Sugar tolerance curve generation system and method

Abstract

The invention provides a sugar tolerance curve generation system and method, and the system comprises a monitoring data collection module which is used for obtaining multi-modal time series data which are collected by monitoring equipment and comprise diet logs; the feeding data interception module is used for intercepting a plurality of panel data sets under the multi-modal time sequence data based on the time of each feeding event in the diet log; the personalized model module is used for estimating to obtain a blood glucose concentration estimation sequence under a preset glucose metabolism kinetic model based on the plurality of panel data sets; based on the blood glucose concentration estimation sequence and a blood glucose concentration sequence in a plurality of panel data sets, fitting under an iterative particle filter algorithm to obtain an optimization model parameter set; based on the optimization model parameter set, obtaining a personalized glucose metabolism model under a preset glucose metabolism kinetic model; and the glucose tolerance curve generation module is used for predicting the blood glucose concentration prediction sequence under the metabolic stimulation condition based on the personalized glucose metabolism model to generate a glucose tolerance curve. The reliability of the sugar tolerance curve can be improved.

Description

Technical Field

[0001] This invention belongs to the field of blood glucose detection technology, and in particular relates to a glucose tolerance curve generation system and method. Background Technology

[0002] In glucose tolerance tests, a pre-defined metabolic stimulus needs to be applied to the subject to obtain a glucose tolerance curve. Existing technologies rely on spectral data and complex deep network structures to predict blood glucose concentration and derive the glucose tolerance curve based on the predicted concentration. However, spectral data is easily affected by external factors such as ambient temperature and contact pressure, making the predicted blood glucose concentration data unstable and reducing the reliability of the glucose tolerance curve. Furthermore, existing technologies use random historical blood glucose sequences to fit and train a time-series model, and then use this model to predict blood glucose concentration. However, the blood glucose concentration prediction results obtained by this existing time-series model cannot simulate the specific blood glucose concentration changes under the application of a pre-defined metabolic stimulus. Therefore, the glucose tolerance curve generated based on its blood glucose concentration prediction results differs significantly from the actual situation, resulting in low reliability. Summary of the Invention

[0003] The present invention aims to provide a system and method for generating sugar tolerance curves to solve the above-mentioned technical problems and improve the reliability of sugar tolerance curves.

[0004] To address the aforementioned technical problems, this invention provides a sugar tolerance curve generation system, comprising: The monitoring data acquisition module is used to acquire multimodal time-series data collected by the monitoring equipment; the multimodal time-series data includes dietary logs. The food intake data extraction module is used to extract data from the multimodal time-series data obtained by the monitoring data acquisition module based on the time corresponding to each food intake event in the food log obtained by the monitoring data acquisition module, and to obtain several panel datasets. The personalized model module is used to estimate blood glucose concentration based on several panel datasets obtained by the food data extraction module under a preset glucose metabolism kinetic model, thereby obtaining a blood glucose concentration estimation sequence; based on the blood glucose concentration estimation sequence and the blood glucose concentration sequences in several panel datasets, it is fitted using an iterative particle filter algorithm to obtain an optimized model parameter set; based on the optimized model parameter set, a personalized glucose metabolism model is obtained under the preset glucose metabolism kinetic model. The glucose tolerance curve generation module is used to perform predictive processing on the personalized glucose metabolism model obtained by the personalized model module under preset metabolic stimulation conditions to obtain a blood glucose concentration prediction sequence, and generate a glucose tolerance curve based on the blood glucose concentration prediction sequence.

[0005] In the above scheme, the collected multimodal time-series data is divided into several panel datasets according to the time corresponding to the eating event. Based on these panel datasets, estimation is performed under the glucose metabolism kinetic model to obtain optimized model parameters. A personalized glucose metabolism model is then constructed based on these optimized model parameters. This scheme divides the multimodal time-series data according to the time corresponding to the eating event, ensuring that each panel dataset corresponds to the multimodal time-series data change process under a metabolic stimulation condition. That is, each panel dataset in this scheme contains the change process of multimodal time-series data after the corresponding eating event, and effectively isolates the mutual interference between different eating events, making the consistency of the multimodal time-series data change process higher. This improves the stability of the obtained personalized glucose metabolism model and enables the obtained blood glucose concentration prediction sequence to adapt to the needs of glucose tolerance test scenarios, thereby improving the reliability of the glucose tolerance curve.

[0006] Furthermore, the food intake data extraction module is used to extract data from the multimodal time-series data obtained by the monitoring data acquisition module based on the time corresponding to each food intake event in the food log obtained by the monitoring data acquisition module, to obtain several panel datasets, including: obtaining several food intake monitoring time periods based on the time corresponding to each food intake event in the food log obtained by the monitoring data acquisition module under a preset monitoring duration; and extracting several panel datasets corresponding to several food intake monitoring time periods based on the multimodal time-series data obtained by the monitoring data acquisition module to obtain several panel datasets.

[0007] In the above scheme, the food intake data extraction module defines a fixed preset monitoring period based on the time of each food intake event recorded in the diet log, thus assigning a corresponding food intake monitoring time period to each food intake event. Subsequently, panel datasets corresponding to all the above-mentioned food intake monitoring time periods are accurately extracted from continuously collected multimodal time-series data. This scheme effectively isolates the mutual interference of metabolic responses between different food intake events by defining an independent food intake monitoring time period for each food intake event. This ensures that each panel dataset contains the dynamic change process of panel data corresponding to a complete food intake event, laying a data foundation for subsequent fitting of a reliable personalized glucose metabolism model, thereby improving the reliability of the glucose tolerance curve.

