Intelligent non-invasive blood glucose detection method suitable for diabetic patients

Abstract

The present application relates to the technical field of blood glucose detection, and in particular to an intelligent non-invasive blood glucose detection method suitable for diabetic patients. The method comprises: obtaining physiological data of different dimensions of a plurality of reference diabetic patients and NIR wavelengths; obtaining decision-related values according to the aggregate distribution characteristics and numerical distribution characteristics of the physiological data of each dimension of the reference patients, and screening and optimizing the dimensions in combination with the differences in the physiological data of each two dimensions of different reference patients and the differences in the decision-related values; merging different categories according to the differences in the wavelengths corresponding to each two categories and the quantity distribution of similar patients of the reference patients to obtain several groups; determining target wavelengths in combination with all the wavelengths corresponding to each group, the physiological data of the optimized dimensions of the reference patients in each group, and the physiological data of the optimized dimensions of the to-be-detected diabetic patients; and detecting the blood glucose of the to-be-detected diabetic patients by using the target wavelengths. The present application improves the accuracy of the blood glucose detection results of diabetic patients.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of blood glucose detection, and particularly relates to an intelligent non-invasive blood glucose detection method suitable for diabetic patients. BACKGROUND

[0002] Diabetes is a chronic metabolic disease characterized by high blood sugar, and the sustained fluctuation of blood sugar level of patients can cause damage to the organs and tissues of patients, and can cause various complications such as cardiovascular disease, retinopathy, neuropathy and kidney disease. Near-infrared (NIR) spectroscopy technology has become a research focus for non-invasive blood glucose detection due to its non-invasive and rapid characteristics, which reflects the blood glucose concentration by using the light absorption or scattering characteristics at different wavelengths. After the near-infrared light contacts the skin surface, part of the light is absorbed by glucose in the blood, and the received light of the photoelectric sensor will present a periodic band similar to the glucose information, and the suitable NIR wavelength is obtained by detecting the absorption spectrum of the blood glucose by invasive method, and then a unified blood glucose prediction model is established according to the linear relationship between blood glucose and NIR absorption rate, so as to realize the detection of blood glucose by using spectral information.

[0003] Due to the significant individual differences in physiological characteristics, skin tissue composition, blood glucose fluctuation range and dynamic regulation mechanism of diabetic patients, the suitable NIR wavelength of the patients is different, and thus the improper wavelength setting will lead to inaccurate blood glucose detection results of the patients. SUMMARY

[0004] In order to solve the problem that the improper wavelength setting of the existing near-infrared spectroscopy technology in the blood glucose detection of diabetic patients will lead to inaccurate blood glucose detection results, the purpose of the present application is to provide an intelligent non-invasive blood glucose detection method suitable for diabetic patients, and the technical scheme adopted is as follows:

[0005] The present application provides an intelligent non-invasive blood glucose detection method suitable for diabetic patients, which comprises the following steps:

[0006] Obtaining physiological data of different dimensions and NIR wavelengths of a predetermined number of diabetic reference patients;

[0007] According to the aggregation distribution characteristics and numerical distribution characteristics of the physiological data of each dimension of all reference patients, the decision correlation value of the physiological data of each dimension is obtained, the difference of the physiological data of each two dimensions of different reference patients and the difference of the decision correlation values are comprehensively evaluated to evaluate the exchange degree of the physiological data of each two dimensions, and the preferred dimensions are screened in combination with all the decision correlation values and all the exchange degrees;

[0008] screening similar patients of each category of reference patients in other categories based on differences between physiological data of preferred dimensions of reference patients of different categories, wherein the wavelength is same for all reference patients within the same category; performing merging processing on different categories according to differences of wavelengths corresponding to each two categories and number distribution of similar patients of reference patients in each two categories to obtain several groups;

[0009] determining target wavelengths in combination of all wavelengths corresponding to each group, physiological data of preferred dimensions of reference patients within each group and physiological data of preferred dimensions of the to-be-detected diabetes patient, and detecting blood glucose of the to-be-detected diabetes patient by using the target wavelengths.

[0010] Preferably, the decision-related value of physiological data of each dimension is obtained according to the aggregation distribution feature and the numerical distribution feature of physiological data of each dimension of all reference patients, and the decision-related value of physiological data of each dimension is obtained according to the aggregation distribution feature and the numerical distribution feature of physiological data of each dimension of all reference patients.

[0011] calculating number differences of all reference patients of each two categories and differences of average values of physiological data of candidate dimensions of reference patients within each two categories;

[0012] obtaining a decision-related value of physiological data of the candidate dimension according to all the number differences and all the average value differences;

[0013] The candidate dimension is any dimension.

[0014] Preferably, the exchange degree of physiological data of each two dimensions is evaluated by comprehensively considering the difference situation of physiological data of each two dimensions of different reference patients and the difference situation of decision-related values, and the exchange degree of physiological data of each two dimensions is evaluated by comprehensively considering the difference situation of physiological data of each two dimensions of different reference patients and the difference situation of decision-related values.

[0015] respectively calculating a first difference between physiological data of each dimension of each two reference patients;

[0016] for any two dimensions:

[0017] obtaining a first feature value of the any two dimensions of each two reference patients according to the first difference corresponding to the any two dimensions of each two reference patients;

[0018] obtaining an exchange degree of physiological data of the any two dimensions according to a difference between all first feature values of the any two dimensions of all reference patients and a difference between decision-related values of physiological data of the any two dimensions, wherein the difference between all first feature values and the difference between decision-related values are negatively correlated with the exchange degree.

[0019] Preferably, the first feature value of the any two dimensions of each two reference patients is obtained according to the first difference corresponding to the any two dimensions of each two reference patients, and the first feature value of the any two dimensions of each two reference patients is obtained according to the first difference corresponding to the any two dimensions of each two reference patients.

