Obesity life cycle management system and method based on artificial intelligence large model

Abstract

The application discloses an obesity whole-cycle management system and method based on an artificial intelligence large model, relates to the technical field of medical artificial intelligence, and comprises a multi-modal data fusion and preprocessing module, an obesity special disease large model core module and a whole-cycle management application module.The multi-modal data fusion and preprocessing module integrates and processes obesity related data of a target patient to generate structured multi-modal patient data.The obesity special disease large model core module processes the data through a large language model trained in a professional manner to output a pre-diagnosis brief containing demand conflict points and an initial individualized intervention scheme containing dietary suggestions and exercise plans.The whole-cycle management application module generates a customized management scheme according to the initial scheme and a patient characteristic portrait.The application provides reliable decision support through multi-modal data processing and professional model analysis, realizes standardized individualized diagnosis and treatment, and improves the whole-cycle management effect.

Description

Technical Field

[0001] This invention relates to the field of medical artificial intelligence technology, and in particular to a full-cycle obesity management system and method based on a large artificial intelligence model. Background Technology

[0002] Obesity has become a major global public health problem. Currently, obesity prevention and control systems face the following technical bottlenecks: First, the lack of individualized tools for early prediction of obesity risk leads to delayed intervention, often only taking action when significant complications arise, missing the optimal intervention window. Second, the diagnosis and treatment process lacks standardized support; treatment plans heavily rely on doctors' personal experience, lacking unified and quantitative decision-making support, resulting in uneven distribution of medical resources and disparities in the level of primary care. Third, traditional post-diagnosis management methods have limitations; paper records and verbal instructions are insufficient for continuous and objective tracking and effective guidance of patients' dietary and exercise behaviors, leading to unsatisfactory intervention results and a high risk of relapse.

[0003] In existing technologies, general-purpose large language models have been attempted for application in the medical field, possessing fundamental capabilities in natural language processing and knowledge reasoning. Meanwhile, traditional health management systems collect physiological parameters through wearable devices and gather patient self-reported data via mobile applications. Electronic health record systems are used to store and manage patients' clinical data. Each of these technologies has a certain foundation in data acquisition, information processing, or human-computer interaction.

[0004] However, existing technological solutions have significant shortcomings. General-purpose large language models are prone to generating inaccurate information in the medical field, exhibiting problems such as insufficient professional knowledge and poor scenario adaptability, failing to meet the high accuracy and reliability requirements of clinical diagnosis and treatment. Traditional health management methods lack deep integration of professional medical knowledge, making it difficult to provide personalized solutions based on medical evidence. Existing systems are often limited to processing single data types, failing to achieve effective fusion and collaborative analysis of multi-source data. These limitations prevent existing technologies from systematically addressing the key issues in the full-cycle management of obesity.

[0005] Therefore, there is an urgent need in this field for an innovative solution that can overcome the above-mentioned technical bottlenecks, and achieve the shift of the focus of obesity management to earlier intervention, the standardization of the diagnosis and treatment process, and the effectiveness of long-term management through a systematic technical architecture, thereby improving the overall quality of medical services and the health level of patients. Summary of the Invention

[0006] The technical problem to be solved by this invention is to address the shortcomings of existing technologies, specifically by providing a full-cycle obesity management system and method based on a large-scale artificial intelligence model, as detailed below: 1) In a first aspect, the present invention provides a full-cycle obesity management system based on a large artificial intelligence model, the specific technical solution of which is as follows: It includes a multimodal data fusion and preprocessing module, a core module for a large-scale obesity disease model, and a full-cycle management application module; The multimodal data fusion and preprocessing module is used to: integrate and structure obesity-related data of target patients to generate structured multimodal patient data; The core module of the obesity-specific big model is used to: process structured multimodal patient data using a large language model trained with knowledge of the obesity domain, and output a pre-diagnosis brief containing identified conflicting needs and an initial personalized intervention plan containing dietary recommendations and exercise plans. Conflicting needs refer to the logical conflict between the target patient's weight loss goal and the target patient's self-reported diet or exercise behavior. The full-cycle management application module is used to generate customized management plans based on the initial personalized intervention plan and the multi-dimensional feature profile of the target patient.

[0007] The beneficial effects of the obesity full-cycle management system based on a large artificial intelligence model provided by this invention are as follows: The multimodal data fusion and preprocessing module integrates and structures electronic medical record data, wearable device data, and patient-reported data, generating standardized and structured multimodal patient data to provide a complete data foundation for early risk identification. This process effectively overcomes the problem of delayed intervention caused by traditional fragmented data collection. The core module of the obesity-specific large-scale model, trained with professional medical knowledge, accurately processes structured multimodal patient data, reliably identifies the contradictions between the target patient's weight loss goals and self-reported behaviors, and generates a pre-diagnosis briefing including contradiction analysis and an initial personalized intervention plan including dietary recommendations and exercise plans. This solves the problem of inaccurate information generated by general-purpose large-scale models in the medical field, providing professional and reliable decision support. The full-cycle management application module generates customized management plans based on the initial personalized intervention plan and the multidimensional feature profile of the target patient, achieving standardization and personalization of treatment plans. This process reduces the reliance on doctors' personal experience in treatment plans, helps improve the service level of primary healthcare institutions, and promotes the balanced allocation of medical resources. Through continuous data collection and analysis, the system achieves full-cycle coverage from early prediction to long-term management. The combination of structured data processing and professional model analysis provides patients with continuous and objective behavioral tracking and guidance, effectively improving the problems of poor patient compliance and easy rebound after intervention in traditional management models, and enhancing the overall effect of obesity management.

