Intelligent recommendation system for postoperative nutrition support scheme of hepatobiliary surgery patient

Abstract

The invention relates to the technical field of nutrition in the hepatobiliary surgery department, and discloses an intelligent recommendation system for a postoperative nutrition support scheme of a patient in the hepatobiliary surgery department. A patient data acquisition module of the system constructs a nutrition demand model based on postoperative clinical indexes and nutrition intake information; the scheme library construction module divides historical nutrition schemes into multiple levels according to the degree of fit and clinical effect scores to form a multi-layer scheme library; the recommendation strategy generation module determines a basic recommendation strategy and an experience recommendation strategy, the basic recommendation strategy is a standardized strategy, and the experience recommendation strategy is a personalized adjustment strategy; the dynamic adjustment module optimizes a strategy in real time through a double-circulation mechanism to cope with body state changes, and a final scheme is determined; the nutrition effect monitoring module tracks a nutrition support process to obtain an actual nutrition state; and the scheme library updating module updates the scheme hierarchy based on the deviation factor between the actual state and the target state, and perfects the data. The system can dynamically adapt to postoperative changes of the patient, and the accuracy and adaptability of a nutrition scheme are improved.

Description

Technical Field

[0001] This invention relates to the field of hepatobiliary surgical nutrition technology, specifically an intelligent recommendation system for postoperative nutritional support programs for hepatobiliary surgical patients. Background Technology

[0002] Hepatobiliary surgery involves vital metabolic organs such as the liver and bile ducts, often leading to complex metabolic disorders and nutritional risks for postoperative patients. As the core organ for metabolism, the liver's functional recovery directly depends on adequate nutrition. Furthermore, damage to or reconstruction of the biliary system can further affect the absorption of substances such as fats and fat-soluble vitamins. Therefore, postoperative nutritional support must address multiple goals simultaneously: metabolic regulation, organ protection, and functional recovery.

[0003] In clinical practice, the development of postoperative nutritional support plans for hepatobiliary surgery often relies on the personal experience of medical staff and traditional guidelines. However, this model has significant limitations. Different patients have significantly different surgical procedures, underlying diseases, and postoperative complication risks. For example, the metabolic needs of patients who have undergone liver cancer resection are vastly different from those who have undergone cholecystectomy. Patients undergoing hepatobiliary surgery with cirrhosis are more prone to ascites and hypoalbuminemia. A single guideline or experience cannot cover all individual differences. Postoperative recovery is a dynamic process. From the early postoperative inflammatory stress phase to the later anabolic phase, patients' nutritional needs change dramatically. For instance, fat intake may need to be restricted within 24 hours postoperatively to reduce the burden on the bile ducts, while protein intake needs to be increased after 72 hours to promote tissue repair. Traditional plans are often adjusted based on fixed time points, making it difficult to match real-time changes in needs.

[0004] Existing nutritional protocols often rely on fragmented case studies, lacking systematic hierarchical classification and effectiveness evaluation. A successful case may only apply to patients of a specific age, weight, or with certain complications, leading to inadequate suitability when directly applied to other patients. Furthermore, protocol updates largely depend on manual summarization, lagging behind clinical practice. When patients experience unforeseen fluctuations in their physical condition, such as a sudden infection causing an increase in metabolic rate, existing protocols cannot quickly draw upon adjustment experiences from similar cases, resulting in a disconnect between nutritional support and actual needs.

[0005] Traditional methods for monitoring and adjusting nutritional support protocols lack a closed-loop mechanism. Healthcare professionals must regularly collect patients' clinical indicators to assess nutritional outcomes, but there is often a time lag between data feedback and protocol adjustments. During this period, the patient's nutritional status may have changed, further exacerbating protocol mismatches. Some medical institutions have attempted to introduce information systems to assist management, but these often remain at the data recording level, failing to achieve end-to-end intelligent management from needs modeling and strategy generation to dynamic optimization. Significant manual intervention is still required, making it difficult to address the complexity and dynamism of postoperative nutritional support in hepatobiliary surgery. Summary of the Invention

[0006] The purpose of this invention is to provide an intelligent recommendation system for postoperative nutritional support programs for hepatobiliary surgery patients, in order to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides an intelligent recommendation system for postoperative nutritional support programs for hepatobiliary surgery patients, the system comprising: The patient data acquisition module is used to build a model based on the patient's postoperative clinical indicators and nutritional intake information in order to obtain a model of the patient's nutritional needs. The protocol library construction module is used to divide the patient's historical nutrition protocols into multiple protocol levels based on protocol suitability and clinical efficacy scores, so as to form a multi-level protocol library. The recommendation strategy generation module is used to determine the basic recommendation strategy for each stage of patient recovery under the target nutritional status based on the patient's target nutritional status in the set patient nutritional needs model, and to screen the multi-layer solution library based on the clinical effect score and the degree of matching of needs to obtain the experience recommendation strategy. The basic recommendation strategy is a standardized strategy to meet the target nutritional status, and the experience recommendation strategy is a personalized adjustment strategy to meet the target nutritional status. The dynamic adjustment module is used to optimize the basic recommendation strategy and the experience recommendation strategy in real time using a dual-loop mechanism to respond to changes in the patient's physical condition after surgery, thereby determining the final nutrition plan. The nutrition effect monitoring module is used to track the patient's nutritional support process based on nutrition implementation instructions in order to obtain the patient's actual nutritional status. The nutrition implementation instructions are generated based on the final nutrition plan and the patient's current clinical indicators. The scheme library update module is used to determine the deviation factor between the actual nutritional state and the target nutritional state, and to update the content of the scheme level based on the deviation factor to improve the multi-level scheme library data.

