CN102831326A

CN102831326A - Mean amplitude of glucose excursions (MAGE) calculation method

Info

Publication number: CN102831326A
Application number: CN2012103278556A
Authority: CN
Inventors: 余学飞; 李彬; 简峻; 彭达明; 沈洁
Original assignee: Southern Medical University
Current assignee: Southern Medical University
Priority date: 2012-09-06
Filing date: 2012-09-06
Publication date: 2012-12-19

Abstract

The invention relates to a method for calculating the average blood sugar fluctuation range. The method is composed of the following steps: reading in the dynamic blood sugar monitoring data obtained from a dynamic blood sugar monitoring instrument, and calculating the blood sugar standard deviation of the obtained data; finding out the obtained data All the extreme points in , get a set of extreme point sequences, and then find out that the absolute value of the blood glucose difference corresponding to two adjacent extreme points in the set is not less than the standard deviation of blood glucose and the adjacent two extreme points are extreme The effective extremum point of the maximum value and the minimum value, and then calculate the average blood sugar fluctuation range according to the calculation formula of the average blood sugar fluctuation range. The method can quickly and accurately analyze the dynamic blood glucose monitoring data, obtain effective parameters, greatly shorten the calculation time while ensuring the accuracy of the average blood glucose fluctuation range calculation, and improve the efficiency of clinical and scientific research work.

Description

A Calculation Method of Average Blood Sugar Fluctuation Range

技术领域 technical field

本发明涉及体内血液特性的测量，具体涉及人体内平均血糖波动幅度参数的处理方法。The invention relates to the measurement of blood characteristics in the body, in particular to a processing method for parameters of the average blood sugar fluctuation range in the human body.

背景技术 Background technique

平均血糖波动幅度（Mean Amplitude of Glycemic Excursions，MAGE）是一种基于动态血糖监测数据建立的反应血糖波动的参考数值。目前的实际临床应用中，医务工作者让患者一次性佩戴监测仪器超过72小时来获取超过864个动态血糖监测数据，而MAGE数值往往不能直接从动态血糖监测系统中直接分析获得，仍必须通过人工筛选方式筛选动态血糖监测数据辅以分析系统进行计算，工作量较大。若获取的动态血糖监测数据增加时，计算工作量将大大增加，大大降低了临床诊断和科研的效率和参数的实效性。此外，为保证MAGE数值的准确性，进行筛选分析的医务工作者需要进行长时间的专业培训，若医务工作者的经验不足导致计算结果与真实值之间存在较大误差，也会降低了MAGE数值的准确性和有效性。Mean Amplitude of Glycemic Excursions (MAGE) is a reference value based on continuous blood glucose monitoring data to reflect blood glucose fluctuations. In the current actual clinical application, medical workers ask patients to wear the monitoring instrument for more than 72 hours at a time to obtain more than 864 continuous blood glucose monitoring data, and the MAGE value often cannot be directly analyzed and obtained from the continuous blood glucose monitoring system, and must still be obtained manually. Screening method Screening continuous blood glucose monitoring data supplemented by an analysis system for calculation, the workload is relatively large. If the acquired continuous blood glucose monitoring data increases, the calculation workload will be greatly increased, greatly reducing the efficiency of clinical diagnosis and scientific research and the effectiveness of parameters. In addition, in order to ensure the accuracy of MAGE values, medical workers who conduct screening analysis need to undergo long-term professional training. If the medical workers have insufficient experience, there will be a large error between the calculation results and the real value, which will also reduce the MAGE value. Accuracy and Validity of Values.

