Chronic kidney disease and diabetes collaborative nutrition monitoring and early warning method and system

Abstract

The invention relates to the technical field of medical health management, in particular to a chronic kidney disease and diabetes collaborative nutrition monitoring and early warning method and system.The method comprises the steps that multi-dimensional physiological indexes, nutrition intake and sleep data of a patient are obtained, and a dynamically-coupled metabolic state topological space is constructed; based on the space, constructing a dual-disease metabolism coupling matrix, analyzing stability characteristics of a metabolism state trajectory, identifying stable and unstable attractors, and calculating a multi-dimensional risk score; a sleep-metabolism synchronism risk prediction model is introduced, and a personalized early warning threshold value is dynamically adjusted; when the risk score exceeds a threshold value, early warning is triggered, and intervention suggestions are generated, the interactive influence of protein metabolism and glycometabolism is quantitatively described from the metabolic mechanism level, and the contradiction of chronic kidney disease and diabetic patients in nutrition management is solved; limitation of a static threshold is broken through, and dynamic risk assessment is realized; sleep factors are creatively integrated, and a new way for improving the metabolic state is provided.

Description

Technical Field

[0001] This invention relates to the field of medical and health management technology, specifically to a method and system for synergistic nutritional monitoring and early warning of chronic kidney disease and diabetes. Background Technology

[0002] Chronic kidney disease (CKD) and diabetes mellitus (DM) are two prevalent chronic diseases that often coexist and influence each other. In clinical practice, there is a significant contradiction in the nutritional management of these two diseases: diabetic patients need to control their carbohydrate intake to stabilize blood sugar, but reducing carbohydrates often leads to increased protein intake; however, CKD patients need to restrict protein intake to reduce the burden on their kidneys. This contradiction presents patients with both diseases with complex nutritional management challenges.

[0003] Currently, clinical nutritional management of chronic kidney disease and diabetes mostly adopts a single-disease approach, that is, developing separate dietary plans for each disease. For example, chronic kidney disease is managed by setting a fixed upper limit for protein intake, while diabetes is managed by using a carbohydrate counting method. This approach ignores the interaction between the two metabolic pathways and makes it difficult to find the optimal balance in managing the two diseases.

[0004] Furthermore, existing nutrition management systems often rely on static thresholds to assess risk, such as fixed daily protein or carbohydrate intake limits. This approach fails to consider individual patient differences and dynamic changes in their condition, leading to either over- or under-warnings. Simultaneously, existing systems generally neglect the profound impact of lifestyle factors such as sleep on metabolism, making it difficult to explain differences in metabolic responses among patients under the same diet and medication conditions.

[0005] Currently, there is a lack of collaborative nutrition monitoring and early warning systems on the market that can simultaneously consider the characteristics of both chronic kidney disease and diabetes, and integrate multi-dimensional physiological data, dietary data, and sleep data. Summary of the Invention

[0006] In view of the shortcomings of the prior art, the purpose of this invention is to provide a method and system for synergistic nutritional monitoring and early warning of chronic kidney disease and diabetes. This method and system can quantitatively describe the interaction between protein metabolism and glucose metabolism at the metabolic mechanism level, realize accurate modeling and risk prediction of complex metabolic relationships, and solve the contradictions in nutritional management of patients with chronic kidney disease and diabetes.

[0007] This invention proposes a method for synergistic nutritional monitoring and early warning of chronic kidney disease and diabetes, including:

[0008] The system acquires multidimensional physiological indicators, nutritional intake data, and sleep data from patients. The multidimensional physiological indicators include blood glucose, renal function, cardiovascular, and electrolyte levels. The nutritional intake data includes protein intake, carbohydrate intake, fat intake, and the distribution of meal times. The sleep data includes sleep duration, sleep quality, and sleep-wake cycles.

[0009] Based on the multi-dimensional physiological indicator data, the nutritional intake data, and the sleep data, a dynamically coupled metabolic state topology space is constructed, including:

[0010] The multidimensional physiological indicator data, the nutritional intake data, and the sleep data are converted into standardized feature vectors.

[0011] Based on the standardized feature vector, the location of the patient's current metabolic state point is determined in the multidimensional topological space;

[0012] A metabolic state trajectory is constructed by continuously monitoring the changes in the positions of the metabolic state points;

[0013] Based on the aforementioned metabolic state topological space, a nonlinear co-metabolic risk measure is established, including:

[0014] Construct a metabolic coupling matrix for two diseases to describe the interaction between protein metabolism and glucose metabolism;

[0015] Analyze the stability characteristics of the metabolic state trajectory to identify stable and unstable attractors in the metabolic system;

[0016] Based on topological distance, rate of change of state and stability analysis, a multi-dimensional risk score is calculated.

[0017] Based on the multidimensional risk score and the sleep data, a sleep-metabolism synchronicity risk prediction model is constructed, including:

[0018] Identify the correlation between sleep patterns and metabolic state;

[0019] Construct a time-dependent probability field describing how risk changes over time;

[0020] Set and dynamically adjust personalized early warning thresholds;

[0021] When the multi-dimensional risk score exceeds the personalized early warning threshold, an early warning message of the corresponding level is triggered, and personalized intervention suggestions are generated.

[0022] Preferably, the acquisition of the patient's multi-dimensional physiological indicators, nutritional intake data, and sleep data specifically includes:

[0023] Obtain real-time blood glucose data using a continuous glucose monitor or electronic blood glucose meter;

[0024] Kidney function indicators are obtained through hospital test reports or home testing devices. These indicators include serum creatinine, blood urea nitrogen, urine albumin to creatinine ratio, and glomerular filtration rate.

[0025] Cardiovascular indicator data, including blood pressure, heart rate, and blood lipid profile, are obtained through smart wearable devices or smart home devices.

[0026] The type and quantity of food are automatically identified using image recognition or voice recognition technology, and the nutritional intake data is calculated and obtained.

[0027] The sleep data is obtained through monitoring via smart bracelets or mattress sensors.

[0028] Preferably, the step of converting the multi-dimensional physiological indicator data, the nutritional intake data, and the sleep data into a standardized feature vector specifically includes:

[0029] Standardize the indicator data of different dimensions;

[0030] Extract time-domain features, frequency-domain features, morphological features, and correlation features;

[0031] Principal component analysis is used to reduce feature dimensionality;

[0032] Weights are assigned to features based on their clinical significance, and an n-dimensional feature vector is constructed.

[0033] Preferably, the construction of the metabolic state trajectory specifically includes:

[0034] Connect the patient's metabolic state points at different time points to form a temporal state trajectory;

[0035] Calculate the direction vector, velocity vector, and acceleration vector of the state trajectory;

[0036] Identify the pattern characteristics of the state trajectory, including stable mode, deterioration mode, improvement mode and fluctuation mode;

[0037] Analyze the fractal dimension of the state trajectory to assess its complexity.