[0008] Preferably, the patient data acquisition module includes: The clinical indicator collection unit is used to extract postoperative physiological parameters of patients based on the electronic medical record system; The nutrition intake recording unit is used to determine the patient's nutrition intake information through diet logs and nutritional preparation usage records; The demand model building unit is used to create a patient nutritional demand model based on input data to simulate the patient's nutritional metabolism process, wherein the input data includes the clinical indicators and the nutritional intake information.

[0009] Preferably, the solution library construction module includes: The protocol evaluation unit assesses the clinical efficacy score and protocol suitability of each nutrition protocol based on historical case data. The hierarchical division unit divides the historical nutrition protocols into a preferred protocol layer, a standard protocol layer, and an alternative protocol layer based on the clinical efficacy score and the protocol suitability to form a multi-layer protocol library.

[0010] Preferably, the recommendation strategy generation module includes: The multi-source data integration unit uses medical information systems and IoT devices to collect postoperative recovery data and nutritional metabolism data of patients. The basic strategy generation unit is used to input the patient's postoperative recovery data and nutritional metabolism data into the prediction algorithm, perform calculation and analysis on the target nutritional status as output data, and determine the basic recommendation strategy based on the analysis results. The basic recommendation strategy is a standardized strategy that meets the target nutritional status.

[0011] Preferably, the recommendation strategy generation module further includes: The protocol sample extraction unit is used to extract matching information samples from each layer of the protocol library based on the clinical efficacy score. The experience strategy generation unit matches the adaptation information samples according to the demand matching degree to determine the experience recommendation strategy.

[0012] Preferably, the dynamic adjustment module tracks the patient's clinical indicators and nutritional metabolic status in real time. If abnormal fluctuations occur, the experience-recommended strategy in the dual-cycle mechanism is activated; otherwise, the basic recommendation strategy is executed to optimize the basic recommendation strategy and the experience-recommended strategy in real time, thereby determining the final nutrition plan.

[0013] Preferably, the nutritional effect monitoring module includes: The real-time data acquisition unit is used to collect patients' physiological indicators through portable monitoring devices; The nutrition plan execution unit is used to analyze the physiological indicators and the final nutrition plan to generate nutrition implementation instructions and send the nutrition implementation instructions to the nutrition support device for execution. The efficacy evaluation unit analyzes the patient's nutritional status using biochemical testing techniques based on the nutritional support results.

[0014] Preferably, the scheme library update module includes: The deviation calculation unit compares and analyzes the actual nutritional status and the target nutritional status to obtain the deviation value; The deviation stage identification unit uses a classification algorithm to determine the nutritional deviation stage based on the deviation value, analyzes the patients in the nutritional deviation stage, and identifies deviation factors. The scheme optimization unit formulates corresponding optimization strategies based on the deviation factor and updates the corresponding scheme content in the multi-layer scheme library to improve the multi-layer scheme library.

[0015] Preferably, the deviation calculation unit is the deviation value between the patient's target nutritional indicators and actual nutritional indicators, and the deviation value between the target recovery period and actual recovery period.

[0016] Preferably, the effect evaluation unit includes: The biochemical indicator detection subunit is used to obtain the patient's serum albumin and bilirubin levels through blood sample analysis; The nutritional status scoring subunit is used to calculate a nutritional risk score based on the serum albumin and bilirubin levels. The comprehensive judgment subunit is used to determine the patient's actual nutritional status by combining the nutritional risk score and the description of clinical symptoms.

[0017] Compared with the prior art, the beneficial effects of the present invention are: This system, through the collaborative operation of multiple modules, effectively solves many challenges in postoperative nutritional support for hepatobiliary surgery. The patient data acquisition module constructs a nutritional requirement model based on clinical indicators and nutritional intake information, accurately capturing the unique postoperative nutritional needs of each individual and eliminating the subjectivity of traditional experience-based assessments. Different patients have different baseline conditions and surgical trauma levels; for example, the metabolic capacity of young patients differs from that of elderly patients, and the glucose metabolism requirements of patients with diabetes differ from those of ordinary patients. The modeling process of this module quantifies these differences, providing a precise basis for subsequent treatment plans.

[0018] The protocol library construction module divides the protocol library into multi-level sections based on suitability and clinical efficacy scores, changing the traditional situation where protocol libraries are fragmented and difficult to access. This hierarchical division allows protocols with different effects and suitability ranges to be categorized, enabling the rapid identification of protocols with a high degree of match to the current patient's needs during recommendation strategy generation, reducing ineffective screening. Simultaneously, this division method also lays the foundation for generating experience-based recommendation strategies, providing a basis for personalized adjustments and avoiding blind modifications.

[0019] The recommendation strategy generation module combines basic and empirical recommendation strategies, ensuring that the baseline of nutritional support aligns with the target nutritional status while allowing for personalized adjustments to accommodate individual differences. The basic recommendation strategy, as a standardized approach, covers fundamental nutritional needs at each postoperative stage, such as energy supply in the early postoperative period and protein supplementation in the mid-term. The empirical recommendation strategy optimizes for specific situations, such as adjusting the fat ratio for patients with insufficient bile secretion and restricting the types of amino acids for patients with liver dysfunction, thus achieving a balance between standardization and flexibility.

[0020] The dual-loop mechanism of the dynamic adjustment module enables real-time optimization of the treatment plan, allowing for timely responses to changes in the patient's postoperative physical condition. Postoperative physical conditions in hepatobiliary surgery patients often fluctuate due to factors such as complications and drug reactions. For example, postoperative infection may cause a sudden increase in metabolic rate, and bile leakage may affect nutrient absorption. The dual-loop mechanism can identify these changes through real-time collection of clinical indicators and quickly adjust key parameters in the nutritional plan, such as energy and nutrient ratios, preventing the plan from lagging behind actual needs.