[0020] calculating a first sum value between the first difference corresponding to one of the arbitrary two dimensions of each two reference patients and a preset first adjustment parameter, and a second sum value between the first difference corresponding to the other of the arbitrary two dimensions of each two reference patients and a preset first adjustment parameter; wherein the preset first adjustment parameter is greater than 0;

[0021] determining a ratio between the first sum value and the second sum value as a first feature value corresponding to the arbitrary two dimensions of the two reference patients.

[0022] Preferably, the screening of the preferred dimensions based on all the decision-related values and all the interchange degrees comprises:

[0023] taking a product between the interchange degree of the candidate dimension and the physiological data of each related dimension of the candidate dimension and the decision-related value of the physiological data of each related dimension of the candidate dimension as a second feature value of each related dimension of the candidate dimension; wherein the related dimension of the candidate dimension is a dimension with an interchange degree greater than a preset interchange degree threshold value with the candidate dimension;

[0024] combining the second feature values of all the related dimensions of the candidate dimension and the interchange degree of the physiological data of the candidate dimension to obtain a preference degree of the candidate dimension, the second feature values and the preference degree being in a negative correlation relationship, and the interchange degree of the physiological data of the candidate dimension and the preference degree being in a positive correlation relationship;

[0025] screening the preferred dimensions according to the size relationship of the preference degrees of all the dimensions.

[0026] Preferably, the screening of the preferred dimensions according to the size relationship of the preference degrees of all the dimensions comprises:

[0027] sequencing all the dimensions according to the preference degrees from large to small to obtain a dimension sequence;

[0028] determining the first preset number of dimensions in the dimension sequence as the preferred dimensions.

[0029] Preferably, the screening of the similar patients of each category of reference patients in other categories based on the differences between the physiological data of the preferred dimensions of the reference patients of different categories comprises:

[0030] obtaining a similarity value between the first patient and the second patient according to the difference between the physiological data of the same preferred dimensions of the first patient and the second patient, the difference between the physiological data of the same preferred dimensions and the similarity value being in a negative correlation relationship;

[0031] if the similarity value is greater than a preset similarity threshold value, determining that the first patient and the second patient are similar patients;

[0032] wherein the first patient and the second patient are two reference patients in different categories.

[0033] Preferably, the different categories are merged to obtain several groups according to the difference of the wavelengths corresponding to each two categories and the quantity distribution of the similar patients of the reference patients in each two categories, including:

[0034] calculating a third sum between the quantity of all similar patients of the reference patient in the first category in the second category and the quantity of all similar patients of the reference patient in the second category in the first category, and obtaining the fusion degree of the arbitrary two categories according to the difference between the wavelengths corresponding to the first category and the second category and the third sum, wherein the difference between the wavelengths and the fusion degree are in a negative correlation, and the third sum and the fusion degree are in a positive correlation, wherein the first category and the second category are any two categories in all categories;

[0035] merging the corresponding categories based on the fusion degree of each two categories to obtain several groups.

[0036] Preferably, the merging the corresponding categories based on the fusion degree of each two categories to obtain several groups includes:

[0037] sorting all categories according to a preset order to obtain a category sequence, wherein the preset order is an order from small to large of the wavelengths or an order from large to small of the wavelengths;

[0038] merging two categories with the largest fusion degree, and judging whether the fusion degree of each category is greater than a preset fusion degree threshold value with any one of the merged categories in sequence according to the order of the categories in the category sequence, if greater, merging the corresponding category with the merged categories, and if less than or equal to, taking it as a new category, and taking each category after the complete merging as a group.

[0039] Preferably, the determining the target wavelength by combining all wavelengths corresponding to each group, the physiological data of the preferred dimension of the reference patients in each group and the physiological data of the preferred dimension of the to-be-detected diabetes patient includes:

[0040] training a wavelength selection model by using all wavelengths corresponding to each group and the physiological data of the preferred dimension of the reference patients in each group to obtain a trained wavelength selection model;

[0041] inputting the physiological data of the preferred dimension of the to-be-detected diabetes patient into the trained wavelength selection model to obtain the target wavelength.

[0042] The present application has at least the following beneficial effects:

[0043] The application first calculates the decision correlation value of the physiological data of each dimension according to the aggregate distribution characteristics and numerical distribution characteristics of the physiological data of each dimension of all reference patients, then evaluates the interchangeability of the physiological data of different dimensions according to the difference of the physiological data of each two dimensions of different reference patients and the difference of the decision correlation values, and obtains the corresponding interchangeability. Since the collected physiological data has many dimensions, the data of some dimensions has little influence on the detection result. In order to reduce the calculation amount, the application screens multiple preferred dimensions from all dimensions by combining the decision correlation value and the interchangeability, that is, the physiological data is screened according to the different characteristics of different dimensions on blood glucose detection, and the similar patients of each category of reference patients in other categories are screened based on the difference between the physiological data of the preferred dimensions of the reference patients of different categories. According to the difference of the corresponding wavelengths of different categories and the quantity distribution of the similar patients of the reference patients, different categories are merged to obtain multiple groups. Further, the near-infrared wavelength of the to-be-detected diabetic patient is selected by combining all the wavelengths corresponding to each group, the physiological data of the preferred dimensions of the reference patients in each group, and the physiological data of the preferred dimensions of the to-be-detected diabetic patient, and the blood glucose of the to-be-detected diabetic patient is detected by using the wavelength. The method provided by the application makes the wavelength selection of the to-be-detected diabetic patient more accurate and reliable, and improves the accuracy of the blood glucose detection result of the diabetic patient. BRIEF DESCRIPTION OF DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, and the advantages thereof, below will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor based on these drawings.