[0038] Preferably, the construction of the dual-disease metabolic coupling matrix describing the interaction between protein metabolism and glucose metabolism specifically includes:

[0039] Calculate the vector of the effect of protein intake on glycemic stability;

[0040] Calculate the vector of the effect of carbohydrate intake on renal functional burden;

[0041] Calculate the cross-sensitivity coefficient between the two metabolic pathways;

[0042] Introducing the time lag effect, we analyzed the short-term and long-term effects of intake patterns on metabolism;

[0043] The coupling matrix parameters are adjusted individually based on the patient's historical data.

[0044] Preferably, the analysis of the stability characteristics of the metabolic state trajectory specifically includes:

[0045] Calculate the regression characteristics of metabolic state points after perturbation;

[0046] Identify stable attractor regions that represent a healthy state;

[0047] Identify unstable attractor regions that represent risk states;

[0048] Identify the saddle point region representing the state bifurcation point;

[0049] Calculate the Lyapunov exponent of the state trajectory to quantify the system stability.

[0050] Preferably, the calculation of the multi-dimensional risk score based on topological distance, rate of change of state, and stability analysis specifically includes:

[0051] Calculate the shortest topological distance from the current metabolic state to the risk boundary;

[0052] Calculate the velocity and acceleration components of the state trajectory toward the risk area;

[0053] Assess the sensitivity of the current state to disturbances;

[0054] A comprehensive risk score is calculated using a weighted fusion method.

[0055] Based on historical data, the weighting coefficients of each dimension of risk score are dynamically adjusted.

[0056] Preferably, the identification of the association between sleep patterns and metabolic states specifically includes:

[0057] Analyze the metabolic response characteristics under conditions of sufficient sleep;

[0058] Analyze the metabolic response characteristics under sleep deprivation;

[0059] Analyze the metabolic response characteristics under sleep disturbance conditions;

[0060] Calculate the regulatory factors of sleep quality on metabolic efficiency;

[0061] Establish a classification model for sleep-metabolism synchronicity.

[0062] Preferably, the setting and dynamic adjustment of personalized early warning thresholds specifically includes:

[0063] Initial warning thresholds are set based on medical guidelines;

[0064] Adjust the baseline thresholds based on disease severity and individual characteristics;

[0065] The threshold sensitivity is dynamically adjusted based on patient feedback.

[0066] Establish a multi-level early warning threshold architecture, including alert level, warning level, intervention level, and emergency level;

[0067] Adjust the threshold fluctuation range based on time factors, such as circadian rhythms and seasonal changes.

[0068] A multidimensional metabolic state-based collaborative nutrition monitoring and early warning system for chronic kidney disease and diabetes includes:

[0069] The data acquisition module is used to acquire multi-dimensional physiological indicator data, nutritional intake data, and sleep data of patients. The multi-dimensional physiological indicator data includes blood glucose, renal function, cardiovascular, and electrolyte indicators. The nutritional intake data includes protein intake, carbohydrate intake, fat intake, and the distribution of meal times. The sleep data includes sleep duration, sleep quality, and sleep-wake cycle.

[0070] A state modeling module is used to construct a dynamically coupled metabolic state topology space based on the multi-dimensional physiological indicator data, the nutritional intake data, and the sleep data; it includes: a feature processing unit, used to convert the multi-dimensional physiological indicator data, the nutritional intake data, and the sleep data into standardized feature vectors; a spatial mapping unit, used to determine the current metabolic state point position of the patient in the multi-dimensional topology space based on the standardized feature vectors; and a trajectory analysis unit, used to construct a metabolic state trajectory by continuously monitoring the changes in the metabolic state point position.

[0071] The risk assessment module is used to establish a nonlinear synergistic metabolic risk measure based on the topological space of the metabolic state; it includes: a coupling analysis unit for constructing a dual-disease metabolic coupling matrix describing the interaction between protein metabolism and glucose metabolism; a stability analysis unit for analyzing the stability characteristics of the metabolic state trajectory and identifying stable and unstable attractors in the metabolic system; and a risk calculation unit for calculating a multi-dimensional risk score based on topological distance, state change rate, and stability analysis.

[0072] The prediction and early warning module is used to construct a sleep-metabolism synchronicity risk prediction model based on the multi-dimensional risk score and the sleep data; it includes: a correlation analysis unit for identifying the correlation pattern between sleep patterns and metabolic states; a time-series modeling unit for constructing a time-dependent probability field describing the change of risk over time; a threshold management unit for setting and dynamically adjusting personalized early warning thresholds; and an early warning triggering unit for triggering corresponding level of early warning information and generating personalized intervention suggestions when the multi-dimensional risk score exceeds the personalized early warning threshold.

[0073] The present invention has the following beneficial effects:

[0074] 1. Solved the core problem of the interaction between metabolic pathways: By establishing a multidimensional metabolic state space and metabolic coupling matrix, the interaction between protein metabolism and glucose metabolism was quantitatively described at the level of metabolic mechanism, and the optimal balance point of the two metabolic needs was found, thus solving the long-standing treatment contradiction in clinical practice.

[0075] 2. Overcoming the limitations of static thresholds: A dynamic risk assessment method based on metabolic state trajectory and attractor theory is introduced, which enables the warning threshold to be dynamically adjusted according to the patient's real-time status, historical response patterns and disease progression stage, greatly improving the accuracy and reliability of the warning.

[0076] 3. Innovative integration of sleep factors: A sleep-metabolism synchronicity model was established, revealing the intrinsic link between sleep patterns and metabolic efficiency. By optimizing sleep strategies, a new approach to improve metabolic status was provided, and the accuracy of predicting blood glucose fluctuations and protein metabolism abnormalities was improved.

[0077] 4. Achieved true closed-loop optimization: A complete closed-loop management mechanism was designed, including risk prediction, early warning triggering, intervention suggestions, effect evaluation, and model optimization. The system can learn from the intervention effect, continuously optimize early warning strategies and intervention suggestions, and improve early warning accuracy and user acceptance. Attached Figure Description

[0078] Fig. 1 This is a flowchart of the method for synergistic nutritional monitoring and early warning of chronic kidney disease and diabetes provided in the embodiments of the present invention;

[0079] Fig. 2 This is a structural block diagram of a collaborative nutrition monitoring and early warning system for chronic kidney disease and diabetes based on a multidimensional metabolic state space, provided in an embodiment of the present invention. Detailed Implementation

[0080] Please refer to Figs. 1-2 The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0081] Reference Fig. 1This invention provides a method for synergistic nutritional monitoring and early warning of chronic kidney disease and diabetes, comprising the following steps:

[0082] We acquired multidimensional physiological data, nutritional intake data, and sleep data from patients. The multidimensional physiological data included blood glucose, renal function, cardiovascular, and electrolyte levels; the nutritional intake data included protein intake, carbohydrate intake, fat intake, and the distribution of meal times; and the sleep data included sleep duration, sleep quality, and sleep-wake cycles.