[0021] The nutritional efficacy monitoring module tracks the nutritional support process, enabling real-time monitoring of the patient's actual nutritional status and overcoming the limitations of traditional post-assessment methods. By continuously collecting data such as changes in the patient's weight, serum albumin levels, and liver enzyme indicators, deviations between the nutritional plan and actual needs can be detected promptly. For example, if a patient experiences unexplained weight loss, it can quickly determine whether it is due to insufficient energy supply or malabsorption, providing a direct basis for dynamic adjustments.

[0022] The protocol library update module updates the protocol hierarchy based on deviation factors between actual and target nutritional status, enabling the protocol library to continuously evolve with clinical practice. Data from each nutritional support practice is incorporated into the protocol library optimization process. Experiences from successful cases are reinforced, and lessons from unsuccessful cases are avoided, making the protocol library increasingly rich and practical. Subsequent patient protocol recommendations will thus benefit from the accumulated practical experience, gradually improving the overall suitability of nutritional support. Attached Figure Description

[0023] Figure 1 This is a schematic diagram illustrating the working principle of the intelligent recommendation system for postoperative nutritional support programs for hepatobiliary surgery patients as described in this invention. Figure 2 A schematic diagram illustrating the working principle of the patient data acquisition module; Figure 3 A diagram illustrating the working principle of the module for building the solution library; Figure 4 A flowchart illustrating the generation of the basic strategy for the recommendation strategy generation module. Detailed Implementation

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

[0025] Please see Figure 1This invention provides an intelligent recommendation system for postoperative nutritional support programs for hepatobiliary surgery patients. The system includes: a patient data acquisition module, a program library construction module, a recommendation strategy generation module, a dynamic adjustment module, a nutritional effect monitoring module, and a program library update module. The specific implementation steps are as follows: The patient data acquisition module models the patient's nutritional needs based on postoperative clinical indicators and nutritional intake information, obtaining a patient nutritional requirement model. The protocol library construction module divides historical nutritional protocols into multiple protocol levels based on protocol suitability and clinical efficacy scores, forming a multi-layered protocol library. The recommendation strategy generation module determines the basic recommended strategies for each stage of patient recovery under the established patient nutritional requirement model, based on the patient's target nutritional status. Simultaneously, it filters the multi-layered protocol library based on clinical efficacy scores and requirement matching to obtain empirical recommended strategies. The basic recommended strategies are standardized strategies to meet the target nutritional status, while the empirical recommended strategies are personalized adjustment strategies to meet the target nutritional status. The dynamic adjustment module uses a dual-loop mechanism to optimize the basic and empirical recommended strategies in real time to address changes in the patient's postoperative physical condition and determine the final nutritional plan. The nutritional effect monitoring module tracks the patient's nutritional support process based on nutritional implementation instructions, obtaining the patient's actual nutritional status. Nutritional implementation instructions are generated based on the final nutritional plan and the patient's current clinical indicators. The protocol library update module identifies the deviation factors between the actual and target nutritional status and updates the protocol levels based on these deviation factors, improving the multi-layered protocol library data.

[0026] Example 1: Please refer to Figure 2 The patient data acquisition module includes a clinical indicator collection unit, a nutrition intake recording unit, and a demand model construction unit. These units work together to complete the construction of the patient's nutritional demand model.

[0027] The clinical indicator collection unit relies on the electronic medical record system to extract postoperative physiological parameters from patients. The electronic medical record system stores the patient's complete medical records from admission to postoperative period. This unit connects to the electronic medical record system through a preset data interface and automatically captures various postoperative physiological parameters at set time intervals. The extracted physiological parameters are diverse, covering vital signs indicators such as hourly recorded body temperature, blood pressure and heart rate monitored every 15 minutes; liver function-related indicators, including daily postoperative measurements of alanine aminotransferase (ALT), aspartate aminotransferase (AST), total bilirubin, and direct bilirubin, which reflect the liver's metabolic and synthetic functions; kidney function indicators, such as creatinine and blood urea nitrogen, used to assess the kidneys' ability to excrete metabolic waste; complete blood count indicators, including white blood cell count, neutrophil percentage, red blood cell count, hemoglobin concentration, and platelet count, which reflect the patient's anti-infection ability and anemia status; in addition, it includes coagulation function indicators such as prothrombin time and activated partial thromboplastin time, as well as electrolyte levels such as potassium, sodium, chloride, and calcium. By continuously extracting these parameters, a dynamic dataset of the patient's postoperative physiological state is formed, which fully presents the changes in the patient's bodily functions.

[0028] The nutrition intake recording unit determines patients' nutritional intake information by integrating dietary logs and nutritional preparation usage records. Dietary logs are recorded through various channels. Medical staff can manually enter patients' daily dietary information through the system terminal during daily rounds, including the names of breakfast, lunch, dinner, and snacks, such as rice porridge, steamed buns, lean meat, and vegetables; the approximate amount of each food, such as one ounce of rice, half a steamed bun, or 50 grams of lean meat; and the cooking method, such as steaming, braising, or frying. Simultaneously, the system allows family members to upload photos and text descriptions of the food to supplement dietary information via a linked mobile application. Nutritional preparation usage records are synchronously obtained from the hospital's pharmacy management system and nurses' workstation execution records, detailing the names of the nutritional preparations used by the patient, such as enteral nutrition emulsions, albumin preparations, and amino acid injections; the dosage, concentration, and specific time and frequency of administration, such as 200 ml three times daily, or 100 ml intravenously every 8 hours. This unit standardizes the collected dietary and pharmaceutical information, converts the food according to the nutritional composition database, calculates the total amount and proportion of protein, fat, and carbohydrates ingested daily, and also counts the intake of micronutrients such as vitamin A, B vitamins, vitamin C, calcium, iron, and zinc, forming a detailed list of the patient's daily nutritional intake.