[0008] Based on the above scheme, the obesity full-cycle management system based on artificial intelligence large model of the present invention can be further improved as follows.

[0009] Furthermore, the full-cycle management application module is also used to dynamically adjust the customized management plan based on the latest monitored physiological parameter data and behavioral data of the target patients, and generate the final personalized intervention plan.

[0010] The beneficial effects of adopting the above-mentioned further approach are as follows: By analyzing physiological parameter data from wearable devices and behavioral data actively reported by patients, the system can accurately assess the effectiveness of customized management plans. When a patient's actual indicators deviate from the expected trajectory, the system promptly adjusts parameters such as dietary recommendations and exercise plans to ensure that intervention measures always match the patient's current state. This dynamic adjustment capability overcomes the shortcomings of traditional management models, such as rigid plans and poor adaptability, and effectively responds to various changes that occur in patients during the management process. The final personalized intervention plan generated is more closely aligned with the patient's actual condition, improving the accuracy and effectiveness of long-term management and providing a reliable guarantee for maintaining intervention effects and preventing rebound.

[0011] Furthermore, the full-cycle management application module is also used for: conducting pre-diagnosis analysis based on pre-diagnosis briefings and generating structured pre-diagnosis reports.

[0012] The beneficial effects of adopting the above-mentioned further approach are as follows: the natural language pre-diagnosis briefing, which contains conflicting needs points, output by the core module of the obesity-specific disease model, is transformed into a structured pre-diagnosis report with standard chapters and fixed fields. The structured pre-diagnosis report presents the patient's basic information, health summary, and identified conflicting needs points in a unified format, making the preliminary analysis results of the patient's condition clearer and more standardized. This structured presentation method facilitates doctors' quick and accurate understanding of the patient's core situation, especially enabling them to intuitively understand the logical conflict between weight loss goals and self-reported behaviors. This provides a systematically organized and standardized information foundation for subsequent diagnostic and treatment decisions, improving the efficiency and accuracy of transmitting pre-diagnosis information to the diagnostic stage.

[0013] Furthermore, the multimodal data fusion and preprocessing module is specifically used to: integrate and structure the electronic medical record data of the target patient, the physiological parameter data collected by wearable devices, and the behavioral data actively reported by the target patient, to generate structured multimodal patient data.

[0014] The beneficial effects of adopting the above-mentioned further approach are: it integrates clinical records from electronic medical records, real-time physiological parameters monitored by equipment, and patient-reported behavioral data to form unified and structured data, providing a complete and reliable data foundation for subsequent analysis. This systematic data processing method solves the problem of the difficulty in collaboratively utilizing multi-source medical data, ensuring the quality and consistency of data input into the core module of the obesity-specific disease model, and laying a solid data foundation for accurate risk assessment and personalized treatment plan generation.

[0015] 2) Secondly, the present invention also provides a method for full-cycle obesity management based on a large artificial intelligence model, the specific technical solution of which is as follows: Integrate and structure obesity-related data of target patients to generate structured multimodal patient data; The large language model, trained with knowledge of the obesity domain, processes structured multimodal patient data and outputs a pre-diagnosis briefing containing identified conflicting needs and an initial personalized intervention plan containing dietary recommendations and exercise plans. Conflicting needs refer to the logical conflict between the target patient's weight loss goals and the target patient's self-reported diet or exercise behavior. A customized management plan is generated based on the initial personalized intervention plan and the multidimensional feature profile of the target patient.

[0016] Based on the above scheme, the obesity full-cycle management method based on artificial intelligence large model of the present invention can be further improved as follows.

[0017] Furthermore, it also includes: dynamically adjusting the customized management plan based on the latest monitored physiological parameter data and behavioral data of the target patients to generate the final personalized intervention plan.

[0018] Furthermore, it also includes: conducting pre-diagnosis analysis based on pre-diagnosis briefings to generate structured pre-diagnosis reports.

[0019] Furthermore, obesity-related data from target patients are integrated and structured to generate structured multimodal patient data, including: The system integrates and structures electronic medical record data of target patients, physiological parameter data collected by wearable devices, and behavioral data actively reported by target patients to generate structured multimodal patient data.

[0020] 3) In a third aspect, the present invention also provides an electronic device, the electronic device including a processor coupled to a memory, the memory storing at least one computer program, the at least one computer program being loaded and executed by the processor, so as to enable the electronic device to implement any of the above-mentioned methods for full-cycle obesity management based on a large artificial intelligence model.

[0021] 4) In a fourth aspect, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements any of the above-mentioned methods for full-cycle obesity management based on a large artificial intelligence model.

[0022] It should be noted that the beneficial effects of the technical solutions of the second to fourth aspects of the present invention and their corresponding possible implementations can be found in the above description of the technical effects of the first aspect and its corresponding possible implementations, and will not be repeated here. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments of the present invention will be briefly introduced below: Figure 1 This is a schematic diagram of the structure of a full-cycle obesity management system based on an artificial intelligence large model according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating a method for full-cycle obesity management based on a large artificial intelligence model, according to an embodiment of the present invention. Figure 3 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0024] The principles and features of the present invention are described below. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0025] The technical solution of the present invention and how the technical solution of the present invention solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of the present invention will now be described with reference to the accompanying drawings.

