Healthy diet index evaluation method and system based on diabetic patient

Abstract

The invention relates to a healthy diet index evaluation method and system based on a diabetic patient. The method comprises the steps that a multi-modal data set of a patient is obtained, the multi-modal data set comprises diet image data, continuous blood glucose monitoring data and medicine use parameters, and mixed food material component analysis processing is conducted on the diet image data to obtain layered nutrition quantification data; performing synergistic effect analysis according to the drug use parameters and the pharmacokinetic model to generate a nutrition absorption correction coefficient; performing time sequence alignment processing on the layered nutrition quantification data and the continuous blood glucose monitoring data, and establishing a blood glucose response prediction model; and generating a diet health assessment report according to the blood glucose response prediction model and the nutrition absorption correction coefficient. According to the method, through the synergistic effect of the multi-modal data, the blood glucose response prediction model and the drug absorption correction coefficient, the diet effect of the diabetic patient can be accurately evaluated, nutrient absorption is optimized, and the scientificity and individuation degree of diet health management of the diabetic patient are effectively improved.

Description

Technical Field

[0001] The present invention belongs to the field of healthy diet assessment, and in particular relates to a healthy diet index assessment method and system based on diabetic patients. Background Art

[0002] With the rapid development of technology in the medical and health field, methods for assessing the healthy diet of diabetic patients have emerged. However, this method usually relies on medical staff to conduct a simple dietary questionnaire survey on patients to understand the types of daily diet and the approximate intake of patients, and combine the patients' blood glucose monitoring data to give general dietary recommendations based on experience. On the one hand, the questionnaire survey relies on patient recollection, and the accuracy and completeness of the data are difficult to guarantee, and it is impossible to accurately obtain detailed information on the ingredients of each meal. On the other hand, this method lacks effective integration and dynamic analysis of dietary data and blood glucose monitoring data, resulting in the lack of personalization and scientificity in the generated dietary health assessment report, which cannot effectively meet the needs of diabetic patients for refined dietary management. In addition, when considering the relationship between drugs and diet, this method ignores the pharmacokinetic factors, fails to deeply analyze the impact of drugs on nutrient absorption and adjust the diet plan accordingly. Summary of the Invention

[0003] Based on this, it is necessary to provide a healthy diet index assessment method and system based on diabetic patients to address the above technical problems, so as to improve the accuracy and personalization of dietary health assessment and enhance the scientificity and effectiveness of dietary management for diabetic patients.

[0004] In a first aspect, the present application provides a method for evaluating a healthy diet index for diabetic patients, the method comprising:

[0005] Obtain a multimodal dataset of patients, including dietary image data, continuous blood glucose monitoring data, and medication usage parameters. Perform mixed food ingredient analysis on the dietary image data to obtain stratified nutritional quantitative data.

[0006] Perform synergistic analysis based on drug usage parameters and pharmacokinetic models to generate nutrient absorption correction factors;

[0007] Perform time series alignment processing on stratified nutritional quantification data and continuous blood glucose monitoring data to establish a blood glucose response prediction model;

[0008] Generate a dietary health assessment report based on the blood glucose response prediction model and nutrient absorption correction factor.

[0009] In one embodiment, mixed food ingredient analysis is performed on the dietary image data to obtain layered nutritional quantitative data, including:

[0010] A deep convolutional attention network is used to extract multi-scale features from food image data. A residual channel attention module is used to separate the outline boundaries of stacked ingredients in food image data to generate superpixel segmentation maps.

[0011] Based on a preset regional recipe database, the graph neural network is used to match the ingredient hierarchy in the superpixel segmentation image, identify hidden added sugar components, and obtain matched ingredient data;

[0012] The weight of the matched food data is calculated based on the near-infrared spectral characteristics to obtain the initial nutritional quantitative index;

[0013] Based on the initial nutritional quantitative index and gray-level co-occurrence matrix, the carbonized layer on the surface of fried ingredients in the dietary image data is compensated and corrected to generate layered nutritional quantitative data.

[0014] In one embodiment, a synergistic analysis is performed based on drug usage parameters and a pharmacokinetic model to generate a nutrient absorption correction factor, including:

[0015] According to the absorption kinetics module based on the pharmacokinetic model and the daily dose of metformin in the drug usage parameters, a vitamin B12 absorption rate attenuation function is constructed to generate a vitamin B12 dynamic absorption coefficient;

[0016] The dietary fiber inhibition coefficient was calculated based on the intestinal metabolism module of the pharmacokinetic model and the patient's intestinal flora detection data, and the abundance ratio of Bifidobacterium / Prevotella;

[0017] The plasma concentration prediction module based on the pharmacokinetic model calculates the real-time vitamin B12 concentration prediction value according to the vitamin B12 dynamic absorption coefficient and the patient's baseline serum concentration;

[0018] When the real-time vitamin B12 concentration prediction value is greater than the preset threshold, an animal protein restriction instruction is generated. The animal protein restriction instruction is used to adjust the protein source allocation weight, increase the recommended proportion of plant protein, and output the protein source allocation weight;

[0019] When the drug use parameters include sulfonylureas, an inverse proportional relationship model between dietary fiber intake and drug effect duration is established through the drug effect duration module of the pharmacokinetic model to generate drug effect duration parameters;

[0020] The nutrient absorption correction factor is generated by combining the dietary fiber inhibition coefficient, protein source allocation weight and drug effect duration parameters.

[0021] In one embodiment, time series alignment processing is performed on stratified nutritional quantification data and continuous blood glucose monitoring data to establish a blood glucose response prediction model, including:

[0022] The LSTM-Transformer hybrid model is used to perform timestamp interpolation processing on the stratified nutritional quantification data and continuous blood glucose monitoring data to generate synchronized time series data.

[0023] According to the individual insulin sensitivity coefficient in the nutrient absorption correction coefficient, a Gaussian kernel blood glucose response function is constructed;

[0024] Obtain the patient's exercise intensity data within 1 hour after a meal and convert the exercise intensity data into metabolic equivalent values;

[0025] Calculate the metabolic compensation factor for the post-meal exercise period based on the metabolic equivalent value. The metabolic compensation factor is used to reduce the glycemic index of the corresponding post-meal exercise period.