[0029] The demand model building unit creates a patient nutritional demand model based on the aforementioned clinical indicators and nutritional intake information to simulate the patient's nutritional metabolic process. This unit employs various data analysis methods and algorithms to process the input data. First, the clinical indicators and nutritional intake information are preprocessed, including data cleaning, removal of outliers and missing values, and unit and format standardization for data from different sources. Then, the preprocessed data is divided into training and validation datasets. During model building, clinical indicators are used as characteristic variables reflecting the patient's metabolic capacity and physical state; for example, the levels of alanine aminotransferase (ALT) and aspartate aminotransferase (AST) reflect the metabolic load on the liver, and hemoglobin concentration reflects the impact of oxygen transport capacity on energy metabolism. Nutritional intake information, such as protein intake and total calorie intake, is used as input variables. By introducing algorithms such as regression analysis and random forest, the model is trained using the training dataset, enabling it to learn the metabolic patterns under different combinations of clinical indicators and nutritional intake conditions. The model's output includes the patient's demand thresholds for different nutrients, such as the minimum daily protein intake, the optimal calorie supply range, and the absorption rate and metabolic conversion rate of nutrients in the body. By repeatedly adjusting and optimizing the model using validation datasets, we can ensure that the model can accurately simulate the nutritional metabolism process of patients at different stages after surgery, such as the characteristics of nutrient consumption in the early postoperative hypermetabolic state, and the changes in nutritional needs as the body recovers and metabolism gradually stabilizes. This provides a precise reference for the subsequent development of personalized nutrition plans.

[0030] Example 2: Please refer to Figure 3 The solution library construction module includes a solution evaluation unit and a hierarchical division unit. Through the systematic evaluation and hierarchical division of historical nutrition solutions, a multi-layered solution library with a clear structure and detailed content is formed.

[0031] The protocol evaluation unit comprehensively assesses the clinical efficacy and protocol suitability of each nutritional protocol based on historical case data. Historical case data covers nutritional support information for all postoperative hepatobiliary surgery patients within the past five years, including the specific nutritional protocols received by patients, such as the types and dosages of daily nutritional preparations, dietary combinations, and timing of nutritional supplementation; basic patient information, such as age, gender, height, weight, preoperative liver function classification, surgical procedure (e.g., hepatectomy, bile duct exploration), and comorbidities (e.g., diabetes, hypertension, cirrhosis); dynamic changes in postoperative clinical indicators, such as daily liver function indicators, blood routine tests, body temperature, and blood pressure; and patient recovery outcomes, such as wound healing, postoperative complications (e.g., infection, bile leakage), length of hospital stay, and nutritional status at discharge.

[0032] The clinical efficacy assessment employs a multi-dimensional weighted calculation method. Specific dimensions include: wound healing, graded into 5 levels based on postoperative wound redness and swelling reduction time, suture removal time, and presence or absence of exudation; liver function recovery speed, graded into 4 levels based on the time it takes for indicators such as alanine aminotransferase (ALT) and bilirubin to return to normal ranges; complication rate, statistically analyzing the occurrence of complications such as infection and ascites within 30 days postoperatively; hospital stay, graded into 3 levels compared to the average hospital stay for similar surgeries; and improvement in nutritional status, graded into 5 levels based on changes in serum albumin, weight, and other indicators compared to preoperative and discharge levels. Each dimension is assigned a score of 1-5 points according to its grade, and weights are assigned based on the impact of each dimension on overall recovery (e.g., liver function recovery speed weighted at 0.3, complication rate weighted at 0.25, etc.). Finally, a weighted sum is used to obtain the clinical efficacy score for each nutritional regimen, with a score range of 1-5 points.

[0033] The assessment of protocol suitability focuses on the degree of matching between the nutritional plan and the patient's individual characteristics. First, key individual characteristic parameters of the patient are extracted, such as age group (≤60 years, >60 years), surgical trauma size (classified as mild, moderate, or severe based on the extent of surgical resection or complexity of the procedure), early postoperative nutritional risk score (e.g., NRS2002 score), and underlying medical conditions. Then, the matching of various elements in the nutritional plan with these characteristic parameters is analyzed. For example, for diabetic patients, the appropriateness of the carbohydrate ratio in the plan is assessed; for elderly patients, whether the protein supply takes into account their digestive and absorptive capacity is assessed; for patients undergoing major surgery, whether the calorie supply meets their high metabolic needs is assessed. A suitability score is calculated based on the number and importance of matching items, with a maximum score of 5 points.

[0034] Based on the aforementioned clinical efficacy scores and protocol fit, the hierarchical division unit divides historical nutrition protocols into a preferred protocol layer, a standard protocol layer, and a candidate protocol layer to form a multi-layered protocol library. During the division process, clear threshold standards are first set: the preferred protocol layer requires a clinical efficacy score ≥ 4.2 and a protocol fit ≥ 4.0; the standard protocol layer requires a clinical efficacy score between 3.0 and 4.1 and a protocol fit between 3.0 and 3.9; the candidate protocol layer includes all protocols that do not meet the requirements of the standard protocol layer, including those with a clinical efficacy score < 3.0, a protocol fit < 3.0, or protocols that meet one indicator but have a low fit in another (e.g., a clinical efficacy score of 4.5 but a fit of only 2.8).

[0035] After the classification was completed, detailed index information was created for each level of protocols, including the characteristic tags of applicable patients (such as "post-hepatectomy, no diabetes, NRS score ≥3"), the core nutritional elements of the protocol (such as daily protein intake, main nutritional preparation types), and the numbers of typical application cases. Simultaneously, dynamic management identifiers were set for each level of protocols to record information such as usage frequency and the date of the most recent application, facilitating subsequent retrieval and updates. This hierarchical classification ensures that the multi-level protocol library includes not only proven and efficient protocols, but also routine protocols applicable to different scenarios and alternative protocols for special circumstances, forming a complete and well-structured protocol reserve system.