[0026] like Figure 1 As shown in the figure, an embodiment of the present invention provides an obesity full-cycle management system based on an artificial intelligence big model, which includes a multimodal data fusion and preprocessing module, an obesity-specific disease big model core module, and a full-cycle management application module. The multimodal data fusion and preprocessing module is used to: integrate and structure the obesity-related data of the target patients to generate structured multimodal patient data. Specifically, it integrates and structures the electronic medical record data of the target patients, the physiological parameter data collected by wearable devices, and the behavioral data actively reported by the target patients to generate structured multimodal patient data.

[0027] Electronic medical record data refers to digitized medical information about patients recorded in medical institutions, including patients' chief complaints, medical history, laboratory test results, and imaging reports. This data comes from the hospital information system's database and is usually in unstructured or semi-structured text format, requiring the extraction of key clinical features for subsequent analysis.

[0028] Physiological parameter data refers to the patient's physiological indicators collected in real time through wearable devices, such as heart rate, steps, sleep duration, weight, and body fat percentage. This data, acquired from the device's sensors in time-series format, reflects the patient's real-time physiological state and daily activity level, providing a foundation for continuous health monitoring.

[0029] Among these, proactively reported behavioral data refers to information about patients' own behavior that they actively provide through voice or text / image interactions, such as food logs, exercise experiences, and self-attributions. This data, input through mobile applications or online platforms, expresses patients' subjective experiences and behavioral habits, helping to understand their personal lifestyles and weight loss motivations.

[0030] Structured multimodal patient data refers to data in a unified format formed by integrating and structuring electronic medical record data, physiological parameter data, and proactively reported behavioral data from different sources. This data is organized in structured forms such as tables or graphs, and contains standardized fields and values, facilitating efficient processing and analysis by artificial intelligence models.

[0031] The specific implementation process of generating structured multimodal patient data includes three main stages: data integration, data preprocessing, and data fusion. Specifically: During the data integration phase, the system acquires electronic medical record data from the electronic health record system via a standardized application programming interface (API), synchronizes physiological parameter data from a wearable device cloud platform via a wireless communication protocol, and receives patient-reported behavioral data via a mobile application interface. All data is transmitted to a central data storage system for temporary storage and preliminary verification. In the data preprocessing phase, for electronic medical record data, natural language processing (NLP) techniques are used for entity recognition and relation extraction, converting unstructured text into structured clinical features. For example, a named entity recognition model is used to extract diagnostic terms and laboratory values, mapping them to the International Classification of Diseases (ICD) standard coding to form standardized clinical feature vectors. For physiological parameter data, time series cleaning and outlier detection are performed, interpolation methods are used to handle missing data, and daily statistical indicators such as mean and variance are calculated to generate a well-organized time series table. For proactively reported behavioral data, speech recognition algorithms are applied to convert speech input into text, or text parsing models are used to extract structured information from free text, such as parsing dietary logs into food types and intake amounts, and classifying exercise sensations into predefined emotion tags. During the data fusion phase, the system associates and integrates preprocessed electronic medical record data, physiological parameter data, and proactively reported behavioral data according to the patient's unique identifier. Through data connection operations, a unified multimodal data table is created, containing fields such as patient identifier, timestamp, clinical characteristics, physiological indicators, and behavioral records. All data fields use consistent units and encoding schemes; for example, weight is uniformly represented in kilograms, and timestamps are uniformly formatted using the international standard time format. The resulting structured multimodal patient data is stored in a relational database or data warehouse, ensuring data integrity and access efficiency, and providing input for subsequent analysis of the core modules of the obesity-specific disease model.

[0032] The core module of the obesity-specific big model is used to process structured multimodal patient data through a large language model trained with knowledge of the obesity domain. It outputs a pre-diagnosis briefing containing identified conflicting needs and an initial personalized intervention plan containing dietary recommendations and exercise plans. Conflicting needs refer to the logical conflict between the target patient's weight loss goals and the target patient's self-reported diet or exercise behavior.

[0033] Among them, the pre-diagnosis briefing is a structured document generated after processing structured multimodal patient data in a large model. This document summarizes the key health information provided by patients through interaction in the pre-diagnosis stage, highlights the identified conflicting needs, and provides doctors with a preliminary analysis summary of the patient's condition, helping to improve diagnostic and treatment efficiency.

[0034] The initial personalized intervention plan refers to a preliminary management plan tailored for patients based on the analysis results of structured multimodal patient data and the core modules of the obesity-specific disease model. This plan includes specific dietary recommendations and exercise plans, aiming to support patients in achieving their weight loss goals through scientific guidance.

[0035] The process of obtaining the pre-diagnosis briefing and initial personalized intervention plan is as follows: 1) The core module of the obesity-specific large-scale model loads a large language model trained with knowledge of the obesity domain. Through pre-training and fine-tuning, this model learns professional knowledge from authoritative medical guidelines, expert consensus, and scientific research literature, enabling it to process medical data and perform clinical reasoning. The model receives structured multimodal patient data generated by the multimodal data fusion and preprocessing module as input. This data is organized in tabular or vector form and includes fields such as patient identifiers, timestamps, clinical characteristics, physiological indicators, and behavioral records.

