Cloud big data-based system and method for insulin pump individualized configuration optimization

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) FIG. 1 is a schematic diagram showing a structure of the system the present invention.

(2) FIG. 2 is a flow chart of the method of the present invention.

DETAILED DESCRIPTION OF THE EMBODIMENTS

(3) The present invention is described in more detail below in combination with embodiments and drawings. The following embodiments are used only to illustrate the present invention, but are not intended to limit the scope of implementation of the present invention.

(4) A cloud big data-based system for insulin pump individualized configuration optimization, as shown in FIG. 1, includes an insulin pump, a real-time continuous glucose monitoring system (CGMS), a smart phone, a glucose monitoring application software installed in the smart phone and a cloud big data server.

(5) The insulin pump includes a syringe pump body with a control module and a wireless transmission module, a replaceable drug container and a subcutaneous indwelling needle. The wireless transmission module of the insulin pump is wirelessly connected to the smart phone via Bluetooth and transmits data with the glucose monitoring application software. During the connection and data transmission between the smart phone and the insulin pump, it can also be that the insulin pump only transmits insulin injection records and time information to the smart phone glucose monitoring application software and executes the high-dose and basal infusion rate commands issued by the glucose monitoring application software. The BOLUS and BASAL are calculated in the smart phone glucose monitoring application software according to the following algorithm, and the storage and cloud data synchronization and update of the relevant parameters IS, CR, TA and GR are also realized in the smart phone glucose monitoring application software, that is, the smart phone glucose monitoring application software replaces the control module of the insulin pump to process the data.

(6) The real-time continuous glucose monitoring system includes a replaceable implantable glucose sensor probe, a reusable signal collector and a signal transmitter. The signal transmitter of the real-time continuous glucose monitoring system is connected to the smart phone through wireless means such as Bluetooth communication and transmits data with the glucose monitoring application software.

(7) The smart phone and the glucose monitoring application software installed in the smart phone have the function of data transmission with the real-time continuous glucose monitoring system and the insulin pump through wireless transmission technology such as Bluetooth, and the function of data upload and download through smart phone data network or wireless network from the cloud big data server. The smart phone can also be other smart devices.

(8) The cloud big data server has the functions of storing, updating, calculating and transmitting user's personal information and historical data. The user's personal information and historical data stored in the cloud big data server include but are not limited to the user's name, gender, age, contact number, serial number of insulin pump products used, records of insulin pump infusion volume, infusion time and infusion rate, blood glucose output value BG and corresponding data measurement time Ts, carbohydrate intake, sleep and exercise recorded by the patient.

(9) The cloud big data server calculates the personalized parameters related to diabetes according to the stored user historical data, and automatically corrects and calculates the parameter output data of insulin pump and implantable glucose sensor and transfers it to the smart phone. The parameters include, but are not limited to, the amount CR of carbohydrate converted by 1 unit of insulin, insulin sensitivity index IS, and insulin retention time TA, glucose release rate GR via metabolism during fasting, injection volume BOLUS of single high-dose injection and basal infusion rate BASAL.

(10) The insulin pump can download the latest user parameters from the cloud big data server through the smart phone, then calculate and recommend the high-dose insulin injection scheme according to the carbohydrate intake input by the user, and recommend the updated basal infusion rate scheme to the user according to the time segmentation of the user's basal infusion rate.

(11) The cloud big data server can archive the individualized parameters related to diabetes and blood glucose historical data of the user, and the historical data of insulin infusion and form an analysis report for the doctor's reference, so as to make the continued treatment plan after the patient stops the insulin pump therapy.

(12) In the cloud big data server, the definition and calculation of the physiological parameters CR, IS, TA, GR and insulin infusion volume are as follows:

(13) CR: Carbohydrate Ratio, the amount of carbohydrate converted by one unit of insulin

(14) IS: insulin sensitivity, insulin sensitivity index

(15) TA: active insulin time, insulin retention time

(16) GR: Glucose Release Rate, the rate of glucose released into the blood by the body through metabolism during fasting

(17) BOLUS: injection volume of single large-dose injection

(18) BASAL: basal infusion rate, usually counted in insulin units per hour (U/h)

(19) According to the given CR, IS, TA, the BOLUS calculation formula is as follows:

(20) $\mathrm{BOLUS}=\frac{\mathrm{CARBS}}{\mathrm{CR}}+\frac{\mathrm{BGcurrent}-\mathrm{BGtarget}}{\mathrm{IS}}-\mathrm{BOLUSprev}\left(1-\frac{\min \left(\mathrm{TI},\mathrm{TA}\right)}{\mathrm{TA}}\right)$

