Methods, systems, equipment, and media for dietary recommendations for chronic kidney disease based on multidisciplinary team (MDT)

Abstract

This invention belongs to the technical field of chronic disease management systems, specifically relating to a method, system, device, and medium for dietary recommendations for chronic kidney disease based on multidisciplinary team (MDT). The system of this invention includes: an input module configured to input information; a triage module configured to determine the constraints to be introduced; and an MDT conflict arbitration and negotiation solution module configured to generate dietary recommendations based on the constraints. y If a solution exists that satisfies all constraints, the solution result is output; otherwise, further optimization algorithms are used to make dietary recommendations. y The system is optimized; the output module is configured to output the solution or optimization results. This invention further provides a method and apparatus for dietary recommendations using the above system. The technical solution of this invention achieves automatic dietary recommendations for CKD patients, possessing advantages such as accuracy, high robustness, dynamic adjustability, and ease of use, and has promising application prospects.

Description

Technical Field

[0001] This invention belongs to the technical field of chronic disease management systems, specifically relating to a method, system, device, and medium for dietary recommendations for chronic kidney disease based on multidisciplinary team (MDT). Background Technology

[0002] Chronic kidney disease (CKD) has become a significant global public health challenge. CKD progresses gradually, eventually leading to end-stage renal disease (ESRD), requiring renal replacement therapy (hemodialysis or peritoneal dialysis) to sustain life. However, dialysis treatment not only imposes a heavy economic burden on patients but also significantly reduces their quality of life. Therefore, delaying CKD progression and postponing the initiation of dialysis have become core goals of CKD clinical management. Dietary and nutritional management is a core non-pharmacological intervention for delaying CKD progression and postponing dialysis; its scientific validity and adherence directly impact patient prognosis.

[0003] However, dietary management for CKD patients faces multiple clinical challenges. First, CKD is often accompanied by multiple chronic diseases such as hypertension, diabetes, and cardiovascular disease, forming a complex "comorbidity network." The nutritional needs of different diseases often conflict: for example, hypertensive patients need to increase potassium intake to help lower blood pressure, while patients in the middle and late stages of CKD need strict potassium restriction due to impaired potassium excretion; diabetic patients emphasize carbohydrate control, while CKD patients need to balance protein intake with calorie balance. This conflict of multiple objectives makes traditional dietary recommendations based on single-disease guidelines difficult to apply.

[0004] Secondly, there is a significant gap in the existing dietary management model. In-hospital management mainly relies on brief outpatient education, lacking a mechanism for continuous monitoring and dynamic adjustment outside the hospital. When patients follow the guidelines at home, they face practical problems such as confusion in food selection, improper cooking methods, and difficulty in estimating intake, resulting in a long-term adherence rate of less than 30%. Managing the diet of CKD patients through digital dietary management tools is a better solution to improve patient adherence. However, existing health apps use static rule engines, which can only handle single-disease scenarios. In cases of multiple coexisting diseases, dietary management for CKD patients is highly challenging, and existing technologies lack appropriate decision-making algorithms to achieve reasonable and effective dietary plan recommendations.

[0005] Therefore, there is an urgent need to develop a dietary recommendation decision-making algorithm for CKD patients that can integrate dietary restrictions for multiple diseases, in order to fill the management gap for CKD patients from in-hospital to out-of-hospital care and improve the accuracy and sustainability of nutritional intervention. Summary of the Invention

[0006] To address the problems of existing technologies, this invention provides a method, system, and medium for dietary recommendations for chronic kidney disease based on a multidisciplinary team (MDT) framework.

[0007] A multidisciplinary team (MDT)-based dietary recommendation system for chronic kidney disease includes: The input module is configured to accept patient diagnostic information and / or personalized needs information. The triage module is configured to determine the constraints that need to be introduced based on the diagnostic information. The MDT conflict arbitration and negotiation solution module is configured to generate dietary recommendations based on the aforementioned constraints. y If a solution exists that satisfies all constraints, the solution result is output; otherwise, further optimization algorithms are used to make dietary recommendations. y The optimization algorithm's objective function is: in, It uses a decision variable vector. The dietary recommendations are shown. Indicates the first The weight of each ingredient; It uses a decision variable vector to represent the patient's preferred dietary needs, which is generated from the patient's historical profile or personalized needs information input by the patient; For slack variables, including and , As a relaxation variable for medical hard constraints, For the relaxation variables of the nutritional soft constraint; M represents the set of diseases or complications of the patient, and m represents a specific disease or complication of the patient. This represents the priority coefficient of the department to which m belongs; The global weight representing the cost of medical violations. The global weight representing the cost of nutritional deviation. The global weight of the cost of taste deprivation; The output module is configured to output the solution or optimization results obtained from the MDT conflict arbitration and negotiation solution module.