[0083] Specifically, this invention acquires patient data through multiple channels: real-time blood glucose data is acquired through a continuous glucose monitor (CGM) or an electronic blood glucose meter, with the preferred acquisition frequency being once every 5 minutes for CGM and measurements taken as needed by the electronic blood glucose meter, such as at key time points like before meals and two hours after meals; renal function index data, including serum creatinine (preferred range 44-133 μmol / L), blood urea nitrogen (preferred range 3.2-7.1 mmol / L), urine albumin to creatinine ratio (preferred range <30 mg / g), and glomerular filtration rate (preferred range >90 mL / min / 1.73 m²), are acquired through smart wearable devices or smart home devices; cardiovascular index data, including blood pressure (preferred range <140 / 90 mmHg), heart rate (preferred range 60-100 beats / min), and lipid profile data, are acquired through smart wearable devices or smart home devices.

[0084] For acquiring nutritional intake data, this invention employs image recognition or voice recognition technology to automatically identify food types and portions. For image recognition, patients take photos of food using their smartphones, and the system uses a deep learning model to automatically identify the food type and estimate the portion size. Then, it calculates the protein, carbohydrate, and fat content based on a food nutrient database. For voice recognition, the system converts the patient's voice description into text, extracts key information, and maps it to a standardized food database. Furthermore, the system also supports patients manually recording their dietary information as a supplement.

[0085] Sleep data is monitored through smart bracelets or mattress sensors, recording indicators such as sleep duration (preferred range 7-9 hours), sleep efficiency (preferred value >85%), deep sleep ratio (preferred range 20-25%), and number of sleep interruptions.

[0086] Based on the aforementioned multi-dimensional physiological indicator data, nutritional intake data, and sleep data, a dynamically coupled metabolic state topological space is constructed. This process includes the following sub-steps:

[0087] Multidimensional physiological indicator data, nutritional intake data, and sleep data are converted into standardized feature vectors. Specifically, the indicator data of different dimensions are first standardized using the Z-score standardization method.

[0088] ,

[0089] in: These are the standardized eigenvalues. These are the original eigenvalues. The characteristic mean, The characteristic standard deviation is denoted as .

[0090] Next, time-domain features (such as mean, standard deviation, and coefficient of variation), frequency-domain features (spectral features extracted via Fast Fourier Transform), morphological features (morphological features of the fluctuation curve), and correlation features (correlation between different indicators) are extracted. For blood glucose data, key time-domain features include average blood glucose level (preferred range 4.4-7.8 mmol / L), blood glucose coefficient of variation (preferred value <36%), and blood glucose fluctuation amplitude (preferred value <3.0 mmol / L). For renal function data, key features include the rate of decline in glomerular filtration rate and the trend of changes in proteinuria.

[0091] To reduce computational complexity, this invention employs Principal Component Analysis (PCA) to reduce feature dimensionality, retaining principal components that contribute up to 95% of the cumulative variance. Finally, features are weighted according to clinical significance to construct an n-dimensional feature vector, where n is typically chosen between 15 and 25 to balance computational complexity and model accuracy.

[0092] Based on standardized feature vectors, the current metabolic state of a patient is determined in a multidimensional topological space. This invention maps the patient's feature vectors to an n-dimensional topological space, where each dimension corresponds to a key feature. In this space, distance has clear clinical significance: closer distances indicate more similar metabolic states, while greater distances indicate greater differences in metabolic states.

[0093] Preferably, the present invention uses Mahalanobis distance to measure distances in the feature space, taking into account the correlation between different features:

[0094] ,

[0095] in: The Mahalanobis distance between two points. and For feature vectors, Let be the characteristic covariance matrix.

[0096] To facilitate visualization and understanding, the system can project high-dimensional space onto three-dimensional or two-dimensional space. For example, the t-SNE algorithm can be used for dimensionality reduction, preserving local structural features in the high-dimensional space.

[0097] By continuously monitoring changes in the location of metabolic state points, a metabolic state trajectory is constructed. The system connects the patient's metabolic state points at different time points to form a temporal state trajectory, and calculates the direction vector, velocity vector, and acceleration vector of the trajectory.

[0098] The formula for calculating the direction vector is:

[0099] ,

[0100] The formula for calculating the velocity vector is:

[0101] ,

[0102] The formula for calculating the acceleration vector is:

[0103] ,

[0104] in: and Let be the state point position vectors at time t and time t-1, respectively. For time intervals, It is a direction vector. For velocity vector, This is the acceleration vector.

[0105] In addition, the system identifies the pattern characteristics of the state trajectory, including stable mode (state point fluctuates within a small range in the healthy area), deterioration mode (state point gradually moves towards the risk area), improvement mode (state point gradually moves towards the healthy area) and fluctuation mode (state point moves back and forth between the healthy area and the risk area).

[0106] To analyze trajectory complexity more deeply, the system calculates the fractal dimension of the trajectory using the box-counting method:

[0107] ,

[0108] in: For fractal dimension, The size required to cover the trajectory is The number of boxes. Fractal dimension reflects the complexity and irregularity of the trajectory. A higher fractal dimension (such as close to 2) indicates that the patient's metabolic state changes are more complex and irregular, which may require more refined monitoring and intervention.

[0109] A nonlinear co-metabolic risk measure is established based on the topological space of metabolic states. This process includes the following sub-steps:

[0110] A metabolic coupling matrix for two diseases, describing the interaction between protein metabolism and glucose metabolism, is constructed. This matrix is ​​one of the core innovations of this invention, as it quantitatively describes the interaction between the two metabolic pathways at the metabolic mechanism level.

[0111] First, calculate the vector of the effect of protein intake on glycemic stability. The system analyzes blood glucose response curves under different protein intake levels, extracting key features such as peak blood glucose, rate of blood glucose rise, and recovery time to construct an influence vector. For example, for patients with mild renal insufficiency (GFR 60-89 mL / min / 1.73 m²), for every 10g increase in protein intake, the average postprandial peak blood glucose increases by 0.3-0.5 mmol / L, and the recovery time is prolonged by 15-25 minutes.