[0045] Figure 1 A flowchart of an intelligent non-invasive blood glucose detection method for diabetic patients provided by an embodiment of the present application;

[0046] Figure 2 A structural block diagram of an intelligent non-invasive blood glucose detection system for diabetic patients provided by an embodiment of the present application. DETAILED DESCRIPTION

[0047] In order to further illustrate the technical means and effects taken by the present application to achieve the predetermined invention purpose, the following will combine the drawings and the preferred embodiments to specifically describe a kind of intelligent non-invasive blood glucose detection method for diabetic patients provided by the present application as follows.

[0048] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0049] The application provides a specific scheme of an intelligent noninvasive blood glucose detection method for diabetic patients.

[0050] An embodiment of an intelligent noninvasive blood glucose detection method for diabetic patients is provided.

[0051] The specific scenario of the embodiment is as follows: when noninvasive blood glucose detection is performed on diabetic patients, the physiological characteristics of different diabetic patients can be quite different, and traditional fixed wavelength or unified blood glucose prediction models often have certain limitations. The method provided in the embodiment can provide different wavelength selections for diabetic patients with different physiological characteristics, thereby improving the accuracy and reliability of blood glucose detection results.

[0052] The embodiment provides an intelligent noninvasive blood glucose detection method for diabetic patients, as shown in Figure 1 The intelligent noninvasive blood glucose detection method for diabetic patients provided in the embodiment includes the following steps.

[0053] In step S1, physiological data of different dimensions and NIR wavelengths of a preset number of diabetic reference patients are acquired.

[0054] Firstly, electronic medical records, age, several symptom degrees, several disease degrees, smoking index, drinking index, etc. of a preset number of diabetes patients are acquired, wherein the symptom degrees are obtained by doctors scoring the degrees of each symptom, and the symptoms include polyuria, weight loss, blurred vision, etc.; the disease degrees are also obtained by doctors scoring the degrees of each disease, and the diseases include hypertension, coronary heart disease and common complications of diabetes patients, etc.; in this embodiment, the smoking index = the number of cigarettes smoked per day x the number of years of smoking, and the drinking index = the amount of alcohol consumed per week x the number of years of drinking. Then, the physiological indexes of these diabetes patients are collected, including skin reflectance (finger pulp), tissue thickness, BMI index, heart rate, blood pressure, and several components of blood; each type of data collected is used as a dimension of physiological data, that is, different dimensions of physiological data of these patients are collected, and the physiological data of these patients will be used for training of the wavelength selection model, therefore, these patients are recorded as reference patients, and all the collected physiological data are normalized to eliminate the influence of dimension, and the normalization of data is prior art, which will not be described in detail here, and the physiological data mentioned later are all normalized data, that is, the physiological data of multiple dimensions of multiple reference patients are collected. It should be noted that the electronic medical records and physiological data of all diabetes patients in this embodiment are collected with permission, and are non-private data. In this embodiment, the preset number is 500, that is, the number of reference patients is 500, and in specific applications, the implementer can set it according to the specific circumstances.

[0055] The absorption spectrum of blood glucose of each reference patient is measured by drawing venous blood, and the glucose absorption peak is obtained as the NIR wavelength corresponding to the patient.

[0056] Up to now, the physiological data of different dimensions of each diabetes reference patient and the NIR wavelength corresponding to each diabetes reference patient have been collected in this embodiment.

[0057] In step S2, the decision-related values of each dimension of physiological data are obtained according to the aggregation distribution characteristics and numerical distribution characteristics of the physiological data of each dimension of all reference patients; the exchange degree of the physiological data of each two dimensions is evaluated according to the difference of the physiological data of each two dimensions of different reference patients and the difference of the decision-related values; and the preferred dimensions are screened in combination with all the decision-related values and all the exchange degrees.

[0058] The embodiment is to measure the blood sugar of a diabetes patient in real time by a non-invasive method, so it is extremely important to select the optimal NIR wavelength representing the blood sugar of the patient, and the method of drawing venous blood not only has damage, but also the optimal NIR wavelength of the diabetes patient can fluctuate. Therefore, the embodiment considers analyzing the relationship between the physiological data of the patient and the NIR wavelength, so as to achieve the purpose of predicting the optimal NIR wavelength of the patient only through the physiological data of the patient, and timely adjusting the NIR wavelength for measuring the blood sugar when the patient has symptom changes or body changes, so as to ensure the accuracy of the blood sugar measurement.

[0059] Firstly, the relationship between the physiological data of the patient and the NIR wavelength needs to be determined, and the random forest model predicts the category of the sample by constructing multiple decision trees, which is suitable for the purpose, so the method adopts the random forest algorithm for training and learning.

[0060] Since not all dimensions of the physiological data collected in step S1 can affect the optimal NIR wavelength, in order to optimize the model, the dimensions should be screened and the different dimensions should be weighted to improve the classification effect of the model. In addition, since the optimal NIR wavelength can have a large range, each wavelength is used as a scheme, which has certain problems in implementation. The differences between patients corresponding to different NIR wavelengths can merge some similar available NIR wavelengths, which can improve the generalization ability of the model and provide multiple choices for the patient.

[0061] The embodiment calculates the decision correlation value of each dimension of the physiological data according to the aggregation distribution characteristics and the numerical distribution characteristics of each dimension of the physiological data of all reference patients, and then screens the preferred dimensions from all dimensions in combination with the difference of each two dimensions of the physiological data of different reference patients, and the physiological data of the preferred dimensions will be used for the selection of the wavelength.