[0008] Preferably, the constraints to be introduced in the triage module include medical hard constraints and nutritional soft constraints. The medical hard constraints include basic hard constraints, which are restrictions on potassium, phosphorus, and protein intake based on the stage of chronic kidney disease, and are expressed as follows: in, This is the set of basic hard constraints for staging chronic kidney disease. Transpose the feature column vector consisting of the potassium content per unit weight of all ingredients in the predefined ingredient library of the system. Transpose the feature column vector consisting of the phosphorus content per unit weight of all ingredients in the predefined ingredient library of the system. This is the transpose of the feature column vector consisting of the protein content per unit weight of all ingredients in the predefined ingredient library of the system. The maximum daily allowable potassium intake threshold is set based on the patient's chronic kidney disease stage and biochemical indicators. The maximum daily allowable intake threshold for phosphorus is set based on the patient's chronic kidney disease stage and biochemical indicators. The maximum daily allowable protein intake threshold is set based on the patient's chronic kidney disease stage and biochemical indicators. The medical hard constraints include complication hard constraints, which are selected from at least one of glycemic control or blood pressure control. The blood glucose control mentioned above involves limiting total carbohydrates and glycemic load, expressed as follows: The blood pressure control involves limiting sodium intake, expressed as follows: in, The maximum daily allowable protein intake threshold is set based on the patient's chronic kidney disease stage and biochemical indicators. Transpose the feature column vector consisting of the glycemic load per unit weight of all ingredients in the predefined food library for the system. This is the maximum permissible daily glycemic load threshold set based on the patient's current blood glucose levels. This represents the set of hard constraints for complications of hypertension. Transpose the feature column vector consisting of the sodium content per unit weight of all ingredients in the predefined food ingredient library of the system. This is the maximum daily sodium intake threshold set based on the patient's current blood pressure. The nutritional soft constraint sets a target calorie range, expressed as follows: The nutritional soft constraints include calorie target constraints and dietary fiber target constraints, expressed as follows: in, This is a daily maximum allowable sodium intake threshold set based on the patient's current blood pressure. This is the maximum daily calorie restriction threshold calculated based on the patient's body mass index and nutritional plan. Transpose of the feature column vector consisting of the calorie values ​​per unit weight of all ingredients in the predefined food library of the system; This is the minimum daily dietary fiber intake target calculated based on the patient's need to improve gut microbiota and assist in the excretion of metabolic waste; This is the maximum recommended daily intake of dietary fiber calculated based on the patient's gastrointestinal tolerance and overall nutritional plan; Transpose of the feature column vector consisting of the dietary fiber content per unit weight of all ingredients in the predefined food library of the system.

[0009] Preferably, in the MDT conflict arbitration and negotiation solution module, the constraints included in the optimization algorithm are: medical hard constraints with relaxation variables, nutritional soft constraints with relaxation variables, and non-negativity constraints. The medical hard constraint with relaxation variables is represented as follows: The nutrient soft constraint with relaxation variables is expressed as: The nonnegativity constraint is expressed as: in, For the corresponding to the first A medical nutrition characteristic matrix for a disease or complication, wherein each row of the matrix represents the content distribution vector of a restricted nutrient in a predefined food library; For the corresponding to the first A column vector of medical intake limits for a disease or complication, where each element in the vector corresponds to the maximum allowable intake threshold of the corresponding restricted nutrient. A feature matrix for nutritional objectives, used to quantify the total amount of basic nutrients in food combinations; To ensure sufficient quantity is recommended, To recommend excess amount, This is the column vector of theoretical target intakes under the soft constraint of nutrition.

[0010] Preferably, the set M includes chronic kidney disease, and also includes at least one of diabetes and hypertension; , , The setting rules are as follows: ; The setting rule is that the nephrology department has the highest coefficient.

