Diabetes blood sugar intelligent monitoring method and system based on artificial intelligence and medium

Abstract

The invention belongs to the technical field of data processing, and particularly relates to an intelligent diabetes blood sugar monitoring method and system based on artificial intelligence and a medium, the method comprises the steps of collecting personal information of a diabetic patient, periodically monitoring physiological data of the diabetic patient, and converting the personal information; classifying the diabetic patients, analyzing the physiological data of different types of diabetic patients to obtain an analysis result, calculating a correlation score of each physiological index and the blood glucose level based on the analysis result, and obtaining a first physiological index based on the correlation score; training the learning data by using a machine learning algorithm to generate a prediction model of a corresponding type; the intelligent wearable device is used for measuring physiological data of a patient to be monitored, the prediction model of the corresponding type is obtained, the prediction model outputs the predicted blood glucose level, and whether the blood glucose level monitored by the intelligent wearable device is accurate or not is judged based on the predicted blood glucose level. The blood glucose monitoring accuracy of the diabetic patient can be improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing, and in particular relates to an artificial intelligence-based diabetes blood sugar intelligent monitoring method, system and medium. Background Art

[0002] Diabetes is a chronic disease that requires patients to monitor their blood sugar levels regularly to control the condition. Traditional blood sugar monitoring methods rely on fingertip blood sampling, which is not only inconvenient for patients, but also cannot continuously monitor blood sugar changes in real time. With the development of artificial intelligence technology, non-invasive smart wearable devices for monitoring blood sugar have emerged, but smart wearable devices have high requirements for wearing methods, and inaccurate wearing methods may lead to inaccurate blood sugar monitoring.

[0003] A similar prior art Chinese patent application with publication number CN107203700A provides a method and device based on continuous blood glucose monitoring, the method comprising: obtaining blood glucose data of at least two users, constructing a first blood glucose data matrix according to a preset data splicing method, obtaining multiple first blood glucose data input vectors and output data corresponding to each first blood glucose data input vector from the first blood glucose data matrix; using a machine learning algorithm, training a preset mathematical model according to multiple groups of first blood glucose data input vectors and output data corresponding to each first blood glucose data input vector, to obtain a blood glucose data prediction model; obtaining existing blood glucose data of a user to be predicted, constructing a second blood glucose data input vector according to a preset data splicing method; inputting the second blood glucose data input vector into the blood glucose data prediction model, and obtaining a predicted value of the blood glucose data of the user to be predicted after the current moment.

[0004] A similar prior art includes a Chinese patent application with publication number CN105653865A, which discloses a blood glucose monitoring method, device and system. The method includes: establishing a food nutrient database; collecting a user's dietary information; fitting a sugar metabolism curve according to the sugar metabolism process; calculating the amount of sugar ingested by the user based on the food nutrient database and the dietary information; calculating the user's sugar change information based on the user's sugar ingestion and the sugar metabolism curve; obtaining the user's historical blood glucose data; and fitting a blood glucose change curve based on the historical blood glucose data and the sugar change information.

[0005] However, the above two documents do not consider the problem that the monitored blood sugar may be inaccurate. Therefore, the present invention provides an artificial intelligence-based diabetes blood sugar intelligent monitoring method, system and medium. Summary of the invention

[0006] The present invention collects physiological data of a large number of diabetic patients, trains a prediction model based on the physiological data to predict blood sugar levels based on physiological indicators, and judges the blood sugar level monitored by the smart wearable device based on the prediction model to ensure that the correct blood sugar level is monitored.

[0007] In order to achieve the above-mentioned purpose of the invention, the present invention provides the following artificial intelligence-based diabetes blood sugar intelligent monitoring method, which is implemented by executing the following steps:

[0008] S1. Collect personal information of diabetic patients, including height, weight, age and medical history data, periodically monitor physiological data of diabetic patients, including at least blood sugar level, heart rate, blood pressure, skin temperature, skin humidity and monitoring time, and convert the personal information;

[0009] S2. Classifying the diabetic patients based on the converted personal information to obtain a plurality of diabetic patients of different types, analyzing the physiological data of the diabetic patients of different types to obtain analysis results, calculating the correlation score between each physiological index and the blood glucose level based on the analysis results, and obtaining a first physiological index whose correlation score is greater than a preset first threshold;

[0010] S3, taking the first physiological indicator of each type of diabetic patient as input data, taking the blood glucose level detected simultaneously with the first physiological indicator as output data, combining the input data and the output data to generate learning data, and using a machine learning algorithm to train the learning data to generate a prediction model of a corresponding type;

[0011] S4. Use a smart wearable device to measure the physiological data of the patient to be monitored, and also collect the personal information of the patient to be monitored, obtain the type of the patient to be monitored based on the personal information, extract the first physiological indicator from the physiological data and input it into the prediction model of the corresponding type, output the predicted blood glucose level by the prediction model, and judge whether the blood glucose level monitored by the smart wearable device is accurate based on the predicted blood glucose level.

[0012] As a preferred technical solution of the present invention, the personal information is converted and processed, and the following steps are further performed:

[0013] S11, obtaining a data type of each data parameter in the personal information, wherein the data type includes a numeric type and a text type, and obtaining a first data parameter whose data type is text data;

[0014] S12, obtaining the first data parameter with the same semantic type as the second data parameter, obtaining the data value of the second data parameter, obtaining the data amount of the different data values, initializing the first code to zero, the number of bits of the first code to the data amount, and sorting each of the data values ​​in order to correspond one by one to each bit of the first code;

[0015] S13. Obtain a data value of the second data parameter, modify the number of bits corresponding to the data value in the first code to one, convert the modified first code into decimal data, and use the decimal data as a substitute value for the data value.