[0062] The embodiment takes one dimension as an example for description, and the method provided by the embodiment can be used for processing other dimensions. Specifically, any dimension is denoted as a candidate dimension, and all reference patients corresponding to the same wavelength are denoted as a category, that is, all reference patients are divided into multiple categories according to the wavelength, the wavelength corresponding to all patients in the same category is equal, and the wavelengths corresponding to patients in different categories are not equal. The average value of the physiological data of the candidate dimension of all reference patients in each category is calculated according to the physiological data of the candidate dimension of all reference patients in each category. It should be noted that each category corresponds to an average value under the candidate dimension. The number difference of all reference patients of each two categories and the average value difference of the physiological data of the candidate dimension of the reference patients in each two categories are calculated respectively. The decision correlation value of the physiological data of the candidate dimension is obtained according to all the number differences and all the average value differences.

[0063] In the present embodiment, a specific calculation formula of the decision correlation value is given, and the decision correlation value of the physiological data of the i-th dimension can be expressed as:

[0064]

[0065] wherein L i represents the decision correlation value of the physiological data of the i-th dimension, A represents the number of categories, represents the number of all reference patients in the s-th category, represents the number of all reference patients in the t-th category, C s represents the average value of the physiological data of the candidate dimension of the reference patient in the s-th category, C t represents the average value of the physiological data of the candidate dimension of the reference patient in the t-th category, norm() represents a normalization function, and || represents an absolute value symbol.

[0066] represents the cumulative sum of the number difference of the reference patients of different categories, represents the difference between the average values of the physiological data of the i-th dimension of the reference patients in different categories, both of which reflect the preliminary decision correlation of the physiological data of the i-th dimension in terms of quantity and density.

[0067] For any two dimensions, if the variation of the physiological data between any patients is correlated, i.e., large for one patient and small for another patient, or one dimension is large and the other is small for one patient, and the opposite for another patient, the correlation of the physiological data of the two dimensions is high. At the same time, if the decision correlation is also similar, the two features can be interchangeable in the wavelength selection decision.

[0068] Based on this, the difference between the physiological data of each dimension of each two reference patients is calculated, and the difference is recorded as a first difference. For any two dimensions: calculate the first sum value between the first difference corresponding to one of the two dimensions of each two reference patients and a preset first adjustment parameter, and the second sum value between the first difference corresponding to the other of the two dimensions of each two reference patients and a preset first adjustment parameter; the ratio between the first sum value and the second sum value is determined as the first feature value of the two reference patients in the two dimensions. According to the difference between all first feature values of all reference patients in the two dimensions and the difference between the decision correlation values of the physiological data of the two dimensions, the degree of interchangeability of the physiological data of the two dimensions is obtained, and the difference between the all first feature values and the difference between the decision correlation values are negatively correlated with the degree of interchangeability.

[0069] The negative correlation indicates that the dependent variable decreases with the increase of the independent variable, and the dependent variable increases with the decrease of the independent variable, which can be a subtraction relationship, a division relationship, etc., and is determined by actual application.

[0070] In this embodiment, specific calculation formulas of the first characteristic value and the interchange degree are given, and can be specifically represented as:

[0071]

[0072] wherein, M i,j represents the interchange degree of the physiological data of the i-th dimension and the physiological data of the j-th dimension, L i represents the decision correlation value of the physiological data of the i-th dimension, L j represents the decision correlation value of the physiological data of the j-th dimension, Di x represents the physiological data of the i-th dimension of the x-th reference patient, Di y represents the physiological data of the i-th dimension of the y-th reference patient, Dj x represents the physiological data of the j-th dimension of the x-th reference patient, Dj y represents the physiological data of the j-th dimension of the y-th reference patient, represents the first characteristic value of the i-th dimension and the j-th dimension of the x-th reference patient and the y-th reference patient, represents the n-th first characteristic value of the i-th dimension and the j-th dimension of all reference patients, represents the r-th first characteristic value of the i-th dimension and the j-th dimension of all reference patients, and λ1 is a preset first adjustment parameter.

[0073] The preset first adjustment parameter is introduced into the calculation formula of the first characteristic value in this embodiment to prevent the denominator from being 0. In this embodiment, the preset first adjustment parameter is 0.01, which can be set according to specific circumstances by the implementer in specific applications. x -Di y |) represents the first difference between the physiological data of the i-th dimension of the x-th reference patient and the y-th reference patient, norm(|Dj x -Dj y |) represents the first difference between the physiological data of the j-th dimension of the x-th reference patient and the y-th reference patient, norm(|Di x -Di y |) represents the first sum, norm(|Dj x -Dj y |) represents the second sum.

[0074] The greater the difference between the decision-related values of the physiological data between the ith dimension and the jth dimension and the greater the difference between the different first characteristic values, the more inconsistent the ith dimension and the jth dimension are in different patients, and the smaller the interchangeability of the ith dimension and the jth dimension, that is, the smaller the interchangeability of the physiological data of the ith dimension and the physiological data of the jth dimension.

[0075] In the multi-dimension with high interchangeability, the higher the decision-related value, the greater the possibility of being selected as the preferred feature to participate in the model establishment, and thus the preferred degree of each dimension is calculated.

[0076] Next, still taking the candidate dimension as an example, specifically, all the dimensions with an interchangeability greater than a preset interchangeability threshold with the candidate dimension are taken as the related dimensions of the candidate dimension. In this embodiment, the preset interchangeability threshold is 0.5, and in specific applications, the implementer can set it according to the specific situation. The product of the interchangeability of the physiological data of the candidate dimension and each related dimension of the candidate dimension and the decision-related value of the physiological data of each related dimension of the candidate dimension is taken as the second characteristic value of each related dimension of the candidate dimension, and it is to be noted that each related dimension of the candidate dimension corresponds to a second characteristic value. The preferred degree of the candidate dimension is obtained in combination with the second characteristic values of all the related dimensions of the candidate dimension and the interchangeability of the physiological data of the candidate dimension, and the second characteristic value is negatively correlated with the preferred degree, and the interchangeability of the physiological data of the candidate dimension is positively correlated with the preferred degree.