[0026] The dynamic blood glucose prediction curve is generated by combining the synchronous time series data, Gaussian kernel blood glucose response function and metabolic compensation factor, and the blood glucose response prediction model is obtained.

[0027] In one embodiment, a dietary health assessment report is generated based on the blood glucose response prediction model and the nutrient absorption correction factor, including:

[0028] According to the vitamin B12 dynamic absorption coefficient, dietary fiber inhibition coefficient and protein source allocation weight in the nutrient absorption correction coefficient, the dynamic blood glucose prediction curve output by the blood glucose response prediction model is adjusted to generate a corrected blood glucose prediction curve;

[0029] Based on the corrected blood glucose prediction curve, the blood glucose fluctuation amplitude and the area under the curve are calculated;

[0030] Perform risk assessment on the blood sugar fluctuation amplitude according to the preset blood sugar fluctuation threshold and generate an early warning signal, which can be any one of a green safety signal, a yellow warning signal, and a red high-risk signal;

[0031] A multi-objective optimization algorithm is used to optimize the diet plan based on the revised blood glucose prediction curve to generate personalized diet recommendations, including a list of ingredient replacements and meal time windows.

[0032] A dietary health assessment report is generated by combining the revised blood sugar prediction curve, early warning signals and personalized dietary recommendations.

[0033] In one embodiment, the multimodal dataset is obtained by:

[0034] Acquire dietary image data containing visible light bands and record the start and end time of eating;

[0035] Continuous blood glucose concentration data collected at 5-minute intervals were obtained. Continuous blood glucose monitoring data were converted from subcutaneous interstitial fluid glucose levels.

[0036] Obtain medication parameters including daily metformin dose and insulin injection regimen, including rapid-acting / long-acting insulin type and injection time;

[0037] Obtain the patient's energy expenditure intensity data within 1 hour after a meal;

[0038] Dietary image data, continuous blood glucose monitoring data, medication usage parameters, and energy expenditure intensity data are timestamp-aligned to generate a multimodal dataset.

[0039] In a second aspect, the present application also provides a healthy diet index assessment system for diabetic patients, the system comprising:

[0040] A data acquisition and processing module is used to obtain a patient's multimodal dataset, which includes dietary image data, continuous blood glucose monitoring data, and medication usage parameters. The module also performs mixed food ingredient analysis on the dietary image data to obtain layered nutritional quantitative data.

[0041] Drug use analysis module, used to perform synergistic analysis based on drug use parameters and drug metabolism kinetics model and generate nutrient absorption correction coefficient;

[0042] The blood glucose response prediction module is used to perform time-series alignment processing on stratified nutritional quantification data and continuous blood glucose monitoring data to establish a blood glucose response prediction model;

[0043] The evaluation report generation module is used to generate a diet health evaluation report based on the blood glucose response prediction model and the nutrient absorption correction coefficient.

[0044] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps in the first aspect when executing the computer program.

[0045] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which implements the steps in the first aspect when executed by a processor.

[0046] The above-mentioned method and system for assessing a healthy diet index for diabetic patients acquires a multimodal dataset consisting of dietary image data, continuous blood glucose monitoring data, and medication usage parameters. The system then performs mixed ingredient analysis on the dietary image data to generate detailed, stratified nutritional quantification data. This not only enriches the data foundation for dietary assessment but also improves the accuracy and detail of the data, providing strong support for subsequent dietary analysis and avoiding the inaccurate and incomplete dietary records often encountered in traditional methods. Furthermore, a synergistic effect analysis based on medication usage parameters and pharmacokinetic models is performed, taking into account individual patient differences and the effects of medications on nutrient absorption. This makes dietary assessments more tailored to the patient's specific circumstances and provides a scientific basis for personalized dietary guidance. Subsequently, the stratified nutritional quantification data and continuous blood glucose monitoring data are time-series aligned to establish a blood glucose response prediction model. This model enables more accurate prediction of the impact of diet on blood glucose, helping patients and physicians to promptly adjust dietary plans, prevent abnormal blood glucose fluctuations, and reduce the risk of complications. This overcomes the limitation of traditional intermittent blood glucose monitoring, which lacks real-time reflection of blood glucose changes. Finally, a dietary health assessment report is generated based on the blood glucose response prediction model and nutrient absorption correction coefficients. This report can comprehensively reflect the impact of diet on patients' blood sugar, nutrient absorption and other aspects, providing a scientific and systematic reference for doctors to formulate treatment plans and patients to manage themselves.

[0047] Compared with traditional healthy diet assessment methods for diabetic patients, this method further improves the accuracy, scientificity and personalization of dietary health assessment for diabetic patients through multimodal data acquisition and processing, drug-nutrition synergy, dynamic blood glucose response modeling and personalized assessment report generation, providing more effective and scientific technical support for the dietary management and treatment of diabetic patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0049] Figure 1 A flow chart of a method for evaluating a healthy diet index for diabetic patients provided by an exemplary embodiment of the present invention;

[0050] Figure 2 A schematic diagram of the structure of a healthy diet index assessment system for diabetic patients provided by an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0052] In one embodiment, Figure 1 As shown, a method for evaluating a healthy diet index for diabetic patients is provided. This embodiment uses the method applied to a terminal as an example. It is understandable that the method can also be applied to a server, or to a system including a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0053] S101: Acquire a multimodal dataset of a patient, which includes dietary image data, continuous blood glucose monitoring data, and medication usage parameters, and perform mixed food ingredient analysis on the dietary image data to obtain layered nutritional quantitative data.

[0054] Specifically, the patient's dietary image data for each meal can be obtained through a mobile terminal application. And the image recognition technology in deep learning can be used to analyze the dietary image data to identify the various ingredients and their proportions in the image. Furthermore, the corresponding ingredients can be converted into detailed nutritional data in combination with the food nutritional database. Among them, the food nutritional database contains detailed nutritional information of various ingredients, such as the content of various nutrients such as carbohydrates, protein, fat, dietary fiber, vitamins and minerals. By matching the identified ingredients with the information in the database, the nutritional components of each ingredient can be accurately calculated, and finally the layered nutritional quantitative data can be obtained. This process not only takes into account the type and intake of ingredients, but also can adjust the impact of different cooking methods and processing levels on nutritional components, thereby providing more accurate and comprehensive basic data for subsequent dietary assessments.