[0016] As a preferred technical solution of the present invention, classifying the diabetic patients based on the standardized personal information includes the following steps:

[0017] S211, presetting a plurality of different classification numbers, classifying the personal information multiple times, obtaining a plurality of different classification results, and calculating a corresponding classification score for each of the classification results;

[0018] S212: performing a subtraction operation on each of the classification scores and a preset first value to obtain a corresponding absolute value of the difference, obtaining a minimum value among the absolute values ​​of the difference, and taking the classification result corresponding to the minimum value as the final classification result.

[0019] As a preferred technical solution of the present invention, calculating the corresponding classification score for each of the classification results includes the following steps:

[0020] S2111. Calculate a first distance corresponding to each piece of data information in the personal information, where the first distance refers to the distance between the data information and other pieces of data information of the same type as the data information, and calculate a first average distance corresponding to all the first distances for each piece of data information;

[0021] S2112. For each type of the personal information, calculate the inter-type distance between each type and other types, and take the type with the smallest inter-type distance as the adjacent type of each type;

[0022] S2113. Calculate a corresponding second distance for each piece of data information in the personal information, where the second distance refers to the distance between the data information and the adjacent type of data information, and calculate a second average distance of multiple second distances corresponding to each piece of data information;

[0023] S2114: Divide a result obtained by subtracting the first average distance from the second average distance by the second average distance to obtain a first score;

[0024] S2115. Calculate an average value of the first scores corresponding to all the data information in the personal information, and use the average value as a classification score corresponding to the classification result.

[0025] As a preferred technical solution of the present invention, calculating the correlation score between each physiological index and the blood glucose level comprises the following steps:

[0026] S221, arbitrarily selecting a diabetic patient from all diabetic patients as a first patient, selecting any monitoring time from the physiological data of the first patient as a first monitoring time, and acquiring the physiological data corresponding to the first monitoring time as first physiological data;

[0027] S222, obtaining a plurality of diabetic patients of the same type as the first patient as second patients, obtaining second physiological data of the second patients, calling each physiological indicator except the blood glucose level in the physiological data a characteristic physiological indicator, and setting an initial correlation score of each characteristic physiological indicator of each type of patient to zero;

[0028] S223, taking the type to which the first patient belongs as the first type, arbitrarily selecting one diabetic patient from each other type as the third patient, and obtaining third physiological data corresponding to the first monitoring time from the physiological data of each third patient;

[0029] S224. Calculate a correlation score of each of the characteristic physiological indicators based on the first physiological data, the second physiological data, and the third physiological data.

[0030] As a preferred technical solution of the present invention, calculating the correlation score of each of the first physiological indicators comprises the following steps:

[0031] Calculate the first difference value of each characteristic physiological indicator between the second physiological data and the first physiological data, and also calculate the second difference value of each characteristic physiological indicator between the third physiological data and the first physiological data. For each characteristic physiological indicator, calculate the first average value of multiple first difference values. If the first average value is less than the second difference value, increase the correlation score corresponding to the characteristic physiological indicator corresponding to the corresponding type of patient by a preset second value. If the first average value is less than the second difference value, the correlation score remains unchanged. Otherwise, increase the preset third value, and the third value is greater than the second value.

[0032] As a preferred technical solution of the present invention, using a machine learning algorithm to train the learning data to generate a corresponding type of prediction model includes the following steps:

[0033] S31, analyzing the complexity of the relationship between the input data and the output data in the learning data, and selecting one from a plurality of machine learning algorithms based on the complexity of the relationship;

[0034] S32, dividing the learning data into a first data set and a second data set, and using the first data set to train the selected machine learning algorithm;

[0035] S32. During the training process, the prediction model is tested using the second data set, and model parameters are adjusted based on the test results to obtain the final prediction model.

[0036] As a preferred technical solution of the present invention, adjusting the model parameters based on the test results includes the following steps:

[0037] S321, setting a corresponding optimization coefficient for the model parameter, setting a selection range for the optimization coefficient, and setting a plurality of different optimization coefficient combinations within the selection range;

[0038] S322: adjusting the model parameters based on each of the optimization coefficient combinations, calculating the corresponding prediction error for the prediction model after adjusting the model parameters, and taking the prediction model corresponding to the minimum prediction error as the final prediction model

[0039] The present invention also provides an artificial intelligence-based diabetes blood sugar intelligent monitoring system, comprising the following modules:

[0040] A conversion unit, used to collect personal information of a diabetic patient, the personal information including height, weight, age and medical history data, periodically monitor physiological data of the diabetic patient, the physiological data including at least blood sugar level, heart rate, blood pressure, skin temperature, skin humidity and monitoring time, and convert the personal information;

[0041] a classification unit, configured to classify the diabetic patients based on the processed personal information, obtain a plurality of diabetic patients of different types, analyze the physiological data of different types to obtain analysis results, calculate a correlation score between each physiological index and the blood glucose level based on the analysis results, and obtain a first physiological index whose correlation score is greater than a preset first threshold;

[0042] A prediction unit, configured to use the first physiological indicator of each type of diabetic patient as input data, use the blood glucose level detected simultaneously with the first physiological indicator as output data, combine the input data and the output data to generate learning data, and use a machine learning algorithm to train the learning data to generate a prediction model of a corresponding type;

[0043] A judgment unit is used to use a smart wearable device to measure the physiological data of the patient to be monitored, and also collect the personal information of the patient to be monitored, obtain the type of the patient to be monitored based on the personal information, extract the first physiological indicator from the physiological data and input the prediction model of the corresponding type, output the predicted blood sugar level by the prediction model, and judge whether the blood sugar level monitored by the smart wearable device is accurate based on the predicted blood sugar level.