[0077] The negative correlation indicates that the dependent variable will decrease with the increase of the independent variable, and the dependent variable will increase with the decrease of the independent variable, which can be a subtraction relationship, a division relationship, etc., and is determined by actual application. The positive correlation indicates that the dependent variable will increase with the increase of the independent variable, and the dependent variable will decrease with the decrease of the independent variable, which can be an addition relationship, a multiplication relationship, etc., and is determined by actual application.

[0078] In this embodiment, a specific calculation formula of the preferred degree is given, and the preferred degree of the ith dimension can be represented as:

[0079]

[0080] wherein, P i represents the preferred degree of the ith dimension, L i represents the decision-related value of the physiological data of the ith dimension, G represents the number of related dimensions of the ith dimension, M i,g represents the interchangeability of the physiological data of the gth related dimension of the ith dimension, L i,g represents the decision-related value of the gth related dimension of the ith dimension.

[0081] M i,g ×L i,g denotes the second eigenvalue of the gth relevant dimension of the ith dimension. The proportion of the decision-related value of the physiological data of the ith dimension, the higher the proportion, the higher the decision relevance in the case of replacing the ith dimension in its relevant dimension, and then the ith dimension should be selected as the preferred feature among all relevant dimensions, that is, the preferred degree of the ith dimension is greater.

[0082] By using the above method, the preferred degree of each dimension can be obtained, and then the preferred dimensions are selected according to the size relationship of the preferred degrees of all dimensions. Specifically, all dimensions are sorted in descending order of preferred degree to obtain a dimension sequence, and the first preset number of dimensions in the dimension sequence are determined as the preferred dimensions, that is, a plurality of preferred dimensions are selected from all dimensions. In this embodiment, the preset number is 50% of the total number of all dimensions. It should be noted that if 50% of the total number of all dimensions is not an integer, it is rounded up and the result is taken as the preset number.

[0083] Step S3, based on the difference between the physiological data of the preferred dimensions of the reference patients of different categories, screening similar patients of each category of reference patients in other categories, wherein the wavelength of all reference patients in the same category is the same; according to the difference between the wavelength corresponding to each two categories and the number distribution of the similar patients of the reference patients in each two categories, the different categories are merged to obtain several groups.

[0084] After obtaining the preferred dimensions, the NIR wavelengths of different patients are analyzed. The NIR wavelength is 780-2526nm, so the range of the best NIR wavelength is also relatively large, and the usefulness of setting such a number of different wavelengths of blood glucose detector is relatively small. If the information sequences of two patients with different best NIR wavelengths are similar, the best NIR wavelengths used by these two types of patients can also be replaced and fused into one wavelength, which can reduce the possible types of classification results when establishing a random forest, and improve the accuracy of wavelength selection and generalization ability.

[0085] Because the data in the physiological data of the preferred dimensions can have some influence on the selection of the NIR wavelength, the difference between the physiological data of the preferred dimensions of the patients of different categories can represent the difference between the NIR wavelengths.

[0086] The dimensions of the physiological data are diverse, and the physiological data of the preferred dimensions of the patients also have differences, so the relationship between the physiological data of the preferred dimensions and the NIR wavelength is diverse. The following four kinds are considered:

[0087] a) The physiological data of different preferred dimensions have the same NIR wavelength;

[0088] b) physiological data of different preferred dimensions have different NIR wavelengths;

[0089] c) physiological data of the same preferred dimension have the same NIR wavelength;

[0090] d) physiological data of the same preferred dimension have different NIR wavelengths.

[0091] For a), patients belonging to the same category have physiological data of preferred dimensions with large differences, indicating that the influence of the preferred dimensions on the NIR wavelength is offset; for b), it indicates that the influence is not completely offset; for c), it indicates that the same NIR wavelength is used under ideal conditions of consistent physiological conditions; for d), it indicates that different NIR wavelengths are fusible. Therefore, only condition d) is considered when fusing the NIR wavelengths.

[0092] The more patients with physiological data of the same preferred dimension in different categories of reference patients, the more different categories should be merged.

[0093] Based on the above characteristics, the embodiment next first screens similar patients of each category of reference patients in other categories based on the differences between the physiological data of the preferred dimensions of the reference patients in different categories. Specifically, two reference patients in different categories are denoted as a first patient and a second patient, respectively, and a similarity value between the first patient and the second patient is obtained according to the differences between the physiological data of the preferred dimensions of the first patient and the second patient, which are negatively correlated with the similarity value.

[0094] The negative correlation indicates that the dependent variable decreases with the increase of the independent variable, and the dependent variable increases with the decrease of the independent variable, which can be a subtraction relationship, a division relationship, etc., and is determined by actual application.

[0095] In the embodiment, a specific calculation formula of the similarity value is given, and the similarity value between the first patient and the second patient can be represented as:

[0096]

[0097] wherein, P 1,2 represents the similarity value between the first patient and the second patient, H represents the number of preferred dimensions, D 1,h represents the physiological data of the hth preferred dimension of the first patient, D 2,h represents the physiological data of the hth preferred dimension of the second patient, and exp() represents an exponential function with a natural constant as the base.

[0098] |D 1,h -D 2,hThis value represents the difference between the physiological data of the first patient and the second patient in the h-th preferred dimension. The larger the value, the greater the difference between the two. When the differences between the physiological data of the first patient and the second patient in all preferred dimensions are large, it indicates that the first patient and the second patient are less similar, that is, the similarity value between the first patient and the second patient is smaller.