(21) where, BGcurrent is the blood glucose value before the high-dose injection read by CGMS; BGtarget is the target blood glucose value; BOLUSprev is the injection volume of the previous high-dose injection; CARBS is the current carbohydrate intake input by the user; TI is the time between current high-dose injection and the midpoint of the previous high-dose injection process, min (TI, TA) is the smaller value of TI and TA, so that when TI is greater than or equal to TA, the residual amount of the previous high-dose injection

(22) $\mathrm{BOLUSprev}\left(1-\frac{\min \left(\mathrm{TI},\mathrm{TA}\right)}{\mathrm{TA}}\right)\mathrm{is}0;$

(23) According to the given GR, the BASAL during fasting for t is calculated as follows:

(24) $\mathrm{BASAL}=\frac{\frac{\mathrm{BGstart}-\mathrm{BGtarget}+\mathrm{GR}\times t}{\mathrm{IS}}-\mathrm{BOLUSprev}\left(1-\frac{\min \left(\mathrm{TI},\mathrm{TA}\right)}{\mathrm{TA}}\right)}{t}$

(25) Where BGstart is the average blood glucose in a period after the fasting starts read by CGMS, and TI is the time between current high-dose injection and the midpoint of the previous high-dose injection process.

(26) The cloud big data server can optimize the physiological parameters CR, IS, and TA by collecting real-time data obtained by users using CGMS and insulin pump for high-dose injection. The specific steps are as follows:

(27) Step A, establishing a regression equation

(28) $\mathrm{BGbefore}-\mathrm{BGafter}=\left(\mathrm{BOLUS}+\mathrm{BOLUSprev}\right)\mathrm{IS}+\left(-\mathrm{CARBS}\right)\frac{\mathrm{IS}}{\mathrm{CR}}+\left(-\mathrm{BOLUSprev}\times \min \left(\mathrm{TI},\mathrm{TA}\right)\right)\frac{\mathrm{IS}}{\mathrm{TA}}$

(29) where, BGbefore is the blood glucose value before the high-dose injection, which is the equal to the BGcurrent in the calculation formula; BGafter is the measured blood glucose value after a period of high-dose injection, such as after 2 hours of the high-dose injection;

(30) Step B, obtaining the following data near each high-dose injection Tstart from the insulin pump and CGMS through the smart phone:

(31) Injection start time Tstart: insulin pump data

(32) Injection end time Tend: insulin pump data

(33) Injection volume BOLUS of high-dose injection at Tstart: insulin pump data

(34) Blood glucose value BGbefore measured by implantable dynamic glucose sensor at Tstart

(35) Blood glucose value BGafter measured by implantable dynamic glucose sensor a period after Tend (for example, 2 hours after Tend)

(36) Carbohydrate intake CARBS input by users near Tstart

(37) Forming a sample record packet for calculation
[Tstart.sub.nTend.sub.nBOLUS.sub.nCARBS.sub.nBGbefore.sub.nBGafter.sub.n]

(38) The data in the last three to six months are used for regression. The subscript number n of historical data variable is arranged in reverse order of Tstart, that is, the closer to the current historical data, the smaller the subscript number;

(39) Step C, constructing the sample matrix:

(40) $G=\left[\begin{array}{c}\Delta {\mathrm{BG}}_1\\ \Delta {\mathrm{BG}}_2\\ \Delta {\mathrm{BG}}_3\\ .Math.\\ \Delta {\mathrm{BG}}_n\end{array}\right]X=\left[\begin{array}{ccc}{\mathrm{BOLUS}}_1^{\prime }& -{\mathrm{CARBS}}_1& -{\mathrm{BOLUS}}_2\times {\mathrm{TI}}_1^{\prime }\\ {\mathrm{BOLUS}}_2^{\prime }& -{\mathrm{CARBS}}_2& -{\mathrm{BOLUS}}_3\times {\mathrm{TI}}_2^{\prime }\\ {\mathrm{BOLUS}}_3^{\prime }& -{\mathrm{CARBS}}_3& -{\mathrm{BOLUS}}_4\times {\mathrm{TI}}_3^{\prime }\\ .Math.& .Math.& .Math.\\ {\mathrm{BOLUS}}_n^{\prime }& -{\mathrm{CARBS}}_n& -{\mathrm{BOLUS}}_{n+1}\times {\mathrm{TI}}_n^{\prime }\end{array}\right]$

(41) where,
ΔBG.sub.n=BGafter.sub.n−BGbefore.sub.n

(42) When TI.sub.n<TAl, BOLUS′.sub.n=BOLUS.sub.n+BOLUS.sub.n+1, TI′.sub.n=TI.sub.n,
TI.sub.n=Tstart.sub.n−(Tstart.sub.n+1+Tend.sub.n+1)/2;

(43) When TI.sub.n>TAu, BOLUS′.sub.n=BOLUS.sub.n, TI′.sub.n=0;

(44) When TAl≤TI.sub.n≤TAu, the sample was abandoned.