[0044] The present invention also provides a storage medium, which stores program instructions, wherein when the program instructions are executed, the device where the storage medium is located is controlled to execute any of the above-mentioned artificial intelligence-based intelligent diabetes blood glucose monitoring methods.

[0045] Compared with the prior art, the beneficial effects of the present invention are at least as follows:

[0046] In the present invention, a large amount of personal data and physiological data of diabetic patients are collected, and the diabetic patients are classified multiple times based on the personal data to obtain different classification results. The corresponding classification scores are calculated for the different classification results to obtain the best classification result, and each type of diabetic patients is analyzed to obtain a first physiological indicator that contributes greatly to the prediction of blood glucose levels. Learning data is generated based on the first physiological indicator and the blood glucose level, and a prediction model corresponding to each type is trained based on the learning data. In the process of training the prediction model, the model parameters are continuously adjusted based on the optimization coefficient to avoid overfitting of the prediction model, thereby improving the prediction accuracy of the prediction model. Based on the blood glucose level predicted by the prediction model, it is judged whether the blood glucose level monitored by the smart wearable device is correct. The present invention can ensure that an accurate blood glucose level is detected. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A flowchart of the steps of the artificial intelligence-based diabetes blood sugar intelligent monitoring method of the present invention;

[0048] Figure 2 This is a structural diagram of the composition of the artificial intelligence-based diabetes blood sugar intelligent monitoring system of the present invention. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0050] It is understood that the terms "first", "second", etc. used in this application may be used herein to describe various elements, but unless otherwise specified, these elements are not limited by these terms. These terms are only used to distinguish a first element from another element. For example, without departing from the scope of this application, a first xx script may be referred to as a second xx script, and similarly, a second xx script may be referred to as a first xx script.

[0051] With the continuous increase in the number of diabetic patients, blood glucose monitoring has become an indispensable part of the daily life of diabetic patients. In order to accurately monitor the blood glucose of diabetic patients, the prior art has realized the non-invasive monitoring of blood glucose levels using smart wearable devices. However, when using smart wearable devices to monitor blood glucose, monitoring errors may occur due to incorrect wearing of the smart wearable devices. Therefore, the present invention provides a method for determining whether the blood glucose level monitored by the smart wearable device is accurate.

[0052] Thus, the present invention provides Figure 1 The artificial intelligence-based diabetes blood sugar intelligent monitoring method shown is implemented by executing the following steps:

[0053] S1. Collect personal information of diabetic patients, including height, weight, age and medical history data, periodically monitor physiological data of diabetic patients, including at least blood sugar level, heart rate, blood pressure, skin temperature, skin humidity and monitoring time, and convert personal information.

[0054] Specifically, a large amount of personal information of diabetic patients is collected, and the physiological data of diabetic patients are periodically monitored. In order to ensure the accuracy of the personal information and physiological data, the personal information and physiological data are measured and acquired using the hospital's standard equipment. After the personal information and physiological data are collected, in order to facilitate the subsequent use of the personal information, the personal information is converted and processed. For example, text-type data in the personal information is converted into digital information to facilitate the subsequent classification of the personal information. The specific steps for converting the personal information will be explained in detail later.

[0055] S2. Classify the diabetic patients based on the converted personal information, obtain a plurality of diabetic patients of different types, analyze the physiological data of the diabetic patients of different types to obtain analysis results, calculate the correlation score between each physiological indicator and the blood glucose level based on the analysis results, and obtain the first physiological indicator whose correlation score is greater than a preset first threshold.

[0056] Specifically, different diabetic patients may have different personal circumstances, so diabetic patients are first classified based on personal information, and similar diabetic patients are classified into one category. Subsequently, the prediction model is trained based on the physiological data of each type of diabetic patient to make the prediction model more targeted and at the same time make the prediction accuracy higher. After obtaining different types of diabetic patients, since we have obtained a lot of physiological data, some physiological indicators may not have much effect on the prediction of blood sugar levels. If these physiological indicators are also used as learning data for training the prediction model, the efficiency of training the prediction model will be reduced, and the prediction accuracy of the prediction model will also be reduced. Therefore, by calculating the correlation score between the physiological indicator and the blood sugar level, a first physiological indicator with a correlation score greater than the first threshold is obtained. The first physiological indicator refers to a physiological indicator that has a strong correlation with the blood sugar level. The specific method of calculating the correlation score will be explained in detail later.

[0057] S3. Use the first physiological indicator of each type of diabetic patient as input data, and the blood glucose level detected simultaneously with the first physiological indicator as output data, combine the input data and the output data to generate learning data, and use a machine learning algorithm to train the learning data to generate a corresponding type of prediction model.