[0112] Secondly, calculate the vector effect of carbohydrate intake on renal function burden. The system analyzed changes in renal filtration burden under different carbohydrate intake patterns, including changes in glomerular filtration rate and urinary protein excretion rate. For example, in patients with mild diabetes (HbA1c 6.5-7.0%), for every 50g increase in carbohydrate intake, the glomerular filtration rate increased by an average of 5-8% in the short term, and the urinary albumin excretion rate increased by 10-15%.

[0113] Then, the cross-sensitivity coefficient matrix S of the two metabolic pathways is calculated, which represents the strength of the effect of one metabolic change on the other:

[0114] ,

[0115] in: This represents the sensitivity coefficient of protein metabolism itself. This represents the coefficient indicating the influence of carbohydrate metabolism on protein metabolism. This represents the coefficient indicating the influence of protein metabolism on carbohydrate metabolism. This represents the sensitivity coefficient of carbohydrate metabolism itself.

[0116] Sensitivity coefficients are obtained by analyzing historical data. For example, a typical sensitivity coefficient matrix for a patient with moderate chronic kidney disease and diabetes might look like this:

[0117] ,

[0118] This indicates that protein metabolism is 0.8 sensitive to its own changes and 0.3 sensitive to changes in carbohydrates; while carbohydrate metabolism is 0.4 sensitive to changes in proteins and 0.7 sensitive to its own changes.

[0119] In addition, the system incorporates a time lag effect to analyze the short-term and long-term impacts of intake patterns on metabolism. For short-term effects, the system primarily focuses on metabolic changes within 2–4 hours postprandial; for long-term effects, the system focuses on the cumulative effects over 7–30 days.

[0120] Finally, the system adjusts the coupling matrix parameters individually based on the patient's historical data. For example, for patients more sensitive to protein metabolism abnormalities, the parameters are increased. and The weighting; for patients with significant blood glucose fluctuations, increase and The weight.

[0121] Analyzing the stability characteristics of metabolic state trajectories helps identify stable and unstable attractors in the metabolic system. This step introduces stability analysis methods from dynamical systems theory.

[0122] First, the regression characteristics of metabolic state points after perturbation are calculated. The system simulates the trajectory of metabolic state changes after dietary perturbation (such as additional intake of 10g protein or 30g carbohydrates). If the state point can return to its original state within a certain time (e.g., 24 hours), the state is considered stable; otherwise, the state is considered unstable.

[0123] Next, the system identifies stable attractor regions representing a healthy state. These regions typically correspond to a set of metabolic states characterized by stable blood glucose levels and normal renal function indicators. For example, for a patient with mild renal insufficiency and mild diabetes, stable attractor regions might correspond to a state space region with an average blood glucose level of 4.4–6.7 mmol / L, a blood glucose variability of <25%, and a urinary albumin to creatinine ratio of <20 mg / g.

[0124] Simultaneously, the system identifies unstable attractor regions representing risk states. These regions typically correspond to a set of metabolic states characterized by significant blood glucose fluctuations and abnormal renal function indicators. For example, state spaces with average blood glucose >8.5 mmol / L, blood glucose variability >40%, and urinary albumin to creatinine ratio >50 mg / g.

[0125] Furthermore, the system identifies saddle point regions that represent bifurcation points in the metabolic state. These regions are key bifurcation points where the metabolic state may evolve towards a healthy or risky direction. In these regions, small interventions can lead to different outcomes, making them a key focus for early warning and intervention.

[0126] Finally, the Lyapunov exponent of the state trajectory is calculated to quantify the system stability:

[0127] ,

[0128] in: The Lyapunov index, For the initial small perturbation, Let t be the magnitude of the disturbance development at time t. A positive Lyapunov exponent indicates that the system has chaotic characteristics and is sensitive to initial conditions; a negative Lyapunov exponent indicates that the system tends to a stable state.

[0129] For patients with chronic kidney disease and diabetes, the Lyapunov index is typically between -0.5 and 0.5. A value >0.2 indicates that the patient's metabolic state is highly sensitive to disturbances, requiring more frequent monitoring and more cautious intervention; a value <-0.2 indicates that the patient's metabolic state is relatively stable, and the monitoring frequency can be appropriately relaxed.

[0130] Based on topological distance, rate of change of state, and stability analysis, a multi-dimensional risk score is calculated. This step comprehensively considers multiple risk factors to form a comprehensive risk assessment system.

[0131] First, calculate the shortest topological distance from the current metabolic state to the risk boundary. The risk boundary is defined by medical experts based on clinical guidelines or learned from historical data using machine learning algorithms. The formula for calculating the topological distance risk score is:

[0132] ,

[0133] in: To score the risk of topological distance, The shortest distance from the current state point to the risk boundary. This is the distance attenuation factor (typically 0.5-1.5). The value range is 0-1, and the larger the value, the higher the risk.

[0134] Secondly, calculate the velocity and acceleration components of the trajectory toward the risk area. The formula for calculating the velocity risk score is:

[0135] ,

[0136] in: For speed risk scoring, The velocity vector of state change. It is a vector pointing from the current state point to the nearest risk boundary. The value range is 0-1, with positive values ​​indicating that the state is moving towards the risk area, and larger values ​​indicating a greater component of the movement velocity in the risk direction. The formula for calculating the acceleration risk score is similar:

[0137] ,

[0138] in: For acceleration risk scoring, This is the acceleration vector for state change.

[0139] Then, assess the sensitivity of the current state to resistance to motion and calculate the stability risk score:

[0140] ,

[0141] in: For stability risk scoring, The Lyapunov index, This is the sensitivity adjustment coefficient (typically 5-10). The value range is 0-1. The larger the value, the more unstable the system and the higher the risk.

[0142] Finally, a comprehensive risk score is calculated using a weighted fusion method:

[0143] ,

[0144] in: For comprehensive risk scoring, , , , These are the weighting coefficients for each risk score, and they satisfy... .

[0145] In the initial stage, the system can use balanced weights, such as , , , As the system collects more data, it will dynamically adjust the weighting coefficients of each dimension of the risk score based on historical data, improving the accuracy of risk assessment. For example, for patients with significant fluctuations in metabolic status, the weighting coefficients will be increased. and The weighting; for patients nearing the risk boundary, increase The weight.

[0146] A sleep-metabolic synchrony risk prediction model was constructed based on multidimensional risk scores and sleep data. This process includes the following sub-steps:

[0147] Identifying the correlation between sleep patterns and metabolic state. This step is another innovation of the present invention, as it is the first time that the correlation between sleep and metabolism has been systematically analyzed in clinical practice.