[0008] Furthermore, the glucose tolerance curve generation module is used to perform predictive processing on the personalized glucose metabolism model obtained by the personalized model module under preset metabolic stimulation conditions to obtain a blood glucose concentration prediction sequence, and generate a glucose tolerance curve based on the blood glucose concentration prediction sequence, including: performing predictive processing on the personalized glucose metabolism model under preset metabolic stimulation conditions to obtain a blood glucose concentration prediction sequence; obtaining a blood glucose concentration extraction sequence based on the blood glucose concentration prediction sequence under a preset extraction time set; and generating a glucose tolerance curve based on the blood glucose concentration extraction sequence using a spline interpolation algorithm.

[0009] In the above scheme, preset metabolic stimulation conditions are input into the personalized glucose metabolism model. By running the personalized glucose metabolism model, a blood glucose concentration prediction sequence is obtained. Subsequently, based on a preset extraction time set, the blood glucose concentration at each time corresponding to the preset extraction time set is extracted from the blood glucose concentration prediction sequence to form a blood glucose concentration extraction sequence. Finally, a spline interpolation algorithm is used to process the blood glucose concentration extraction sequence to generate a glucose tolerance curve. This scheme, by applying uniform preset metabolic stimulation conditions to the personalized glucose metabolism model, can fundamentally eliminate the interference of individual preparation state, activity level, and other factors on blood glucose concentration, ensuring that the generated glucose tolerance curve is obtained under a uniform benchmark. Furthermore, by processing the blood glucose concentration extracted at each time point using a spline interpolation algorithm, this scheme can eliminate the random fluctuations that may exist in the blood glucose concentration prediction sequence output by the personalized glucose metabolism model, generating a smoother glucose tolerance curve that better conforms to physiological changes, further improving the reliability of the glucose tolerance curve.

[0010] Furthermore, in the personalized module, the construction of the preset glucose metabolism kinetic model includes: acquiring time-series data of carbohydrate content, the time corresponding to the eating event, calorie expenditure during exercise, and blood glucose concentration; acquiring a preset monitoring duration to obtain a monitoring time series, and performing a blood glucose concentration prediction step based on each monitoring time in the monitoring time series to obtain several estimated blood glucose concentrations, and obtaining a blood glucose concentration estimation sequence based on the several estimated blood glucose concentrations; the blood glucose concentration prediction step includes: obtaining the gastrointestinal carbohydrate content based on the carbohydrate content, the time corresponding to the eating event, and the current monitoring time; and obtaining the blood glucose concentration time series data. The blood insulin level is obtained by analyzing the blood glucose concentration at the current monitoring time based on the sequence data; the peripheral tissue insulin level is obtained based on the blood insulin level; the estimated blood glucose concentration is obtained based on the gastrointestinal carbohydrate content, the peripheral tissue insulin level, and the exercise calorie expenditure at the current monitoring time based on the time-series data; the number of times the blood glucose concentration prediction step is executed is obtained, and it is determined whether the number of executions has reached the preset monitoring duration. If the number of executions has not reached the preset monitoring duration, the blood glucose concentration prediction step is executed based on the next monitoring time; otherwise, the execution of the blood glucose concentration prediction step is stopped, and several estimated blood glucose concentrations are obtained.

[0011] The above scheme dynamically calculates gastrointestinal carbohydrate content by integrating carbohydrate content, the time corresponding to the eating event, and the current monitoring time to simulate the digestion and absorption process. Next, based on the current actual blood glucose concentration, blood insulin levels are estimated to reflect the pancreas's response to the eating event. Then, based on the blood insulin levels, peripheral tissue insulin levels are further estimated to describe the distribution of insulin transport after the eating event. Finally, by integrating gastrointestinal carbohydrate content, peripheral tissue insulin levels, and current exercise calories burned, a comprehensive blood glucose concentration estimate is calculated to fully simulate the absorption, utilization, and regulation of carbohydrates in the body. This process is repeated at each monitoring time point until the entire monitoring duration is covered, ultimately outputting a blood glucose concentration estimate sequence. Therefore, the glucose metabolism kinetic model constructed by this scheme can reflect the complex coupling relationship between eating, exercise, and blood glucose. Furthermore, by estimating blood glucose concentration over a preset monitoring time series, it can capture the dynamic process of blood glucose concentration evolution over time, ensuring that the obtained blood glucose concentration estimate sequence reflects the change in blood glucose concentration after the eating event, laying the foundation for generating a reliable glucose tolerance curve.

[0012] Further, obtaining the gastrointestinal carbohydrate content based on the carbohydrate content, the time corresponding to the eating event, and the current monitoring time includes: obtaining the eating time difference based on the time corresponding to the eating event and the current monitoring time; and obtaining the gastrointestinal carbohydrate content based on the eating time difference and the carbohydrate content, under preset carbohydrate ingestion parameters and preset carbohydrate expulsion parameters.

[0013] The above scheme dynamically calculates the gastrointestinal carbohydrate content by taking advantage of the time difference between eating and coupling it with carbohydrate content, preset carbohydrate ingestion parameters, and preset carbohydrate emptying parameters. The preset carbohydrate ingestion parameter represents the speed at which carbohydrates enter the gastrointestinal tract. The larger the carbohydrate ingestion parameter, the faster the carbohydrate content in the gastrointestinal tract increases. The preset carbohydrate emptying parameter represents the speed at which carbohydrates are emptied from the gastrointestinal tract. The larger the carbohydrate emptying parameter, the faster the blood glucose concentration increases. This scheme achieves a quantitative simulation of the carbohydrate digestion and absorption process after a eating event.

[0014] Further, obtaining the blood insulin level based on the blood glucose concentration at the current monitoring time according to the blood glucose concentration time-series data includes: obtaining the blood glucose-stimulated insulin secretion based on the blood glucose concentration under a preset Hill function response curve; and obtaining the blood insulin level based on the blood glucose concentration and the blood glucose-stimulated insulin secretion under preset hepatic insulin clearance rate parameters and preset peripheral tissue insulin uptake efficiency parameters.