[0058] Specifically, in order to train a prediction model that can accurately predict blood sugar levels, the first physiological indicator of each type of diabetic patient that is strongly correlated with the blood sugar level is used as input data, and the blood sugar level that is detected simultaneously with the first physiological indicator, that is, the same detection time is used as output data, and the input data and the output data are combined to generate learning data. The learning data is trained using a machine learning algorithm to generate a corresponding type of prediction model. The specific training method will be explained in detail later. Subsequently, the physiological data detected in real time can be input into the prediction model, and the prediction model outputs the corresponding predicted blood sugar level. Based on the predicted blood sugar level, it is determined whether the blood sugar level detected using the smart wearable device is correct.

[0059] S4. Use a smart wearable device to measure the physiological data of the patient to be monitored, and also collect the personal information of the patient to be monitored, obtain the type of the patient to be monitored based on the personal information, extract the first physiological indicator from the physiological data and input it into a prediction model of the corresponding type, and output the predicted blood glucose level from the prediction model, and judge whether the blood glucose level monitored by the smart wearable device is accurate based on the predicted blood glucose level.

[0060] Specifically, a smart wearable device is used to measure the physiological data of the patient to be monitored, and the personal information of the patient to be monitored is collected. The personal information here can be collected in advance before using the smart wearable device. The type of the patient to be monitored is determined based on the personal information, and then a first physiological indicator is extracted from the physiological data, and the first physiological indicator is input into a prediction model. The prediction model outputs a predicted blood glucose level, and based on the predicted blood glucose level, it is determined whether the blood glucose level monitored by the smart wearable device is accurate. For example, a difference value can be preset, and the predicted blood glucose level and the blood glucose level monitored by the smart wearable device are subtracted to obtain the absolute value of the corresponding difference value. If the absolute value is less than the difference, it means that the blood glucose level monitored by the smart wearable device is accurate. If the absolute value is greater than the difference, it means that the blood glucose level monitored by the smart wearable device is inaccurate. At this time, the smart wearable device can be notified to make an abnormal sound to prompt the patient to be monitored to remeasure.

[0061] Furthermore, the personal information is converted and processed, including the following steps:

[0062] S11, obtaining a data type of each data parameter in the personal information, the data type including a numeric type and a text type, and obtaining a first data parameter whose data type is text data;

[0063] S12, obtaining a first data parameter with the same semantic type as the second data parameter, obtaining a data value of the second data parameter, obtaining a data amount of different data values, initializing the first code to zero, the number of bits of the first code to be the data amount, and sorting each data value in order to correspond one to one with each bit of the first code;

[0064] S13. Obtain a data value of a second data parameter, modify the number of bits corresponding to the data value in the first code to 1, convert the modified first code into decimal data, and use the decimal data as a substitute value for the data value.

[0065] Specifically, since some data in personal information, such as gender and medical record data, may be text-type data, when classifying personal information, some are digital-type data and some are text-type data, which are not convenient to classify. Therefore, the text data in the personal information is converted and processed. First, the first data parameter of the text data is obtained. For example, the first data parameters of the text data include male, female, type 1 diabetes, type 2 diabetes, gestational diabetes, special diabetes, etc. Obtaining the first data parameter with the same semantic type means that male and female are data parameters belonging to the semantic type of gender, type 1 diabetes, type 2 diabetes, gestational diabetes, and special diabetes are data parameters belonging to the semantic type of diabetes type. Obtaining the first data parameter with the same semantic type, for example, gender male and gender female are two second data parameters, diabetes type is type 1 diabetes, and diabetes type is type 2 diabetes. , diabetes type is gestational diabetes, and diabetes type is special diabetes are four second data parameters. Taking the semantic type of diabetes type as an example, there are 4 different parameter values. You can initialize a first code with 4 bits, sort the 4 types of diabetes and correspond one-to-one with the 4 bits of the first code, and obtain the data value of the diabetes type of the second data parameter. For example, if the data value is type 1 diabetes, the corresponding data value is set to 1000, if the data value is type 2 diabetes, the corresponding data value is set to 0100, if the data value is gestational diabetes, the corresponding data value is set to 0010, and if the data value is special diabetes, the corresponding data value is set to 0001. Then, the data value is converted into decimal data, and the decimal data replaces the original text data. Through the above method, the text type data is converted into digital type data, which is convenient for the subsequent classification of personal information.

[0066] Furthermore, classifying the diabetic patients based on the standardized personal information includes the following steps:

[0067] S211, presetting a plurality of different classification numbers, classifying the personal information multiple times, obtaining a plurality of different classification results, and calculating a corresponding classification score for each classification result;

[0068] S212: Perform a subtraction operation on each classification score and a preset first value to obtain a corresponding absolute value of the difference, obtain a minimum value among the absolute values ​​of the difference, and use the classification result corresponding to the minimum value as the final classification result.

[0069] Specifically, in order to accurately classify personal information, firstly, a plurality of different classification numbers are preset, for example, the preset classification numbers are 3, 4 and 5, and classification algorithms such as K-means can be used to classify personal information into 3 types, 4 types and 5 types, and 3 different classification results are obtained. A corresponding classification score is also calculated for each classification result. The classification score represents the accuracy of the classification. If the personal information is divided into too many types, personal information with great similarity may belong to different types respectively. If the number of types of personal information is too small, personal information with little similarity may be classified into the same type. Therefore, the accuracy of the classification is judged by calculating the classification score. The specific calculation method will be explained in detail later. Then, the classification score of each classification result is subtracted from the preset first value to obtain the corresponding absolute value of the difference. The absolute value of the difference refers to the absolute value of the difference. The smaller the absolute value of the difference, the better the classification effect of the current classification result and the higher the classification accuracy. Therefore, the classification result corresponding to the minimum absolute value of the difference is taken as the final classification result.