[0011] Preferred options also include: The feedback input module is configured to take in the patient's actual intake vector. , The decision variables are represented by a vector. The weight update module is configured to update , renew The specific steps include: Calculate compliance rewards using the following formula. : Calculate the updated formula as follows: Value: in, For the updated , Before the update , For learning rate, The gradient of the objective function; Use the updated Perform a simulation to obtain the simulation solution. ,examine Whether an absolute hard constraint has been violated, wherein the absolute hard constraint includes: Not violated , This represents an extreme physiological red line that cannot be crossed in clinical practice. Less than or equal to the safety threshold; If any absolute hard constraint is violated, this update will be rejected, and the status quo will be maintained. constant; If no absolute hard constraint is violated, The value is updated to .

[0012] Preferred options also include: The doctor intervention module is configured to output a history of recommended results for doctors to view, and to receive the highest priority medical orders input by doctors; based on the medical orders, it parses out the instructions to forcibly modify the parameters of the objective function or constraint conditions in the optimization algorithm.

[0013] Preferably, the input module includes a first large language model unit; The first large language model unit is used to convert the unstructured dialogue content input by the patient into structured parameters that can be recognized by the triage module and the MDT conflict arbitration and negotiation solution module; The output module includes the second largest language model unit; The second large language model unit is used to process the diet recommendation results. y Convert it into natural language interpretation.

[0014] This invention also provides a diet recommendation device for chronic kidney disease based on multidisciplinary team (MDT), comprising: Input devices for recording input information via voice interaction, text interaction, and / or food weighing; The decision-making unit integrates the aforementioned MDT-based dietary recommendation system for chronic kidney disease. Output devices are used to display dietary recommendations via voice and / or screen.

[0015] This invention also provides a method for dietary recommendations for chronic kidney disease based on multidisciplinary team (MDT) approaches, comprising the following steps performed using the aforementioned MDT-based dietary recommendation system for chronic kidney disease: Step 1: Enter the patient's diagnostic information and / or personalized needs information; Step 2: Based on the diagnostic information, determine the constraints that need to be introduced; Step 3: Develop dietary recommendations based on the constraints. y If a solution exists that satisfies all constraints, the solution result is output; otherwise, further optimization algorithms are used to make dietary recommendations. y Optimization; Step 4: Output the dietary recommendations.

[0016] The present invention also provides a computer-readable storage medium having stored thereon a computer program for implementing the above-described MDT-based dietary recommendation method for chronic kidney disease.

[0017] This invention proposes a multi-objective dietary recommendation optimization algorithm based on dynamic weights of biochemical indicators. In existing technologies, CKD often co-occurs with multiple chronic diseases such as hypertension, diabetes, and cardiovascular disease, forming a complex "comorbidity network," making it difficult to automatically recommend diets to patients simply by setting medical rules. For example, existing systems cannot effectively resolve the triangular problem in CKD dietary management between "multiple comorbidity conflicts" (e.g., hyperkalemia requiring restricted fruits and vegetables vs. hypertension requiring a high-potassium diet), "strong constraints of clinical rules," and "soft constraints of dynamically changing patient needs." To address this problem, the algorithm of this invention constructs an MDT conflict arbitration and negotiation solution module, mapping the "consultation-conflict-compromise-decision" process in clinical MDT to a mathematical "multi-objective mixed integer programming" solution process. This achieves coordination and arbitration of conflicting treatment goals among different specialties (nephrology, endocrinology, and cardiovascular), ensuring the compliance and safety of the dietary plan recommendation process, and enabling real-world clinicians to intervene and adjust the recommended direction.

[0018] Furthermore, while strictly adhering to the "hard constraints" of medical indicators, we maximize the satisfaction of patients' taste preferences and other "soft constraints" to resolve the contradiction between "disease control" and "quality of life".

[0019] In the optimal approach, dynamic weight adjustments allow for tailored recommendations based on the patient's individual needs. Furthermore, in special circumstances, doctors can adjust dietary recommendations with the highest level of authority to meet the specific needs of individual cases.

[0020] Based on the above, the present invention also lowers the usage threshold for elderly patients through a multimodal intelligent hardware and software system, and realizes a closed-loop data monitoring process from recommendation to intake execution.

[0021] In summary, the technical solution of this invention realizes automatic dietary recommendations for CKD patients, which has advantages such as accuracy, high robustness, dynamic adjustment, and ease of use, and has good application prospects.

[0022] Obviously, based on the above description of the present invention, and according to common technical knowledge and conventional methods in the field, various other modifications, substitutions or alterations can be made without departing from the basic technical concept of the present invention.