[0015] The above scheme is based on the current blood glucose concentration and uses a preset Hill function response curve to simulate the pancreas's response, calculates the amount of insulin secreted by blood glucose stimulation, and finally obtains the amount of blood insulin, which can reflect the dynamic characteristics of insulin secretion after a eating event.

[0016] Further, obtaining the peripheral tissue insulin level based on the blood insulin level includes: obtaining the blood insulin increment based on the blood insulin level under a preset peripheral tissue insulin uptake efficiency parameter; and obtaining the peripheral tissue insulin level based on the blood insulin increment under a preset peripheral tissue insulin degradation rate parameter.

[0017] The above scheme obtains the increase in blood insulin based on the blood insulin level, and then obtains the peripheral tissue insulin level based on the increase in blood insulin, which can reflect the process of insulin transport and distribution from the blood to the peripheral tissues after a feeding event.

[0018] Further, the step of obtaining the estimated blood glucose concentration based on the time-series data of the gastrointestinal carbohydrate content, the peripheral tissue insulin level, and the exercise calorie expenditure at the current monitoring time includes: obtaining hepatic gluconeogenesis based on the peripheral tissue insulin level under preset gluconeogenesis parameters and preset peripheral insulin gluconeogenesis parameters; obtaining gastrointestinal absorption glucose fluctuations based on the gastrointestinal carbohydrate content under preset carbohydrate emptying parameters; obtaining peripheral tissue non-insulin dependent glucose based on the hepatic gluconeogenesis; obtaining peripheral tissue insulin dependent glucose based on the peripheral tissue insulin level under preset insulin utilization rate parameters; obtaining renal glucose filtration based on the blood glucose concentration under a preset glucose filtration threshold; obtaining exercise expenditure based on the exercise calorie expenditure; and obtaining the estimated blood glucose concentration based on the hepatic gluconeogenesis, the gastrointestinal absorption glucose fluctuations, the peripheral tissue non-insulin dependent glucose, the peripheral tissue insulin dependent glucose, the renal glucose filtration, and the exercise expenditure.

[0019] In the above scheme, the preset gluconeogenesis parameter represents the degree of change in hepatic gluconeogenesis caused by changes in blood glucose; the larger the gluconeogenesis parameter, the greater the reduction in hepatic gluconeogenesis. The preset peripheral insulin gluconeogenesis parameter represents the degree of change in hepatic gluconeogenesis caused by changes in peripheral insulin concentration; the larger the peripheral insulin gluconeogenesis parameter, the greater the reduction in hepatic gluconeogenesis. By integrating the obtained gastrointestinal carbohydrate content and peripheral tissue insulin levels, the six key carbohydrate inflow and outflow pathways are calculated: hepatic gluconeogenesis, gastrointestinal absorption of blood glucose fluctuations, peripheral tissue non-insulin-dependent glucose, peripheral tissue insulin-dependent glucose, and exercise-induced glucose consumption. The estimated blood glucose concentration is obtained through the above carbohydrate inflow and outflow pathways. As can be seen, this scheme decomposes the complex glucose metabolism process into multiple physiological characteristics, including hepatic gluconeogenesis, gastrointestinal absorption and blood glucose fluctuations, peripheral tissue non-insulin-dependent glucose metabolism, peripheral tissue insulin-dependent glucose metabolism, and exercise-induced glucose expenditure. All of these are controlled by corresponding preset parameters. Therefore, when the preset glucose metabolism kinetic model is subsequently optimized, the resulting personalized glucose metabolism model can better reflect the carbohydrate metabolism characteristics of the target subject after the eating event, thus improving the reliability of the subsequently obtained personalized glucose metabolism model in glucose tolerance test scenarios.

[0020] This invention also provides a method for generating a glucose tolerance curve, applied to any of the glucose tolerance curve generation systems described above. The method includes: acquiring multimodal time-series data collected by a monitoring device; the multimodal time-series data includes a diet log; based on the time corresponding to each eating event in the diet log, segmenting the multimodal time-series data to obtain several panel datasets; estimating the blood glucose concentration using a preset glucose metabolism kinetic model based on the panel datasets obtained by the eating data segmentation module; fitting the blood glucose concentration estimates and the blood glucose concentration sequences from the panel datasets using an iterative particle filter algorithm to obtain an optimized model parameter set; obtaining a personalized glucose metabolism model based on the optimized model parameter set under the preset glucose metabolism kinetic model; performing prediction processing under preset metabolic stimulation conditions based on the personalized glucose metabolism model to obtain a predicted blood glucose concentration sequence; and generating a glucose tolerance curve based on the predicted blood glucose concentration sequence.

[0021] Furthermore, the step of extracting several panel datasets from the multimodal time-series data based on the time corresponding to each eating event in the diet log includes: obtaining several eating monitoring time periods based on the time corresponding to each eating event in the diet log under a preset monitoring duration; and extracting several panel datasets corresponding to the eating monitoring time periods from the multimodal time-series data obtained by the monitoring data acquisition module to obtain several panel datasets.

[0022] In the above scheme, the multimodal time series data is divided according to the time corresponding to the eating event. This ensures that each panel dataset contains the change process of the multimodal time series data after the corresponding eating event, effectively isolating the mutual interference between different eating events, making the change process of the multimodal time series data more consistent, thereby improving the stability of the obtained personalized glucose metabolism model, and enabling the obtained blood glucose concentration prediction sequence to adapt to the needs of glucose tolerance test scenarios, thus improving the reliability of the glucose tolerance curve. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of a sugar tolerance curve generation system provided in an embodiment of the present invention; Figure 2 A flowchart illustrating the steps of a method for generating a sugar tolerance curve according to an embodiment of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Please see Figure 1 This embodiment provides a sugar tolerance curve generation system, including: The monitoring data acquisition module is used to acquire multimodal time-series data collected by the monitoring equipment; the multimodal time-series data includes dietary logs. The food intake data extraction module is used to extract data from the multimodal time-series data obtained by the monitoring data acquisition module based on the time corresponding to each food intake event in the food log obtained by the monitoring data acquisition module, and to obtain several panel datasets. The personalized model module is used to estimate blood glucose concentration based on several panel datasets obtained by the food data extraction module under a preset glucose metabolism kinetic model, thereby obtaining a blood glucose concentration estimation sequence; based on the blood glucose concentration estimation sequence and the blood glucose concentration sequences in several panel datasets, it is fitted using an iterative particle filter algorithm to obtain an optimized model parameter set; based on the optimized model parameter set, a personalized glucose metabolism model is obtained under the preset glucose metabolism kinetic model. The glucose tolerance curve generation module is used to perform predictive processing on the personalized glucose metabolism model obtained by the personalized model module under preset metabolic stimulation conditions to obtain a blood glucose concentration prediction sequence, and generate a glucose tolerance curve based on the blood glucose concentration prediction sequence.