[0070] Furthermore, the corresponding classification score is calculated for each classification result, including the following steps:

[0071] S2111. Calculate a first distance corresponding to each piece of data information in the personal information, where the first distance refers to the distance between other pieces of data information of the same type as the data information, and calculate a first average distance for all first distances corresponding to each piece of data information;

[0072] S2112. For each type of personal information, calculate the inter-type distance between each type and other types, and take the type with the smallest inter-type distance as the adjacent type of each type;

[0073] S2113. Calculate a corresponding second distance for each piece of data information in the personal information, where the second distance refers to the distance between the data information and adjacent types of data information, and calculate a second average distance of multiple second distances corresponding to each piece of data information;

[0074] S2114, dividing the result obtained by subtracting the first average distance from the second average distance by the second average distance as a first score;

[0075] S2115. Calculate the average of the first scores corresponding to all data information in the personal information, and use the average as the classification score of the corresponding classification result.

[0076] Specifically, for each classification result, taking the classification into three types as an example, for each data information in each type, the first distance is calculated. The first distance refers to the distance between other data information and data information belonging to the same type as the data information. For example, the three types of data are A: {a1, a2, a3, a4}, B: {b1, b2, b3, b4, b5}, C: {c1, c2, c3}. For data information a1, the first distance between a1 and a2, a3, and a4 is calculated, and then the average distance of the three first distances is calculated as the first average distance. The smaller the first average distance, the higher the aggregation degree of the data within the type, and the better the classification effect. The inter-type distance between A and B and A and C is also calculated. Based on the inter-type distance, the type closest to A is obtained as the adjacent type of A. Assuming that the adjacent type of A is B, the second distance of each data information in types a1 and B is calculated, and the second average distance of the second distance is calculated. The second average distance represents the inter-class distance. The larger the inter-class distance, the greater the difference in data between types, and the better the classification effect. The second average distance minus the first average distance represents the difference between the distance between each data information and its nearest classification and the distance between the same type. A positive value represents a center closer to its own type than to the center of the adjacent type. The result value obtained by subtraction is divided by the second average distance as the first score, ensuring that the first score is between -1 and 1, so that the first score is not affected by the distance dimension and can be compared across different types of data. The above method is used to calculate the first score of each data information, and then the average of all first scores is calculated. The average value is used as the classification score of the corresponding classification results divided into 3 types. The classification score is the average of the first scores of all data information, taking into account the classification quality of each data information in the personal information. By calculating the average value, a single value is obtained to quantify the classification effect of the personal information. The closer this classification score is to 1, the better the classification effect, so the above first value is set to 1.

[0077] It should be noted that the distance D between two pieces of data information can be calculated using the following formula: where x ai refers to the i-th data parameter in data information a1, x bi refers to the i-th data parameter in the data information b1, and n refers to the number of data parameters included in the data information.

[0078] It should also be noted that to calculate the inter-type distance between each type and other types, you can first calculate the average value of each data parameter in the data information of each type, combine the average values ​​of each data parameter as the data center of each type, and then use the above formula to calculate the distance between the two data centers as the inter-type distance.

[0079] Furthermore, calculating the correlation score between each physiological index and blood sugar level includes the following steps:

[0080] S221, arbitrarily selecting a diabetic patient from all diabetic patients as a first patient, selecting any monitoring time from the physiological data of the first patient as a first monitoring time, and acquiring physiological data corresponding to the first monitoring time as first physiological data;

[0081] S222, obtaining a plurality of diabetic patients of the same type as the first patient as second patients, obtaining second physiological data of the second patients, calling each physiological indicator except the blood glucose level in the physiological data a characteristic physiological indicator, and setting an initial correlation score of each characteristic physiological indicator of each type of patient to zero;

[0082] S223, taking the type to which the first patient belongs as the first type, arbitrarily selecting one diabetic patient from each other type as a third patient, and obtaining third physiological data corresponding to the first monitoring time from the physiological data of each third patient;

[0083] S224: Calculate a correlation score of each characteristic physiological index based on the first physiological data, the second physiological data, and the third physiological data.

[0084] Specifically, as shown in Table 1 below, in the form of physiological data, using physiological indicators that are strongly correlated with blood sugar levels as input data can not only improve the training efficiency of the model but also improve the prediction accuracy of the model. In order to calculate physiological indicators that are strongly correlated with blood sugar levels, diabetic patients have been classified based on personal information. A diabetic patient is arbitrarily selected as the first patient. For example, patient E is selected as the first patient. Since the physiological data of each diabetic patient is measured periodically, a physiological data is arbitrarily selected as the first physiological data. Assume that physiological data 1 is selected as the first physiological data. Multiple physiological data of each patient correspond to different monitoring times. When calculating the correlation score, only the physiological data of one of the monitoring times needs to be used. Therefore, the monitoring time of this physiological data is obtained as The first monitoring time, the first monitoring time is monitoring time 1, and multiple diabetic patients of the same type as the first patient are taken as the second patients. The first patient belongs to type C, and other patients E, G, and H in type C are taken as the second patients. The physiological data 1 corresponding to the first monitoring time of patients E, G, and H are obtained as the second physiological data. Then, the difference value between patients of the same type can be calculated based on the first physiological data and the second physiological data. Patient A in type A is obtained as the third patient, and the physiological data 1 corresponding to the third patient is obtained as the third physiological data. The difference value between patients of different types is calculated based on the first physiological data and the second physiological data. The correlation score of each characteristic physiological indicator is calculated based on the first physiological data, the second physiological data, and the third physiological data. The specific calculation method will be explained in detail later.