(45) TAu is the upper limit allowed by TA, and TAl is the lower limit allowed by TA.

(46) Step D, if for each n, X.sub.n,3=0, then:

(47) $X=\left[\begin{array}{c}{\mathrm{BOLUS}}_1^{\prime }\\ {\mathrm{BOLUS}}_2^{\prime }\\ {\mathrm{BOLUS}}_3^{\prime }\\ .Math.\\ {\mathrm{BOLUS}}_n^{\prime }\end{array}\begin{array}{c}-{\mathrm{CARBS}}_1\\ -{\mathrm{CARBS}}_2\\ -{\mathrm{CARBS}}_3\\ .Math.\\ -{\mathrm{CARBS}}_n\end{array}\right]C=\left[\begin{array}{c}\mathrm{IS}\\ \mathrm{IS}/\mathrm{CR}\end{array}\right]$

(48) Otherwise, the sample matrix remains unchanged;

(49) Step E, solving the overdetermined equation G=XC,

(50) The weighted least square method is used to solve: Ĉ=(X.sup.TWX).sup.−1X.sup.TWG;

(51) Step F, eliminating the abnormal data: calculate the residual error: {circumflex over (ε)}=G−XĈ, eliminate the data items whose residual error is greater than the threshold, and then repeat the regression algorithm A-F until there are no data items whose residual error is greater than the threshold.

(52) Step G, calculating the updated physiological parameters IS, CR, and TA according to the results of the regression algorithm:
=Ĉ.sub.1.1
=Ĉ.sub.1.1/Ĉ.sub.2.1

(53) If Ĉ.sub.3.1 exists, then =Ĉ.sub.1.1/Ĉ.sub.3.1, otherwise =TA;

(54) Step H, finally, using the obtained , and to correct the currently set IS, CR and TA with a certain correction ratio γ, wherein the range of γ values is 0<γ<1,
IS:=(1−γ)×IS+γ×
CR:=(1−γ)×CR+γ×
TA:=(1−γ)×TA+γ×

(55) the above is used as a setting parameter for high-dose injection of insulin pump from now on;

(56) TAl and TAu are revised at the same time:
TAl:=TA×τ%, wherein 0<τ<100;
TAu:=TA×υ%, wherein 100<υ<150;

(57) Storing and updating physiological parameters IS, CR, TA, TAl and TAu to the cloud big data server and pushing them to mobile application and insulin pump.

(58) The cloud big data server can optimize the value of physiological parameter GR in different time periods and the corresponding basal infusion rate BASAL by collecting real-time data obtained by users using CGMS and insulin pump in real time. The specific steps are as follows:

(59) Step A, first, segmenting the 24 hours a day according to the basal infusion rate established by the user with reference to the doctor's recommendations and their own situation. The GR and BASAL values in each time period need to be set and calculated independently. For each time period and 2 hours before the each time period, if the user eats something, and a high-dose injection accompanied with or without food is performed, the data obtained at 2 hours after the meal or the high-dose injection needs to be excluded from this time period, the data of the time period is updated to only include the data of the longer continuous time remaining after the removal of the data obtained at 2 hours after the meal/high-dose injection. For example, if the time period is set to: (1) 6:00-13:00, (2) 13:00-20:00, (3) 20:00—the next day 6:00, and the user eats and injects insulin at 7:00, 12:00, 18:00, then the time period (1) is adjusted to 9:00-12:00, the time period (2) is adjusted to 14:00-18:00, and the time period (3) remains unchanged.

