulin infusion control system according to claim 1, characterized in that, The cPID algorithm is calculated based on the current blood glucose value predicted by the prediction model of the MPC algorithm, and the calculation formula is as follows: in: K P It is the gain coefficient of the proportional part; K I It is the gain coefficient of the integral part; K D It is the gain coefficient of the differential part; G B It is the target blood glucose level; G MPC(t) This represents the blood glucose level at the current moment, predicted by the MPC prediction model. C represents a constant; cPID(t) represents the infusion instruction sent to the insulin infusion system.

3. The closed-loop artificial pancreas insulin infusion control system according to claim 1, characterized in that, The cPID algorithm is calculated by performing risk transformation on the current blood glucose value predicted by the prediction model of the MPC algorithm, and the calculation formula is as follows: in: K P It is the gain coefficient of the proportional part; K I It is the gain coefficient of the integral part; K D It is the gain coefficient of the differential part; r MPC(t) This represents the risk of blood glucose after risk conversion based on the current blood glucose value predicted by the MPC prediction model; C represents a constant. rcPID(t) represents the infusion instruction sent to the insulin infusion system.

4. The closed-loop artificial pancreas insulin infusion control system according to claim 1, characterized in that, The current insulin infusion rate in the prediction model of the cMPC algorithm is calculated using the PID algorithm, and the prediction model is as follows: x t+1 =Ash t +BI PID(t) G t =Cx t in: x t+1 Indicates the state parameters at the next moment. x t This represents the state parameters at the current moment. I PID(t) This represents the insulin infusion volume at the current moment, calculated using the PID algorithm. G t This indicates the current blood glucose concentration; The parameter matrix is ​​as follows: C＝[1 0 0] b1,b2,b 3, K is the prior value.

5. The closed-loop artificial pancreas insulin infusion control system according to claim 1, characterized in that, The current insulin infusion rate in the prediction model of the cMPC algorithm is calculated using the rPID algorithm, and the prediction model is as follows: x t+1 =Ash t +BI rPID(t) G t =Cx t in: x t+1 Indicates the state parameters at the next moment. x t This represents the state parameters at the current moment. I rPID(t) This represents the insulin infusion volume at the current moment, calculated using the rPID algorithm. G t This indicates the current blood glucose concentration; The parameter matrix is ​​as follows: C＝[1 0 0] b1,b2,b 3, K is the prior value.

6. The closed-loop artificial pancreas insulin infusion control system according to claim 5, characterized in that, The rPID algorithm, based on the PID algorithm, transforms the asymmetric blood glucose in the original physical space into a blood glucose risk that is approximately symmetric in the risk space.

7. The closed-loop artificial pancreas insulin infusion control system according to claim 1, characterized in that, The blood glucose risk conversion method also includes one or more of the following processing methods: ① The component that is proportional to the estimated plasma insulin concentration is deducted from the predicted plasma insulin concentration; ② Subtract the amount of insulin that has not yet taken effect in the body; ③ An autoregressive method was used to compensate for the sensing delay of blood glucose and interstitial fluid glucose concentrations.

8. The closed-loop artificial pancreas insulin infusion control system according to claim 1, characterized in that, The hybrid artificial pancreas algorithm includes a cPID algorithm and a cMPC algorithm. The cPID algorithm calculates the first insulin infusion volume I1, and the cMPC algorithm calculates the second insulin infusion volume I2. The hybrid artificial pancreas algorithm optimizes the calculation of the first insulin infusion volume I1 and the second insulin infusion volume I2 to obtain the final insulin infusion volume I3.

9. The closed-loop artificial pancreas insulin infusion control system according to claim 8, characterized in that, The final insulin infusion volume I3 is optimized by averaging the first insulin infusion volume I1 and the second insulin infusion volume I2: ① Calculate the average value of the first insulin infusion volume I1 and the second insulin infusion volume I2. ② Average value The parameters are then incorporated into the cPID and cMPC algorithms and adjusted accordingly. ③ Based on the current blood glucose level, the cPID algorithm and the cMPC algorithm after parameter adjustment, recalculate the first insulin infusion volume I1 and the second insulin infusion volume I2; ④ Repeat steps ① to ③ until I1 = I2, and the final insulin infusion volume I3 = I1 = I2.

10. The closed-loop artificial pancreas insulin infusion control system according to claim 8, characterized in that, The final insulin infusion volume I3 is optimized using a weighted average of the first insulin infusion volume I1 and the second insulin infusion volume I2: ① Calculate the weighted value of the first insulin infusion volume I1 and the second insulin infusion volume I2. Where α and β are the weighting coefficients of the first insulin infusion volume I1 and the second insulin infusion volume I2, respectively; ②Weighted values The parameters are then incorporated into the cPID and cMPC algorithms and adjusted accordingly. ③ Based on the current blood glucose value, the cPID algorithm and the cMPC algorithm after parameter adjustment, recalculate the first insulin infusion volume I1 and the second insulin infusion volume I2; ④ Repeat steps ① to ③ until I1 = I2, and the final insulin infusion volume I3 = I1 = I2.

11. The closed-loop artificial pancreas insulin infusion control system according to claim 8, characterized in that, The final insulin infusion volume I3 is obtained by comparing the first insulin infusion volume I1 and the second insulin infusion volume I2 with the statistical analysis result I4 of historical data:

12. The closed-loop artificial pancreas insulin infusion control system according to claim 1, characterized in that, Two of the detection module, the program module, and the infusion module are interconnected to form an integrated structure, and are respectively attached to different locations on the skin along with the third module.

13. The closed-loop artificial pancreas insulin infusion control system according to claim 1, characterized in that, The detection module, the program module, and the infusion module are connected to form an integrated structure and are attached to the same location on the skin.

Images