[0055] S102: Perform synergistic analysis based on drug usage parameters and drug metabolism kinetics model to generate nutrient absorption correction coefficient.

[0056] Specifically, medication usage parameters such as medication type, dosage, and time of use can be obtained by the patient manually entering them into the mobile terminal application or through communication with medical devices, hospital information systems, etc. Pharmacokinetic models are then used to analyze the drug's metabolic processes within the patient's body and the interactions between the drug and food components. For example, certain glucose-lowering medications may delay carbohydrate absorption or interact with certain food components, thereby affecting the drug's efficacy and nutrient absorption. Synergy analysis can be used to calculate a nutrient absorption correction factor, which reflects the extent to which medication use affects nutrient absorption. For example, if a patient is taking a glucose-lowering medication that may slow carbohydrate absorption, the nutrient absorption correction factor can be used to adjust the timing and extent of the carbohydrate's effect on blood sugar accordingly.

[0057] S103: Perform time series alignment processing on the stratified nutritional quantification data and the continuous blood glucose monitoring data to establish a blood glucose response prediction model.

[0058] Specifically, continuous blood glucose monitoring data can be obtained in real time through a continuous blood glucose monitoring device worn by the patient. The continuous blood glucose monitoring data can reflect the dynamic changes in the patient's blood glucose. Accurately aligning the stratified nutritional quantification data with the continuous blood glucose monitoring data in a time series can ensure that the dietary data of each meal corresponds to the corresponding blood glucose change data. Subsequently, machine learning algorithms such as neural networks and support vector machines can be used to train the aligned data to establish a blood glucose response prediction model. The model can predict the impact of the diet on the patient's blood glucose based on the nutritional component data of the diet, including the magnitude of the increase, speed and downward trend of blood glucose. For example, through training with a large amount of historical data, the blood glucose response prediction model can learn the specific impact patterns of different nutrient combinations on blood glucose, thereby accurately predicting the changes in the patient's blood glucose when the patient consumes a new diet, providing patients and doctors with a scientific basis for decision-making.

[0059] S104: Generate a dietary health assessment report based on the blood glucose response prediction model and the nutrient absorption correction coefficient.

[0060] Specifically, the Glycemic Response Prediction Model, based on a patient's current dietary data and blood glucose monitoring data, can predict changes in blood glucose levels after consuming specific foods, such as the magnitude and rate of blood glucose rise, as well as the downward trend. The Nutrient Absorption Correction Factor (NACF) reflects the impact of medication use on nutrient absorption. Combining the Glycemic Response Prediction Model with the NACF adjusts the model's predictions to better align with the patient's actual physiological responses, enabling a comprehensive assessment of the patient's diet and generating a dietary health assessment report. This report includes, but is not limited to, a prediction of the impact of a diet on blood glucose, an assessment of nutritional balance, and dietary recommendations. For example, if a particular meal is predicted to cause a significant increase in blood glucose, the report can identify dietary risks and recommend dietary adjustments, such as reducing carbohydrate intake and increasing the proportion of vegetables and protein. The report can also provide a personalized diet plan based on the patient's nutritional needs and medication use, helping them better control their blood glucose and maintain their health.

[0061] The method first acquires a multimodal dataset consisting of dietary image data, continuous glucose monitoring data, and medication use parameters, providing a comprehensive data foundation for subsequent dietary assessment. Mixed ingredient composition analysis is then performed on the dietary image data, enabling precise analysis of the various dietary ingredients and their corresponding nutrients, improving the accuracy and comprehensiveness of the dietary data. Secondly, a synergistic analysis based on medication use parameters and pharmacokinetic models is performed, taking into account individual patient differences and the impact of medications on nutrient absorption. This makes the dietary assessment more tailored to the patient's specific circumstances and provides a scientific basis for subsequent personalized dietary guidance. Furthermore, time-series alignment of the stratified nutrient quantification data and continuous glucose monitoring data is performed to establish a glucose response prediction model. This model accurately captures the link between dietary intake and glucose dynamics, further enhancing the scientific and reliable nature of glucose prediction. Finally, a dietary health assessment report is generated based on the glucose response prediction model and nutrient absorption correction coefficients. This report not only assesses the patient's current dietary health status but also provides personalized dietary improvement recommendations, further enhancing the scientific and personalized nature of dietary health management for diabetic patients.

[0062] In an exemplary embodiment, the multimodal dataset is obtained by the following steps:

[0063] Acquire dietary image data containing visible light bands and record the start and end time of eating;

[0064] Continuous blood glucose concentration data collected at 5-minute intervals were obtained. Continuous blood glucose monitoring data were converted from subcutaneous interstitial fluid glucose levels.

[0065] Obtain medication parameters including daily metformin dose and insulin injection regimen, including rapid-acting / long-acting insulin type and injection time;

[0066] Obtain the patient's energy expenditure intensity data within 1 hour after a meal;

[0067] Dietary image data, continuous blood glucose monitoring data, medication usage parameters, and energy expenditure intensity data are timestamp-aligned to generate a multimodal dataset.

[0068] Specifically, a mobile terminal application can capture images of each patient's meals. These images, covering the visible light band, can intuitively reflect information such as the type, freshness, and cooking method, thereby ensuring that rich details and characteristics of the food are captured. Furthermore, the pre-meal image can be automatically associated with the current time information as the meal start time, and the post-meal image or meal stop instruction can automatically record the meal stop time. Furthermore, a continuous blood glucose monitoring device can monitor the glucose concentration in the patient's subcutaneous interstitial fluid in real time, collecting data at set intervals, such as 5 minutes, to obtain continuous blood glucose monitoring data. Furthermore, the mobile terminal application can also capture medication information entered by the patient, such as daily metformin dosage and insulin injection regimen. Energy expenditure intensity is one of the important factors affecting blood glucose levels in diabetic patients. In this embodiment, wearable devices such as smart bracelets and smart watches can be used to obtain energy expenditure intensity data within one hour after a meal. Finally, by accurately matching the multi-source data obtained above in chronological order, it is possible to ensure that each meal's dietary data corresponds to the corresponding blood glucose changes, medication use, and energy expenditure data, thereby generating a multimodal dataset. This dataset provides a comprehensive and accurate data basis for subsequent comprehensive analysis and evaluation, and helps to improve the scientific nature and reliability of the entire healthy diet index assessment method.