[0036] Example 3: Please refer to Figure 4 The recommendation strategy generation module includes a multi-source data integration unit, a basic strategy generation unit, a scheme sample extraction unit, and an experience strategy generation unit. These units work together to generate basic recommendation strategies and experience recommendation strategies.

[0037] The multi-source data integration unit utilizes medical information systems and IoT devices to collect postoperative recovery and nutritional metabolism data from patients. The data provided by the medical information system includes the patient's complete electronic medical record, covering preoperative diagnosis, surgical records, postoperative orders, and various examination reports (such as imaging results and pathology reports); nursing data recorded at the nurses' workstation, such as daily intake and output, wound care, and activity level assessments; and biochemical test results from the laboratory, such as regular liver and kidney function tests, electrolytes, and blood glucose levels. IoT devices include smart bracelets worn by patients, which collect real-time data such as heart rate, steps, sleep duration and quality; bedside smart monitors that continuously monitor blood oxygen saturation and respiratory rate; and sensors connected to nutritional infusion equipment that record information such as the infusion rate and remaining volume of enteral or parenteral nutrition preparations. This unit processes the data from different sources. First, it cleans the data to remove duplicate records and obvious errors. Then, it standardizes the format, unifying the timestamps of different devices and systems into the same format, converting the units of physiological indicators into standard units, and finally integrating them according to the time series to form a complete dataset of postoperative recovery and nutritional metabolism of patients, which comprehensively reflects the changes in the patients' physical condition.

[0038] The basic strategy generation unit inputs postoperative recovery data and nutritional metabolism data from patients into the prediction algorithm, uses the target nutritional status as output data for computational analysis, and determines the basic recommended strategy based on the analysis results. The target nutritional status is set according to the rehabilitation standards for patients after hepatobiliary surgery, including daily calorie intake targets (e.g., 25-30 kcal / kg body weight), protein intake targets (e.g., 1.2-1.5 g / kg body weight), fat energy contribution ratio (e.g., 25%-30%), carbohydrate energy contribution ratio (e.g., 50%-55%), as well as serum albumin target values ​​(e.g., ≥35 g / L) and bilirubin target values ​​(e.g., ≤20 μmol / L). The prediction algorithm employs a deep learning model based on a long short-term memory network. This model uses postoperative recovery data (such as postoperative days, trends in liver function indicators, infection indicators, etc.) and nutritional metabolism data (such as changes in blood glucose after previous nutritional intake, nutrient absorption rate, etc.) as input features. Through multi-layer neural network operations, it predicts the probability of patients reaching their target nutritional status under different nutritional supply conditions. Finally, it selects a nutritional plan framework that enables patients to reach their target nutritional status with a high probability, forming a standardized basic recommendation strategy. This strategy includes daily total calories, the ratio of the three macronutrients, recommended types and basic dosages of nutritional preparations, and dietary recommendations.

[0039] The protocol sample extraction unit extracts suitable information samples from each layer of the protocol library based on clinical efficacy scores. This unit first identifies the current patient's key characteristics, such as surgical type (partial hepatectomy, liver transplantation, etc.), postoperative days, major clinical indicators (such as current ALT levels, nutritional risk score), and underlying diseases. Then, in the preferred protocol layer, standard protocol layer, and alternative protocol layer of the multi-layer protocol library, historical nutritional protocols with the top 30% clinical efficacy scores and patient characteristics somewhat similar to the current patient are selected as candidate samples. For each candidate sample, key suitable information is extracted, including the patient characteristics to which the protocol applies, daily nutrient supply, specific usage methods of the nutritional preparation (such as infusion rate and dilution ratio), dietary adjustment recommendations, and trends in clinical indicators after protocol implementation, forming a suitable information sample set.

[0040] The experience-based strategy generation unit matches the adaptation information samples based on the demand matching degree to determine the experience-based recommendation strategy. The demand matching degree is calculated by determining the degree of fit between the sample plan and the current nutritional needs of the patient, and the calculation formula is as follows: in, This indicates the degree of matching between requirements, with a value ranging from 0 to 1. The closer the value is to 1, the higher the degree of matching. This refers to the total daily nutrient intake in the sample protocol. This represents the current total target nutrient requirement for the patient. The mean clinical indicators of patients to whom the sample protocol applies; These are the current clinical indicator values ​​for the patient; and These are weighting coefficients, representing the importance of nutrient supply and clinical indicators, respectively, and are set based on clinical experience. , .

[0041] This unit calculates the matching degree between each adaptation information sample and the current patient's needs according to the above formula, and selects the 3-5 sample plans with the highest matching degree for fusion analysis. During the analysis, the similarities and differences of these sample plans are compared. For consistent content (such as the recommended use of a certain nutritional preparation), it is directly retained. For differences (such as different protein supply), adjustments are made based on the current patient's specific situation (such as digestive and absorptive capacity, tolerance to the preparation), ultimately forming an empirical recommendation strategy. This strategy adds more personalized content to the framework of the basic recommendation strategy, such as adjusting the proportion of branched-chain amino acids for patients with elevated liver function indicators; replacing a certain nutritional preparation with an alternative preparation based on the patient's history of allergy; and refining dietary recommendations based on the patient's dietary habits, making the recommendation strategy more closely aligned with the individual needs of the patient.