[0099] The higher the similarity value, the more similar the two patients' conditions are. Therefore, if the similarity value between the first patient and the second patient is greater than the preset similarity threshold, the first patient and the second patient are determined to be similar patients. In this embodiment, the preset similarity threshold is 0.5. In specific applications, the implementer can set it according to the specific situation.

[0100] For any two categories, the higher the similarity among all patients, the stronger the correlation between the two categories, and the more likely the two categories can be merged. Furthermore, the closer the NIR wavelengths corresponding to the two categories are, belonging to similar bands, the more likely the two categories can be merged. Therefore, the fusion potential between each pair of categories is evaluated based on the differences in wavelengths corresponding to each pair of categories and the distribution of the number of similar patients among the reference patients in each pair of categories.

[0101] The following explanation uses any two categories as examples. Other categories can be processed using the method provided in this embodiment.

[0102] Specifically, any two categories are designated as Category 1 and Category 2. A third sum is calculated between the number of reference patients in Category 1 and all similar patients in Category 2, and the number of reference patients in Category 2 and all similar patients in Category 1. The fusion degree between any two categories is obtained based on the difference between the wavelengths corresponding to Category 1 and Category 2 and the third sum. The difference between the wavelengths is negatively correlated with the fusion degree, and the third sum is positively correlated with the fusion degree.

[0103] Among them, a positive correlation means that the dependent variable increases as the independent variable increases, and the dependent variable decreases as the independent variable decreases. It can be an additive relationship, a multiplicative relationship, etc., which is determined by practical application. A negative correlation means that the dependent variable decreases as the independent variable increases, and the dependent variable increases as the independent variable decreases. It can be a subtractive relationship, a division relationship, etc., which is determined by practical application.

[0104] In this embodiment, a specific formula for calculating the fusion degree is given. The fusion degree between the s-th category and the t-th category can be expressed as:

[0105]

[0106] Among them, Q s,trepresents the fusion degree of the s-th category and the t-th category, l s represents the wavelength corresponding to the s-th category, l t represents the wavelength corresponding to the t-th category, X1represents the number of reference patients in the s-th category, R s,x1 (t) represents the number of all reference patients of the t-th category for the X1th reference patient in the s-th category, X2represents the number of reference patients in the t-th category, R t,x2 (s) represents the number of all reference patients of the s-th category for the X1th reference patient in the t-th category, exp() represents an exponential function with a natural constant as a base, norm() represents a normalization function, and || represents an absolute value symbol.

[0107] | l s - l t represents the difference between the wavelengths corresponding to the s-th category and the t-th category, and the greater the absolute value, the greater the difference between the two wavelengths. represents a third sum. When the wavelength difference between the s-th category and the t-th category is smaller and the third sum is greater, it means that the s-th category and the t-th category are more suitable for fusion, i.e., the fusion degree of the s-th category and the t-th category is higher.

[0108] By using the above method, the fusion degree of each two categories can be obtained, and the greater the fusion degree, the more suitable the corresponding two categories are for merging processing. Therefore, the corresponding categories are merged based on the fusion degree of each two categories, and a plurality of groups are obtained.

[0109] Specifically, all categories are sorted in a preset order to obtain a category sequence, wherein the preset order is an order from small to large wavelength or an order from large to small wavelength; in this embodiment, the preset order is an order from small to large wavelength, i.e., all categories are sorted in an order from small to large wavelength to obtain a category sequence. Then, the two categories with the largest fusion degree are merged, and then whether the fusion degree of each category is greater than a preset fusion degree threshold in the order of the categories in the category sequence is determined in sequence, if greater, the corresponding category is merged with the merged categories, and if less than or equal to, it is taken as a new category; all categories in the category sequence are determined in sequence and the categories that need to be merged are merged, and each category after all merging is completed is taken as a group, i.e., a plurality of groups are obtained. In this embodiment, the fusion degree threshold is 0.5, which can be set according to specific circumstances by the implementer in specific applications.

[0110] Step S4, determine the target wavelength in combination with all wavelengths corresponding to each group, the physiological data of the preferred dimension of the reference patient in each group and the physiological data of the preferred dimension of the diabetes patient to be detected, and detect the blood glucose of the diabetes patient to be detected by using the target wavelength.

[0111] Next, this embodiment will construct a wavelength selection model, train the wavelength selection model by using all wavelengths corresponding to each group and the physiological data of the preferred dimension of the reference patient in each group, obtain the best wavelength of the diabetes patient to be detected by using the trained wavelength selection model, and detect the blood glucose of the diabetes patient to be detected by using the best wavelength.

[0112] Specifically, each reference patient is taken as a sample, resampling is performed from the sample by using the Bootstrap method, and a training sample subset with the same sample amount is randomly selected with replacement, a single decision tree is constructed, the random forest model is constituted by repeating several times, that is, the wavelength selection model of the patient, the input of the model is all physiological data of the preferred dimension corresponding to each sample, and the output of the model is the wavelength.

[0113] The physiological data of the preferred dimension of the diabetes patient to be detected is collected, normalized, input into the trained wavelength selection model, the target wavelength is obtained, and the blood glucose of the diabetes patient to be detected is detected by using the target wavelength. It should be noted that since there can be multiple wavelengths in a group, the target wavelength obtained finally can be more than one, and when the blood glucose of the diabetes patient to be detected is detected, one wavelength can be selected from all target wavelengths for blood glucose detection, thereby realizing the non-invasive detection of the blood glucose of the diabetes patient and improving the convenience and accuracy of the patient in detecting the blood glucose.