[0069] In an exemplary embodiment, mixed food ingredient analysis processing is performed on the dietary image data to obtain layered nutritional quantitative data, including:

[0070] A deep convolutional attention network is used to extract multi-scale features from food image data. A residual channel attention module is used to separate the outline boundaries of stacked ingredients in food image data to generate superpixel segmentation maps.

[0071] Based on a preset regional recipe database, the graph neural network is used to match the ingredient hierarchy in the superpixel segmentation image, identify hidden added sugar components, and obtain matched ingredient data;

[0072] The weight of the matched food data is calculated based on the near-infrared spectral characteristics to obtain the initial nutritional quantitative index;

[0073] Based on the initial nutritional quantitative index and gray-level co-occurrence matrix, the carbonized layer on the surface of fried ingredients in the dietary image data is compensated and corrected to generate layered nutritional quantitative data.

[0074] The Deep Convolutional Attention Network (DCNN) is an image processing model that combines the feature extraction capabilities of convolutional neural networks with an attention mechanism to automatically focus on key image regions. Specifically, when processing food image data, the network's convolutional layers can slide convolution kernels of varying sizes across the image, capturing food features at different scales. For example, small kernels can extract detailed food features, such as surface texture and subtle decorations, while large kernels can capture overall shape and structural features, such as the food's outline and spatial layout. Finally, multiple layers of feature maps are generated, each corresponding to different scales and semantic information. Subsequently, a residual channel attention module is used to enhance the recognition of food outlines and boundaries while preserving image details, thereby separating the outlines and boundaries of stacked ingredients in food image data. Schematically, this module calculates attention weights across different channels to highlight the boundaries of ingredients, effectively separating stacked ingredients. Finally, a superpixel segmentation map is generated, which segments the food image into multiple superpixel regions, each roughly corresponding to an ingredient or portion of an ingredient. This provides a clear framework for subsequent ingredient recognition and ingredient analysis. And considering that eating habits and food ingredient preferences vary greatly in different regions, a preset regional recipe database has been constructed. This database collects a large amount of local common recipe information, including ingredient combinations, cooking methods, and the hierarchical relationship between ingredients. Through graph neural networks, graph structure data can be used for learning and reasoning, matching the ingredient areas in the superpixel segmentation map with the ingredient hierarchy in the recipe database, and identifying the various ingredients in the image and their combinations. In the matching process, by analyzing the combination and cooking methods of ingredients, combined with the information in the database, it is possible to identify which ingredients or foods may contain hidden added sugars and record them in the matched ingredient data.

[0075] Specifically, near-infrared spectral features can be used to estimate the proportion of ingredients in an image. This is achieved by comparing the characteristic spectra of different ingredients with a database of known near-infrared spectra. The weight of each ingredient in the image is then calculated based on the intensity and distribution of the spectra. An initial nutritional quantification index can be generated based on the weights and the corresponding nutrient data for the ingredients. This nutrient data can be sourced from a publicly available food nutrition database, which contains detailed nutritional information for various ingredients, such as carbohydrates, protein, fat, dietary fiber, vitamins, and minerals. By multiplying the weight of each ingredient with its nutrient data and summing the results, an initial nutritional quantification index for the entire food image can be obtained. The grayscale co-occurrence matrix (GLM) reflects the correlation and distribution of grayscale values ​​between two pixels in an image. For fried ingredients, the formation of a charred layer on their surface can alter their appearance and nutritional content. By analyzing the GLM, charred areas on the surface of fried ingredients can be identified, and the initial nutritional quantification index can be adjusted based on the degree of charring. This compensation correction process takes into account the loss and changes in nutrients caused by the burnt layer, such as the degradation of certain nutrients or the formation of new compounds, to more accurately reflect the actual nutritional intake and ultimately generate stratified nutritional quantitative data. This data not only includes the nutritional composition information of various ingredients in the diet, but also considers the nutritional impact of factors such as the ingredient structure, hidden added sugars, and deep-frying burn, providing a comprehensive and accurate data foundation for subsequent healthy diet assessments.

[0076] In an exemplary embodiment, a synergistic analysis is performed based on drug usage parameters and a pharmacokinetic model to generate a nutrient absorption correction factor, including:

[0077] According to the absorption kinetics module based on the pharmacokinetic model and the daily dose of metformin in the drug usage parameters, a vitamin B12 absorption rate attenuation function is constructed to generate a vitamin B12 dynamic absorption coefficient;

[0078] The dietary fiber inhibition coefficient was calculated based on the intestinal metabolism module of the pharmacokinetic model and the patient's intestinal flora detection data, and the abundance ratio of Bifidobacterium / Prevotella;

[0079] The plasma concentration prediction module based on the pharmacokinetic model calculates the real-time vitamin B12 concentration prediction value according to the vitamin B12 dynamic absorption coefficient and the patient's baseline serum concentration;

[0080] When the real-time vitamin B12 concentration prediction value is greater than the preset threshold, an animal protein restriction instruction is generated. The animal protein restriction instruction is used to adjust the protein source allocation weight, increase the recommended proportion of plant protein, and output the protein source allocation weight;

[0081] When the drug use parameters include sulfonylureas, an inverse proportional relationship model between dietary fiber intake and drug effect duration is established through the drug effect duration module of the pharmacokinetic model to generate drug effect duration parameters;

[0082] The nutrient absorption correction factor is generated by combining the dietary fiber inhibition coefficient, protein source allocation weight and drug effect duration parameters.