[0026] In the above embodiments, the collected multimodal time-series data is divided into several panel datasets according to the time corresponding to the eating event. Based on these panel datasets, estimation is performed under the glucose metabolism kinetic model to obtain optimized model parameters. A personalized glucose metabolism model is then constructed based on these optimized model parameters. This embodiment divides the multimodal time-series data according to the time corresponding to the eating event, ensuring that each panel dataset corresponds to the multimodal time-series data change process under a metabolic stimulation condition. That is, each panel dataset in this embodiment contains the change process of multimodal time-series data after the corresponding eating event, and effectively isolates the mutual interference between different eating events, making the consistency of the multimodal time-series data change process higher. This improves the stability of the obtained personalized glucose metabolism model and enables the obtained blood glucose concentration prediction sequence to adapt to the needs of glucose tolerance test scenarios, thereby improving the reliability of the glucose tolerance curve.

[0027] It should be noted that the monitoring equipment includes blood glucose monitoring devices and triaxial accelerometers.

[0028] In one embodiment, the collected multimodal time-series data includes interstitial fluid glucose levels, a diet log, and an exercise log. The exercise log is collected using a triaxial accelerometer, while the interstitial fluid glucose levels and the diet log are collected using a blood glucose monitoring device. The diet log records the type and estimated weight of food consumed at each eating event, and calculates and records carbohydrate intake using a built-in nutrient database. The exercise log records the type, duration, and intensity of exercise, and records energy expenditure. This embodiment performs preprocessing operations such as timestamp alignment, outlier removal, and missing value imputation on the three sets of data (interstitial fluid glucose levels, diet log, and exercise log), and uses the preprocessed data as multimodal time-series data.

[0029] Furthermore, the food intake data extraction module is used to extract data from the multimodal time-series data obtained by the monitoring data acquisition module based on the time corresponding to each food intake event in the food log obtained by the monitoring data acquisition module, to obtain several panel datasets, including: obtaining several food intake monitoring time periods based on the time corresponding to each food intake event in the food log obtained by the monitoring data acquisition module under a preset monitoring duration; and extracting several panel datasets corresponding to several food intake monitoring time periods based on the multimodal time-series data obtained by the monitoring data acquisition module to obtain several panel datasets.

[0030] In the above embodiment, the food intake data extraction module defines a fixed preset monitoring period based on the time of each food intake event recorded in the diet log, thus assigning a corresponding food intake monitoring time period to each food intake event. Subsequently, panel datasets corresponding to all the aforementioned food intake monitoring time periods are accurately extracted from continuously collected multimodal time-series data. This embodiment effectively isolates the mutual interference of metabolic responses between different food intake events by defining an independent food intake monitoring time period for each food intake event. This ensures that each panel dataset contains the dynamic change process of panel data corresponding to a complete food intake event, laying a data foundation for subsequently fitting a reliable personalized glucose metabolism model, thereby improving the reliability of the glucose tolerance curve.

[0031] In one embodiment, the eating events in the diet log include breakfast, lunch, and dinner. After a eating event is detected, a feeding monitoring time is generated every 5 minutes for a preset monitoring period of 3 hours, resulting in several feeding monitoring time periods. Based on 14 days of multimodal time-series data obtained by the monitoring data acquisition module, panel datasets corresponding to several of the feeding monitoring time periods are extracted, resulting in several panel datasets. The time corresponding to each eating event is the average of the start time and end time of the eating event.

[0032] Furthermore, the glucose tolerance curve generation module is used to perform predictive processing on the personalized glucose metabolism model obtained by the personalized model module under preset metabolic stimulation conditions to obtain a blood glucose concentration prediction sequence, and generate a glucose tolerance curve based on the blood glucose concentration prediction sequence, including: performing predictive processing on the personalized glucose metabolism model under preset metabolic stimulation conditions to obtain a blood glucose concentration prediction sequence; obtaining a blood glucose concentration extraction sequence based on the blood glucose concentration prediction sequence under a preset extraction time set; and generating a glucose tolerance curve based on the blood glucose concentration extraction sequence using a spline interpolation algorithm.

[0033] In the above embodiment, preset metabolic stimulation conditions are input into the personalized glucose metabolism model. By running the personalized glucose metabolism model, a blood glucose concentration prediction sequence is obtained. Subsequently, based on a preset extraction time set, the blood glucose concentration at each time corresponding to the preset extraction time set is extracted from the blood glucose concentration prediction sequence to form a blood glucose concentration extraction sequence. Finally, a spline interpolation algorithm is used to process the blood glucose concentration extraction sequence to generate a glucose tolerance curve. This embodiment, by applying uniform preset metabolic stimulation conditions to the personalized glucose metabolism model, can fundamentally eliminate the interference of individual preparation state, activity level, and other factors on blood glucose concentration, ensuring that the generated glucose tolerance curve is obtained under a uniform benchmark. Furthermore, by processing the blood glucose concentration extracted at each time point using a spline interpolation algorithm, this embodiment can eliminate the random fluctuations that may exist in the blood glucose concentration prediction sequence output by the personalized glucose metabolism model, generating a smoother glucose tolerance curve that better conforms to physiological changes, further improving the reliability of the glucose tolerance curve.