[0036] 2) The large language model uses an embedding layer to convert structured multimodal patient data into numerical vector representations. For textual data, such as chief complaints in electronic medical records or self-reported content in behavioral data, the model applies word embedding techniques to map them into high-dimensional vectors; for numerical data, such as weight or heart rate in physiological parameter data, the model performs standardization and converts them into feature vectors. All vectors are fused into a unified multimodal representation vector through concatenation or attention mechanisms, denoted as [vector name missing]. ,in d represents the dimension of the vector.

[0037] 3) The model performs the task of identifying conflicting needs. Conflicting needs refer to the logical conflict between the target patient's weight loss goals and their self-reported dietary or exercise behaviors. The model uses predefined rules and a machine learning classifier to analyze the multimodal representation vectors. For example, the model compares a patient's self-reported daily calorie intake with the calorie deficit required for their weight loss goal. If the intake exceeds the target value, it is marked as a contradiction. This process is implemented using logistic regression or a neural network classifier, outputting a list of contradictions. Each of them This indicates a specific contradiction.

[0038] 4) Model generates a pre-diagnosis briefing. The pre-diagnosis briefing is output in a structured text format, including basic patient information, a health summary, and identified conflicting needs. The model uses sequence generation techniques, such as a Transformer-based decoder, to transform multimodal representation vectors. List of contradictions As input, it generates coherent natural language summaries. For example, the model output paragraphs describe the patient's current state and insert contradictions such as "The patient's weight loss goal is to lose 5 kg per month, but he / she states that his / her daily calorie intake exceeds the recommended value by 20%".

[0039] Simultaneously, the model generates an initial personalized intervention plan. This plan includes dietary recommendations and an exercise program, calculated based on the patient's multimodal data and obesity characteristics. The model uses a multi-task learning framework to handle nutritional assessment and exercise capacity assessment separately. For dietary recommendations, the model calculates daily calorie requirements by referencing the patient's weight, body fat percentage, and dietary preferences. Use formula Here, BMR is the basal metabolic rate, activity_factor is the activity factor, and deficit is the calorie deficit. The model then generates a food recommendation list to ensure nutritional balance. For exercise plans, the model analyzes the patient's physiological parameters such as heart rate and steps, assesses exercise tolerance, and generates weekly exercise frequency and intensity suggestions, such as the duration of aerobic exercise. and number of sets for strength training All outputs are transformed into readable text using template filling or natural language generation techniques, forming a complete initial personalized intervention plan.

[0040] 5) The core module of the obesity-specific disease model outputs pre-diagnosis reports and initial personalized intervention plans to the full-cycle management application module for subsequent use. The entire process ensures a continuous data flow, traceable model inference, and outputs that meet clinical standards.

[0041] The large language model can be an LLaMA-2 model, a ChatGLM model, etc., and can be set according to the actual situation.

[0042] The full-cycle management application module is used to generate customized management plans based on the initial personalized intervention plan and the multi-dimensional feature profile of the target patient.

[0043] The process of obtaining the multidimensional feature profile of the target patient is as follows: 1) The full-cycle management application module receives structured multimodal patient data from the multimodal data fusion and preprocessing module, as well as pre-diagnosis briefings and initial personalized intervention plans from the core module of the obesity-specific disease model. These data together serve as the input source for constructing a multidimensional feature profile.

[0044] 2) The system initiates a feature extraction and quantification module. This module parses the input structured multimodal patient data, extracts key feature indicators, and quantifies them. For example, it extracts laboratory indicators such as body mass index, fasting blood glucose, and blood lipid profile from electronic medical record data; calculates average daily steps, resting heart rate variability, and sleep efficiency from physiological parameter data; and quantifies daily fruit and vegetable intake frequency, high-calorie meal frequency, and moderate-intensity exercise duration from actively reported behavioral data. These features are organized into an initial feature vector. ,in , where m represents the number of features.

[0045] 3) The core module of the obesity-specific large-scale model is invoked to process the initial feature vector. Deep analysis and interpretation are performed. The large model, based on its built-in medical knowledge graph and inference engine, executes two key tasks. The first task is etiological classification, where the model matches patient characteristics with predefined categories of obesity etiologies. For example, a classifier calculates the probability that a patient belongs to metabolic obesity, behavioral obesity, or other classifications. This classification result... It can be represented as The second task is risk assessment. The model integrates a patient's medical history, physiological parameters, and current behavior to predict their risk level for developing complications. For example, a regression model can be used to estimate the risk score of developing type 2 diabetes within the next year. Its calculation process can be abstracted as follows: .

[0046] 4) After completing feature extraction and large-scale model analysis, the system enters the profile synthesis stage. All generated information constitutes the quantized initial feature vector. Etiological classification results Risk assessment score The conflicting needs identified in the pre-diagnosis briefing are aggregated and mapped into a standardized profiling framework. This framework typically includes several dimensions, such as basic information, physiological and metabolic, behavioral and psychological, and risk prediction. Each dimension contains a series of specific feature values, labels, or scores.

[0047] 5) The system integrates all data items across these dimensions to generate a unified multidimensional profile of the target patient. This profile typically exists as a structured data object within the system, such as a JSON document containing key-value pairs or a database record, fully encapsulating the patient's obesity status. It is then invoked by the full-cycle management application module to generate and dynamically adjust customized management plans.