[0083] Specifically, the absorption kinetics module of the pharmacokinetic model describes the drug absorption process in the body, including the rate and extent of absorption. Metformin is a commonly used hypoglycemic drug, but long-term use may affect vitamin B12 absorption. By analyzing the relationship between the daily metformin dose and vitamin B12 absorption, a decay function can be constructed that reflects the change in vitamin B12 absorption rate with metformin dose. This function takes the daily metformin dose as an input variable and outputs a dynamic absorption coefficient for vitamin B12. This coefficient reflects the actual absorption efficiency of vitamin B12 under a patient's specific medication conditions. Furthermore, the intestinal metabolism module of the pharmacokinetic model simulates the metabolism of drugs and nutrients in the intestine and considers the impact of the intestinal flora on drug and nutrient absorption. Patient intestinal flora data provides information on the abundance of different bacterial species in the intestine. Bifidobacterium and Prevotella are two important bacterial species in the intestine, and the abundance ratio between them is related to the metabolism and utilization efficiency of dietary fiber. This ratio can be analyzed in conjunction with the intestinal metabolism module to calculate the dietary fiber inhibition coefficient. This coefficient reflects the potential inhibitory effect of dietary fiber on nutrient absorption, given the patient's intestinal flora. Furthermore, the plasma concentration prediction module of the pharmacokinetic model utilizes pharmacokinetic principles, combined with the dynamic absorption coefficient of vitamin B12 and the patient's initial baseline serum concentration, to predict the real-time concentration of vitamin B12 in the patient's plasma at different time points. This predicted value helps understand the dynamic changes in vitamin B12 in the patient's body and provides a basis for subsequent nutrient absorption adjustments.

[0084] Specifically, when the real-time predicted vitamin B12 concentration exceeds a preset threshold, indicating a patient's potential risk of vitamin B12 overdose, animal protein restriction instructions can be generated, reducing the recommended animal protein intake while increasing the recommended proportion of plant protein. By adjusting the protein source allocation weight, patients can be guided to rationally balance their animal and plant protein intake in their diet to avoid potential health problems associated with vitamin B12 overdose. Furthermore, when medication usage parameters include sulfonylureas, the duration of effect module of the pharmacokinetic model establishes an inversely proportional relationship between dietary fiber intake and duration of effect, generating a duration of effect parameter. Sulfonylureas are insulin secretagogues, and their duration of effect is affected by multiple factors. Dietary fiber intake affects the rate of drug metabolism and absorption efficiency in the intestine, thereby affecting the duration of effect. By establishing this inversely proportional relationship model, the effect of dietary fiber intake on duration of effect can be quantified, generating a duration of effect parameter. This parameter reflects the expected duration of effect of sulfonylureas at a specific dietary fiber intake level. Finally, a nutrient absorption correction factor is generated by combining the dietary fiber inhibition coefficient, the protein source allocation weight, and the duration of effect parameter. This nutrient absorption correction coefficient integrates the impact of the above-mentioned multiple factors on nutrient absorption. It can fully reflect the nutrient absorption characteristics of individual patients. It can adjust the results of dietary nutrition assessment to make the assessment more accurate and personalized.

[0085] In an exemplary embodiment, performing time series alignment processing on the stratified nutritional quantification data and the continuous blood glucose monitoring data to establish a blood glucose response prediction model includes:

[0086] The LSTM-Transformer hybrid model is used to perform timestamp interpolation processing on the stratified nutritional quantification data and continuous blood glucose monitoring data to generate synchronized time series data.

[0087] According to the individual insulin sensitivity coefficient in the nutrient absorption correction coefficient, a Gaussian kernel blood glucose response function is constructed;

[0088] Obtain the patient's exercise intensity data within 1 hour after a meal and convert the exercise intensity data into metabolic equivalent values;

[0089] Calculate the metabolic compensation factor for the post-meal exercise period based on the metabolic equivalent value. The metabolic compensation factor is used to reduce the glycemic index of the corresponding post-meal exercise period.

[0090] The dynamic blood glucose prediction curve is generated by combining the synchronous time series data, Gaussian kernel blood glucose response function and metabolic compensation factor, and the blood glucose response prediction model is obtained.

[0091] Specifically, the LSTM (Long Short-Term Memory) network is a recurrent neural network that effectively processes time series data and can capture long-term dependencies within the data. The Transformer model, through its self-attention mechanism, can dynamically capture relationships between data without relying on a recurrent structure. A hybrid model combining these two methods can first perform timestamp interpolation on hierarchical nutritional quantitative data and continuous blood glucose monitoring data. This involves inserting estimated values ​​for missing time points, aligning data from different sources at the same time interval. After processing, synchronized time series data is generated, providing a unified data foundation for subsequent model development. The individual insulin sensitivity coefficient reflects a patient's insulin sensitivity and is a crucial component of the nutrient absorption correction factor. The Gaussian kernel function is a commonly used kernel function with excellent mathematical properties and flexibility, allowing for flexible shape adjustment to adapt to the blood glucose response characteristics corresponding to different individual insulin sensitivity coefficients. By incorporating the individual insulin sensitivity coefficient into the Gaussian kernel function, a functional model that describes the blood glucose response to dietary intake can be constructed. This functional model can predict the dynamic changes in blood glucose levels after dietary intake based on individual patient differences, providing a theoretical basis for subsequent blood glucose prediction.

[0092] Specifically, exercise intensity data can be collected through wearable devices such as smart bracelets and smart watches. These wearable devices can monitor a patient's exercise status, step count, heart rate, and other information, and calculate exercise intensity using corresponding algorithms. The metabolic equivalent value (MEV) is a metric that indicates exercise intensity, defined as the ratio of energy expenditure during exercise to energy expenditure at rest. Converting exercise intensity data into MEVs can more accurately quantify the impact of exercise on energy expenditure and blood glucose regulation. The MEVs can also be weighted, taking into account the cumulative effect of different exercise intensities over time and integrating them with exercise duration. This yields a factor that quantifies the effect of exercise compensation, known as the metabolic compensation factor. This MEV is used to attenuate the glycemic index corresponding to the post-meal exercise period, reflecting the inhibitory effect of exercise on blood glucose levels. It can be used to adjust the Glycemic Index in a blood glucose response prediction model, enabling the model to more accurately reflect the impact of exercise on blood glucose. Finally, synchronized time-series data can be used as input to the model, combined with a Gaussian kernel blood glucose response function, to predict blood glucose trends in the absence of exercise intervention. The prediction results are then adjusted based on a metabolic compensation factor to attenuate the glycemic index corresponding to the post-meal exercise period, generating a dynamic blood glucose prediction curve that accounts for the effects of exercise. This curve visually illustrates how blood glucose levels change over time after a patient consumes a specific diet and engages in appropriate exercise. Through multiple iterations and optimizations, a blood glucose response prediction model was ultimately developed. This model can accurately predict blood glucose changes based on a patient's diet, exercise, and medication use, providing a scientific basis for dietary management and treatment of diabetic patients.