[0023] The following detailed embodiments further illustrate the above-described content of the present invention. However, this should not be construed as limiting the scope of the present invention to the following examples. All technologies implemented based on the above-described content of the present invention fall within the scope of the present invention. Attached Figure Description

[0024] Figure 1 This is a flowchart illustrating Embodiment 1 of the present invention. Detailed Implementation

[0025] It should be noted that the algorithms for data acquisition, transmission, storage and processing steps not specifically described in the embodiments, as well as the hardware structures and circuit connections not specifically described, can all be implemented using content already disclosed in the prior art.

[0026] Example 1: A Dietary Recommendation System and Method for Chronic Kidney Disease Based on Multidisciplinary Team (MDT) The dietary recommendation system for chronic kidney disease in this embodiment is as follows: Figure 1 As shown, it includes: The input module is configured to input the patient's diagnostic information and / or personalized needs information; the input module may integrate a first large language model unit; the first large language model unit is used to convert the unstructured dialogue content input by the patient (which may be in the form of text, pictures, voice or video, etc.) into structured parameters that can be recognized by the subsequent triage module and the MDT conflict arbitration and negotiation solution module.

[0027] The triage module is configured to determine the constraints that need to be introduced based on the diagnostic information.

[0028] The MDT conflict arbitration and negotiation solution module is configured to generate dietary recommendations based on the aforementioned constraints. y If a solution exists that satisfies all constraints, the solution result is output; otherwise, further optimization algorithms are used to make dietary recommendations. y Optimization.

[0029] The output module is configured to output the solution or optimization results obtained from the MDT conflict arbitration and negotiation solution module; the output module includes a second large language model unit; the second large language model unit is used to convert the dietary recommendation results into natural language explanations.

[0030] The feedback input module is configured to take in the patient's actual intake vector. , The decision variable vector is used. In a preferred embodiment, the feedback input module may integrate a first large language model unit; the first large language model unit is used to convert the unstructured dialogue content input by the patient (which may be in the form of text, images, voice or video, etc.) into structured parameters (e.g., true intake vector) that can be recognized by the subsequent weight update module. ).

[0031] The weight update module is configured to update the global weights of the flavor deprivation cost. .

[0032] The doctor intervention module is configured to output a history of recommended results for doctors to view, and to receive the highest priority medical orders input by doctors; based on the medical orders, it parses out the instructions to forcibly modify the parameters of the objective function or constraint conditions in the optimization algorithm.

[0033] The method for making dietary recommendations using the above system includes the following steps: Step 1: Input the patient's diagnostic information and / or personalized needs information into the input module; Step 2: Based on the diagnostic information, the triage module determines the constraints that need to be introduced. Step 3: The MDT conflict arbitration and negotiation solution module generates dietary recommendations based on the constraints. y If a solution exists that satisfies all constraints, the solution result is output; otherwise, further optimization algorithms are used to make dietary recommendations. y Optimization; Step 4: The output module outputs the dietary recommendation results.

[0034] As an alternative, a global weighting of the flavor deprivation cost can also be applied through a weighting update module. Update the model; or, the doctor can adjust the model's recommended direction by inputting the highest-level medical orders through the doctor intervention module. As an example of application, the doctor intervention module can modify at least one of the following: adjusting the upper limit threshold of medical hard constraints (…). Adjust the coefficient representing departmental priority in the objective function (); ).

[0035] Specifically, since the information input by patients may be unstructured, such as laboratory reports (creatinine, urea nitrogen, blood potassium, etc.), text input, or personalized requests for complaints input by voice, the unstructured dialogue content can first be converted into structured parameters that can be recognized by the subsequent triage module and MDT conflict arbitration and negotiation solution module through the first major language model unit.

[0036] The triage module requires the introduction of both medical hard constraints and nutritional soft constraints. A multi-label classification function is used to determine which medical hard constraints need to be introduced. For example, if the patient has chronic kidney disease, then restrictions based on the stage of chronic kidney disease need to be introduced; if the patient also has diabetes, then blood glucose control needs to be introduced; if the patient also has hypertension, then blood pressure control needs to be introduced.

[0037] Meanwhile, the triage module uses a completeness judgment function to check whether the key indicators used to determine the constraints to be introduced are missing. If they are missing, multiple rounds of inquiry to supplement the data are triggered (achieved through interaction with the patient via the output and input modules).