[0048] The full-cycle management application module, which generates a customized management plan based on the initial personalized intervention plan and the multi-dimensional feature profile of the target patient, is a systematic process of plan integration and personalized adaptation. This process deeply integrates basic recommendations with the patient's specific situation, outputting a directly executable and highly personalized health management plan. The specific implementation process is as follows: 1) The full-cycle management application module simultaneously reads two key input data. The first input is the initial personalized intervention plan output by the core module of the obesity-specific disease model, which includes a general dietary recommendation and exercise plan framework. The second input is the constructed multi-dimensional feature profile of the target patient, which comprehensively describes the patient's physiological state, behavioral habits, etiological classification, and risk prediction.

[0049] 2) The system initiates a scheme parsing and feature mapping process. The module performs structured parsing of the initial personalized intervention plan, decomposing it into configurable parameter units. For example, dietary recommendations are parsed into daily calorie targets. Nutrient ratios (such as protein content) Carbohydrate percentage ) and food category recommendation list At the same time, the exercise plan will be analyzed into exercise types. Weekly frequency Duration of each session and intensity index .

[0050] 3) The system matches and adjusts the parsed parameter units with the corresponding feature values ​​in the target patient's multidimensional feature profile. This process relies on a set of preset clinical rules and adaptation algorithms. Specifically: For the customized dietary portion, the system references behavioral and psychological dimension data from the multidimensional profile. For example, if the profile indicates that the patient has "behavioral obesity due to eating habits" and exhibits the behavioral characteristic of "frequent late-night snacks," the system will add targeted behavioral intervention strategies to the initial plan, such as setting reminders to fast after dinner and setting daily calorie targets. Distributing the food proportionally among the three meals, the calculation formula may be adjusted to... , , Additionally, if the patient's profile indicates a risk of hypertension, the system will automatically reinforce low-sodium dietary recommendations by adding them to the food recommendation list. Filter high-sodium foods.

[0051] For customizing exercise plans, the system uses data from the physiological and metabolic dimensions of the multidimensional feature profile. For example, if the risk assessment in the profile indicates a high risk of joint weight-bearing, the system will incorporate high-impact exercise types into the initial plan. Replace running with low-impact exercises (such as swimming or elliptical training). Simultaneously, based on the patient's age, resting heart rate, and daily activity level in the profile, use the formula... Precisely calculate its personalized target heart rate range, where This is an estimate of the maximum heart rate. is the resting heart rate, and k is the intensity coefficient based on the risk level.

[0052] 4) All personalized parameters are re-integrated to generate a complete and detailed customized management plan. This plan not only includes specific numerical goals and a list of recommendations, but also incorporates motivational information and potential obstacle coping strategies extracted from multi-dimensional feature profiles, forming a comprehensive and highly actionable management document centered on the patient. This customized management plan is then stored and prepared for distribution to patients and clinicians, entering the implementation and dynamic tracking phase.

[0053] Optionally, in the above technical solution, the full-cycle management application module is also used to: dynamically adjust the customized management plan based on the latest monitored physiological parameter data and behavioral data of the target patient, and generate a final personalized intervention plan, specifically: 1) The system initiates data monitoring and collection. The full-cycle management application module continuously receives the latest data streams from two main channels. One is physiological parameter data automatically uploaded by wearable devices, such as daily weight. Body fat percentage Mean resting heart rate and sleep duration These data are transmitted in time-series format. Secondly, they come from behavioral data proactively reported by patients through mobile applications, such as actual calorie intake recorded in daily food logs. And the completion status of the exercise, including the type of exercise and the duration.

[0054] 2) The system performs data integration and trend analysis. Newly arrived physiological and behavioral data are temporarily stored and preprocessed, including data cleaning, unit standardization, and handling of missing values. Subsequently, the system merges these latest data points with the patient's historical data to calculate the trends of key indicators. For example, the system calculates the moving average of weight over the past week. and rate of change Simultaneously, the system matches patients' self-reported exercise completion status with the exercise plan in the customized management program to calculate the exercise adherence rate. .

[0055] 3) The system invokes the core module of the obesity-specific disease model for effect evaluation and deviation assessment. The system inputs the latest data trends, calculated indicators, and the current customized management plan into the model. Based on its built-in medical knowledge, the model performs reasoning and executes two key tasks. The first task is to evaluate the intervention effect and determine whether the patient's progress is on the expected trajectory. The second task is to identify any deviations from the expected indicators. For example, the model will analyze the five-day moving average weight. The ideal weight curve value was higher than the model's prediction, and the compliance rate was higher. If the progress falls below a preset threshold, it is considered a "progress deviation".

[0056] 4) Based on deviation assessment and cause analysis of the large model, the system startup plan is adjusted and calculated. This process relies on a pre-set adjustment rule library. The adjustments are made to specific parameters in the customized management plan, specifically: Regarding dietary adjustments, if weight loss has stalled and the cause is calorie intake... Continuously exceeding the target value of the plan The system may initiate adaptive adjustments. For example, while ensuring nutritional balance, it may fine-tune the daily calorie target, setting a new target value. According to the formula Calculation, where This is a small-scale calorie reduction calculated based on the patient's metabolic characteristics and the degree of deviation. It may also optimize the food recommendation list, increasing the number of food categories that provide a greater feeling of fullness.