为了解决MAGE数值自动计算的问题，已有一些自动计算MAGE数值方法的报道，如，PeterA.Baghurst,”Calculating the Mean Amplitude of Glycemic Excursion from ContinuousGlucose Monitoring Data-An Automated Algorithm”,DiabetesTechnology&Therapeutics,vol.13,pp.296-302,2011和Gert Fritzsche,Klaus-DieterKohnert,Peter Heinke,Lutz Vogt,and Eckhard Salzsieder."The use of a computerprogram to calculate the Mean Amplitude of Glycemic Excursions",DiabetesTechnology&Therapeutics,vol.13,pp.319-325,2011。上述文献中，前者提供了一种基于自动识别动态血糖数据曲线中峰值和谷值点的计算方法，该方法由于对动态血糖数据曲线进行了平滑处理，容易导致数据失真，计算结果的准确性会受到影响，此外，该方法采用一阶差分方式识别曲线拐点，然而，动态血糖数据曲线中存在许多血糖值连续相等的数据，因此受到这些数据的影响，一阶差分方式识别曲线拐点也会导致计算误差；后者主要介绍了一种计算MAGE数值的计算分析程序，该文章的研究重点更多集中在计算的可行性和有效性评估，而对计算方法介绍较为简略。另外，这些报道中对于动态血糖监测数据中的有效血糖波动的判断仍然过分依赖肉眼观察和经验判断，因此仍然存在主观性的偏差。In order to solve the problem of automatic calculation of MAGE values, there have been some reports on automatic calculation of MAGE values, such as, PeterA. Baghurst, "Calculating the Mean Amplitude of Glycemic Excursion from Continuous Glucose Monitoring Data-An Automated Algorithm", Diabetes Technology & Therapeutics, vol.13, pp.296-302, 2011 and Gert Fritzsche, Klaus-Dieter Kohnert, Peter Heinke, Lutz Vogt, and Eckhard Salzsieder. "The use of a computer program to calculate the Mean Amplitude of Glycemic Excursions", Diabetes Technology, pp. 193 -325, 2011. In the above literature, the former provides a calculation method based on automatic identification of peak and valley points in the dynamic blood glucose data curve. This method is easy to cause data distortion due to the smoothing of the dynamic blood glucose data curve, and the accuracy of the calculation results will be affected. Affected, in addition, this method uses the first-order difference method to identify the inflection point of the curve. However, there are many data with continuous equal blood glucose values in the dynamic blood glucose data curve, so affected by these data, the first-order difference method to identify the inflection point of the curve will also lead to calculation Error; the latter mainly introduces a calculation and analysis program for calculating the MAGE value. The research focus of this article is more on the calculation feasibility and effectiveness evaluation, and the introduction to the calculation method is relatively brief. In addition, in these reports, the judgment of effective blood glucose fluctuations in continuous blood glucose monitoring data still relies too much on visual observation and empirical judgment, so there are still subjective biases.

发明内容 Contents of the invention

鉴于现有技术存在上述不足，本发明所要解决的技术问题是提供一种平均血糖波动幅度的计算方法，该方法可自动识别动态血糖监测数据中的有效血糖波动，快捷、准确地计算出平均血糖波动幅度。In view of the above deficiencies in the prior art, the technical problem to be solved by the present invention is to provide a method for calculating the average blood sugar fluctuation range, which can automatically identify the effective blood sugar fluctuations in the dynamic blood sugar monitoring data, and quickly and accurately calculate the average blood sugar volatility.

本发明解决上述问题的技术方案如下：The technical scheme that the present invention solves the above problems is as follows:

1.一种平均血糖波动幅度的计算方法，该方法包括以下步骤：1. A calculation method for the average blood sugar fluctuation range, the method comprising the following steps:

1)读入从动态血糖监测仪器中获取的动态血糖监测数据，计算所获取的数据的血糖标准差；计算所获数据中所有的极值点，得到一极值点序列的集合A；1) Read in the continuous blood glucose monitoring data obtained from the continuous blood glucose monitoring instrument, calculate the blood glucose standard deviation of the obtained data; calculate all the extreme points in the obtained data, and obtain a set A of extreme point sequences;

2）利用差分进化算法求解下式（1）所示的目标函数，统计集合A中符合以下两个条件的极值点，得到集合B:2) Use the differential evolution algorithm to solve the objective function shown in the following formula (1), count the extreme points in the set A that meet the following two conditions, and obtain the set B:

a.相邻的两个极值点对应的血糖值差的绝对值≥血糖标准差；a. The absolute value of the blood glucose value difference corresponding to two adjacent extreme points ≥ blood glucose standard deviation;

b.相邻的两个极值点一个是极大值，另一个是极小值；b. One of the two adjacent extreme points is a maximum value and the other is a minimum value;

$\underset{K K,, {n no}_{11},, {n no}_{22},, . . . . . .,, {n no}_{K K}}{arg arg max max} {Z Z}_{K K} (({n no}_{11},, {n no}_{22},, . . . . . .,, {n no}_{K K})) = = \underset{K K,, {n no}_{11},, {n no}_{22},, . . . . . .,, {n no}_{K K}}{arg arg max max} {Σ Σ}_{k k = = 11}^{K K - - 11} | | f f (({t t}_{{n no}_{k k + + 11}})) - - f f (({t t}_{{n no}_{k k}})) | | - - - - - - ((11))$

式（1）中，元素n₁,n₂,...,n_k均为集合A中的元素,Z_K(n₁,n₂,...,n_k)为由元素n₁,n₂,...,n_k所组成的子集{n₁,n₂,...,n_k}对应的有效血糖波动的波动幅度之和，常数K为满足上述条件的极值点的个数，