[0042] Example 4: The dynamic adjustment module tracks the patient's clinical indicators and nutritional metabolic status in real time. It continuously acquires various patient data by establishing a real-time data transmission channel with the clinical indicator collection unit and the multi-source data integration unit. Tracked clinical indicators include hourly updated body temperature, blood pressure, and heart rate; daily measured liver function indicators such as alanine aminotransferase (ALT), aspartate aminotransferase (AST), and bilirubin; and blood glucose and white blood cell count monitored every 6 hours. Nutritional metabolic status is comprehensively reflected through real-time data from the nutritional intake recording unit and metabolic-related indicators (such as ketone body levels and nitrogen balance). The system sets normal fluctuation ranges for each indicator; for example, the normal range for body temperature is set to 36.0-37.5℃, systolic blood pressure to 90-140 mmHg, diastolic blood pressure to 60-90 mmHg, ALT to 5-40 U / L, and blood glucose to 3.9-6.1 mmol / L. When an indicator exceeds the set range, it is considered an abnormal fluctuation. For example, if a patient's body temperature rises to 38.7℃ on the third day after surgery, or if blood glucose levels exceed 7.8 mmol / L twice consecutively, the abnormal fluctuation response mechanism is triggered. At this time, the dynamic adjustment module activates the experience-recommended strategy in the dual-loop mechanism. If all indicators are within the normal range, the basic recommendation strategy continues to be executed. The inner loop of the dual-loop mechanism quickly adjusts the experience-recommended strategy, calling the corresponding adjustment rule library based on the type of abnormal fluctuation. For example, when blood glucose rises abnormally, the corresponding entry in the rule library is "reduce the proportion of carbohydrate supply by 5%-10% and increase dietary fiber intake." The system automatically extracts this rule to modify the experience-recommended strategy. The outer loop, in conjunction with the framework of the basic recommendation strategy, checks whether the adjusted experience-recommended strategy meets the overall requirements of the target nutritional status. For example, whether the adjusted carbohydrate proportion is still within the 50%-55% total calorie percentage set by the basic recommendation strategy. If it exceeds the range, further fine-tuning is performed to ensure that personalized adjustments do not deviate from the standardized nutritional goals. Through the rapid response of the inner loop and the compliance verification of the outer loop, the recommended strategy is optimized in real time, ultimately determining the final nutritional plan suitable for the patient's current state.

[0043] The real-time data acquisition unit of the nutritional effect monitoring module collects patients' physiological indicators through various portable monitoring devices. These devices include a dynamic blood pressure monitor worn on the patient's wrist, which automatically measures and records blood pressure data every 30 minutes; a non-invasive blood glucose monitoring patch attached to the abdomen, which monitors the glucose concentration of interstitial fluid in real time and uploads data every 5 minutes; a pulse oximeter clipped to the finger, which continuously monitors blood oxygen levels; and a portable rapid liver function tester, which allows patients to obtain real-time data on bilirubin and albumin through daily finger-prick blood sampling. All devices are connected to the system via the hospital's internal wireless network, with data transmission latency controlled within 10 seconds, ensuring that the system can promptly obtain the patient's latest physiological status.

[0044] After receiving the final nutrition plan and real-time physiological data, the nutrition protocol execution unit performs a comprehensive analysis. During the analysis, the plan is refined based on the patient's current physiological indicators. For example, if the patient's real-time blood glucose is slightly higher than the upper limit of normal, the carbohydrate intake of the next meal is reduced by 20 grams based on the final nutrition plan of "500 grams of carbohydrates per day"; if the patient's blood pressure is low, the amount of sodium supplementation is increased in the nutritional preparation. Based on the analysis results, specific nutrition implementation instructions are generated, including the type and quantity of food for each meal (e.g., breakfast: 200ml millet porridge, 1 boiled egg, 100g of cold spinach salad), details of the use of nutritional preparations (e.g., 500ml of enteral nutrition emulsion, continuously infused at a rate of 50ml / h, starting at 8:00 AM), and the time and food selection for snacks (e.g., snack at 3:00 PM: 150ml of yogurt, half an apple), etc. These instructions are sent to the terminal system of the nutrition preparation center in the form of electronic work orders, and are simultaneously synchronized to the nurse's workstation and the patient's bedside display device. The nutrition preparation center prepares meals and nutritional preparations according to the instructions, and the nurses execute the infusion operations according to the instructions and record the execution status.

[0045] The biochemical index detection subunit of the efficacy evaluation unit collects patients' blood samples periodically. The collection frequency is determined according to the patient's postoperative recovery stage: once daily for days 1-3 postoperatively, once every two days for days 4-7, and once every three days after one week. The collected blood samples are sent to the laboratory for testing serum albumin and bilirubin levels using a fully automated biochemical analyzer. After testing, the results are automatically synchronized to the system's biochemical index database, generating dynamic change curves of the patient's postoperative serum albumin and bilirubin levels.

[0046] The nutritional status scoring subunit calculates the nutritional risk score based on serum albumin and bilirubin levels. The normal reference value for serum albumin is set at 35-50 g / L. When it is below 35 g / L, 1 point is deducted for every 1 g / L decrease. The normal reference value for bilirubin is 5-21 μmol / L. When it is above 21 μmol / L, 1 point is deducted for every 5 μmol / L increase. The sum of the two scores is the nutritional risk score, with a total score range of 0-20 points.

[0047] The comprehensive assessment subunit combines nutritional risk scores and clinical symptom descriptions to determine the patient's actual nutritional status. Clinical symptom descriptions are entered by healthcare professionals through the system interface and include the patient's appetite (e.g., "poor appetite, only 50% of recommended intake" or "good appetite, able to complete 90% of recommended intake"), weight changes (e.g., "weight loss of 0.5 kg in the past three days" or "weight stable"), mental state (e.g., "listless" or "fairly alert"), and gastrointestinal symptoms such as bloating or diarrhea. When the nutritional risk score is 0-5 and the clinical symptom description shows good appetite, stable weight, and no significant discomfort, the actual nutritional status is assessed as "good"; a score of 6-10, or the presence of decreased appetite or slight weight loss, is assessed as "fair"; a score of 11-15, accompanied by significant decreased appetite, weight loss exceeding 1 kg, and poor mental state, is assessed as "poor"; and a score of 16-20, accompanied by severe gastrointestinal symptoms or significant weight loss, is assessed as "poor". This comprehensive assessment fully reflects the patient's actual nutritional status after nutritional support.