[0057] Regarding adjustments to the exercise plan, if the system identifies low adherence due to inappropriate exercise intensity, it will adjust the exercise parameters. For example, it will adjust the duration of each session. Reduce the duration from 40 minutes to 30 minutes, or decrease the intensity of the exercise. The lower limit of the corresponding target heart rate range is lowered. The adjustment formula may be: ,in, It is an adjustment factor calculated based on patient tolerance and compliance.

[0058] 5) The system generates and outputs the final personalized intervention plan. All adjusted parameters are reintegrated to form an updated final personalized intervention plan that matches the patient's latest condition. This plan records the reasons for the adjustments and the specific goals after the adjustments, and is pushed to the patient and doctor through the application interface, thus completing a dynamic optimization cycle. This process will be repeated periodically as management continues.

[0059] In another feasible approach, the process of generating the final personalized intervention plan includes: 1) The system continuously receives physiological parameter data from wearable devices collected in real time from the target patient, including heart rate, weight, body fat percentage, and sleep duration. It also receives behavioral data proactively reported by the target patient via a mobile application, such as dietary logs and exercise completion status. This data is cleaned, denoised, and standardized. The system calculates the moving average and trend of key indicators, such as the rate of weight change. And integrate them into a unified time series dataset to prepare for dynamic analysis, specifically: The system establishes a continuous connection with various wearable devices through a pre-configured data interface protocol. These devices transmit real-time collected physiological parameter data to the system's data receiving end via a wireless network at a set sampling frequency. The transmitted data includes parameters such as heart rate, weight, body fat percentage, and sleep duration, with each data point accompanied by a device identifier, timestamp, and measurement value. Simultaneously, the system receives behavioral data proactively reported by target patients through an interactive interface provided by a mobile application. Patients record dietary logs using text and images, including food types and intake; they describe exercise completion status, such as exercise type and duration, using selection buttons or free text. These behavioral data are automatically appended with a patient identifier and timestamp upon submission. After data reception, the system initiates a data cleaning and preprocessing procedure. For physiological parameter data, outlier detection and processing are first performed, using a threshold judgment method based on statistical distribution to identify and remove measurements that significantly exceed reasonable ranges. For example, records with resting heart rates below 40 beats / min or above 120 beats / min are marked as abnormal. For missing data, an appropriate interpolation method is selected based on the data type, such as using linear interpolation of adjacent time series data. Simultaneously, all physiological parameter data are uniformly converted into standard units of measurement, such as converting weight to kilograms and duration to minutes. For behavioral data proactively reported by patients, the system uses natural language processing technology for structured transformation. Through a pre-trained information extraction model, key information is extracted from free text, such as converting the text description "running for half an hour in the evening" into a structured exercise type of "running" and a duration of 30 minutes. The extracted information is also standardized, mapping food names to standard food codes according to nutritional standards and categorizing exercise types into a pre-defined exercise category system.

[0060] After data cleaning and standardization, the system calculates the statistical characteristics and trends of key indicators. The system employs a sliding window analysis method, using a seven-day calculation period to calculate the moving average for each physiological parameter. For example, it calculates the seven-day moving average of daily weight and, based on this, the rate of weight change, using the following formula: in, This represents the current day's moving average weight. This represents the moving average weight seven days prior. The same method can be applied to other physiological parameters and behavioral data to calculate their trend indicators.

[0061] Finally, the system integrates all processed data into a unified time-series dataset. This dataset is organized chronologically, with each time point containing complete physiological and behavioral data, as well as calculated statistical characteristics and trend indicators. The dataset uses a columnar storage format to ensure efficient data retrieval and access, providing complete and well-organized data preparation for subsequent dynamic analysis.

[0062] 2) Based on the processed physiological and behavioral data, the core module of the obesity-specific disease model is invoked to evaluate the intervention effect; the model analyzes the implementation of the current customized management plan, compares the deviation of actual indicators from the expected trajectory, such as calculating the calorie intake deviation. It identifies key factors leading to deviations, such as insufficient exercise adherence or overeating; and outputs assessment results, including deviation indicators and potential cause analysis, specifically: The system first performs data matching and time alignment between the processed physiological and behavioral data and the currently implemented customized management plan. The system then reads the expected target trajectory set in the customized management plan, including the planned weight loss curve and the target daily calorie intake. And the expected exercise completion indicators. Simultaneously, the system extracts actual monitoring values ​​for the corresponding time periods from the processed dataset, including the actual calorie intake recorded by the patient through their diet log. The system records the actual amount of exercise performed. Then, it calls the core module of the obesity-specific disease model, inputting the matched dataset. The model activates its built-in medical inference engine to perform multi-dimensional deviation calculations. The model first quantifies the degree of deviation of key indicators, such as calculating calorie intake deviation, using the formula... Simultaneously calculate the deviation from the expected weight change, and then calculate the actual weight change rate. Rate of change compared to the expected rate of change of the plan Comparisons were made. For exercise indicators, the exercise compliance rate was calculated. After completing the quantitative calculations, the large model performs causal relationship analysis and identifies key factors. Based on its training in the obesity domain, the model analyzes the correlation between various deviation indicators to determine the dominant factors causing the deviations. For example, the model might analyze that stagnant weight loss is due to an imbalance in calorie intake. The high level is likely due to the high rate of adherence to exercise. Too low a level, or the result of both factors combined. The model also incorporates physiological parameter data, such as analyzing the relationship between resting heart rate changes and exercise adherence, to distinguish between a problem of willingness to implement and a problem of physical tolerance. Finally, the large model integrates the analysis results to generate a structured assessment. This result includes clear deviation indicators, such as deviation from target weight progress or deviation from dietary plan implementation, and provides a detailed analysis of the potential causes for each deviation indicator. The assessment results are output in a machine-readable, standardized format, providing a direct basis for subsequent protocol adjustment decisions, completing the transformation from data to clinical insights.