为子集{n₁,n₂,...,n_k}中n_k对应的血糖值；In formula (1), elements n ₁ , n ₂ ,...,n _k are all elements in set A, and Z _K (n ₁ ,n ₂ ,...,n _k ) is composed of elements n ₁ ,n ₂ ,...,n _k The subset {n ₁ ,n ₂ ,...,n _k } corresponding to the sum of the effective blood sugar fluctuations, the constant K is the number of extreme points that meet the above conditions number,

is the blood glucose value corresponding to n _k in the subset {n ₁ ,n ₂ ,...,n _k };

3）先计算集合B中各相邻极值点对应的血糖值差的绝对值，再计算所有绝对值的平均值，即得平均血糖波动幅度。3) First calculate the absolute value of the blood glucose value difference corresponding to each adjacent extreme point in set B, and then calculate the average value of all absolute values to obtain the average blood glucose fluctuation range.

为进一步保证最后计算的准确性，本发明在步骤1）中还包括数据预处理操作，具体过程为：In order to further ensure the accuracy of the final calculation, the present invention also includes a data preprocessing operation in step 1), the specific process is:

读入从动态血糖监测仪器中获取的动态血糖监测数据后，先删除血糖值相等的相邻两点中的后一血糖值数据，再计算所述的血糖标准差和所有的极值点。After reading in the continuous blood glucose monitoring data obtained from the continuous blood glucose monitoring instrument, first delete the next blood glucose value data among two adjacent points with equal blood glucose values, and then calculate the blood glucose standard deviation and all extreme points.

由于本发明采用差分进化算法求解无约束条件非线性规划问题目标函数，保证了有效血糖波动幅度的准确识别，避免了在计算过程中引入主观性因素影响，在保证计算准确性和一致性的同时，利用计算机辅助计算，大大提高了计算效率，具有较高的实效性。Since the present invention adopts differential evolution algorithm to solve the objective function of non-constrained nonlinear programming problem, it ensures the accurate identification of effective blood sugar fluctuation range, avoids the influence of subjectivity factors in the calculation process, and ensures the accuracy and consistency of calculation. , the use of computer-aided calculation greatly improves the calculation efficiency and has high effectiveness.

附图说明 Description of drawings

图1为本发明所述一种平均血糖波动幅度的计算方法的一个具体实施例的流程图。FIG. 1 is a flow chart of a specific embodiment of a method for calculating the average blood sugar fluctuation range of the present invention.

图2为一例就医者24小时动态血糖变化曲线图。Figure 2 is a 24-hour ambulatory blood glucose change curve of a patient seeking medical treatment.

图3为本发明所述一种平均血糖波动幅度的计算方法中滤除动态血糖监测数据曲线中的血糖值连续相等的数据段的示意图。Fig. 3 is a schematic diagram of filtering out data segments with consecutive equal blood glucose values in a dynamic blood glucose monitoring data curve in a method for calculating the average blood glucose fluctuation range according to the present invention.

图4为本发明所述一种平均血糖波动幅度的计算方法中求得所有有效极值点的后的数据曲线。Fig. 4 is a data curve after obtaining all effective extreme points in a method for calculating the average blood sugar fluctuation range according to the present invention.

图5为本发明所述一种平均血糖波动幅度的计算方法中所述的计算方法与传统人工计算方法对于不同数据分组的相关性分析结果。FIG. 5 is a correlation analysis result of different data groupings between the calculation method described in the present invention and the traditional manual calculation method.

图6为本发明所述一种平均血糖波动幅度的计算方法中所述的计算方法与传统人工计算方法对于不同组别数据的Bland-Altman评估图。Fig. 6 is a Bland-Altman evaluation chart of different groups of data between the calculation method described in the present invention and the traditional manual calculation method.

具体实施方式 Detailed ways

下面结合附图和实施例对本发明做进一步的详细描述。The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments.

例1example 1

参见图1，以下以一例就医者24小时动态血糖变化曲线图为例详细描述本发明所述平均血糖波动幅度的计算方法：Referring to Fig. 1, the calculation method of the average blood sugar fluctuation range of the present invention will be described in detail by taking a 24-hour dynamic blood sugar change curve of a medical patient as an example:

1）从就医者所用的从动态血糖监测仪器中导出与图2相对应监测数据进行分析可知，血糖值为时间变量t的函数,可表示为f(t)，且其中变量t为等时间间隔集合{t₁,t₂,...,t_L}。如图2可见，曲线中存在许多如图中箭头所指的血糖值连续相等的数据段，即f(t_x)＝f(t_x+1)，其中x表示记录血糖值的序号。因此，为保证平均血糖波动幅度计算的准确性，需先去除血糖值相等的相邻两点中的后一血糖值数据，去除过程如图3所示。然后再计算处理后的数据的血糖标准差；利用离散序列极值点算法对处理后的数据进行求离散序列极值点操作，从而获得该数据的极值点序列的集合{1,2,...,N}；1) From the analysis of monitoring data corresponding to Figure 2 derived from the continuous blood glucose monitoring instrument used by the doctor, it can be known that the blood glucose value is a function of the time variable t, which can be expressed as f(t), and the variable t is an equal time interval Set {t ₁ ,t ₂ ,...,t _L }. It can be seen from Fig. 2 that there are many data segments in which the blood glucose values are continuously equal as indicated by the arrows in the graph, that is, f(t _x )=f(t _x+1 ), where x represents the serial number of the recorded blood glucose values. Therefore, in order to ensure the accuracy of the calculation of the average blood sugar fluctuation range, it is necessary to first remove the data of the next blood sugar value among two adjacent points with equal blood sugar values, and the removal process is shown in Figure 3 . Then calculate the blood glucose standard deviation of the processed data; use the discrete sequence extreme point algorithm to calculate the discrete sequence extreme point operation on the processed data, so as to obtain the set of extreme point sequences of the data {1,2,. ..,N};

2）以序列集合{1,2,...,N}作为搜索空间，利用差分进化算法求解目标函数的解，即求得所有有效极值点，具体步骤如下：2) Use the sequence set {1,2,...,N} as the search space, and use the differential evolution algorithm to solve the solution of the objective function, that is, to obtain all effective extreme points. The specific steps are as follows:

2.1）通常函数的极大值和极小值应交错出现，因此，当找出所有有效极值点时，血糖数据中有效血糖波动的波动幅度之和最大，用公式表示为：2.1) Normally, the maximum and minimum values of the function should appear alternately. Therefore, when all effective extreme points are found, the sum of the fluctuation ranges of the effective blood sugar fluctuations in the blood sugar data is the largest, expressed as:

式（1）中，元素n₁,n₂,...,n_k均为集合{1,2,...,N}中的元素,Z_K(n₁,n₂,...,n_k)为由元素n₁,n₂,...,n_k所组成的子集{n₁,n₂,...,n_k}对应的有效血糖波动的波动幅度之和，常数K为有效极值点的个数，

为子集{n₁,n₂,...,n_k}中n_k对应的血糖值。子集{n₁,n₂,...,n_k}满足条件：In formula (1), elements n ₁ , n ₂ ,...,n _k are elements in the set {1,2,...,N}, Z _K (n ₁ ,n ₂ ,..., n _k ) is the sum of the fluctuation amplitudes of effective blood sugar fluctuations corresponding to the subset {n ₁ , n ₂ ,...,n _k } composed of elements n ₁ , n ₂ ,...,n _k , and the constant K is the number of effective extreme points,

is the blood glucose value corresponding to n _k in the subset {n ₁ ,n ₂ ,...,n _k }. The subset {n ₁ ,n ₂ ,...,n _k } satisfies the condition:

$\{\begin{matrix} 11 \leq \leq {n no}_{11} < < {n no}_{22} < < . . . . . . < < {n no}_{K K} \leq \leq N N \\ {((- - 11))}^{{n no}_{k k + + 11} - - {n no}_{k k}} = = - - 11,, k k = = 1,2 1,2,, . . . . . .,, K K - - 11 \\ | | f f (({t t}_{{n no}_{k k + + 11}})) - - f f (({t t}_{{n no}_{k k}})) | | &GreaterEqual; &Greater Equal; SD SD,, k k = = 1,2 1,2,, . . . . . .,, K K - - 11 \end{matrix} - - - - - - ((22))$

式（2）中，SD为步骤1）中所述的血糖标准差，常数K为有效极值点的个数，2≤K≤N。In formula (2), SD is the standard deviation of blood glucose described in step 1), and the constant K is the number of effective extreme points, 2≤K≤N.

结合式（1）和式（2）可知，平均血糖波动幅度的计算即为以式（2）为约束条件对式（1）为目标函数的非线性规划问题的求解，若该问题对于常数K存在一个可行解

和一个最大值

那么，也一定存在一个对应于常数K-1的最大值

且若K^*为该问题的可行解常数K的最大值，则可行解即为平均血糖波动幅度计算所需的所有有效极值点序列集合。Combining formula (1) and formula (2), it can be seen that the calculation of the average blood sugar fluctuation range is the solution of the nonlinear programming problem with formula (1) as the objective function with formula (2) as the constraint condition, if the problem is for the constant K there is a feasible solution

and a maximum

Then, there must also be a maximum value corresponding to the constant K-1

and If K ^* is the maximum value of the feasible solution constant K of the problem, then the feasible solution That is, the set of all effective extreme point sequences required for the calculation of the average blood sugar fluctuation range.