[0148] First, the system analyzed the metabolic response characteristics under sufficient sleep conditions. For patients with sufficient sleep (usually defined as 7-9 hours / night and sleep efficiency >85%), the system recorded metabolic indicators such as blood glucose stability and changes in renal function. For example, under sufficient sleep conditions, the average postprandial blood glucose peak decreased by 0.5-1.0 mmol / L, the blood glucose recovery time was shortened by 15-25%, and the urinary albumin excretion rate decreased by 8-12%.

[0149] Secondly, the metabolic response characteristics under sleep deprivation were systematically analyzed. For patients with sleep deprivation (<6 hours / night or sleep efficiency <70%), common metabolic changes included: increased fasting blood glucose by 0.3-0.7 mmol / L, increased blood glucose variability by 15-25%, decreased insulin sensitivity by 10-20%, and increased urinary albumin excretion rate by 10-15%.

[0150] Then, the metabolic response characteristics under sleep disturbance states were systematically analyzed. Sleep disturbances include frequent sleep interruptions (>5 times / night), low proportion of deep sleep (<15%), and abnormal sleep phases (such as circadian rhythm disorders). These states are usually associated with more serious metabolic abnormalities, such as elevated fasting blood glucose of 0.5-1.2 mmol / L, increased blood glucose variability by 20-35%, and increased urinary albumin excretion rate by 15-25%.

[0151] Based on the above analysis, the system calculates the regulatory factors of sleep quality on metabolic efficiency. :

[0152] ,

[0153] in: As a sleep regulator, The overall sleep quality score is between 0 and 1, with 1 indicating the best sleep quality. The sleep impact factor (typical value is 0.2-0.4). The value range is 0.6-1.0, and the smaller the value, the greater the negative impact of sleep on metabolism.

[0154] Finally, the system establishes a sleep-metabolism synchronicity classification model, classifying patients' sleep-metabolism relationship patterns into the following categories:

[0155] Normal synchronous type: Normal sleep and stable metabolism;

[0156] Sleep deprivation-related: Sleep deprivation leads to metabolic abnormalities;

[0157] Metabolic factors affecting sleep: Abnormal metabolism leads to decreased sleep quality;

[0158] Disordered Cycle Type: Sleep and metabolism influence each other, forming a vicious cycle;

[0159] Identifying which patient pattern they belong to helps in developing targeted intervention strategies. For example, patients with normal synchronous sleep patterns should focus on dietary management, while patients with sleep deprivation-related sleep patterns should prioritize improving sleep quality.

[0160] Construct a time-dependent probability field describing how risk changes over time. This step takes into account the temporal dynamics of risk and improves the timeliness of early warning.

[0161] First, the system divides the day into multiple time windows: the fasting period in the morning (6:00-8:00), the late breakfast period (8:00-10:30), the early lunch period (10:30-12:00), the late lunch period (12:00-14:30), the early dinner period (14:30-18:00), the late dinner period (18:00-20:30), and the night period (20:30-6:00). Specific risk assessment logic is then constructed for each time window.

[0162] Secondly, based on historical data, the system calculates the risk probability distribution P(R|t) for different time windows, representing the probability of having a risk level of R within time window t. For example, for patients with poorly controlled blood sugar, the high-risk probability in the late breakfast period is usually higher than at other times; while for patients with significant proteinuria, the high-risk probability may be even higher during the night.

[0163] The system then identifies high-risk and low-risk time windows. High-risk time windows typically represent the probability of risk. During periods of high risk, closer monitoring and preventative intervention are needed; low-risk time windows represent the probability of risk. During certain time periods, monitoring requirements can be appropriately relaxed.

[0164] Furthermore, the system analyzes the periodic patterns of risk probability changes over time, such as intraday, weekly, and monthly variations. These periodic patterns reflect the influence of patients' lifestyle habits and physiological rhythms, which helps to predict risks more accurately.

[0165] Finally, based on historical time series patterns, the system predicts the risk probability distribution for future time windows. This indicates that the current risk status is... hour, The conditional probability distribution of risk status after a certain time. For example, the system can predict the high probability of a patient being at high risk 2 hours after dinner and issue an early warning.

[0166] Set and dynamically adjust personalized alert thresholds. This step ensures the accuracy and personalization of alerts, avoiding both over-alerts and missed alerts.

[0167] First, the system sets an initial warning threshold based on medical guidelines. For example, for patients with moderate chronic kidney disease (GFR 30-59 mL / min / 1.73 m²) and moderate diabetes mellitus (HbA1c 7.0-8.0%), the initial high-risk threshold can be set to... The medium-risk threshold is The low-risk threshold is .

[0168] Secondly, the system adjusts the baseline threshold based on disease severity and individual characteristics. For patients with more severe disease (e.g., GFR < 30 mL / min / 1.73 m² or HbA1c > 8.0%), the risk threshold is lowered (e.g., the high-risk threshold is adjusted to...). For patients with stable conditions, the risk threshold should be appropriately increased.

[0169] The system then dynamically adjusts the threshold sensitivity based on patient feedback. If a patient frequently receives warnings but no adverse event actually occurs, the system gradually increases the threshold; if the patient's feedback on warnings is not timely, the system appropriately decreases the threshold. The threshold adjustment adopts a gradual strategy, with each adjustment not exceeding 10% to ensure system stability.

[0170] The system is configured with a multi-level early warning threshold architecture, including alert level, warning level, intervention level, and emergency level:

[0171] Alert Level: A value >0.4 indicates that the patient should be aware of potential risks;

[0172] Warning level: If the value is >0.6, it is recommended that the patient take preventative measures.

[0173] Intervention level: A value >0.75 indicates the need for immediate intervention.

[0174] Emergency Level: A value >0.9 indicates a need to seek medical assistance.

[0175] Finally, the system adjusts the threshold fluctuation range based on time factors, such as circadian rhythms and seasonal changes. For example, during patient sleep, the warning threshold is appropriately increased to reduce unnecessary disturbances; during seasonal transitions, the threshold is appropriately decreased to increase monitoring sensitivity.

[0176] When the multi-dimensional risk score exceeds the personalized early warning threshold, the corresponding level of early warning information is triggered, and personalized intervention suggestions are generated.

[0177] The system uses different notification methods for different levels of alerts: reminder alerts are sent via APP push notifications; warning alerts are sent via APP push notifications and sound alerts; intervention alerts are sent via APP push notifications, sound alerts, and SMS notifications; and emergency alerts are sent via APP push notifications, sound alerts, SMS notifications, and automated phone calls.

[0178] The warning information includes the risk type, risk level, possible causes, and recommended interventions. For example, for a detected risk to kidney function caused by high protein intake, the system will prompt, "Protein intake has exceeded the tolerance threshold, which may increase the burden on the kidneys. It is recommended to reduce protein intake at dinner and increase the proportion of high-quality carbohydrates."