[0063] 3) Based on the evaluation results, a multi-objective optimization algorithm is initiated to adjust and calculate the customized management plan; the algorithm considers the patient's multi-dimensional feature profile and real-time data to optimize dietary recommendations and exercise plan parameters, such as recalculating daily calorie targets. ,in It is an adaptive adjustment based on the degree of deviation and metabolic characteristics; it also adjusts the exercise intensity. and frequency To ensure the treatment plan matches the patient's current condition, specifically: The system first reads the evaluation results output by the core module of the obesity-specific disease model, including deviation data for various indicators and identified key factors. Simultaneously, the system acquires a multi-dimensional feature profile of the target patient, extracting metabolic characteristic parameters, etiological classification information, and basic physiological data. These data serve as input parameters for the multi-objective optimization algorithm. The system initializes the computational environment for the multi-objective optimization algorithm. The algorithm sets three main optimization objectives: the first objective is to promote the regression of weight indicators to the expected trajectory; the second objective is to ensure that nutritional intake meets basic health needs; and the third objective is to maintain the safety and feasibility of the exercise program. The algorithm also defines a set of constraints, including a lower limit for daily calorie intake, a range of nutrient ratios, and a safe threshold for exercise intensity. These constraints are set based on the patient's metabolic characteristics and health status. Then, the algorithm performs iterative optimization calculations. For adjustments to dietary recommendations, the algorithm calculates a new daily calorie target based on the degree of calorie intake deviation and the patient's basal metabolic rate. The specific calculation formula is as follows: Among them, adjustment amount The calculation takes into account the deviation between actual intake and recommended values. and the patient's metabolic characteristics ,Right now This function ensures that the adjustment amount corrects deviations without exceeding the patient's metabolic tolerance.

[0064] When adjusting the exercise plan, the algorithm also optimizes the exercise intensity. and motion frequency Two parameters. The algorithm establishes a mapping relationship between exercise parameters and physiological responses. Based on the patient's recent physiological parameter data trends and performance in behavioral data, it searches for the optimal parameter combination within a safe boundary. For example, for patients with low exercise compliance, the exercise frequency may be appropriately reduced. However, moderately increasing the intensity of a single exercise session In order to maintain the achievement of the total energy consumption target.

[0065] During the optimization process, the algorithm employs a Pareto optimality search strategy to balance the trade-offs among multiple objective functions. Each iteration's candidate solutions undergo a feasibility check to ensure that the adjusted parameters meet medical safety standards. The algorithm continues to run until it finds the optimal solution set that satisfies all constraints and achieves a balance among the objective functions.

[0066] Finally, the system selects the most suitable parameter combination from the optimal solution set for the current patient's condition, outputs adjusted dietary recommendations and exercise plan parameters, and completes the optimization calculation of the customized management plan. These optimized parameters provide a precise numerical basis for generating the final personalized intervention plan.

[0067] 4) Integrate the optimized parameters into the customized management plan to generate the final personalized intervention plan. This plan includes updated dietary recommendations, exercise plans, and behavioral intervention strategies, and is output in a structured document format. The system automatically pushes the final personalized intervention plan to patients and doctors, and updates the records in the full-cycle management application module to complete the dynamic adjustment cycle. Specifically: The system first creates a scheme update workspace, integrating the optimized parameters output by the multi-objective optimization algorithm with the existing framework of the current customized management scheme. The system then reads the optimized dietary recommendation parameters, including the new daily calorie target. And updated nutrient ratios; simultaneously read optimized exercise plan parameters, including adjusted exercise intensity. and motion frequency These numerical parameters are systematically replaced in the corresponding data fields of the customized management plan. The replacement process preserves the original logical structure and textual description framework of the plan, updating only the numerical content and specific recommendations that need adjustment. The system then activates the document generation engine, transforming the integrated plan content into a structured output document based on a preset final personalized intervention plan template. This template defines the standardized chapter layout of the plan, including chapters on dietary recommendations, exercise plans, and behavioral intervention strategies. The document generation engine converts the numerical parameters into easily understandable textual descriptions, such as daily calorie targets. Convert this into specific meal distribution recommendations, and adjust the exercise intensity accordingly. The protocol is converted to heart rate zone guidance. Simultaneously, the system inserts explanatory notes into the protocol, explaining the basis for the parameter adjustments and the expected effects. Then, through an integrated push notification service, the system automatically distributes the generated structured document as the final personalized intervention protocol to all relevant parties. The system pushes the protocol document to the mobile application used by the target patient via a secure messaging interface and creates a notification reminder in the multidisciplinary collaborative diagnosis and treatment subsystem of the doctor's workstation. The push process includes a protocol version identifier and release timestamp to ensure that all parties receive the latest version of the intervention protocol. Finally, the system updates the final personalized intervention protocol to the current effective version in the database of the full-cycle management application module, overwriting previous customized management protocol records. The system also establishes a new protocol execution baseline, recording the time point and reason for this adjustment, preparing for the next round of dynamic monitoring and adjustment cycles, thus completing the full dynamic adjustment process.