[0038] In a preferred embodiment, the medical hard constraints include baseline hard constraints, which are restrictions on potassium, phosphorus, and protein intake based on the stage of chronic kidney disease, and are expressed as follows: in, This is the set of basic hard constraints for staging chronic kidney disease. Transpose the feature column vector consisting of the potassium content per unit weight of all ingredients in the predefined ingredient library of the system. Transpose the feature column vector consisting of the phosphorus content per unit weight of all ingredients in the predefined ingredient library of the system. This is the transpose of the feature column vector consisting of the protein content per unit weight of all ingredients in the predefined ingredient library of the system. The maximum daily allowable potassium intake threshold is set based on the patient's chronic kidney disease stage and biochemical indicators. The maximum daily allowable intake threshold for phosphorus is set based on the patient's chronic kidney disease stage and biochemical indicators. The maximum daily allowable protein intake threshold is set based on the patient's chronic kidney disease stage and biochemical indicators.

[0039] The medical hard constraints include complication hard constraints, which are selected from glycemic control, blood pressure control, or combinations thereof. In other embodiments, complication hard constraints may also include other possible constraints that need to be met for other complications.

[0040] The blood glucose control mentioned above involves limiting total carbohydrates and glycemic load, expressed as follows: The blood pressure control involves limiting sodium intake, expressed as follows: in, The maximum daily allowable protein intake threshold is set based on the patient's chronic kidney disease stage and biochemical indicators. Transpose the feature column vector consisting of the glycemic load per unit weight of all ingredients in the predefined food library for the system. This is the maximum permissible daily glycemic load threshold set based on the patient's current blood glucose levels. This represents the set of hard constraints for complications of hypertension. Transpose the feature column vector consisting of the sodium content per unit weight of all ingredients in the predefined food ingredient library of the system. This is the maximum daily sodium intake threshold set based on the patient's current blood pressure.

[0041] The nutritional soft constraint sets a target calorie range, expressed as follows: in, This is a daily maximum allowable sodium intake threshold set based on the patient's current blood pressure. This is the maximum daily calorie restriction threshold calculated based on the patient's body mass index and nutritional plan. Transpose of the feature column vector consisting of the calorie values ​​per unit weight of all ingredients in the system's predefined ingredient library. This is the minimum daily dietary fiber intake target calculated based on the patient's need to improve gut microbiota and assist in the excretion of metabolic waste; This is the maximum recommended daily intake of dietary fiber calculated based on the patient's gastrointestinal tolerance and overall nutritional plan; Transpose of the feature column vector consisting of the dietary fiber content per unit weight of all ingredients in the predefined food library of the system.

[0042] In the MDT conflict arbitration and negotiation solution module, the objective function of the optimization algorithm is: in, It uses a decision variable vector. The dietary recommendations shown are based on a predefined food library. , Indicates the first The weight of each ingredient; It uses a decision variable vector to represent the patient's preferred dietary needs, which is generated from the patient's historical profile or personalized needs information input by the patient; For slack variables, including and , As a relaxation variable for medical hard constraints, M represents the relaxation variable of the nutritional soft constraint; M represents the set of diseases or complications of a patient (e.g., a set of patients M includes chronic kidney disease, diabetes, and hypertension), and m represents a disease or complication of a patient (e.g., m could represent one of chronic kidney disease, diabetes, and hypertension). This represents the priority coefficient of the department to which m belongs; The global weight representing the cost of medical violations. The global weight representing the cost of nutritional deviation. This is the global weight for the taste deprivation cost. For most embodiments, , , The setting rules are as follows: , The setting rule prioritizes the nephrology department for the highest coefficient. Furthermore, The settings can be adjusted according to individual patient differences. For example, if a patient's current blood potassium level is 6.5 (critical), the priority coefficient of the department to which the chronic kidney disease belongs will be adjusted. Setting a larger value means that the opinion of the nephrologist is absolutely inviolable; if the patient's blood sugar is well controlled but their blood pressure is extremely high, then the priority coefficient for the department dealing with hypertension is... Set to a priority level greater than that of the department to which diabetes belongs. .

[0043] The constraints included in the optimization algorithm are: 1) Medical hard constraints with slack variables are represented as: This indicates the degree to which medical indicators are forcibly violated. Under normal circumstances, it should be 0, only... y Non-zero values ​​are allowed when there is no solution.

[0044] 2) The nutrient soft constraint with slack variables is expressed as: in, Indicates the recommended insufficient / gap value: When "actual value < target value", this value is positive, indicating how much is still needed to meet the target. This indicates the recommended excess / surplus value: When the actual value > the target value, this value is positive, indicating how much the target has been exceeded. In the objective function... It is the sum of the absolute values ​​of these two components: because The variable is defined as a non-negative number, and in mathematical solutions, it is directly equal to: .