(60) Step B, collecting sample data in each valid time period:

(61) Tstart: the start time of the time period

(62) BGstart: the average value of blood glucose in the previous short period (for example, 15 min) of the time period

(63) BGend: the average value of blood glucose in the last short period (for example, 15 min) of the time period

(64) BASAL: basal infusion rate during this time period

(65) t: duration of this time period

(66) IS: insulin sensitivity index

(67) TA: insulin retention time

(68) 0$\mathrm{RESIDUAL}=\mathrm{BOLUSprev}\left(1-\frac{\min \left(\mathrm{TI},\mathrm{TA}\right)}{\mathrm{TA}}\right):$
residual insulin in the body at the beginning of the time period

(69) SNR: signal-to-noise ratio of dynamic blood glucose data

(70) Forming a sample record package for calculation
[Tstart.sub.nBGstart.sub.nBGend.sub.nBASAL.sub.nt.sub.nRESIDUAL.sub.nSNR.sub.n] and system parameters IS and TA;

(71) The data of the same period in the last three to six months are used for regression, and the subscript number n of historical data variables is arranged in reverse order of Tstart, that is, the closer to the current historical data, the smaller the subscript number.

(72) Step C, for each effective time period, considering the effectiveness of ingesting insulin, during which the body releases total glucose ΔBG into the blood through metabolism:
ΔBG.sub.n=BGend.sub.n−BGstart.sub.n+[BASAL.sub.n×t.sub.n+RESIDUAL.sub.n]×IS

(73) establishing regression equation Δ BG=GR×t;

(74) Step D, for each valid time period, using the regression method to calculate the updated value of GR

(75) $=\frac{{.Math.}_{n=1}^N{t}_n\times \Delta {\mathrm{BG}}_n\times {\mathrm{SNR}}_n\times w\left(T_n^{\prime }\right)}{{.Math.}_{n=1}^N{t}_n^2\times {\mathrm{SNR}}_n\times w\left(T_n^{\prime }\right)}$

(76) where, w(T′) is the time-related weight, T′.sub.n=Tcurrent−Tstart.sub.1, Tcurrent.sub.n is the current time, that is, the time of the latest historical data Tstart.sub.1;

(77) The closer to the nearest sample, the greater the weight; for example,

(78) $w=\{\begin{array}{c}0.6T^{\prime }\le 1\mathrm{month}\\ 0.31\mathrm{month}<T^{\prime }\le 2\mathrm{months}\\ 0.12\mathrm{months}<T^{\prime }\le 3\mathrm{months}\end{array}$

(79) Step E, using the obtained to correct the currently set GR with a certain correction ratio γ
GR:=(1−γ)×GR+γ×

(80) The range of γ values is 0<γ<1;

(81) Step F, using the modified GR and the historical sample packet [BGstart.sub.n t.sub.n RESIDUAL.sub.n] of the time period to calculate value that should be set in this time period after correction according to the formula:

(82) ${{\prime }^{}}_n=\frac{\frac{{\mathrm{BGstart}}_n-\mathrm{BGtarget}+\mathrm{GR}\times t_n}{t_n}-{\mathrm{RESIDUAL}}_n}{t_n}$

(83) Step G, weighting all the calculated ′ in time to calculate the current BASAL correction value :

(84) $=\frac{{.Math.}_{n=1}^N{n}_{\prime }^{}\times w^{\prime }\left(T_n^{\prime }\right)}{{.Math.}_{n=1}^N{w}^{\prime }\left(T_n^{\prime }\right)}$

(85) where, w(T′) is the time-related weight, and the closer the sample is to the current time, the greater the weight is;

(86) $\mathrm{For}\mathrm{example}:w^{\prime }=\{\begin{array}{c}0.6T^{\prime }\le 1\mathrm{month}\\ 0.31\mathrm{month}<T^{\prime }\le 2\mathrm{months}\\ 0.12\mathrm{months}<T^{\prime }\le 3\mathrm{months}\end{array}$

(87) Step H, if the difference between the calculated value and the current BASAL value exceeds the threshold, using the obtained to correct the currently set BASAL with a certain correction ratio γ:
BASAL:=(1−γ)×BASAL+γ×

(88) The range of γ values is 0<γ<1;

(89) Storing and updating BASAL as the setting parameter of the basal injection rate of the insulin pump, and storing it to the cloud big data server together with the physiological parameter GR and transfer it to the mobile application and the insulin pump.