[0085]

[0086] Table 1

[0087] Further, calculating the correlation score of each first physiological indicator includes the following steps:

[0088] The first difference value of each characteristic physiological indicator between the second physiological data and the first physiological data is calculated, and the second difference value of each characteristic physiological indicator between the third physiological data and the first physiological data is also calculated. For each characteristic physiological indicator, the first average value of multiple first difference values ​​is calculated. If the first average value is less than the second difference value, the correlation score corresponding to the characteristic physiological indicator corresponding to the corresponding type of patient is increased by a preset second value. If the first average value is less than the second difference value, the correlation score remains unchanged. Otherwise, the preset third value is increased, and the third value is greater than the second value.

[0089] Specifically, for each characteristic physiological indicator, the corresponding first difference value and second difference value are calculated, the first difference value is the difference value between the characteristic physiological indicators of patients of the same type, and the second difference value is the difference value between the characteristic physiological indicators of patients of different types. The first average value of the first difference is calculated. If the first average value is less than the second difference value, it means that the characteristic physiological indicator has significantly different values ​​between different patients who do not belong to the same patient type, but has similar values ​​between different patients of the same patient type, which means that the contribution of the characteristic growth indicator to the prediction of blood glucose levels of this type of patients may not be that great, so the correlation score corresponding to the corresponding characteristic growth indicator is added to the first value, and the first value can be set to 0.1. If the first average value is less than the second difference value, it means that the average difference of the characteristic physiological indicator between different patients of the same patient type is the same as the difference between different patients who do not belong to the same patient type, which means that the The contribution of the characteristic growth index to predicting the blood sugar level of this type of patients is almost the same as its contribution to the blood sugar level of other types of patients, so the correlation corresponding to the non-corresponding characteristic growth index needs to be changed numerically, otherwise it means that the average difference value of the growth index between the same type is greater than the difference value between different types of patients, which means that the contribution of the characteristic growth index to predicting the blood sugar level of this type of patients may be relatively large, so the correlation score corresponding to the corresponding characteristic growth index is added to the second value, and the second value can be set to 0.3 or 0.5. In this embodiment, there are a total of 3 patient types, and there are two corresponding second difference values. The above method is also used to determine the size of the average difference value and other second difference values, and modify the corresponding correlation score. Through the above method, the correlation score of each characteristic physiological indicator corresponding to different types of patients can be quickly calculated, which is convenient for subsequent acquisition of physiological indicators with strong correlation with blood sugar levels through correlation scores.

[0090] Furthermore, the learning data is trained using a machine learning algorithm to generate a corresponding type of prediction model, including the following steps:

[0091] S31, analyzing the complexity of the relationship between input data and output data in the learning data, and selecting one from multiple machine learning algorithms based on the complexity of the relationship;

[0092] S32, dividing the learning data into a first data set and a second data set, and using the first data set to train the selected machine learning algorithm;

[0093] S32. During the training process, the prediction model is tested using the second data set, and the model parameters are adjusted based on the test results to obtain a final prediction model.

[0094] Specifically, since there are many machine learning algorithms, such as decision trees, random forest models or neural network models, different analysis methods such as linear relationships and nonlinear relationships are used to analyze the relationship complexity between input data and output data, and a neural network model is selected based on the relationship complexity to train the first data set. The first data set is a training set, and the second data set is a test set. During the training process, in order to avoid overfitting, the model parameters are adjusted until a prediction model with the smallest prediction error is obtained. The specific method of adjusting the model parameters will be explained in detail later.

[0095] Further, the model parameters are adjusted based on the test results, including the following steps:

[0096] S321, setting corresponding optimization coefficients for model parameters, setting a selection range for the optimization coefficients, and setting a plurality of different optimization coefficient combinations within the selection range;

[0097] S322: Adjust the model parameters based on each optimization coefficient combination, calculate the corresponding prediction error for the prediction model after adjusting the model parameters, and use the prediction model corresponding to the minimum prediction error as the final prediction model

[0098] Specifically, an optimization coefficient is preset, and a selection range is set for the optimization coefficient. For example, if the optimization coefficient includes a first coefficient and a second coefficient, corresponding minimum and maximum values ​​are set for the first coefficient and the second coefficient, and a plurality of optimization coefficient combinations of different first coefficients and second coefficients are obtained within the range of the minimum and maximum values. The model parameters are adjusted based on each optimization coefficient combination, and for the prediction model after the model parameters are adjusted, the prediction error is calculated based on the second data set, and the prediction model corresponding to the minimum prediction error is obtained as the final prediction model. Through the above method, the model parameters can be adjusted based on the optimization coefficients to prevent the model from overfitting, thereby improving the prediction accuracy of the prediction model.