[0114] The embodiment first calculates the decision correlation values of the physiological data of each dimension according to the aggregate distribution characteristics and the numerical distribution characteristics of the physiological data of each dimension of all reference patients, then evaluates the interchangeability of the physiological data of different dimensions according to the difference of the physiological data of each two dimensions of different reference patients and the difference of the decision correlation values, and obtains the corresponding interchangeability. Since the collected physiological data has many dimensions, the data of some dimensions have little influence on the detection result. In order to reduce the calculation amount, the embodiment screens the physiological data from all dimensions by combining the decision correlation values and the interchangeability, screens the similar patients of each category of reference patients in other categories based on the difference between the physiological data of the preferred dimensions of the reference patients of different categories, and combines different categories to obtain a plurality of groups according to the difference of the corresponding wavelengths of different categories and the quantity distribution of the similar patients of the reference patients. The embodiment further selects the near-infrared wavelength for the to-be-detected diabetic patient by combining all the wavelengths corresponding to each group, the physiological data of the preferred dimensions of the reference patients in each group, and the physiological data of the preferred dimensions of the to-be-detected diabetic patient, and detects the blood glucose of the to-be-detected diabetic patient by using the wavelength. The method provided in the embodiment makes the wavelength selection of the to-be-detected diabetic patient more accurate and reliable, and improves the accuracy of the blood glucose detection result of the diabetic patient.

[0115] An embodiment of an intelligent non-invasive blood glucose detection system suitable for diabetic patients:

[0116] As shown in Figure 2 , the figure shows a structural block diagram of an intelligent non-invasive blood glucose detection system suitable for diabetic patients, which includes a data acquisition module, a preferred dimension screening module, a category merging module, and a blood glucose detection module.

[0117] The data acquisition module is configured to acquire physiological data of different dimensions of a preset number of diabetic reference patients and NIR wavelengths.

[0118] The preferred dimension screening module is configured to obtain the decision correlation values of the physiological data of each dimension according to the aggregate distribution characteristics and the numerical distribution characteristics of the physiological data of each dimension of all reference patients, evaluate the interchangeability of the physiological data of each two dimensions according to the difference of the physiological data of each two dimensions of different reference patients and the difference of the decision correlation values, and screen the preferred dimensions by combining all the decision correlation values and all the interchangeabilities.

[0119] The category merging module is configured to screen similar patients of each category among other categories based on the differences between the physiological data of the preferred dimensions of the reference patients of different categories, wherein the wavelengths of all the reference patients in the same category are the same; and to perform merging processing on different categories to obtain several groups according to the differences of the wavelengths corresponding to each two categories and the number distribution of the similar patients of the reference patients in each two categories.

[0120] The blood glucose detection module is configured to determine a target wavelength by combining all the wavelengths corresponding to each group, the physiological data of the preferred dimensions of the reference patients in each group, and the physiological data of the preferred dimensions of the diabetes patient to be detected, and to detect the blood glucose of the diabetes patient to be detected by using the target wavelength.

[0121] It should be understood that, Figure 2 It should be understood that, The system and the modules thereof can be implemented in various ways. For example, in some embodiments, the system and the modules thereof can be implemented in hardware, software, or a combination of software and hardware. The hardware portion can be implemented with dedicated logic; the software portion can be stored in a memory and executed by a suitable instruction execution system, such as a microprocessor or a dedicated design hardware. Those skilled in the art can understand that the above-mentioned method and system can be implemented using computer executable instructions and / or included in processor control code, such as provided on a carrier medium, such as a disk, CD or DVD-ROM, a programmable memory, such as a read-only memory (firmware), or a data carrier, such as an optical or electronic signal carrier. The system and the modules thereof of the present specification can not only be implemented in hardware circuitry, such as very large scale integrated circuits or gate arrays, semiconductors, such as logic chips, transistors, or programmable hardware devices, such as field programmable gate arrays, programmable logic devices, etc., but also in software, for example, executed by various types of processors, and also in a combination of the above-mentioned hardware circuitry and software (for example, firmware).

[0122] Further details of the above-mentioned modules can be referred to other places in the present specification, and will not be described here.

[0123] In other embodiments, an intelligent non-invasive blood glucose monitoring device suitable for diabetic patients is also provided, including a memory and a processor. The memory stores executable program code, and the processor calls and runs the executable program code from the memory, causing the device to execute the aforementioned intelligent non-invasive blood glucose monitoring method for diabetic patients. Specifically, the device may be a chip, component, or module. The chip may include a connected processor and memory; wherein the memory stores instructions, and when the processor calls and executes the instructions, the chip can execute the intelligent non-invasive blood glucose monitoring method for diabetic patients provided in the above embodiments.

[0124] In other embodiments, a computer program product is also provided, which, when run on a computer, causes the computer to perform the aforementioned related steps to implement the intelligent non-invasive blood glucose detection method for diabetic patients provided in the above embodiments.

[0125] In other embodiments, a computer-readable storage medium is also provided, which stores computer program code. When the computer program code is run on a computer, the computer performs the above-described method steps to implement the intelligent non-invasive blood glucose detection method for diabetic patients provided in the above embodiments.

[0126] The systems, electronic devices, computer program products, and computer-readable storage media provided are all used to execute the corresponding methods provided above. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects of the corresponding methods provided above, and will not be repeated here.