[0034] In one embodiment, the preset metabolic stimulation conditions include carbohydrate intake, exercise input, and initial state. The carbohydrate intake is defined as an instantaneous Gaussian pulse input with an integral area of ​​75 grams of carbohydrates, simulating drinking glucose water. The exercise input is defined as setting the exercise energy expenditure value to 0 throughout the entire time period, simulating the requirement of sitting still. The initial state is defined as setting the initial blood glucose value to the average fasting blood glucose concentration of the target subject recently. In this embodiment, the above preset metabolic stimulation conditions are input into a personalized glucose metabolism model for prediction processing, resulting in a blood glucose concentration prediction sequence consisting of blood glucose concentrations at 5-minute intervals over the next 3 hours. Based on the blood glucose concentration prediction sequence, a blood glucose concentration extraction sequence is obtained under preset extraction time sets of 0h, 0.5h, 1h, 2h, and 3h. Based on the blood glucose concentration extraction sequence, a glucose tolerance curve is generated using a spline interpolation algorithm. Optionally, the area under the curve, peak value, and other characteristic parameters can be further calculated based on the glucose tolerance curve, and a risk assessment report can be output.

[0035] Furthermore, in the personalized module, the construction of the preset glucose metabolism kinetic model includes: acquiring time-series data of carbohydrate content, the time corresponding to the eating event, calorie expenditure during exercise, and blood glucose concentration; acquiring a preset monitoring duration to obtain a monitoring time series, and performing a blood glucose concentration prediction step based on each monitoring time in the monitoring time series to obtain several estimated blood glucose concentrations, and obtaining a blood glucose concentration estimation sequence based on the several estimated blood glucose concentrations; the blood glucose concentration prediction step includes: obtaining the gastrointestinal carbohydrate content based on the carbohydrate content, the time corresponding to the eating event, and the current monitoring time; and obtaining the blood glucose concentration time series data. The blood insulin level is obtained by analyzing the blood glucose concentration at the current monitoring time based on the sequence data; the peripheral tissue insulin level is obtained based on the blood insulin level; the estimated blood glucose concentration is obtained based on the gastrointestinal carbohydrate content, the peripheral tissue insulin level, and the exercise calorie expenditure at the current monitoring time based on the time-series data; the number of times the blood glucose concentration prediction step is executed is obtained, and it is determined whether the number of executions has reached the preset monitoring duration. If the number of executions has not reached the preset monitoring duration, the blood glucose concentration prediction step is executed based on the next monitoring time; otherwise, the execution of the blood glucose concentration prediction step is stopped, and several estimated blood glucose concentrations are obtained.

[0036] The above embodiment dynamically calculates gastrointestinal carbohydrate content by integrating carbohydrate content, the time corresponding to the eating event, and the current monitoring time to simulate the digestion and absorption process. Next, based on the current actual blood glucose concentration, the blood insulin level is estimated to reflect the pancreas's response to the eating event. Then, based on the blood insulin level, the peripheral tissue insulin level is further estimated to describe the distribution of insulin transport after the eating event. Finally, by integrating gastrointestinal carbohydrate content, peripheral tissue insulin level, and current exercise calories burned, a comprehensive blood glucose concentration estimate is calculated to fully simulate the absorption, utilization, and regulation of carbohydrates in the body. This process is repeated at each monitoring time until the entire monitoring duration is covered, ultimately outputting a blood glucose concentration estimate sequence. Therefore, the glucose metabolism kinetic model constructed in this embodiment can reflect the complex coupling relationship between eating, exercise, and blood glucose. Furthermore, by estimating blood glucose concentration over a preset monitoring time series, it can capture the dynamic process of blood glucose concentration evolution over time, ensuring that the obtained blood glucose concentration estimate sequence reflects the change in blood glucose concentration after the eating event, laying the foundation for generating a reliable glucose tolerance curve.

[0037] Further, obtaining the gastrointestinal carbohydrate content based on the carbohydrate content, the time corresponding to the eating event, and the current monitoring time includes: obtaining the eating time difference based on the time corresponding to the eating event and the current monitoring time; and obtaining the gastrointestinal carbohydrate content based on the eating time difference and the carbohydrate content, under preset carbohydrate ingestion parameters and preset carbohydrate expulsion parameters.

[0038] The above embodiments dynamically calculate the gastrointestinal carbohydrate content by taking advantage of the time difference between eating and coupling carbohydrate content, preset carbohydrate ingestion parameters, and preset carbohydrate emptying parameters. The preset carbohydrate ingestion parameters represent the rate at which carbohydrates enter the gastrointestinal tract. The larger the carbohydrate ingestion parameters, the faster the carbohydrate content in the gastrointestinal tract increases. The preset carbohydrate emptying parameters represent the rate at which carbohydrates are emptied from the gastrointestinal tract. The larger the carbohydrate emptying parameters, the faster the blood glucose concentration increases. This achieves a quantitative simulation of the carbohydrate digestion and absorption process after an eating event.

[0039] In one embodiment, the feeding time difference is obtained based on the time corresponding to the feeding event and the current monitoring time. Based on the difference in eating time and the carbohydrate content Under preset carbohydrate intake parameters With preset carbohydrate emptying parameters Below, the carbohydrate content of the gastrointestinal tract was obtained. And the carbohydrate content in the gastrointestinal tract The dynamic changes are caused by the alteration of carbohydrate levels in the gastrointestinal tract due to food intake. Changes in carbohydrate levels caused by gastrointestinal emptying It consists of two parts, and the corresponding formulas for the gastrointestinal carbohydrate content are as follows: ; Among them, carbohydrate content by As a unit, For shape parameters, The peak value of carbohydrates entering the gastrointestinal tract after eating is determined, and in hour, The area under the curve is always 1.