[0099] According to another aspect of the embodiment of the present invention, Figure 2 As shown, a diabetes blood sugar intelligent monitoring system based on artificial intelligence is also provided, including a collection unit, an extraction unit, a classification unit and a storage unit, which are used to implement the diabetes blood sugar intelligent monitoring method based on artificial intelligence as described above. The specific functions of each module are as follows:

[0100] A conversion unit is used to collect personal information of diabetic patients, including height, weight, age and medical history data, periodically monitor physiological data of diabetic patients, including at least blood sugar level, heart rate, blood pressure, skin temperature, skin humidity and monitoring time, and convert the personal information;

[0101] a classification unit, configured to classify the diabetic patients based on the processed personal information, obtain a plurality of diabetic patients of different types, analyze the physiological data of different types to obtain analysis results, calculate a correlation score between each physiological index and the blood glucose level based on the analysis results, and obtain a first physiological index having a correlation score greater than a preset first threshold;

[0102] A prediction unit, configured to use a first physiological indicator of each type of diabetic patient as input data, a blood glucose level detected simultaneously with the first physiological indicator as output data, combine the input data and the output data to generate learning data, and use a machine learning algorithm to train the learning data to generate a prediction model of a corresponding type;

[0103] The judgment unit is used to use the smart wearable device to measure the physiological data of the patient to be monitored, and also collect the personal information of the patient to be monitored, obtain the type of the patient to be monitored based on the personal information, extract the first physiological indicator from the physiological data and input it into the corresponding type of prediction model, the prediction model outputs the predicted blood sugar level, and judges whether the blood sugar level monitored by the smart wearable device is accurate based on the predicted blood sugar level.

[0104] According to another aspect of an embodiment of the present invention, a storage medium is also provided, which stores program instructions, wherein when the program instructions are executed, the device where the storage medium is located is controlled to execute any of the above-mentioned artificial intelligence-based intelligent diabetes blood glucose monitoring methods.

[0105] In summary, the present invention proposes an artificial intelligence-based diabetes blood sugar intelligent monitoring method, system and medium, the method includes collecting personal information of diabetic patients, periodically monitoring the physiological data of diabetic patients, converting and processing the personal information; classifying diabetic patients, analyzing the physiological data of different types of diabetic patients to obtain analysis results, calculating the correlation scores between each physiological index and blood sugar level based on the analysis results, and obtaining the first physiological index based on the correlation score; using a machine learning algorithm to train the learning data to generate a corresponding type of prediction model; using a smart wearable device to measure the physiological data of the patient to be monitored, obtaining a corresponding type of prediction model, outputting a predicted blood sugar level from the prediction model, and judging whether the blood sugar level monitored by the smart wearable device is accurate based on the predicted blood sugar level. The present invention can ensure the accuracy of blood sugar monitoring of diabetic patients.

[0106] It should be understood that, although each step in the flow chart of each embodiment of the present invention is shown in sequence according to the indication of the arrow, these steps are not necessarily performed in sequence according to the order indicated by the arrow. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0107] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The above-mentioned program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0108] The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0109] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

[0110] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An artificial intelligence-based intelligent blood sugar monitoring method for diabetes, characterized in that: The steps include: S1. Collect personal information of diabetic patients, including height, weight, age and medical history data, periodically monitor physiological data of diabetic patients, including at least blood sugar level, heart rate, blood pressure, skin temperature, skin humidity and monitoring time, and convert the personal information; S2. Classifying the diabetic patients based on the converted personal information to obtain a plurality of diabetic patients of different types, analyzing the physiological data of the diabetic patients of different types to obtain analysis results, calculating the correlation score between each physiological index and the blood glucose level based on the analysis results, and obtaining a first physiological index whose correlation score is greater than a preset first threshold; S3, taking the first physiological indicator of each type of diabetic patient as input data, taking the blood glucose level detected simultaneously with the first physiological indicator as output data, combining the input data and the output data to generate learning data, and using a machine learning algorithm to train the learning data to generate a prediction model of a corresponding type; S4. Use a smart wearable device to measure the physiological data of the patient to be monitored, and also collect the personal information of the patient to be monitored, obtain the type of the patient to be monitored based on the personal information, extract the first physiological indicator from the physiological data and input it into the prediction model of the corresponding type, and output the predicted blood glucose level by the prediction model, and judge whether the blood glucose level monitored by the smart wearable device is accurate based on the predicted blood glucose level.

2. The method according to claim 1, characterized in that: The personal information is converted and processed, including the following steps: S11, obtaining a data type of each data parameter in the personal information, the data type including a numeric type and a text type, and obtaining a first data parameter whose data type is text data; S12, obtaining the first data parameter with the same semantic type as the second data parameter, obtaining the data value of the second data parameter, obtaining the data amount of the different data values, initializing the first code to zero, the number of bits of the first code to the data amount, and sorting each of the data values ​​in order to correspond one by one to each bit of the first code; S13. Obtain a data value of the second data parameter, modify the number of bits corresponding to the data value in the first code to one, convert the modified first code into decimal data, and use the decimal data as a substitute value for the data value.

3. The method according to claim 1, characterized in that Classifying the diabetic patients based on the standardized personal information includes the following steps: S211, presetting a plurality of different classification numbers, classifying the personal information multiple times, obtaining a plurality of different classification results, and calculating a corresponding classification score for each of the classification results; S212: performing a subtraction operation on each of the classification scores and a preset first value to obtain a corresponding absolute value of the difference, obtaining a minimum value among the absolute values ​​of the difference, and taking the classification result corresponding to the minimum value as the final classification result.