[0127] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent, non-invasive blood glucose detection method suitable for diabetic patients, characterized in that, The method includes the following steps: Acquire physiological data and NIR wavelengths from a predetermined number of diabetic reference patients across different dimensions; Based on the clustering and numerical distribution characteristics of the physiological data of all reference patients for each dimension, the decision-relevance value of the physiological data for each dimension is obtained. Specifically, this includes: calculating the difference in the number of all reference patients between each pair of categories and the difference in the average value of the physiological data of the candidate dimensions of the reference patients within each pair of categories; summing the absolute values ​​of all the said differences in number to obtain a first summation value; summing the absolute values ​​of the differences in all the said average values ​​to obtain a second summation value; and normalizing the product of the first summation value and the second summation value to obtain the decision-relevance value of the physiological data for the candidate dimensions; wherein the candidate dimension can be any dimension. The degree of interchangeability of physiological data in each pair of dimensions is evaluated by comprehensively considering the differences in physiological data and decision-related values ​​among different reference patients. This includes: calculating the first difference between physiological data in each dimension for each pair of reference patients; for any two dimensions: obtaining the first feature value of each pair of reference patients in any two dimensions based on the first difference corresponding to each pair of reference patients in those two dimensions; and obtaining the degree of interchangeability of physiological data in any two dimensions based on the differences between all first feature values ​​of all reference patients in those two dimensions and the differences between decision-related values ​​of physiological data in those two dimensions. The differences between all first feature values ​​and the differences between decision-related values ​​are both negatively correlated with the degree of interchangeability. Optimal dimensions are selected by combining all the aforementioned decision-related values ​​and all the aforementioned degree of interchangeability. Based on the differences in physiological data between preferred dimensions of reference patients in different categories, similar patients in other categories are screened for reference patients in each category, wherein all reference patients in the same category have the same wavelength; according to the differences in wavelengths corresponding to each pair of categories and the distribution of the number of similar patients in each pair of categories, different categories are merged to obtain several groups; The target wavelength is determined by combining all wavelengths corresponding to each group, the physiological data of the preferred dimension of the reference patients within each group, and the physiological data of the preferred dimension of the diabetic patients to be tested. Specifically, this includes: training a wavelength selection model using all wavelengths corresponding to each group and the physiological data of the preferred dimension of the reference patients within each group to obtain a trained wavelength selection model; inputting the physiological data of the preferred dimension of the diabetic patients to be tested into the trained wavelength selection model to obtain the target wavelength; and using the target wavelength to detect the blood glucose of the diabetic patients to be tested.

2. The intelligent non-invasive blood glucose detection method for diabetic patients according to claim 1, characterized in that, The step of obtaining the first feature value of each pair of reference patients in any two dimensions based on the first difference corresponding to each pair of reference patients in any two dimensions includes: Calculate the first sum between the first difference and the preset first adjustment parameter for each pair of reference patients in one of the two dimensions, and the second sum between the first difference and the preset first adjustment parameter for each pair of reference patients in the other of the two dimensions; wherein the preset first adjustment parameter is greater than 0; The ratio between the first sum and the second sum is determined as the first feature value corresponding to the two reference patients in any two dimensions.

3. The intelligent non-invasive blood glucose detection method for diabetic patients according to claim 1, characterized in that, The process of combining all the decision-related values ​​and all the exchange rates to select the preferred dimensions includes: The product of the degree of interchangeability between the candidate dimension and the physiological data of each related dimension and the decision relevance value of the physiological data of each related dimension of the candidate dimension is denoted as the second feature value of each related dimension of the candidate dimension; where the related dimension of the candidate dimension is the dimension whose degree of interchangeability with the candidate dimension is greater than a preset degree of interchangeability threshold. By combining the second feature values ​​of all relevant dimensions of the candidate dimension and the decision-related values ​​of the physiological data of the candidate dimension, the preference level of the candidate dimension is obtained. The second feature value is negatively correlated with the preference level, and the decision-related values ​​of the physiological data of the candidate dimension are positively correlated with the preference level. Optimal dimensions are selected based on the relative importance of all dimensions.

4. The intelligent non-invasive blood glucose detection method for diabetic patients according to claim 3, characterized in that, Optimal dimensions are selected based on the relative importance of all dimensions, including: Sort all dimensions in descending order of preference to obtain the dimension sequence; The first preset number of dimensions in the dimension sequence are determined as preferred dimensions.

5. The intelligent non-invasive blood glucose detection method for diabetic patients according to claim 1, characterized in that, The differences in physiological data between preferred dimensions of reference patients in different categories are used to filter similar patients in other categories for each category, including: Based on the differences between the physiological data of the first patient and the second patient in the same preferred dimension, a similarity value is obtained between the first patient and the second patient, and the differences between the physiological data in the same preferred dimension are negatively correlated with the similarity value; If the similarity value is greater than the preset similarity threshold, then the first patient and the second patient are determined to be similar patients. The first and second patients are two reference patients from different categories.

6. The intelligent non-invasive blood glucose detection method for diabetic patients according to claim 1, characterized in that, The process involves merging different categories based on the wavelength differences between each pair of categories and the distribution of similar patients among the reference patients in each pair of categories to obtain several groups, including: Calculate the third sum between the number of all similar patients in the second category of the reference patient in the first category and the number of all similar patients in the first category of the reference patient in the second category; obtain the fusion degree between any two categories based on the difference between the wavelengths corresponding to the first and second categories and the third sum, wherein the difference between the wavelengths is negatively correlated with the fusion degree and the third sum is positively correlated with the fusion degree; wherein the first category and the second category are any two categories from all categories; Based on the fusion degree of each pair of categories, the corresponding categories are merged to obtain several groups.

7. The intelligent non-invasive blood glucose detection method for diabetic patients according to claim 6, characterized in that, The process of merging the corresponding categories based on the fusion degree of each pair of categories yields several groups, including: All categories are sorted according to a preset order to obtain a category sequence, wherein the preset order is either from wavelength in ascending order or from wavelength in descending order; The two categories with the highest fusion degree are merged. The fusion degree of each category with any category in the merged category sequence is determined sequentially according to the order of the categories. If the fusion degree is greater than the preset fusion degree threshold, the corresponding category is merged with the merged category. If the fusion degree is less than or equal to the threshold, it is treated as a new category. Each category after all the merging is completed is treated as a group.

Images