[0040] Further, obtaining the blood insulin level based on the blood glucose concentration at the current monitoring time according to the blood glucose concentration time-series data includes: obtaining the blood glucose-stimulated insulin secretion based on the blood glucose concentration under a preset Hill function response curve; and obtaining the blood insulin level based on the blood glucose concentration and the blood glucose-stimulated insulin secretion under preset hepatic insulin clearance rate parameters and preset peripheral tissue insulin uptake efficiency parameters.

[0041] The above embodiments, based on the current blood glucose concentration, use a preset Hill function response curve to simulate the pancreas's response, calculate the amount of insulin secreted by blood glucose stimulation, and finally obtain the amount of blood insulin, which can reflect the dynamic characteristics of insulin secretion after a feeding event.

[0042] In one embodiment, blood insulin levels Insulin secretion is stimulated by blood sugar Increase in blood insulin And the amount of insulin consumed by the liver The impact, This represents the current blood glucose concentration, and the corresponding formula is as follows: .

[0043] Blood sugar stimulates insulin secretion The Hill function response curve is used to characterize the effect of blood glucose stimulation on pancreatic insulin secretion. The corresponding formula is: ; in, The maximum steady-state insulin secretion rate is... The maximum secretory capacity of cells under sustained high glucose stimulation. Let be the rate constant of secretion in the second phase, and The larger the size, the faster the insulin secretion rate reaches its maximum. The half-maximal effect concentration reflects Cellular sensitivity to glucose The synergy coefficient reflects the synergy in the glucose sensing process and defines the steepness of the Hill function response curve. The first-phase secretion sensitivity was quantified. The sensitivity of cells to the rate of increase in blood glucose concentration, and The larger, The stronger the initial insulin pulse released by the cell.

[0044] The amount of insulin consumed by the liver Subject to preset hepatic insulin clearance rate parameters The impact, The higher the parameter value, the greater the rate of insulin consumption by the liver. The corresponding formula is: .

[0045] Blood insulin increase Subject to preset peripheral tissue insulin uptake efficiency parameters The impact, The parameters describe the rate at which insulin is transferred from the blood to peripheral tissues, and The higher the parameter value, the higher the efficiency of insulin uptake by peripheral tissues. The corresponding formula is: .

[0046] in, This represents the basal level of insulin in the blood before eating.

[0047] Further, obtaining the peripheral tissue insulin level based on the blood insulin level includes: obtaining the blood insulin increment based on the blood insulin level under a preset peripheral tissue insulin uptake efficiency parameter; and obtaining the peripheral tissue insulin level based on the blood insulin increment under a preset peripheral tissue insulin degradation rate parameter.

[0048] The above embodiments obtain the increase in blood insulin based on the blood insulin level, and obtain the peripheral tissue insulin level based on the increase in blood insulin, which can reflect the process of insulin transfer and distribution from the blood to the peripheral tissues after a feeding event.

[0049] In one embodiment, peripheral tissue insulin levels Increased blood insulin and peripheral tissue consumption The two sources correspond to the following formula: ; in, To preset parameters for the rate of insulin degradation in peripheral tissues, The higher the parameter value, the faster the rate at which peripheral tissues degrade insulin.

[0050] Further, the step of obtaining the estimated blood glucose concentration based on the time-series data of the gastrointestinal carbohydrate content, the peripheral tissue insulin level, and the exercise calorie expenditure at the current monitoring time includes: obtaining hepatic gluconeogenesis based on the peripheral tissue insulin level under preset gluconeogenesis parameters and preset peripheral insulin gluconeogenesis parameters; obtaining gastrointestinal absorption glucose fluctuations based on the gastrointestinal carbohydrate content under preset carbohydrate emptying parameters; obtaining peripheral tissue non-insulin dependent glucose based on the hepatic gluconeogenesis; obtaining peripheral tissue insulin dependent glucose based on the peripheral tissue insulin level under preset insulin utilization rate parameters; obtaining renal glucose filtration based on the blood glucose concentration under a preset glucose filtration threshold; obtaining exercise expenditure based on the exercise calorie expenditure; and obtaining the estimated blood glucose concentration based on the hepatic gluconeogenesis, the gastrointestinal absorption glucose fluctuations, the peripheral tissue non-insulin dependent glucose, the peripheral tissue insulin dependent glucose, the renal glucose filtration, and the exercise expenditure.

[0051] In the above embodiments, the preset gluconeogenesis parameter represents the degree of change in hepatic gluconeogenesis caused by changes in blood glucose levels; the larger the gluconeogenesis parameter, the greater the reduction in hepatic gluconeogenesis. The preset peripheral insulin gluconeogenesis parameter represents the degree of change in hepatic gluconeogenesis caused by changes in peripheral insulin concentration; the larger the peripheral insulin gluconeogenesis parameter, the greater the reduction in hepatic gluconeogenesis. By integrating the obtained gastrointestinal carbohydrate content and peripheral tissue insulin levels, six key carbohydrate inflow and outflow pathways are calculated, namely hepatic gluconeogenesis, gastrointestinal absorption of blood glucose fluctuations, peripheral tissue non-insulin-dependent glucose, peripheral tissue insulin-dependent glucose, renal glucose filtration, and exercise consumption. The estimated blood glucose concentration is obtained through the above carbohydrate inflow and outflow pathways. As can be seen, this embodiment decomposes the complex glucose metabolism process into multiple physiological links, including hepatic gluconeogenesis, gastrointestinal absorption and blood glucose fluctuations, peripheral tissue non-insulin-dependent glucose metabolism, peripheral tissue insulin-dependent glucose metabolism, renal glucose filtration, and exercise-induced glucose expenditure. All of these links are controlled by corresponding preset parameters. Therefore, when the preset glucose metabolism kinetic model is subsequently optimized, the personalized glucose metabolism model obtained in this embodiment can better reflect the carbohydrate metabolism characteristics of the target subject after the eating event, thus improving the reliability of the subsequently obtained personalized glucose metabolism model in the glucose tolerance test scenario.