2.2）利用差分进化算法求解通过对求解步骤2.1）所得有约束条件的非线性规划问题目标函数，从而得到可行解

具体步骤为：2.2) Use the differential evolution algorithm to solve the objective function of the nonlinear programming problem with constraints obtained in the solution step 2.1), so as to obtain a feasible solution

The specific steps are:

2.2.1）利用惩罚函数算法，令式（3）为惩罚函数并初始化惩罚系数，结合式（2）将式（1）转化为无约束条件的非线性规划问题目标函数：2.2.1) Using the penalty function algorithm, let formula (3) be the penalty function and initialize the penalty coefficient, and combine formula (2) to transform formula (1) into an unconstrained nonlinear programming problem objective function:

$\{\begin{matrix} {g g}_{k k (({n no}_{11},, {n no}_{22},, . . . . . .,, {n no}_{K K}))} - - SD SD - - | | {f f}_{{n no}_{k k + + 11}} - - {f f}_{{n no}_{k k}} | | \\ {h h}_{k k (({n no}_{11},, {n no}_{22},, . . . . . .,, {n no}_{K K}))} = = {((- - 11))}^{{n no}_{K K + + 11} - - {n no}_{k k}} + + 11,, k k = = 1,2 1,2,, . . . . . .,, K K - - 11 \end{matrix} - - - - - - ((33))$

$\underset{{n no}_{11},, {n no}_{22},, . . . . . .,, {n no}_{K K}}{arg arg min min} {Y Y}_{K K} (({n no}_{11},, {n no}_{22},, . . . . . .,, {n no}_{K K}))$

$= = \underset{K K,, {n no}_{11},, {n no}_{22},, . . . . . .,, {n no}_{K K}}{arg arg min min} [[- - {Σ Σ}_{k k = = 11}^{K K - - 11} | | {f f}_{{n no}_{k k + + 11}} - - {f f}_{{n no}_{k k}} | | + + {Σ Σ}_{k k = = 11}^{K K - - 11} {μ μ}_{k k} max max {{00,, {g g}_{k k}}}^{22} + + {Σ Σ}_{k k = = 11}^{K K - - 11} {λ λ}_{k k} {h h}_{k k}^{22}]] - - - - - - ((44))$

式（4）中，μ_k和λ_k为惩罚系数，且均趋向于+∞，由式（4）可知，当式（4）的最小值Y_K ^*＞0时，该无约束条件的非线性规划问题目标函数无可行解。In formula (4), μ _k and λ _k are penalty coefficients, and they both tend to +∞. It can be seen from formula (4) that when the minimum value Y _K ^* of formula (4) > 0, the unconstrained The objective function of the linear programming problem has no feasible solution.

2.2.2）利用差分进化算法求解步骤2.2.1）所得无约束条件非线性规划问题目标函数：2.2.2) Use the differential evolution algorithm to solve the objective function of the unconstrained nonlinear programming problem obtained in step 2.2.1):

2.2.2A）初始化差分进化算法中所涉及的参数：包括种群数量NP、最大迭代次数G_max以及迭代计数器G＝1、解空间最大维度N、交叉概率CR；2.2.2A) Initialize the parameters involved in the differential evolution algorithm: including the population size NP, the maximum number of iterations G _max and the iteration counter G=1, the maximum dimension N of the solution space, and the crossover probability CR;

2.2.2B）随机创建一个初始化种群并将该种群均匀分布于整个解的搜索空间，种群的表示形式如下式：2.2.2B) Randomly create an initialization population and evenly distribute the population in the search space of the entire solution. The representation of the population is as follows:

${N N}_{}^{i i} = = ((1,1 1,1,, . . . . . .,, 11)) + + NINT NINT [[(({ρ ρ}_{i i 11},, {ρ ρ}_{i i 22},, . . . . . .,, {ρ ρ}_{iN i})) \cdot &Center Dot; ((N N - - 11))]] - - - - - - ((55))$

式（5）中，i为当前试验向量的索引，为目标向量，ρ_i1为[0,1]间的随机数，NINT[B]表示对向量B做四舍五入取整操作；In formula (5), i is the index of the current test vector, is the target vector, ρ _i1 is a random number between [0,1], and NINT[B] means rounding and rounding the vector B;

2.2.2C）从K＝2维搜索空间开始，搜索整个搜索空间，直至可行解的搜索空间大小等于极值点的集合大小，具体步骤为：2.2.2C) Starting from the K=2-dimensional search space, search the entire search space until the size of the search space of the feasible solution is equal to the size of the set of extreme points. The specific steps are:

2.2.2D）在当前代中搜索目标函数的可行解，对于每一代G的目标向量在同一代中随机抽取其他三个不同的个体向量

和

其中r₂，r₂，r₂是在[1,NP]区间中随机抽取、互不相等的三个整数，且均与当前试验向量的索引i不相等。那么通过差分策略得到一个与之对应的变异向量

2.2.2D) Search for a feasible solution of the objective function in the current generation, for the objective vector of each generation G Randomly sample other three different individual vectors in the same generation

and

Among them, r ₂ , r ₂ , and r ₂ are three unequal integers randomly selected in the [1,NP] interval, and all of them are unequal to the index i of the current test vector. Then get a corresponding mutation vector through the difference strategy

${V V}_{i i}^{G G} = = {N N}_{{r r}_{11}}^{G G} + + NINT NINT [[F f \times \times (({N N}_{{r r}_{22}}^{G G} - - {N N}_{{r r}_{33}}^{G G}))]] - - - - - - ((66))$

式（6）中，F为一个[0,1]区间内取值的差分向量缩放因子；In formula (6), F is a difference vector scaling factor with a value in the interval [0,1];

2.2.2E）在步骤2.2.2D）所得到的变异向量

和对应的目标向量

间进行交叉操作生成一个试验向量

2.2.2E) The mutation vector obtained in step 2.2.2D)

and the corresponding target vector

A test vector

${u u}_{ji the ji}^{G G} = = \{\begin{matrix} {v v}_{ji the ji}^{G G},, & rand rand [[0,1 0,1]] \leq \leq CRorj CRorj = = {j j}_{rand rand},, \\ {n no}_{ji the ji}^{G G},, & otherwise otherwise,, \end{matrix} j j = = 1,2 1,2,, . . . . . .,, n no . . - - - - - - ((77))$

式（7）中CR∈[0,1]是一个由差分进化法使用者所设定的交叉概率，j_rand是在[1,NP]区间内随机选择的索引，该索引用于保证

能在

中至少获得一个参数值。目标向量的整数参数按照从小到大顺序进行排列；In formula (7), CR∈[0,1] is a crossover probability set by the user of the differential evolution method, and j _rand is an index randomly selected in the interval [1,NP], which is used to ensure

can be in

Get at least one parameter value in . target vector The integer parameters of are arranged in ascending order;

2.2.2F）采用贪婪搜索策略对步骤2.2.2E）所得到的试验向量

与目标向量

进行比较选择，若试验向量

所决定的函数值

小于目标向量所决定的函数值

则下一代的目标向量由当前代的试验向量取代，如否，则将当前带的目标向量保留至下一代,用公式（8）表示如下：2.2.2F) Apply the greedy search strategy to the test vector obtained in step 2.2.2E)

with the target vector

For comparison selection, if the test vector

The determined function value

less than target vector The determined function value

Then the target vector of the next generation is replaced by the test vector of the current generation, if not, the target vector of the current band is retained to the next generation, expressed as follows by formula (8):

${N N}_{i i}^{G G + + 11} = = \{\begin{matrix} {U u}_{i i}^{G G},, ifY ifY (({U u}_{i i}^{G G})) < < Y Y (({N N}_{i i}^{G G})),, \\ {N N}_{i i}^{G G},, otherwise otherwise . . \end{matrix} - - - - - - ((88));;$

2.2.2G）令迭代计数器递增1，若当前迭代计数器数值G小于最大迭代次数G_max，则重复步骤2.2.2D)、2.2.2E)、2.2.2F),在下一代中搜索可行解，若当前迭代计数器数值G等于最大迭代次数G_max，则跳出迭代并计算所有有效血糖波动的波动幅度之和的最大值

若

不大于0，则令解空间维度递增1，并重复步骤2.2.2D)、2.2.2E)、2.2.2F)，若

大于0，则差分进化计算结束，此时的即为有约束条件非线性规划问题目标函数的最大值，对应的解即为有约束条件非线性规划问题目标函数的可行解

即求得所有有效极值点序列集合，如图4所示；2.2.2G) Increase the iteration counter by 1, if the current iteration counter value G is less than the maximum number of iterations G _max , then repeat steps 2.2.2D), 2.2.2E), 2.2.2F), and search for a feasible solution in the next generation, if The current iteration counter value G is equal to the maximum number of iterations G _max , then jump out of the iteration and calculate the maximum value of the sum of the fluctuation ranges of all valid blood sugar fluctuations

like

is not greater than 0, increase the solution space dimension by 1, and repeat steps 2.2.2D), 2.2.2E), 2.2.2F), if

is greater than 0, the differential evolution calculation ends, and at this time That is, the maximum value of the objective function of the nonlinear programming problem with constraints, and the corresponding solution is the feasible solution of the objective function of the nonlinear programming problem with constraints