[0093] In an exemplary embodiment, a dietary health assessment report is generated based on the blood glucose response prediction model and the nutrient absorption correction factor, including:

[0094] According to the vitamin B12 dynamic absorption coefficient, dietary fiber inhibition coefficient and protein source allocation weight in the nutrient absorption correction coefficient, the dynamic blood glucose prediction curve output by the blood glucose response prediction model is adjusted to generate a corrected blood glucose prediction curve;

[0095] Based on the corrected blood glucose prediction curve, the blood glucose fluctuation amplitude and the area under the curve are calculated;

[0096] Perform risk assessment on the blood sugar fluctuation amplitude according to the preset blood sugar fluctuation threshold and generate an early warning signal, which can be any one of a green safety signal, a yellow warning signal, and a red high-risk signal;

[0097] A multi-objective optimization algorithm is used to optimize the diet plan based on the revised blood glucose prediction curve to generate personalized diet recommendations, including a list of ingredient replacements and meal time windows.

[0098] A dietary health assessment report is generated by combining the revised blood sugar prediction curve, early warning signals and personalized dietary recommendations.

[0099] Specifically, the nutrient absorption correction factor reflects the impact of factors such as medication use, intestinal flora, and dietary fiber intake on nutrient absorption. Applying this factor to the blood glucose response prediction model can more accurately reflect an individual patient's blood glucose response. For example, the dynamic absorption coefficient of vitamin B12 affects energy metabolism, the dietary fiber inhibition coefficient slows carbohydrate absorption, and the weighting of protein sources affects protein digestion and blood glucose response. By incorporating these factors, the original blood glucose prediction curve can be adjusted to generate a revised blood glucose prediction curve that better reflects the patient's actual situation. The difference between the maximum and minimum values ​​of the revised blood glucose prediction curve can be used to calculate the blood glucose fluctuation amplitude. This blood glucose fluctuation amplitude reflects the stability of blood glucose levels. The area under the curve (AUC) can be calculated by integrating the revised blood glucose prediction curve. This AUC represents the overall blood glucose level over a specific timeframe. The blood glucose fluctuation amplitude and AUC can comprehensively reflect blood glucose control, providing a quantitative basis for subsequent risk assessment and dietary optimization.

[0100] Specifically, the preset blood sugar fluctuation threshold can be set based on medical standards and the patient's specific circumstances. When blood sugar fluctuations are within a safe range, a green safety signal is generated, indicating good blood sugar control. When blood sugar fluctuations approach or exceed the threshold, a yellow warning signal is generated, prompting the patient to adjust their diet and exercise. When blood sugar fluctuations significantly exceed the threshold, a red high-risk signal is generated, alerting the patient to potential health risks and requiring prompt action. Furthermore, a multi-objective optimization algorithm can be used to optimize the diet plan based on the revised blood sugar prediction curve. This algorithm comprehensively considers multiple objectives, such as blood sugar control, nutritional balance, and patient taste preferences, and generates personalized dietary recommendations by adjusting the ingredient types, intake amounts, and mealtimes within the diet plan. This personalized dietary recommendation can include an ingredient substitution list and a mealtime window. The ingredient substitution list provides patients with healthier ingredient alternatives, such as replacing refined grains with whole grains or replacing some animal protein with plant protein. The mealtime window can recommend appropriate mealtimes based on the patient's blood sugar fluctuation patterns and lifestyle to help better control blood sugar. Combining the revised blood sugar prediction curve, warning signals, and personalized dietary recommendations, a dietary health assessment report is generated. This assessment report comprehensively reflects the impact of a patient's diet on their blood sugar levels, presented through intuitive charts and text. This helps patients and doctors clearly understand the relationship between diet and blood sugar, and guides patients in scientific dietary management. For example, the report can compare the revised blood sugar prediction curve with the original curve, highlighting the effects of dietary adjustments, listing changes in early warning signals, emphasizing trends in blood sugar control, and detailing the specifics of personalized dietary recommendations, such as ingredient replacements and adjusted meal times, to provide patients with actionable guidance.

[0101] Based on the same inventive concept, Figure 2 As shown, the embodiment of the present application further provides a healthy diet index evaluation system 200 for diabetic patients, the system comprising:

[0102] The data acquisition and processing module 201 is used to obtain a multimodal dataset of the patient, which includes dietary image data, continuous blood glucose monitoring data, and medication usage parameters, and to perform mixed food ingredient analysis on the dietary image data to obtain layered nutritional quantitative data;

[0103] a drug use analysis module 202 for performing synergistic analysis based on drug use parameters and a drug metabolism kinetic model to generate a nutrient absorption correction coefficient;

[0104] The blood glucose response prediction module 203 is used to perform time series alignment processing on the stratified nutritional quantification data and the continuous blood glucose monitoring data to establish a blood glucose response prediction model;

[0105] The assessment report generating module 204 is used to generate a diet health assessment report based on the blood glucose response prediction model and the nutrient absorption correction coefficient.

[0106] In the above-mentioned healthy diet index assessment system 200 for diabetic patients, the data acquisition and processing module 201 can acquire a multimodal dataset containing the patient's diet image data, continuous blood glucose monitoring data, and medication usage parameters, and perform mixed food ingredient analysis on the diet image data. This allows for comprehensive and accurate acquisition of various nutritional information in the patient's diet, providing a rich data foundation for subsequent assessments. The medication usage analysis module 202 can perform synergistic analysis based on medication usage parameters and pharmacokinetic models, deeply exploring the potential relationship between medication and nutrient absorption. This avoids the dietary assessment bias caused by traditional assessment methods that fail to consider medication factors, making the dietary assessment more tailored to the patient's actual situation. The blood glucose response prediction module 203 can perform time-series alignment processing on the stratified nutritional quantitative data and continuous blood glucose monitoring data, accurately capturing the complex relationship between dietary intake and dynamic changes in blood glucose, thereby comprehensively and accurately predicting the patient's blood glucose response and improving the reliability of blood glucose predictions. The assessment report generation module 204 can generate a diet health assessment report based on the blood glucose response prediction model and nutrient absorption correction coefficient, providing scientific, comprehensive, and personalized diet health assessment recommendations for diabetic patients, helping them better manage their diet and control blood glucose.