[0052] In one embodiment, blood glucose concentration is estimated. Dynamic changes are influenced by hepatic gluconeogenesis Gastrointestinal absorption and blood glucose fluctuations Peripheral tissue non-insulin dependent Peripheral tissue insulin dependence Renal glucose filtration and exercise blood consumption Key physiological processes are affected, among which This represents the calories burned during exercise and the corresponding estimated blood glucose concentration. The formula is as follows: ; Hepatic gluconeogenesis The corresponding formula is as follows: ; in, Hepatic gluconeogenesis at basal state Pre-meal blood glucose levels under basal conditions To preset gluconeogenesis parameters, This is to preset peripheral insulin gluconeogenesis parameters.

[0053] Gastrointestinal absorption of blood glucose fluctuations The corresponding formula is as follows: ; in, for Convert to Conversion factor, For the body's blood volume, The target object's weight.

[0054] Peripheral tissue non-insulin dependent The corresponding formula is as follows: ; in, The Michaelis–Menten constant term for blood glucose utilization.

[0055] Peripheral tissue insulin dependence Subject to insulin utilization rate parameter The effect of this is expressed by the following formula: .

[0056] Renal glucose filtration The corresponding formula is as follows: ; in, For filtration rate, This is the blood glucose filtration threshold.

[0057] Exercise consumption The corresponding formula is as follows: ; in, This is the conversion factor for exercise consumption.

[0058] Please see Figure 2 This embodiment also provides a method for generating a sugar tolerance profile, applied to any of the sugar tolerance profile generation systems described above, the method comprising: Step S1: Acquire multimodal time-series data collected by the monitoring equipment; the multimodal time-series data includes dietary logs; Step S2: Based on the time corresponding to each eating event in the diet log, extract data from the multimodal time series data to obtain several panel datasets; Step S3: Based on several panel datasets obtained by the food data extraction module, blood glucose concentration is estimated under a preset glucose metabolism kinetic model to obtain a blood glucose concentration estimation sequence. Based on the blood glucose concentration estimation sequence and the blood glucose concentration sequences in several panel datasets, the model is fitted using an iterative particle filter algorithm to obtain an optimized model parameter set. Based on the optimized model parameter set, a personalized glucose metabolism model is obtained under the preset glucose metabolism kinetic model. Step S4: Based on the personalized glucose metabolism model, perform prediction processing under preset metabolic stimulation conditions to obtain a blood glucose concentration prediction sequence, and generate a glucose tolerance curve based on the blood glucose concentration prediction sequence.

[0059] Furthermore, the step of extracting several panel datasets from the multimodal time-series data based on the time corresponding to each eating event in the diet log includes: obtaining several eating monitoring time periods based on the time corresponding to each eating event in the diet log under a preset monitoring duration; and extracting several panel datasets corresponding to the eating monitoring time periods from the multimodal time-series data obtained by the monitoring data acquisition module to obtain several panel datasets.

[0060] In the above embodiments, dividing the multimodal time-series data according to the time corresponding to the eating event ensures that each panel dataset contains the change process of the multimodal time-series data after the corresponding eating event, effectively isolating the mutual interference between different eating events, making the change process of the multimodal time-series data more consistent, thereby improving the stability of the obtained personalized glucose metabolism model, and enabling the obtained blood glucose concentration prediction sequence to adapt to the needs of glucose tolerance test scenarios, thus improving the reliability of the glucose tolerance curve.

[0061] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A sugar tolerance curve generation system, characterized in that, include: The monitoring data acquisition module is used to acquire multimodal time-series data collected by the monitoring equipment; The multimodal time-series data includes dietary logs; The food intake data extraction module is used to extract data from the multimodal time-series data obtained by the monitoring data acquisition module based on the time corresponding to each food intake event in the food log obtained by the monitoring data acquisition module, and to obtain several panel datasets. The personalized model module is used to estimate blood glucose concentration sequences based on several panel datasets obtained by the food data extraction module under a preset glucose metabolism kinetic model. Based on the blood glucose concentration estimation sequence and the blood glucose concentration sequences in several panel datasets, an iterative particle filter algorithm is used for fitting to obtain an optimized model parameter set; based on the optimized model parameter set, a personalized glucose metabolism model is obtained under a preset glucose metabolism kinetic model. The glucose tolerance curve generation module is used to perform predictive processing on the personalized glucose metabolism model obtained by the personalized model module under preset metabolic stimulation conditions to obtain a blood glucose concentration prediction sequence, and generate a glucose tolerance curve based on the blood glucose concentration prediction sequence.

2. The sugar tolerance curve generation system as described in claim 1, characterized in that, The food intake data extraction module is used to extract data from the multimodal time-series data obtained by the monitoring data acquisition module based on the time corresponding to each food intake event in the food log obtained by the monitoring data acquisition module, resulting in several panel datasets, including: Based on the time corresponding to each eating event in the diet log obtained by the monitoring data acquisition module, several eating monitoring time periods are obtained under a preset monitoring duration. Based on the multimodal time-series data obtained by the monitoring data acquisition module, several panel datasets corresponding to the feeding monitoring time periods are extracted to obtain several panel datasets.

3. The sugar tolerance curve generation system as described in claim 1, characterized in that, The glucose tolerance curve generation module is used to perform predictive processing on the personalized glucose metabolism model obtained by the personalized model module under preset metabolic stimulation conditions to obtain a blood glucose concentration prediction sequence, and generate a glucose tolerance curve based on the blood glucose concentration prediction sequence, including: Based on the personalized glucose metabolism model, a prediction process is performed under preset metabolic stimulation conditions to obtain a blood glucose concentration prediction sequence. Based on the blood glucose concentration prediction sequence, a blood glucose concentration extraction sequence is obtained under a preset extraction time set; Based on the extracted blood glucose concentration sequence, a glucose tolerance curve is generated using a spline interpolation algorithm.