That is to obtain all effective extremum point sequence sets, as shown in Figure 4;

3）根据步骤2.2.2G）所有有效极值点序列集合，计算平均血糖波动幅度，计算过程如下：3) Calculate the average blood sugar fluctuation range according to step 2.2.2G) of all valid extremum point sequences, and the calculation process is as follows:

a)若

对应的是动态血糖监测数据的第一个局部极小值点，则平均血糖波动幅度的计算公式为：a) if

Corresponding to the first local minimum point of the continuous blood glucose monitoring data, the calculation formula of the average blood glucose fluctuation range is:

b)若

对应的是动态血糖监测数据的第一个局部极大值点，则平均血糖波动幅度的计算公式为：b) if

Corresponding to the first local maximum point of the continuous blood glucose monitoring data, the calculation formula of the average blood glucose fluctuation range is:

式（9）和式（10）中，MAGE₊为以第一个有效波动方向（从波谷到波峰或从波峰到波谷）为计算方向的平均血糖波动幅度，MAGE_-为以该计算方向的另一侧（从波峰到波谷或从波谷到波峰）为计算方向的平均血糖波动幅度，MAGE_a是MAGE₊和MAGE_-的平均值。In formulas (9) and (10), MAGE ₊ is the average blood sugar fluctuation range based on the first effective fluctuation direction (from trough to peak or from peak to trough) as the calculation direction, and MAGE _- is the other calculation direction based on this calculation direction. One side (peak to trough or trough to peak) is the mean magnitude of blood glucose fluctuation in the calculated direction, and MAGE _a is the average of MAGE ₊ and _MAGE- .

例2（方法的准确性和和可行性验证）Example 2 (method accuracy and feasibility verification)

下表1所示为从76例就医者的动态血糖监测数据分别利用传统人工计算方法和本发明所述的计算方法计算得到的平均血糖波动幅度，共分3组，其中，正常成人27例，2型糖尿病成人患者25例，正常孕妇24例，MAGE_O表示采用传统人工计算方法计算所得平均血糖波动幅，MAGE_C表示采用本发明所述的计算方法计算所得平均血糖波动幅度。由于采用传统人工计算方法所得计算结果的准确性会受医务工作者的主观性影响，因此本表中MAGE_O数值均为长期从事相关工作的工作人员采用传统人工计算方法获得，具有较高的准确性。The following table 1 shows the average blood sugar fluctuation range calculated from the continuous blood glucose monitoring data of 76 cases of medical patients by using the traditional manual calculation method and the calculation method of the present invention respectively, and is divided into 3 groups, of which 27 normal adults, 25 adult patients with type 2 diabetes and 24 normal pregnant women. MAGE _O represents the average blood sugar fluctuation range calculated by traditional manual calculation method, and MAGE _C represents the average blood sugar fluctuation range calculated by the calculation method of the present invention. Since the accuracy of calculation results obtained by traditional manual calculation methods will be affected by the subjectivity of medical workers, the values of MAGE _O in this table are all obtained by traditional manual calculation methods by staff who have been engaged in relevant work for a long time, and have high accuracy. sex.

现将表1中3组的本发明所述的计算方法所得计算结果与传统人工计算方法所得计算结果进行相关性分析，分析结果如图5所示，正常成人组的相关系数γ＝0.994(P＜0.01)，参见图5a），2型糖尿病成人患者组的相关系数γ＝0.997(P＜0.01)，参见图5b），正常孕妇组的相关系数γ＝0.998(P＜0.01)，参见图5c），76例就医者的相关系数γ＝0.997(P＜0.01)，参见图5d），由此可知，本发明所述的计算方法所得计算结果MAGE_C和传统人工计算方法所得计算结果MAGE_O对于不同就医者分组均显著相关。Now carry out correlation analysis with the calculation result obtained by the calculation method of the present invention of 3 groups in table 1 and the calculation result obtained by traditional manual calculation method, the analysis result is as shown in Figure 5, and the correlation coefficient γ=0.994(P of normal adult group) <0.01), see Figure 5a), the correlation coefficient γ=0.997 (P<0.01), see Figure 5b) of type 2 diabetes adult group, the correlation coefficient γ=0.998 (P<0.01), see Figure 5c ), the correlation coefficient γ=0.997 (P<0.01) for 76 cases of medical patients, see Figure 5d), it can be seen that the calculation result MAGE _C obtained by the calculation method of the present invention and the calculation result MAGE _O obtained by the traditional manual calculation method are relatively There were significant correlations among different groups of medical seekers.