(90) The cloud big data server can optimize the basal infusion rate BASAL in different time periods by collecting real-time data obtained by users using CGMS and insulin pump in real time. Other specific steps are as follows:

(91) Step A, first, segmenting the 24 hours a day according to the basal infusion rate established by the user with reference to the doctor's recommendations and his/her own situation. The BASAL value in each time period needs to be set and calculated independently. For each time period and 2 hours before the each time period, if the user eats something, and a high-dose injection accompanied with or without food is performed, the data obtained at 2 hours after the meal or high-dose injection needs to be excluded from this time period, the data in this time period is updated to include only the data of the long continuous time remaining after the removal of the data obtained at 2 hours after the eating/high-dose injection; Step B, collecting sample data in each valid time period:

(92) Tstart: the start time of the time period

(93) BGstart: the average value of blood glucose in the previous short period of the time period

(94) BGend: the average value of blood glucose in the last short period of the time period

(95) BASAL: basal infusion rate during this time period

(96) t: duration of this time period

(97) IS: insulin sensitivity index

(98) TA: insulin retention time

(99) $\mathrm{RESIDUAL}=\mathrm{BOLUSprev}\left(1-\frac{\min \left(\mathrm{TI},\mathrm{TA}\right)}{\mathrm{TA}}\right):$
residual insulin in the body at the beginning of the time period

(100) Forming a sample record package for calculation
[Tstat.sub.nBGstat.sub.nBGend.sub.nBASAL.sub.nt.sub.nRESIDUAL.sub.n] and system parameters IS and TA;

(101) The data of the same period in the last three to six months are used for regression, and the subscript number n of historical data variables is arranged in reverse order of Tstart, that is, the closer to the current historical data, the smaller the subscript number.

(102) Step C, for the n.sup.th time period, using the historical sample packet of the n.sup.th time period to calculate value that should be set in this time period after correction according to the formula:

(103) $n_{\prime }^{}={\mathrm{BASAL}}_n+\frac{{\mathrm{RESIDUAL}}_n}{t_n}-\frac{\mathrm{BGtarget}-{\mathrm{BGend}}_n}{\mathrm{IS}\times t_n}$

(104) Step D, weighting all the calculated , in time to calculate the current BASAL correction value BA:

(105) $=\frac{{.Math.}_{n=1}^N{n}_{\prime }^{}\times w^{\prime }\left(T_n^{\prime }\right)}{{.Math.}_{n=1}^N{w}^{\prime }\left(T_n^{\prime }\right)}$

(106) where, w′(T′) is the time-related weight, and the closer the sample is to the current time, the greater the weight is;

(107) Step E, if the difference between the calculated value and the current BASAL value exceeds the threshold, using the obtained to correct the currently set BASAL with a certain correction ratio γ:
BASAL:=(1−γ)×BASAL+γ×

(108) The range of γ values is 0<γ<1.

(109) A cloud big data-based method for insulin pump individualized configuration optimization, as shown in FIG. 2, includes the following steps:

(110) Step 1, when the system is started, the smart phone application obtains the cloud big data server data to determine whether the user is using the insulin injection system for the first time. If yes, the user is prompted to set the parameters IS, CR, TA, GR, time segmentation and the basal injection rate or to continue to use the default setting. If no, the updated above parameters are downloaded from the cloud big data server.

(111) Step 2, if the user inputs the high-dose injection command manually, the insulin pump enters the high-dose mode. Otherwise, the insulin pump maintains in the basic injection mode.

(112) Step 3, in the basic injection mode, insulin injection is performed according to the preset basic rate of the current time period, and the blood glucose data monitored by CGMS are uploaded to the cloud server on a regular basis. After the end of this time period or after the operation of the insulin pump, checking whether there is an update of the GR and basic rate of the cloud big data server. If yes, updating the local storage parameters, and then repeating step 2. If no, repeating step 2 directly.

(113) Step 4, in the high-dose mode, the insulin pump prompts the user to manually enter the carbohydrate intake CARBS and confirm the target blood glucose value to be achieved through the smart phone application, while obtaining the current blood glucose value BGcurrent measured by CGMS.

(114) Step 5, the required high-dose injection volume is calculated using the previously set or obtained parameter values:
BOLUS=CARBS/CR+(BGcurrent−BGtarget)/IS−BOLUSprev[1−min(TI,TA)/TA];

(115) Step 6, the patient is prompted to confirm the infusion volume and high-dose infusion time, and the injection stop time is calculated as Tend=Tstart+T.sub.BOLUS, where =infusion volume/bolus-rate. Bolus-rate is the user-defined high-dose insulin infusion rate.

(116) Step 7, the insulin injection information Tstart, Tend, BOLUS, CARBS and CGMS blood glucose monitoring data are uploaded to the cloud big data server.

(117) Step 8, the high-dose injection is performed until the Tend is reached.

(118) Step 9, detecting whether there are physiological parameters updated in the cloud. If yes, updating the local storage parameters, and then repeating step 2. If no, repeating step 2 directly.