[0045] 3) The nonnegativity constraint is expressed as: in, For the corresponding to the first A medical nutrition characteristic matrix for a disease or complication, wherein each row of the matrix represents the content distribution vector of a restricted nutrient in a predefined food library; For the corresponding to the first A column vector of medical intake limits for a disease or complication, where each element in the vector corresponds to the maximum allowable intake threshold of the corresponding restricted nutrient. A feature matrix for nutritional objectives, used to quantify the total amount of basic nutrients in food combinations; To ensure sufficient quantity is recommended, To recommend excess amount, This is the column vector of theoretical target intakes under the soft constraint of nutrition.

[0046] The core logic of the above MDT conflict arbitration and negotiation resolution module workflow is as follows: 1. Construction of multi-source hard constraints: Based on the patient's biochemical indicators and chief complaint, identify the patient's disease (or complication) and incorporate their respective constraint equations into the calculation. The algorithm first attempts to solve the "intersection of feasible regions".

[0047] 2. Soft and hard constraint game: When the intersection of feasible regions is empty (i.e., a medical conflict occurs) or the intersection cannot meet nutritional needs, the algorithm introduces slack variables and optimizes through optimization algorithms.

[0048] 3. Minimize cost solution: The goal of the optimization algorithm solver is no longer simply to "satisfy all conditions", but to "minimize the total penalty cost".

[0049] The costs stem from: the risk of violating medical rules, the health damage caused by inadequate nutrition, and the psychological deprivation of patients' tastes.

[0050] The weights of the costs differ: Risk of violating medical regulations Health damage from inadequate nutrition The psychological deprivation caused by actions that go against the patient's tastes.

[0051] The following is an example of the operation process of the MDT conflict arbitration and negotiation solution module for a specific patient: Scenario A, no conflict: a solution exists. y This makes all The system directly outputs recipes that are both safe and delicious.

[0052] Scenario B: A "trilemma" conflict occurs, where the medical hard constraints of chronic kidney disease and hypertension cannot be satisfied simultaneously. In this case, the feasible region is empty. It will not be all zeros. Optimization algorithm actions: 1. Solver Calculation Minimum value.

[0053] 2. Because the risk of death from high potassium levels is greater than the risk of chronic hypertension, the system is set... .

[0054] 3. Results: Solver selection increased the relaxation variables for hypertension. (Temporarily sacrificing blood pressure control) to maintain the relaxation variable for chronic kidney disease. (Maintain the red line for blood potassium).

[0055] 4. Combining the second major language model unit, the output module generates a recommendation reminder when outputting the recommendation results: "Given that your current blood potassium is high, the DASH diet rule should be suspended and a low-potassium diet should be prioritized. It is recommended to increase the dosage of antihypertensive medication to compensate for the lack of dietary control." In the weight update module, update The specific steps include: Step A: Calculate the compliance reward using the following formula. : if A low value indicates low patient compliance, suggesting that the "taste deprivation cost" is underestimated, the algorithm is too harsh, and the global weight of the taste deprivation cost should be increased. .

[0056] Step B: Calculate the updated formula as follows Value: in, For the updated , Before the update , For learning rate, The gradient of the objective function.

[0057] Step C, using the updated Perform a simulation to obtain the simulation solution. ,examine Whether an absolute hard constraint has been violated, wherein the absolute hard constraint includes: Not violated , Less than or equal to the safety threshold. , Safety thresholds can be preset according to authoritative medical guidelines.

[0058] It should be noted that the absolute hard constraint in this invention represents an extreme physiological red line that cannot be crossed in clinical practice. Taking chronic kidney disease complicated with hyperkalemia as an example: the upper limit of conventional medical hard constraints. (The clinically recommended daily potassium intake is 1500 mg.) Moderate relaxation is permissible when multiple diseases are conflicting; however, there is an absolute lethal threshold. (e.g., a single-day extreme value of 2000 mg). Therefore, the system's preset absolute safety threshold for potassium ions is... 500 mg. When the simulation solution lead to At a dose of 500 mg, the system determines that an absolute hard constraint has been violated, thus preventing potentially lethal medical recommendations at the underlying algorithm level. If an absolute hard constraint is violated, the update is rejected, and the current state is maintained. Unchanged. Meanwhile, combined with the second major language model unit, the output module can output a warning: "User preferences have approached the limits of medical safety and cannot be further compromised."

[0059] If no absolute hard constraint is violated, The value is updated to .