再通过Bland-Altman图评估本发明所述的计算方法和传统人工计算方法的一致性，所得结果如图6所示，不同就医者采用两种不同方法所得计算结果的差值分别为，正常成人组（-0.00242±0.0136）mmol*L-1，2型糖尿病成人患者组（0.01533±0.0275）mmol*L-1，正常孕妇组（0.0034±0.0191）mmol*L-1，76例就医者（0.00526±0.0218）mmol*L-1。95%一致性界限利用mean±1.96SD公式求得，分别为正常成人组（0.0243，-0.0292）；2型糖尿病成人患者组（0.0692，-0.0386）；正常孕妇组（0.0409，-0.0341）；76例就医者（0.0480，-0.0374）。分析图6中不同的Bland-Altman图可知，落在一致性界限外的点分别为正常成人组3个，占11.1%（3/27）；2型糖尿病成人患者组1个，占4%（1/25）；正常孕妇组1个，占4.2%（1/24）；76例就医者6个，占7.9%（6/76）。分析结果显示本发明所述的平均血糖波动幅度计算方法（MAGEc）与传统人工计算方法（MAGEo）具有较好的一致性。Evaluate the consistency of the calculation method of the present invention and the traditional artificial calculation method by Bland-Altman figure again, the obtained result is as shown in Figure 6, and the difference of the calculation result obtained by two kinds of different methods for different seekers is respectively, normal adult group (-0.00242±0.0136) mmol*L-1, adult patients with type 2 diabetes (0.01533±0.0275) mmol*L-1, normal pregnant women group (0.0034±0.0191) mmol*L-1, 76 patients (0.00526 ±0.0218) mmol*L-1. The limit of 95% agreement was obtained by mean±1.96SD formula, which were normal adult group (0.0243, -0.0292); type 2 diabetic adult patient group (0.0692, -0.0386); normal pregnant woman group (0.0409, -0.0341); 76 cases of medical treatment (0.0480, -0.0374). Analyzing the different Bland-Altman diagrams in Figure 6, it can be seen that there are 3 points falling outside the consistency limit in the normal adult group, accounting for 11.1% (3/27); 1 point in the type 2 diabetic adult group, accounting for 4% ( 1/25); 1 in the normal pregnant group, accounting for 4.2% (1/24); 6 of the 76 cases of medical treatment, accounting for 7.9% (6/76). The analysis results show that the calculation method (MAGEc) of the present invention has a good consistency with the traditional manual calculation method (MAGEo).

表1本发明所述计算方法与传统人工计算方法数据对比Table 1 Calculation method of the present invention and traditional manual calculation method data comparison

Claims

1. the computing method of a MAGE, this method may further comprise the steps:

1) reads in the dynamic glucose monitor data that from dynamic blood sugar monitoring instrument device, obtains, calculate the blood sugar standard deviation of the data of being obtained; Calculate the extreme point all in the data that obtains, obtain the set A of an extreme point sequence;

2) utilize differential evolution algorithm to find the solution the objective function shown in the following formula (1), meet the extreme point of following two conditions in the statistics set A, obtain set B:

Absolute value >=blood sugar the standard deviation of the blood sugar value difference that two a. adjacent extreme points are corresponding;

B. one adjacent of two extreme point are maximum value, and another is a minimal value;

\underset{K, n_{1}, n_{2}, . . ., n_{K}}{\arg \max} Z_{K} (n_{1}, n_{2}, . . ., n_{K}) = \underset{K, n_{1}, n_{2}, . . ., n_{K}}{\arg \max} Σ_{k = 1}^{K - 1} | f (t_{n_{k + 1}}) - f (t_{n_{k}}) | - - - (1)

In the formula (1), element n ₁, n ₂..., n _kBe the element in the set A, Z _K(n ₁, n ₂..., n _k) be by element n ₁, n ₂..., n _kSubclass { the n that is formed ₁, n ₂..., n _kThe fluctuating range sum of corresponding effective blood glucose fluctuation, constant K is to satisfy the number of the extreme point of above-mentioned condition,

Be subclass { n ₁, n ₂..., n _kMiddle n _kCorresponding blood glucose value;

The absolute value of the blood sugar value difference that 3) each adjacent extreme point is corresponding among the first set of computations B calculates all average absolute again, promptly gets MAGE.

2. the computing method of a kind of MAGE according to claim 1 is characterized in that, wherein step 1) also comprises following data pre-treatment step:

After reading in the dynamic glucose monitor data that from dynamic blood sugar monitoring instrument device, obtains, earlier the back blood glucose value data in the deletion blood glucose value equate adjacent 2 are calculated described blood sugar standard deviation and all extreme points again.