[0048] Example 5: The scheme library update module includes a deviation calculation unit, a deviation stage identification unit, and a scheme optimization unit. By comparing and analyzing the actual nutrient status with the target nutrient status, it realizes dynamic updating of the multi-layer scheme library.

[0049] The deviation calculation unit compares and analyzes the actual nutritional status and the target nutritional status to obtain the deviation value. The actual nutritional status is provided by the nutritional effect monitoring module, including the patient's serum albumin level, bilirubin level, nutritional risk score, weight change trend, and the total daily intake of various nutrients. The target nutritional status is preset according to the patient's postoperative recovery stage and individual characteristics. For example, the target serum albumin is 30-35 g / L in the first week after surgery and 35-40 g / L in the second week. The target bilirubin needs to decrease by 15%-20% per week. The target weight change is no more than ±1 kg per week. The daily target nutrient intake includes 1.2-1.5 g of protein per kg of body weight and 25-30 kcal per kg of body weight. The deviation value calculation includes two parts: First, the deviation between the patient's target nutritional indicators and actual nutritional indicators. For example, the deviation of serum albumin is the target value minus the actual measured value. If the target value is 35 g / L and the actual value is 32 g / L, the deviation value is 3 g / L. The deviation of bilirubin is the actual value minus the target value. If the target value is 20 μmol / L and the actual value is 25 μmol / L, the deviation value is 5 μmol / L. The deviation of nutrient intake is the percentage of the difference between the actual intake and the target intake relative to the target intake. For example, if the target protein intake is 80 g / day and the actual intake is 60 g / day, the deviation value is -25%. Second, the deviation between the target recovery period and the actual recovery period. The target recovery period is set according to the type of surgery and the patient's preoperative condition. For example, after partial hepatectomy, the target recovery period is 14 days. The actual recovery period is the number of days from the day of surgery to reaching the discharge criteria (such as stable liver function indicators, ability to eat independently, and no complications). The difference between the two is the recovery period deviation value. If the actual recovery period is 16 days, the deviation value is 2 days.

[0050] The deviation stage identification unit uses a classification algorithm to determine the nutritional deviation stage based on the deviation value, analyzes patients in the nutritional deviation stage, and identifies deviation factors. The classification algorithm adopts a decision tree model, using various deviation values ​​as input features, and determines the division threshold through training with historical data. The nutritional deviation stage is divided into mild deviation, moderate deviation, and severe deviation: Mild deviation refers to a single nutritional indicator deviation value within the allowable range (e.g., serum albumin deviation value <2g / L, nutrient intake deviation value <±10%), and a recovery period deviation value ≤2 days; Moderate deviation refers to two or more nutritional indicator deviation values ​​exceeding the allowable range, or a single indicator deviation value being large (e.g., serum albumin deviation value 2-4g / L, bilirubin deviation value 5-10μmol / L), and a recovery period deviation value of 3-5 days; Severe deviation refers to multiple indicators seriously deviating from the target (e.g., serum albumin deviation value >4g / L, nutrient intake deviation value >±20%), or a recovery period deviation value >5 days, accompanied by an increased risk of complications. Once the deviation phase is identified, a thorough analysis of relevant patient data at that phase is conducted. This includes clinical indicator curves (such as postoperative blood glucose fluctuations and the rate of decline in liver function indicators), records of nutritional regimen execution (such as whether recommended intake was followed and changes in formulations), individual patient responses (such as intolerance symptoms to certain nutritional formulations and dietary preferences), and external factors (such as whether nursing procedures were standardized and the level of family cooperation). The specific causes of the deviation, i.e., deviation factors, are identified through analysis. These factors might include: "poor taste of the nutritional formulation leading to insufficient intake," "postoperative infection causing accelerated metabolism and increased nutrient consumption," or "failure to consider the patient's history of diabetes in the treatment plan, resulting in blood glucose fluctuations affecting nutrient absorption."

[0051] The program optimization unit formulates corresponding optimization strategies based on deviation factors and updates the content of the corresponding programs in the multi-layered program library. The optimization strategies vary depending on the deviation factor: if the deviation factor is "insufficient nutrient intake," the optimization strategies include adjusting the flavor of nutritional preparations (e.g., adding fruit-flavored options), dividing a single intake into multiple small doses, and combining it with appetite-stimulating measures; if the deviation factor is "metabolic abnormalities leading to increased consumption," the strategy is to increase the ratio of calories and protein supply and increase the content of easily absorbed nutrients such as branched-chain amino acids; if the deviation factor is "the program does not consider underlying diseases," the strategy is to adjust the program composition for specific underlying diseases, such as using low-GI carbohydrates for diabetic patients and reducing sodium content for hypertensive patients. Based on the patient characteristics corresponding to the deviation factor (e.g., age, underlying diseases, type of surgery), the relevant program levels in the multi-layered program library are located. For example, programs related to diabetic patients are mainly distributed in the standard program level and the alternative program level. The program content in these levels is then updated, such as adding a sub-program of "diabetic patient-specific carbohydrate ratio" in the standard program level and replacing ineffective preparation types in the alternative program level. The updates include the nutrient composition of the regimen, recommended formulation types, methods of consumption, precautions, etc., and the clinical efficacy scores and suitability parameters of the regimen are updated simultaneously. If the deviation value of a certain regimen is reduced after optimization and applied to similar patients, its clinical efficacy score will be improved, which may promote it from the standard regimen level to the preferred regimen level.