[0107] Furthermore, the data acquisition processing module 201 includes a diet image processing subunit, which is used to:

[0108] A deep convolutional attention network is used to extract multi-scale features from food image data. A residual channel attention module is used to separate the outline boundaries of stacked ingredients in food image data to generate superpixel segmentation maps.

[0109] Based on a preset regional recipe database, the graph neural network is used to match the ingredient hierarchy in the superpixel segmentation image, identify hidden added sugar components, and obtain matched ingredient data;

[0110] The weight of the matched food data is calculated based on the near-infrared spectral characteristics to obtain the initial nutritional quantitative index;

[0111] Based on the initial nutritional quantitative index and gray-level co-occurrence matrix, the carbonized layer on the surface of fried ingredients in the dietary image data is compensated and corrected to generate layered nutritional quantitative data.

[0112] Furthermore, the drug use analysis module 202 includes:

[0113] Vitamin B12 analysis subunit for:

[0114] According to the absorption kinetics module based on the pharmacokinetic model and the daily dose of metformin in the drug usage parameters, a vitamin B12 absorption rate attenuation function is constructed to generate a vitamin B12 dynamic absorption coefficient;

[0115] The plasma concentration prediction module based on the pharmacokinetic model calculates the real-time vitamin B12 concentration prediction value according to the vitamin B12 dynamic absorption coefficient and the patient's baseline serum concentration;

[0116] When the real-time vitamin B12 concentration prediction value is greater than the preset threshold, an animal protein restriction instruction is generated. The animal protein restriction instruction is used to adjust the protein source allocation weight, increase the recommended proportion of plant protein, and output the protein source allocation weight;

[0117] The dietary fiber analysis subunit is used to calculate the dietary fiber inhibition coefficient based on the abundance ratio of Bifidobacterium / Prevotella through the intestinal metabolism module of the pharmacokinetic model and the patient's intestinal flora detection data;

[0118] A sulfonylurea drug analysis subunit is used to establish an inverse proportional relationship model between dietary fiber intake and drug effect duration through the drug effect duration module of the pharmacokinetic model when the drug use parameters include sulfonylurea drugs, and generate drug effect duration parameters;

[0119] The nutrient absorption correction subunit is used to generate a nutrient absorption correction coefficient by combining the dietary fiber inhibition coefficient, the protein source distribution weight and the drug effect duration parameter.

[0120] Furthermore, the blood glucose response prediction module 203 includes:

[0121] Data analysis subunit, used to:

[0122] The LSTM-Transformer hybrid model is used to perform timestamp interpolation processing on the stratified nutritional quantification data and continuous blood glucose monitoring data to generate synchronized time series data.

[0123] According to the individual insulin sensitivity coefficient in the nutrient absorption correction coefficient, a Gaussian kernel blood glucose response function is constructed;

[0124] Obtain the patient's exercise intensity data within 1 hour after a meal and convert the exercise intensity data into metabolic equivalent values;

[0125] Calculate the metabolic compensation factor for the post-meal exercise period based on the metabolic equivalent value. The metabolic compensation factor is used to reduce the glycemic index of the corresponding post-meal exercise period.

[0126] The model building subunit is used to combine the synchronous time series data, the Gaussian kernel blood glucose response function and the metabolic compensation factor to generate a dynamic blood glucose prediction curve and obtain a blood glucose response prediction model.

[0127] Furthermore, the evaluation report generating module 204 includes:

[0128] Blood glucose prediction subunit, used to:

[0129] According to the vitamin B12 dynamic absorption coefficient, dietary fiber inhibition coefficient and protein source allocation weight in the nutrient absorption correction coefficient, the dynamic blood glucose prediction curve output by the blood glucose response prediction model is adjusted to generate a corrected blood glucose prediction curve;

[0130] Based on the corrected blood glucose prediction curve, the blood glucose fluctuation amplitude and the area under the curve are calculated;

[0131] Perform risk assessment on the blood sugar fluctuation amplitude according to the preset blood sugar fluctuation threshold and generate an early warning signal, which can be any one of a green safety signal, a yellow warning signal, and a red high-risk signal;

[0132] The diet optimization subunit is used to optimize the diet plan based on the corrected blood glucose prediction curve using a multi-objective optimization algorithm and generate personalized diet recommendations. The personalized diet recommendations include a list of ingredient replacements and a meal time window.

[0133] The report output subunit is used to generate a dietary health assessment report by combining the revised blood glucose prediction curve, early warning signals and personalized dietary recommendations.

[0134] Furthermore, the system also includes a multimodal data acquisition module for:

[0135] Acquire dietary image data containing visible light bands and record the start and end time of eating;

[0136] Continuous blood glucose concentration data collected at 5-minute intervals were obtained. Continuous blood glucose monitoring data were converted from subcutaneous interstitial fluid glucose levels.

[0137] Obtain medication parameters including daily metformin dose and insulin injection regimen, including rapid-acting / long-acting insulin type and injection time;

[0138] Obtain the patient's energy expenditure intensity data within 1 hour after a meal;

[0139] Dietary image data, continuous blood glucose monitoring data, medication usage parameters, and energy expenditure intensity data are timestamp-aligned to generate a multimodal dataset.

[0140] In an exemplary embodiment, the present invention further provides a computer device comprising a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the steps of the method for assessing a healthy diet index for diabetic patients described herein. A multi-core processor is preferred to improve the system's parallel processing capabilities. The memory provides sufficient temporary storage space to support program execution and data processing. The memory capacity should be large enough to accommodate large amounts of supply information and computing tasks.

[0141] In an exemplary embodiment, the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a method for evaluating a healthy diet index for diabetic patients according to the present application. The computer-readable storage medium may include: a read-only memory (ROM), a random access memory (RAM), a solid-state drive (SSD), or an optical disc. Among them, the random access memory may include a resistance random access memory (ReRAM) and a dynamic random access memory (DRAM).

[0142] The above-described embodiments merely represent several implementation methods of the embodiments of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person skilled in the art may make various modifications and improvements without departing from the concept of the embodiments of the present application, and these modifications and improvements fall within the scope of protection of the embodiments of the present application.