4. The sugar tolerance curve generation system as described in claim 1, characterized in that, The construction of the preset glucose metabolism kinetic model in the personalized module includes: Acquire time-series data on carbohydrate content, the time corresponding to the eating event, calories burned during exercise, and blood glucose concentration; A preset monitoring duration is obtained to get a monitoring time series. A blood glucose concentration prediction step is performed based on each monitoring time in the monitoring time series to get several blood glucose estimated concentrations. Based on several blood glucose estimated concentrations, a blood glucose concentration estimation sequence is obtained. The blood glucose concentration prediction step includes: Based on the carbohydrate content, the time corresponding to the eating event, and the current monitoring time, the gastrointestinal carbohydrate content is obtained; Based on the blood glucose concentration time-series data, the blood insulin level is obtained at the current monitoring time. Based on the blood insulin level, the peripheral tissue insulin level is obtained; Based on the gastrointestinal carbohydrate content, the peripheral tissue insulin level, and the exercise calorie expenditure time series data at the current monitoring time, the blood glucose concentration is estimated. The number of times the blood glucose concentration prediction step is executed is obtained, and it is determined whether the number of executions has reached the preset monitoring time. If the number of executions has not reached the preset monitoring time, the blood glucose concentration prediction step is executed based on the next monitoring time; otherwise, the execution of the blood glucose concentration prediction step is stopped, and several blood glucose estimated concentrations are obtained.

5. The sugar tolerance curve generation system as described in claim 4, characterized in that, The process of obtaining gastrointestinal carbohydrate content based on the carbohydrate content, the time corresponding to the eating event, and the current monitoring time includes: Based on the time corresponding to the feeding event and the current monitoring time, the feeding time difference is obtained; Based on the time difference between eating and the carbohydrate content, the gastrointestinal carbohydrate content is obtained under preset carbohydrate intake parameters and preset carbohydrate emptying parameters.

6. The sugar tolerance curve generation system as described in claim 4, characterized in that, The step of obtaining blood insulin levels based on the blood glucose concentration at the current monitoring time according to the time-series blood glucose concentration data includes: Based on the blood glucose concentration, the amount of insulin secreted by blood glucose stimulation is obtained under the preset Hill function response curve; Based on the blood glucose concentration and the amount of insulin secreted by blood glucose stimulation, the blood insulin level is obtained under preset parameters for liver insulin clearance rate and peripheral tissue insulin uptake efficiency.

7. The sugar tolerance curve generation system as described in claim 4, characterized in that, The process of obtaining peripheral tissue insulin levels based on the blood insulin levels includes: Based on the blood insulin level, the increase in blood insulin is obtained under a preset peripheral tissue insulin uptake efficiency parameter; Based on the increase in blood insulin, the amount of insulin in peripheral tissue is obtained under a preset parameter for the rate of insulin degradation in peripheral tissue.

8. The sugar tolerance curve generation system as described in claim 4, characterized in that, The process of obtaining an estimated blood glucose concentration based on the gastrointestinal carbohydrate content, peripheral tissue insulin levels, and exercise calorie expenditure time-series data at the current monitoring time includes: Based on the peripheral tissue insulin levels, liver gluconeogenesis is obtained under preset gluconeogenesis parameters and preset peripheral insulin gluconeogenesis parameters. Based on the carbohydrate content in the gastrointestinal tract, the fluctuation of blood glucose absorbed in the gastrointestinal tract is obtained under preset carbohydrate emptying parameters; Based on the aforementioned hepatic gluconeogenesis, peripheral tissue non-insulin-dependent gluconeogenesis was obtained; Based on the amount of insulin in the peripheral tissue, the insulin dependence of the peripheral tissue is obtained under a preset insulin utilization rate parameter. Based on the blood glucose concentration, renal blood glucose filtration is obtained at a preset blood glucose filtration threshold. Based on the calories burned during exercise, the exercise expenditure is obtained; The estimated blood glucose concentration is obtained based on the hepatic gluconeogenesis, the fluctuation of blood glucose absorbed by the gastrointestinal tract, the non-insulin-dependent peripheral tissue, the insulin-dependent peripheral tissue, the renal glucose filtration, and the exercise expenditure.

9. A method for generating a sugar tolerance curve, characterized in that, The method applied to the sugar tolerance curve generation system as described in any one of claims 1 to 8 includes: Acquire multimodal time-series data collected by monitoring equipment; the multimodal time-series data includes dietary logs; Based on the time corresponding to each eating event in the diet log, several panel datasets are obtained by extracting data from the multimodal time series data. Based on several panel datasets obtained by the food data extraction module, blood glucose concentration estimation sequences are obtained under a preset glucose metabolism kinetic model. Based on the blood glucose concentration estimation sequence and the blood glucose concentration sequences in several panel datasets, the optimized model parameter set is obtained by fitting the iterative particle filter algorithm. Based on the optimized model parameter set, a personalized glucose metabolism model is obtained under the preset glucose metabolism kinetic model; Based on the personalized glucose metabolism model, a prediction process is performed under preset metabolic stimulation conditions to obtain a blood glucose concentration prediction sequence, and a glucose tolerance curve is generated based on the blood glucose concentration prediction sequence.

10. The method for generating a sugar tolerance curve as described in claim 9, characterized in that, Based on the time corresponding to each eating event in the diet log, the multimodal time-series data is segmented to obtain several panel datasets, including: Based on the time corresponding to each eating event in the diet log, several eating monitoring time periods are obtained under a preset monitoring duration. Based on the multimodal time-series data obtained by the monitoring data acquisition module, several panel datasets corresponding to the feeding monitoring time periods are extracted to obtain several panel datasets.

Images