[0179] Personalized intervention recommendations are generated based on risk factors and the patient’s historical response, including dietary adjustment recommendations (such as adjusting the intake and ratio of protein and carbohydrates), meal timing recommendations (such as optimizing the time distribution of meals throughout the day), activity adjustment recommendations (such as appropriate post-meal activity), and sleep optimization recommendations (such as improving the sleep environment).

[0180] The system also implements a closed-loop feedback mechanism to evaluate the effectiveness of interventions and continuously optimize early warning and intervention strategies. By collecting patient feedback and monitoring changes in physiological indicators after intervention, the system can learn which interventions are most effective for specific patients, continuously improving the accuracy of early warnings and the effectiveness of interventions.

[0181] Reference Fig. 2 The present invention also provides a synergistic nutritional monitoring and early warning system for chronic kidney disease and diabetes based on a multidimensional metabolic state space, including a data acquisition module 1, a state modeling module 2, a risk assessment module 3, and a prediction and early warning module 4.

[0182] Data acquisition module 1 is used to acquire multi-dimensional physiological indicator data, nutritional intake data, and sleep data from patients. This module contains several sub-modules, each responsible for acquiring different types of data:

[0183] The physiological indicator acquisition submodule 11 is responsible for collecting blood glucose, renal function, cardiovascular, and electrolyte indicators. This submodule connects to medical devices such as continuous glucose monitors, electronic blood glucose meters, and smart blood pressure monitors via standard interfaces and can import data from hospital laboratory reports.

[0184] The nutrition information collection submodule 12 is responsible for collecting data such as protein intake, carbohydrate intake, fat intake, and the distribution of meal times. This submodule integrates image recognition and voice recognition technologies, which can automatically identify food types and estimate portion sizes, while also supporting manual recording by patients.

[0185] The sleep monitoring submodule 13 is responsible for collecting data such as sleep duration, sleep quality, and sleep-wake cycles. This submodule automatically records the patient's sleep data by integrating with smart bracelets, smartwatches, or smart mattresses.

[0186] The data preprocessing submodule 14 is responsible for cleaning, standardizing and extracting features from the raw data, providing high-quality data input for subsequent analysis.

[0187] State modeling module 2 is used to construct a dynamically coupled metabolic state topology space based on multi-dimensional physiological indicator data, nutritional intake data, and sleep data. This module includes the following sub-modules:

[0188] The feature processing unit 21 is used to convert multi-dimensional physiological indicator data, nutritional intake data, and sleep data into standardized feature vectors. This unit implements functions such as data standardization, feature extraction, and dimensionality reduction.

[0189] The spatial mapping unit 22 is used to determine the location of the patient's current metabolic state point in a multidimensional topological space based on standardized feature vectors. This unit implements high-dimensional space construction and state point mapping functions.

[0190] The trajectory analysis unit 23 is used to construct a metabolic state trajectory by continuously monitoring changes in the position of metabolic state points. This unit calculates the direction, velocity, and acceleration characteristics of the trajectory, identifies trajectory patterns, and analyzes trajectory complexity.

[0191] Risk assessment module 3 is used to establish a nonlinear synergistic metabolic risk measure based on the metabolic state topological space. This module contains the following sub-modules:

[0192] The coupling analysis unit 31 is used to construct a dual-disease metabolic coupling matrix describing the interaction between protein metabolism and glucose metabolism. This unit analyzes the interactive effects of the two metabolic pathways, constructs the coupling matrix, and continuously optimizes the matrix parameters based on patient feedback.

[0193] The stability analysis unit 32 is used to analyze the stability characteristics of metabolic state trajectories and identify stable and unstable attractors in the metabolic system. This unit implements functions such as state stability assessment, attractor identification, and Lyapunov index calculation.

[0194] Risk calculation unit 33 is used to calculate multi-dimensional risk scores based on topological distance, rate of change of state, and stability analysis. This unit integrates multiple risk factors, generates a comprehensive risk score, and dynamically adjusts the weights of each factor.

[0195] Prediction and early warning module 4 is used to construct a sleep-metabolic synchrony risk prediction model based on multi-dimensional risk scores and sleep data. This module includes the following sub-modules:

[0196] The correlation analysis unit 41 is used to identify the correlation patterns between sleep patterns and metabolic states. This unit analyzes the metabolic response characteristics under different sleep states, calculates sleep regulatory factors, and establishes a sleep-metabolism synchronicity classification model.

[0197] The time-series modeling unit 42 is used to construct a time-dependent probability field describing how risk changes over time. This unit divides time windows, calculates the risk probability distribution, identifies high-risk time periods, and predicts future risk changes.

[0198] The threshold management unit 43 is used to set and dynamically adjust personalized alert thresholds. This unit optimizes alert thresholds based on medical guidelines, disease severity, patient feedback, and time factors to ensure the accuracy and personalization of alerts.

[0199] The early warning triggering unit 44 is used to trigger the corresponding level of early warning information and generate personalized intervention suggestions when the multi-dimensional risk score exceeds the personalized early warning threshold. This unit realizes functions such as multi-level early warning triggering, intervention suggestion generation, and closed-loop feedback optimization.

[0200] In addition, the system includes a user interface module 5 and a data storage module 6. The user interface module 5 provides an intuitive interactive interface, displaying metabolic status and risk trends, and sending early warning information. The data storage module 6 securely stores patient health data and system configuration data, supporting data backup and recovery.

[0201] This embodiment presents a specific clinical application case to demonstrate the application effect of the present invention in a real-world environment.

[0202] Mr. Zhang, a 58-year-old patient, is 175cm tall and weighs 80kg, with a BMI of 26.1kg / m², classifying him as overweight. He also suffers from chronic kidney disease (CKD stage 3a, GFR 52mL / min / 1.73m²) and type 2 diabetes (8-year duration, HbA1c 7.8%). Traditional single-disease management methods struggle to balance the conflicting needs of protein restriction and glycemic control, resulting in significant fluctuations in his blood glucose levels and a continuous decline in renal function.

[0203] After using the collaborative nutrition monitoring and early warning system of this invention for one month, Mr. Zhang found that the system had collected a large amount of multidimensional data, including continuous blood glucose monitoring data (every 5 minutes), weekly kidney function test data, daily blood pressure and heart rate data, detailed dietary records and sleep monitoring data.

[0204] The system constructed Mr. Zhang's personalized metabolic state topological space and calculated his dual-disease metabolic coupling matrix as follows:

[0205] ,

[0206] This indicates that Mr. Zhang's protein metabolism is sensitive to its own changes by 0.85 and to changes in carbohydrate metabolism by 0.25; while carbohydrate metabolism is sensitive to changes in protein metabolism by 0.45 and to its own changes by 0.75. This matrix shows that Mr. Zhang's protein metabolism is highly sensitive, requiring more precise protein intake management.