[0060] Global weight After the update, the algorithm will slightly reduce the mandatory nature of medical / nutritional aspects (within a safe range) and assign a higher weight to "tasty" in the next calculation, thereby finding a "compromise solution" that the patient is more willing to implement.

[0061] Example 2: MDT-based dietary recommendation device for chronic kidney disease The device in this embodiment includes: Input devices for recording input information via voice interaction, text interaction, and / or food weighing; The decision-making unit integrates the MDT-based dietary recommendation system for chronic kidney disease described in Example 1. Output devices are used to display dietary recommendations via voice and / or screen.

[0062] The input device, decision unit, and output device can be implemented according to existing technologies. The following are examples of preferred solutions for each part of the device: Input devices can be selected from microphone, high-precision electronic scale, and touch screen. Microphone can perform voice input, high-precision electronic scale can realize the input of food weighing records, and touch screen can realize text input or control the hardware terminal.

[0063] Output devices can be selected from speakers or touch screens. Speakers can output voice for interaction, while touch screens can output text, images, and videos.

[0064] The decision-making unit can be a cloud server or a locally deployed hardware smart terminal (such as Kunpeng Orange Pie).

Claims

1. A chronic kidney disease diet recommendation system based on MDT, characterized by, include: The input module is configured to accept patient diagnostic information and / or personalized needs information. The triage module is configured to determine the constraints that need to be introduced based on the diagnostic information. The MDT conflict arbitration and negotiation solving module is configured to solve the dietary recommendation result based on the constraint conditions, output a solving result if there is a solution that meets all the constraint conditions, and further optimize the dietary recommendation result by an optimization algorithm if there is no solution y The objective function of the optimization algorithm is: y ​ wherein, is a decision variable vector representing a dietary recommendation result, represents the weight of the th food material; is a patient preference dietary demand represented by a decision variable vector, which is generated by patient historical profiling or patient inputted personalized demand information; is a slack variable, including and , is a slack variable of medical hard constraints, is a slack variable of nutritional soft constraints; M represents a set of diseases or complications of a patient, and m represents a certain disease or complication of a patient; represents a priority coefficient of a department to which m belongs; represents a global weight of a medical violation cost, represents a global weight of a nutritional deviation cost, is a global weight of a taste deprivation cost; The output module is configured to output the solution or optimization results obtained from the MDT conflict arbitration and negotiation solution module.

2. The MDT-based chronic kidney disease diet recommendation system according to claim 1, characterized in that, The triage module needs to introduce constraints including medical hard constraints and nutritional soft constraints. The medical hard constraints include basic hard constraints, which are restrictions on potassium, phosphorus, and protein intake based on the stage of chronic kidney disease, and are expressed as follows: wherein, a set of hard constraints for staging of chronic kidney disease, a transpose of a feature column vector of potassium content per unit weight of all food ingredients in a system predefined food ingredient library, a transpose of a feature column vector of phosphorus content per unit weight of all food ingredients in a system predefined food ingredient library, a transpose of a feature column vector of protein content per unit weight of all food ingredients in a system predefined food ingredient library, a single day maximum potassium element allowed intake threshold value set according to the patient's chronic kidney disease stage and biochemical indicators, a single day maximum phosphorus element allowed intake threshold value set according to the patient's chronic kidney disease stage and biochemical indicators, a single day maximum protein allowed intake threshold value set according to the patient's chronic kidney disease stage and biochemical indicators; The medical hard constraints include complication hard constraints, which are selected from at least one of glycemic control or blood pressure control. The blood glucose control mentioned above involves limiting total carbohydrates and glycemic load, expressed as follows: The blood pressure control involves limiting sodium intake, expressed as follows: wherein, a single-day maximum protein intake threshold set according to the patient's chronic kidney disease stage and biochemical indicators, a transpose of a feature column vector consisting of the glycemic load per unit weight of all food materials in the system's predefined food material library, a single-day maximum glycemic load intake threshold set according to the patient's current blood glucose indicators, a set of hard constraints for hypertension complications, a transpose of a feature column vector consisting of the sodium content per unit weight of all food materials in the system's predefined food material library, a single-day maximum sodium element intake threshold set according to the patient's current blood pressure indicators; The nutritional soft constraint sets a target calorie range, expressed as follows: The nutritional soft constraints include calorie target constraints and dietary fiber target constraints, expressed as follows: wherein, a single-day maximum sodium intake threshold set according to the patient's current blood pressure indicators, a single-day maximum caloric intake threshold calculated based on the patient's body mass index and nutritional plan, a transpose of a feature column vector of caloric values per unit weight of all food items in the system's predefined food item library; a single-day minimum dietary fiber intake target value calculated based on the patient's need for improving gut microecology and assisting in metabolic waste excretion; a single-day maximum dietary fiber recommended intake value calculated based on the patient's gastrointestinal tolerance and overall nutritional plan; a transpose of a feature column vector of dietary fiber content values per unit weight of all food items in the system's predefined food item library.