[0052] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0053] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An intelligent recommendation system for postoperative nutritional support programs for hepatobiliary surgery patients, characterized in that: include: The patient data acquisition module is used to build a model based on the patient's postoperative clinical indicators and nutritional intake information in order to obtain a model of the patient's nutritional needs. The protocol library construction module is used to divide the patient's historical nutrition protocols into multiple protocol levels based on protocol suitability and clinical efficacy scores, so as to form a multi-level protocol library. The recommendation strategy generation module is used to determine the basic recommendation strategy for each stage of patient recovery under the target nutritional status based on the patient's target nutritional status in the set patient nutritional needs model, and to screen the multi-layer solution library based on the clinical effect score and the degree of matching of needs to obtain the experience recommendation strategy. The basic recommendation strategy is a standardized strategy to meet the target nutritional status, and the experience recommendation strategy is a personalized adjustment strategy to meet the target nutritional status. The dynamic adjustment module is used to optimize the basic recommendation strategy and the experience recommendation strategy in real time using a dual-loop mechanism to respond to changes in the patient's physical condition after surgery, thereby determining the final nutrition plan. The nutrition effect monitoring module is used to track the patient's nutritional support process based on nutrition implementation instructions in order to obtain the patient's actual nutritional status. The nutrition implementation instructions are generated based on the final nutrition plan and the patient's current clinical indicators. The scheme library update module is used to determine the deviation factor between the actual nutritional state and the target nutritional state, and to update the content of the scheme level based on the deviation factor to improve the multi-level scheme library data.

2. The intelligent recommendation system for postoperative nutritional support programs for hepatobiliary surgery patients according to claim 1, characterized in that, The patient data acquisition module includes: The clinical indicator collection unit is used to extract postoperative physiological parameters of patients based on the electronic medical record system; The nutrition intake recording unit is used to determine the patient's nutrition intake information through diet logs and nutritional preparation usage records; The demand model building unit is used to create a patient nutritional demand model based on input data to simulate the patient's nutritional metabolism process, wherein the input data includes the clinical indicators and the nutritional intake information.

3. The intelligent recommendation system for postoperative nutritional support programs for hepatobiliary surgery patients according to claim 2, characterized in that, The solution library construction module includes: The protocol evaluation unit assesses the clinical efficacy score and protocol suitability of each nutrition protocol based on historical case data. The hierarchical division unit divides the historical nutrition protocols into a preferred protocol layer, a standard protocol layer, and an alternative protocol layer based on the clinical efficacy score and the protocol suitability to form a multi-layer protocol library.

4. The intelligent recommendation system for postoperative nutritional support programs for hepatobiliary surgery patients according to claim 3, characterized in that, The recommendation strategy generation module includes: The multi-source data integration unit uses medical information systems and IoT devices to collect postoperative recovery data and nutritional metabolism data of patients. The basic strategy generation unit is used to input the patient's postoperative recovery data and nutritional metabolism data into the prediction algorithm, perform calculation and analysis on the target nutritional status as output data, and determine the basic recommendation strategy based on the analysis results. The basic recommendation strategy is a standardized strategy that meets the target nutritional status.

5. The intelligent recommendation system for postoperative nutritional support programs for hepatobiliary surgery patients according to claim 4, characterized in that, The recommendation strategy generation module also includes: The protocol sample extraction unit is used to extract matching information samples from each layer of the protocol library based on the clinical efficacy score. The experience strategy generation unit matches the adaptation information samples according to the demand matching degree to determine the experience recommendation strategy.

6. The intelligent recommendation system for postoperative nutritional support programs for hepatobiliary surgery patients according to claim 5, characterized in that, The dynamic adjustment module tracks the patient's clinical indicators and nutritional metabolic status in real time. If abnormal fluctuations occur, the experience-recommended strategy in the dual-cycle mechanism is activated; otherwise, the basic recommendation strategy is executed to optimize the basic recommendation strategy and the experience-recommended strategy in real time, thereby determining the final nutrition plan.

7. The intelligent recommendation system for postoperative nutritional support programs for hepatobiliary surgery patients according to claim 6, characterized in that, The nutritional effect monitoring module includes: The real-time data acquisition unit is used to collect patients' physiological indicators through portable monitoring devices; The nutrition plan execution unit is used to analyze the physiological indicators and the final nutrition plan to generate nutrition implementation instructions and send the nutrition implementation instructions to the nutrition support device for execution. The efficacy evaluation unit analyzes the patient's nutritional status using biochemical testing techniques based on the nutritional support results.

8. The intelligent recommendation system for postoperative nutritional support programs for hepatobiliary surgery patients according to claim 7, characterized in that, The solution library update module includes: The deviation calculation unit compares and analyzes the actual nutritional status and the target nutritional status to obtain the deviation value; The deviation stage identification unit uses a classification algorithm to determine the nutritional deviation stage based on the deviation value, analyzes the patients in the nutritional deviation stage, and identifies deviation factors. The scheme optimization unit formulates corresponding optimization strategies based on the deviation factor and updates the corresponding scheme content in the multi-layer scheme library to improve the multi-layer scheme library.

9. The intelligent recommendation system for postoperative nutritional support programs for hepatobiliary surgery patients according to claim 8, characterized in that, The deviation calculation unit calculates the deviation between the patient's target nutritional indicators and actual nutritional indicators, as well as the deviation between the target recovery period and actual recovery period.

10. The intelligent recommendation system for postoperative nutritional support programs for hepatobiliary surgery patients according to claim 7, characterized in that, The effect evaluation unit includes: The biochemical indicator detection subunit is used to obtain the patient's serum albumin and bilirubin levels through blood sample analysis; The nutritional status scoring subunit is used to calculate a nutritional risk score based on the serum albumin and bilirubin levels. The comprehensive judgment subunit is used to determine the patient's actual nutritional status by combining the nutritional risk score and the description of clinical symptoms.

Images