Claims

1. A method for evaluating the healthy diet index of diabetic patients, characterized in that: The method comprises: Acquiring a multimodal dataset of the patient, the multimodal dataset comprising dietary image data, continuous blood glucose monitoring data, and medication usage parameters, and performing mixed food ingredient analysis on the dietary image data to obtain layered nutritional quantitative data; Performing synergistic analysis based on the drug usage parameters and the pharmacokinetic model to generate a nutrient absorption correction factor; Performing time series alignment processing on the stratified nutritional quantification data and the continuous blood glucose monitoring data to establish a blood glucose response prediction model; A diet health assessment report is generated based on the blood glucose response prediction model and the nutrient absorption correction coefficient.

2. The method according to claim 1, characterized in that The mixed food ingredient analysis processing is performed on the diet image data to obtain layered nutritional quantitative data, including: A deep convolutional attention network is used to extract multi-scale features from the food image data, and a residual channel attention module is used to separate the outline boundaries of stacked ingredients in the food image data to generate a superpixel segmentation map; According to a preset regional recipe database, the ingredient hierarchy in the superpixel segmentation image is matched based on a graph neural network to identify invisible added sugar components and obtain matched ingredient data; Calculating the weight of the matched food data based on near-infrared spectral characteristics to obtain an initial nutritional quantitative index; Based on the initial nutritional quantification index and the gray-level co-occurrence matrix, compensation correction is performed on the surface of the fried food in the dietary image data to generate the layered nutritional quantification data.

3. The method according to claim 1, characterized in that The synergistic analysis is performed based on the drug usage parameters and the drug metabolism kinetics model to generate a nutrient absorption correction coefficient, including: According to the absorption kinetics module based on the pharmacokinetic model and the daily dose of metformin in the drug usage parameters, a vitamin B12 absorption rate attenuation function is constructed to generate a vitamin B12 dynamic absorption coefficient; Calculating the dietary fiber inhibition coefficient based on the intestinal metabolism module of the pharmacokinetic model and the intestinal flora detection data of the patient and the abundance ratio of Bifidobacterium / Prevotella; The plasma concentration prediction module based on the pharmacokinetic model calculates a real-time vitamin B12 concentration prediction value according to the vitamin B12 dynamic absorption coefficient and the patient's baseline serum concentration; When the real-time vitamin B12 concentration prediction value is greater than a preset threshold, an animal protein restriction instruction is generated, wherein the animal protein restriction instruction is used to adjust the protein source allocation weight, increase the recommended proportion of plant protein, and output the protein source allocation weight; When the drug usage parameters include sulfonylurea drugs, an inverse proportional relationship model between dietary fiber intake and drug effect duration is established through the drug effect duration module of the pharmacokinetic model to generate a drug effect duration parameter; The nutrient absorption correction coefficient is generated by combining the dietary fiber inhibition coefficient, the protein source distribution weight and the drug effect duration parameter.

4. The method according to claim 1, wherein The step of performing time series alignment processing on the stratified nutritional quantification data and the continuous blood glucose monitoring data to establish a blood glucose response prediction model includes: Using an LSTM-Transformer hybrid model to perform timestamp interpolation processing on the stratified nutritional quantification data and the continuous blood glucose monitoring data to generate synchronized time series data; constructing a Gaussian kernel blood glucose response function according to the individual insulin sensitivity coefficient in the nutrient absorption correction coefficient; Obtaining the patient's exercise intensity data within 1 hour after a meal, and converting the exercise intensity data into a metabolic equivalent value; Calculating a metabolic compensation factor for the post-meal exercise period based on the metabolic equivalent value, wherein the metabolic compensation factor is used to attenuate the glycemic index corresponding to the post-meal exercise period; The synchronized time series data, the Gaussian kernel blood glucose response function and the metabolic compensation factor are combined to generate a dynamic blood glucose prediction curve to obtain the blood glucose response prediction model.

5. The method according to claim 3, characterized in that Generating a diet health assessment report based on the blood glucose response prediction model and the nutrient absorption correction coefficient includes: Adjusting the dynamic blood glucose prediction curve output by the blood glucose response prediction model according to the vitamin B12 dynamic absorption coefficient, the dietary fiber inhibition coefficient, and the protein source distribution weight in the nutrient absorption correction coefficient to generate a corrected blood glucose prediction curve; Calculating the blood glucose fluctuation amplitude and the area under the curve based on the corrected blood glucose prediction curve; Performing risk assessment on the blood sugar fluctuation amplitude according to a preset blood sugar fluctuation threshold and generating an early warning signal, wherein the early warning signal is any one of a green safety signal, a yellow early warning signal, and a red high-risk signal; A multi-objective optimization algorithm is used to optimize the diet plan based on the corrected blood glucose prediction curve to generate personalized diet recommendations, including a list of ingredient replacements and a meal time window; The diet health assessment report is generated by combining the revised blood sugar prediction curve, the early warning signal and the personalized diet recommendation.

6. The method according to claim 1, characterized in that The multimodal dataset is obtained by the following steps: Acquire the dietary image data containing the visible light band and record the start and end time of eating; Obtaining the continuous blood glucose concentration data collected at 5-minute intervals, wherein the continuous blood glucose monitoring data is obtained by converting the subcutaneous interstitial fluid glucose level; Obtaining the drug usage parameters including the daily metformin dosage and the insulin injection regimen, wherein the insulin injection regimen includes the type of rapid-acting / long-acting insulin and the injection time point; Acquiring the energy expenditure intensity data of the patient within 1 hour after a meal; The dietary image data, the continuous blood glucose monitoring data, the drug use parameters and the energy expenditure intensity data are timestamp aligned to generate the multimodal dataset.

7. A healthy diet index evaluation system for diabetic patients, characterized in that: The system comprises: a data acquisition and processing module, configured to acquire a multimodal dataset of a patient, the multimodal dataset comprising dietary image data, continuous blood glucose monitoring data, and medication usage parameters, and to perform mixed food ingredient analysis on the dietary image data to obtain layered nutritional quantitative data; a drug use analysis module, configured to perform synergistic analysis based on the drug use parameters and a drug metabolism kinetic model to generate a nutrient absorption correction coefficient; A blood glucose response prediction module, configured to perform time series alignment processing on the stratified nutritional quantification data and the continuous blood glucose monitoring data to establish a blood glucose response prediction model; An assessment report generating module is used to generate a diet health assessment report based on the blood glucose response prediction model and the nutrient absorption correction coefficient.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images