[0207] By analyzing his metabolic state trajectory, the system identified Mr. Zhang's main risk patterns: excessively high blood glucose peak after breakfast (averaging 11.2 mmol / L), excessive protein intake at dinner (averaging 42g, exceeding the recommended upper limit of 35g), and insufficient sleep (averaging 6.1 hours / night), which led to poor blood glucose control the following day.

[0208] Based on these analyses, the system generates personalized collaborative management recommendations: adjust the carbohydrate composition of breakfast (reduce simple carbohydrates and increase dietary fiber), transfer some of the protein from dinner to lunch (control dinner protein intake to 25-30g), and improve sleep hygiene (fix sleep time and relax before bed).

[0209] The system has set a personalized alert threshold for Mr. Zhang: Alert Level. >0.35, Warning Level >0.55, intervention level >0.7, Emergency Level >0.85. The threshold setting took into account Mr. Zhang's disease severity and individual response characteristics.

[0210] After using this system for three months, Mr. Zhang's health indicators improved significantly: HbA1c decreased from 7.8% to 7.0%, blood glucose fluctuation coefficient decreased from 38% to 26%, urine albumin to creatinine ratio decreased from 45 mg / g to 28 mg / g, and GFR remained stable at 51-53 mL / min / 1.73 m². The early warning accuracy rate reached 89%, and patient compliance was 92%, significantly higher than traditional management methods.

[0211] This embodiment introduces an improved method for sleep-metabolism synchronicity analysis, which further enhances the accuracy of risk prediction.

[0212] In this embodiment, the system employs a more detailed sleep stage analysis, dividing sleep into four stages: light sleep (N1), moderate sleep (N2), deep sleep (N3), and rapid eye movement (REM) sleep. The correlation between the proportion of each stage and metabolic status is analyzed. The study found that the proportion of deep sleep (N3) is positively correlated with insulin sensitivity and negatively correlated with fasting blood glucose upon waking; while the proportion of REM sleep is correlated with blood glucose variability.

[0213] The system introduces an improved sleep quality scoring formula:

[0214] ,

[0215] in: Assess sleep quality. Total sleep duration Ideal sleep duration (usually 7.5 hours) and The ratio of deep sleep to REM sleep, respectively. and These represent the ideal percentages of deep sleep (20%) and REM sleep (25%), respectively. The number of times you wake up at night. The maximum number of awakenings allowed (set to 5).

[0216] In addition, the system incorporates sleep-metabolic phase analysis to assess the synchronicity between sleep phases and metabolic rhythms. Ideally, a patient's sleep-wake cycle should be synchronized with the circadian rhythm of the endocrine system to optimize metabolic function. The system calculates the deviation between the midpoint of sleep time and the ideal midpoint time (2:00-3:00 AM) as the basis for phase synchronicity scoring.

[0217] With these improvements, the system can more accurately assess the impact of sleep on metabolism and provide more refined intervention recommendations, such as strategies to optimize the proportion of deep sleep (e.g., moderate exercise, avoiding evening alcohol consumption) or methods to adjust sleep phase (e.g., exposure to natural light in the morning, avoiding the use of electronic devices at night).

[0218] This embodiment presents the clinical validation results of the multidimensional risk score, demonstrating the effectiveness of the present invention.

[0219] In a clinical study involving 120 patients with chronic kidney disease and diabetes, the multidimensional risk scoring method of this invention was compared with the traditional single-threshold assessment method. The study period was 6 months, and the primary endpoint was the accuracy of predicting abnormal glycemic events (glucose <3.9 mmol / L or >13.9 mmol / L) and events of worsening renal function (GFR decrease >20% or urinary albumin to creatinine ratio increase >50%).

[0220] The results showed that the multidimensional risk scoring method of this invention had a sensitivity of 87%, a specificity of 83%, and a positive predictive value of 79% in predicting abnormal blood glucose events, which was significantly better than the sensitivity (71%), specificity (75%), and positive predictive value (65%) of traditional methods. In predicting events of renal function deterioration, the method of this invention had a sensitivity of 82%, a specificity of 85%, and a positive predictive value of 76%, which was also better than the sensitivity (68%), specificity (72%), and positive predictive value (61%) of traditional methods.

[0221] More importantly, the method of this invention can predict risk events 24-72 hours in advance, while the warning time of traditional methods is usually within 0-24 hours. This advantage of early warning gives patients more time to take preventive measures, significantly reducing the incidence of adverse events.

[0222] Furthermore, the study found that the method of this invention performs exceptionally well in reducing false positives and false negatives. Compared to traditional methods, the false positive rate was reduced by 62% (from 24% to 9%), and the false negative rate was reduced by 58% (from 18% to 7.5%). This significantly improves patient trust and adherence to the system, making patients more willing to follow the system's recommendations.

[0223] In summary, this invention provides a method and system for synergistic nutritional monitoring and early warning of chronic kidney disease and diabetes. This method and system can quantitatively describe the interaction between protein metabolism and glucose metabolism at the metabolic mechanism level, achieve accurate modeling and risk prediction of complex metabolic relationships, resolve the contradictions in nutritional management of patients with chronic kidney disease and diabetes, significantly improve the accuracy and timeliness of early warning, provide patients with personalized nutritional management plans, and improve health outcomes.

[0224] Although embodiments of the present invention have been described above in conjunction with the accompanying drawings, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many modifications under the guidance of the present invention without departing from the spirit of the present invention, and these modifications all fall within the protection scope of the present invention.

Claims

1. A method for synergistic nutritional monitoring and early warning of chronic kidney disease and diabetes, characterized in that, include: The system acquires multidimensional physiological indicators, nutritional intake data, and sleep data from patients. The multidimensional physiological indicators include blood glucose, renal function, cardiovascular, and electrolyte levels. The nutritional intake data includes protein intake, carbohydrate intake, fat intake, and the distribution of meal times. The sleep data includes sleep duration, sleep quality, and sleep-wake cycles. Based on the multi-dimensional physiological indicator data, the nutritional intake data, and the sleep data, a dynamically coupled metabolic state topology space is constructed, including: The multidimensional physiological indicator data, the nutritional intake data, and the sleep data are converted into standardized feature vectors. Based on the standardized feature vector, the location of the patient's current metabolic state point is determined in the multidimensional topological space; A metabolic state trajectory is constructed by continuously monitoring the changes in the positions of the metabolic state points; Based on the aforementioned metabolic state topological space, a nonlinear co-metabolic risk measure is established, including: Construct a metabolic coupling matrix for two diseases to describe the interaction between protein metabolism and glucose metabolism; Analyze the stability characteristics of the metabolic state trajectory to identify stable and unstable attractors in the metabolic system; Based on topological distance, rate of change of state and stability analysis, a multi-dimensional risk score is calculated. Based on the multidimensional risk score and the sleep data, a sleep-metabolism synchronicity risk prediction model is constructed, including: Identify the correlation between sleep patterns and metabolic state; Construct a time-dependent probability field describing how risk changes over time; Set and dynamically adjust personalized early warning thresholds; When the multi-dimensional risk score exceeds the personalized early warning threshold, an early warning message of the corresponding level is triggered, and personalized intervention suggestions are generated.