4. The method according to claim 3, characterized in that Calculating the corresponding classification score for each of the classification results includes the following steps: S2111. Calculate a first distance corresponding to each piece of data information in the personal information, where the first distance refers to the distance between the data information and other pieces of data information of the same type as the data information, and calculate a first average distance corresponding to all the first distances for each piece of data information; S2112. For each type of the personal information, calculate the inter-type distance between each type and other types, and take the type with the smallest inter-type distance as the adjacent type of each type; S2113. Calculate a corresponding second distance for each piece of data information in the personal information, where the second distance refers to the distance between the data information and the adjacent type of data information, and calculate a second average distance of multiple second distances corresponding to each piece of data information; S2114: Divide a result obtained by subtracting the first average distance from the second average distance by the second average distance to obtain a first score; S2115. Calculate an average value of the first scores corresponding to all the data information in the personal information, and use the average value as a classification score corresponding to the classification result.

5. The method according to claim 1, characterized in that Calculating the correlation score between each physiological index and the blood sugar level comprises the following steps: S221, arbitrarily selecting a diabetic patient from all diabetic patients as a first patient, selecting any monitoring time from the physiological data of the first patient as a first monitoring time, and acquiring the physiological data corresponding to the first monitoring time as first physiological data; S222, obtaining a plurality of diabetic patients of the same type as the first patient as second patients, obtaining second physiological data of the second patients, calling each physiological indicator except the blood glucose level in the physiological data a characteristic physiological indicator, and setting an initial correlation score of each characteristic physiological indicator of each type of patient to zero; S223, taking the type to which the first patient belongs as the first type, arbitrarily selecting one diabetic patient from each other type as the third patient, and obtaining third physiological data corresponding to the first monitoring time from the physiological data of each third patient; S224. Calculate a correlation score of each of the characteristic physiological indicators based on the first physiological data, the second physiological data, and the third physiological data.

6. The method according to claim 5, characterized in that Calculating the correlation score of each of the first physiological indicators comprises the following steps: Calculate the first difference value of each characteristic physiological indicator between the second physiological data and the first physiological data, and also calculate the second difference value of each characteristic physiological indicator between the third physiological data and the first physiological data. For each characteristic physiological indicator, calculate the first average value of multiple first difference values. If the first average value is less than the second difference value, increase the correlation score corresponding to the characteristic physiological indicator corresponding to the corresponding type of patient by a preset second value. If the first average value is less than the second difference value, the correlation score remains unchanged. Otherwise, increase the preset third value, and the third value is greater than the second value.

7. The method according to claim 1, characterized in that Using a machine learning algorithm to train the learning data to generate a corresponding type of prediction model includes the following steps: S31, analyzing the complexity of the relationship between the input data and the output data in the learning data, and selecting one from a plurality of machine learning algorithms based on the complexity of the relationship; S32, dividing the learning data into a first data set and a second data set, and using the first data set to train the selected machine learning algorithm; S32. During the training process, the prediction model is tested using the second data set, and model parameters are adjusted based on the test results to obtain the final prediction model.

8. The method according to claim 7, characterized in that Adjust the model parameters based on the test results, including the following steps: S321, setting a corresponding optimization coefficient for the model parameter, setting a selection range for the optimization coefficient, and setting a plurality of different optimization coefficient combinations within the selection range; S322. Adjust the model parameters based on each combination of the optimization coefficients, calculate the corresponding prediction error for the prediction model after the model parameters are adjusted, and use the prediction model corresponding to the minimum prediction error as the final prediction model.

9. An artificial intelligence-based diabetes blood sugar intelligent monitoring system, used to implement the artificial intelligence-based diabetes blood sugar intelligent monitoring method according to any one of claims 1 to 8, characterized in that: Includes the following modules: A conversion unit, used to collect personal information of a diabetic patient, the personal information including height, weight, age and medical history data, periodically monitor physiological data of the diabetic patient, the physiological data including at least blood sugar level, heart rate, blood pressure, skin temperature, skin humidity and monitoring time, and convert the personal information; a classification unit, configured to classify the diabetic patients based on the processed personal information, obtain a plurality of diabetic patients of different types, analyze the physiological data of different types to obtain analysis results, calculate a correlation score between each physiological index and the blood glucose level based on the analysis results, and obtain a first physiological index whose correlation score is greater than a preset first threshold; A prediction unit, configured to use the first physiological indicator of each type of diabetic patient as input data, use the blood glucose level detected simultaneously with the first physiological indicator as output data, combine the input data and the output data to generate learning data, and use a machine learning algorithm to train the learning data to generate a prediction model of a corresponding type; A judgment unit is used to use a smart wearable device to measure the physiological data of the patient to be monitored, and also collect the personal information of the patient to be monitored, obtain the type of the patient to be monitored based on the personal information, extract the first physiological indicator from the physiological data and input the prediction model of the corresponding type, output the predicted blood sugar level by the prediction model, and judge whether the blood sugar level monitored by the smart wearable device is accurate based on the predicted blood sugar level.

10. A storage medium, characterized in that: The storage medium stores program instructions, wherein when the program instructions are executed, the device where the storage medium is located is controlled to execute the artificial intelligence-based diabetes blood glucose intelligent monitoring method according to any one of claims 1 to 8.

Images