3. The MDT-based chronic kidney disease diet recommendation system according to claim 1, characterized in that, In the MDT conflict arbitration and negotiation solution module, the constraints included in the optimization algorithm are: medical hard constraints with relaxation variables, nutritional soft constraints with relaxation variables, and non-negativity constraints. The medical hard constraint with relaxation variables is represented as follows: The nutrient soft constraint with relaxation variables is expressed as: The nonnegativity constraint is expressed as: in, For the corresponding to the first A medical nutrition characteristic matrix for a disease or complication, wherein each row of the matrix represents the content distribution vector of a restricted nutrient in a predefined food library; For the corresponding to the first A column vector of medical intake limits for a disease or complication, where each element in the vector corresponds to the maximum allowable intake threshold of the corresponding restricted nutrient. A feature matrix for nutritional objectives, used to quantify the total amount of basic nutrients in food combinations; To ensure sufficient quantity is recommended, To recommend excess volume, This is the column vector of theoretical target intakes under the soft constraint of nutrition.

4. The MDT-based dietary recommendation system for chronic kidney disease according to claim 1 or 3, characterized in that, The set M includes chronic kidney disease, and also includes at least one of diabetes and hypertension; , , The setting rules are as follows: ; The setting rule is that the nephrology department has the highest coefficient.

5. The MDT-based dietary recommendation system for chronic kidney disease according to claim 1, characterized in that, Also includes: The feedback input module is configured to take in the patient's actual intake vector. , The decision variables are represented by a vector. The weight update module is configured to update , renew The specific steps include: Calculate compliance rewards using the following formula. : Calculate the updated formula as follows: Value: in, For the updated , Before the update , For learning rate, The gradient of the objective function; Use the updated Perform a simulation to obtain the simulation solution. ,examine Whether an absolute hard constraint has been violated, wherein the absolute hard constraint includes: Not violated , This represents an extreme physiological red line that cannot be crossed in clinical practice. Less than or equal to the safety threshold; If any absolute hard constraint is violated, this update will be rejected, and the status quo will be maintained. constant; If no absolute hard constraint is violated, The value is updated to .

6. The MDT-based dietary recommendation system for chronic kidney disease according to claim 1, characterized in that, Also includes: The doctor intervention module is configured to output a history of recommended results for doctors to view, and to receive the highest priority medical orders input by doctors. Based on the parsing of the medical order, a forced modification instruction is obtained for the objective function parameters or constraint parameters in the optimization algorithm.

7. The MDT-based dietary recommendation system for chronic kidney disease according to claim 1, characterized in that, The input module includes the first large language model unit; The first large language model unit is used to convert the unstructured dialogue content input by the patient into structured parameters that can be recognized by the triage module and the MDT conflict arbitration and negotiation solution module; The output module includes the second largest language model unit; The second large language model unit is used to process the diet recommendation results. y Convert it into natural language interpretation.

8. A diet recommendation device for chronic kidney disease based on multidisciplinary team (MDT), characterized in that, include: Input devices for recording input information via voice interaction, text interaction, and / or food weighing; The decision-making unit integrates the MDT-based dietary recommendation system for chronic kidney disease as described in any one of claims 1-7; Output devices are used to display dietary recommendations via voice and / or screen.

9. A dietary recommendation method for chronic kidney disease based on multidisciplinary team (MDT), characterized in that, The following steps are performed using the MDT-based dietary recommendation system for chronic kidney disease as described in any one of claims 1-7: Step 1: Enter the patient's diagnostic information and / or personalized needs information; Step 2: Based on the diagnostic information, determine the constraints that need to be introduced; Step 3: Develop dietary recommendations based on the constraints. y If a solution exists that satisfies all constraints, the solution result is output; otherwise, further optimization algorithms are used to make dietary recommendations. y Optimization; Step 4: Output dietary recommendations.

10. A computer-readable storage medium, characterized in that, It stores a computer program for implementing the MDT-based dietary recommendation method for chronic kidney disease as described in claim 9.

Images