2. The method according to claim 1, characterized in that, The acquisition of patients' multi-dimensional physiological indicators, nutritional intake data, and sleep data specifically includes: Obtain real-time blood glucose data using a continuous glucose monitor or electronic blood glucose meter; Kidney function indicators are obtained through hospital test reports or home testing devices. These indicators include serum creatinine, blood urea nitrogen, urine albumin to creatinine ratio, and glomerular filtration rate. Cardiovascular indicator data, including blood pressure, heart rate, and blood lipid profile, are obtained through smart wearable devices or smart home devices. The type and quantity of food are automatically identified using image recognition or voice recognition technology, and the nutritional intake data is calculated and obtained. The sleep data is obtained through monitoring via smart bracelets or mattress sensors.

3. The method according to claim 1, characterized in that, The process of converting multi-dimensional physiological indicator data, nutritional intake data, and sleep data into standardized feature vectors specifically includes: Standardize the indicator data of different dimensions; Extract time-domain features, frequency-domain features, morphological features, and correlation features; Principal component analysis is used to reduce feature dimensionality; Weights are assigned to features based on their clinical significance, and an n-dimensional feature vector is constructed.

4. The method according to claim 1, characterized in that, The construction of the metabolic state trajectory specifically includes: Connect the patient's metabolic state points at different time points to form a temporal state trajectory; Calculate the direction vector, velocity vector, and acceleration vector of the state trajectory; Identify the pattern characteristics of the state trajectory, including stable mode, deterioration mode, improvement mode and fluctuation mode; Analyze the fractal dimension of the state trajectory to assess its complexity.

5. The method according to claim 1, characterized in that, The construction of the dual-disease metabolic coupling matrix describing the interaction between protein metabolism and glucose metabolism specifically includes: Calculate the vector of the effect of protein intake on glycemic stability; Calculate the vector of the effect of carbohydrate intake on renal functional burden; Calculate the cross-sensitivity coefficient between the two metabolic pathways; Introducing the time lag effect, we analyzed the short-term and long-term effects of intake patterns on metabolism; The coupling matrix parameters are adjusted individually based on the patient's historical data.

6. The method according to claim 1, characterized in that, The analysis of the stability characteristics of the metabolic state trajectory specifically includes: Calculate the regression characteristics of metabolic state points after perturbation; Identify stable attractor regions that represent a healthy state; Identify unstable attractor regions that represent risk states; Identify the saddle point region representing the state bifurcation point; Calculate the Lyapunov exponent of the state trajectory to quantify the system stability.

7. The method according to claim 1, characterized in that, The calculation of the multi-dimensional risk score based on topological distance, state change rate, and stability analysis specifically includes: Calculate the shortest topological distance from the current metabolic state to the risk boundary; Calculate the velocity and acceleration components of the state trajectory toward the risk area; Assess the sensitivity of the current state to disturbances; A comprehensive risk score is calculated using a weighted fusion method. Based on historical data, the weighting coefficients of each dimension of risk score are dynamically adjusted.

8. The method according to claim 1, characterized in that, The specific association patterns between sleep patterns and metabolic states include: Analyze the metabolic response characteristics under conditions of sufficient sleep; Analyze the metabolic response characteristics under sleep deprivation; Analyze the metabolic response characteristics under sleep disturbance conditions; Calculate the regulatory factors of sleep quality on metabolic efficiency; Establish a classification model for sleep-metabolism synchronicity.

9. The method according to claim 1, characterized in that, The setting and dynamic adjustment of personalized early warning thresholds specifically includes: Initial warning thresholds are set based on medical guidelines; Adjust the baseline thresholds based on disease severity and individual characteristics; The threshold sensitivity is dynamically adjusted based on patient feedback. Establish a multi-level early warning threshold architecture, including alert level, warning level, intervention level, and emergency level; Adjust the threshold fluctuation range based on time factors, such as circadian rhythms and seasonal changes.

10. A synergistic nutritional monitoring and early warning system for chronic kidney disease and diabetes based on a multidimensional metabolic state space, characterized in that, include: The data acquisition module is used to acquire multi-dimensional physiological indicator data, nutritional intake data, and sleep data of patients. The multi-dimensional physiological indicator data includes blood glucose, renal function, cardiovascular, and electrolyte indicators. The nutritional intake data includes protein intake, carbohydrate intake, fat intake, and the distribution of meal times. The sleep data includes sleep duration, sleep quality, and sleep-wake cycle. A state modeling module is used to construct a dynamically coupled metabolic state topology space based on the multi-dimensional physiological indicator data, the nutritional intake data, and the sleep data; it includes: a feature processing unit, used to convert the multi-dimensional physiological indicator data, the nutritional intake data, and the sleep data into standardized feature vectors; a spatial mapping unit, used to determine the current metabolic state point position of the patient in the multi-dimensional topology space based on the standardized feature vectors; and a trajectory analysis unit, used to construct a metabolic state trajectory by continuously monitoring the changes in the metabolic state point position. The risk assessment module is used to establish a nonlinear synergistic metabolic risk measure based on the topological space of the metabolic state; it includes: a coupling analysis unit for constructing a dual-disease metabolic coupling matrix describing the interaction between protein metabolism and glucose metabolism; a stability analysis unit for analyzing the stability characteristics of the metabolic state trajectory and identifying stable and unstable attractors in the metabolic system; and a risk calculation unit for calculating a multi-dimensional risk score based on topological distance, state change rate, and stability analysis. The prediction and early warning module is used to construct a sleep-metabolism synchronicity risk prediction model based on the multi-dimensional risk score and the sleep data; it includes: a correlation analysis unit for identifying the correlation pattern between sleep patterns and metabolic states; a time-series modeling unit for constructing a time-dependent probability field describing the change of risk over time; a threshold management unit for setting and dynamically adjusting personalized early warning thresholds; and an early warning triggering unit for triggering corresponding level of early warning information and generating personalized intervention suggestions when the multi-dimensional risk score exceeds the personalized early warning threshold.

Images