[0068] Optionally, in the above technical solution, the full-cycle management application module is also used for: performing pre-diagnosis analysis based on the pre-diagnosis briefing and generating a structured pre-diagnosis report, specifically: 1) The pre-diagnosis intelligent pre-consultation subsystem receives a pre-diagnosis briefing generated by the core module of the obesity-specific disease model. This pre-diagnosis briefing is a text document containing basic patient information, a health summary, and identified points of conflict in needs.

[0069] 2) The system initiates an information extraction and standardization module, which uses natural language processing technology to perform in-depth analysis of the pre-diagnosis briefing. First, named entity recognition is performed to automatically extract key entity information from the text, such as the patient's age, gender, chief complaint, self-reported dietary behavior, self-reported exercise habits, and specific weight loss goals. These extracted entities are categorized and assigned predefined labels, such as "Entity Type: Weight Loss Goal, Entity Value: 5 kg weight loss per month".

[0070] 3) The system analyzes and categorizes the identified conflicting needs. A conflicting need refers to a logical clash between the target patient's weight loss goals and their self-reported dietary or exercise behaviors. The system uses a rule engine or a lightweight classification model to map each conflicting need to a standardized conflict type library. For example, if the pre-diagnosis briefing contains the description "the patient expects rapid weight loss but their daily diet includes many high-calorie snacks," the system will categorize it as the standard type "goal and dietary behavior mismatch" and may assign it a confidence score based on keywords (such as "high-calorie"), with a value range of 0 to 1.

[0071] 4) After completing information extraction and contradiction analysis, the system enters the clinical context enrichment stage. In this stage, the system associates the extracted entities and standardized contradictions with a lightweight medical knowledge base. For example, when the system identifies a patient's body mass index (BMI) and "sedentary" behavior, it automatically associates relevant potential risk warnings from the knowledge base, such as "prolonged sedentary behavior is associated with an increased risk of metabolic syndrome." This process adds clinically meaningful annotations to the original information.

[0072] 5) The system synthesizes and outputs a structured pre-diagnosis report. All processed information—including standardized basic patient information, an extracted list of health summary entities, categorized conflicting needs and their confidence levels, and associated clinical risk warnings—is populated into a pre-defined report template. This template defines the sections and fields of the structured pre-diagnosis report, such as "Patient Identifier," "Chief Complaint Summary," "Conflict Identification," and "Preliminary Risk Assessment." The final generated structured pre-diagnosis report typically uses a machine-readable data format, such as JSON or XML, while also providing a document view for easy browsing by physicians. This report is then stored in the system and prepared for presentation to physicians during the consultation phase, completing the full pre-diagnosis process from interactive consultation to structured analysis report.

[0073] In another possible approach, the specific process for generating a structured pre-diagnosis report is as follows: 1) Receive a pre-diagnosis briefing output from the core module of the obesity-specific disease model. This briefing contains identified conflicting needs described in natural language. Utilize a pre-trained information extraction model to perform multiple rounds of parsing and entity linking on the briefing. First, extract the patient's chief complaint and behavioral habit entities using named entity recognition technology. Then, standardize these entities. Finally, accurately map the standardized entities to professional concepts in the medical ontology database to construct a patient-centered basic fact network. This network represents the semantic relationships between entities in a graph structure. Specifically: The system first receives the pre-diagnosis briefing data stream from the core module of the obesity specialty large model through the application programming interface. The pre-diagnosis briefing is a document presented in the form of natural language text, which contains descriptions of the identified requirement contradictions. The system temporarily stores the received text data in the memory buffer and unifies the encoding format to ensure the input consistency for subsequent processing steps. The system loads a pre-trained information extraction model, which is based on a deep neural network architecture and is optimized specifically for medical texts. The model performs multiple rounds of parsing on the pre-diagnosis briefing. In the first round of parsing, the model uses named entity recognition technology to scan the full text, identify and label the patient's chief complaint entities, such as discomfort symptoms and duration, as well as behavioral habit entities, such as dietary preferences and exercise frequency. These identified entities are extracted in the form of original text fragments. The system standardizes the extracted entities. This step is achieved by querying a predefined medical term dictionary. For example, mapping the patient's expression of eating less and more frequently to the standardized term of regular small amounts and multiple times of eating, and mapping not wanting to move to lack of exercise. At the same time, the system unifies the units of numerical entities, such as converting catties to kilograms. The standardized entities are transformed into a standardized medical term expression. After completing entity standardization, the system performs entity linking operations, precisely mapping these standardized entities to professional concepts in the medical ontology library. The medical ontology library contains definitions of concepts, attributes, and relationships related to the field of obesity. The system finds a unique matching concept node in the ontology library for each entity. For example, linking the standardized high-fat diet behavior to the inappropriate diet behavior concept node in the ontology library. Finally, the system constructs a patient-centered ground truth network based on these mapping relationships. This network is represented in the form of a graph structure, where nodes represent the linked medical concepts and edges represent the semantic relationships between concepts. For example, the patient node is connected to the high-calorie intake node through the exhibits edge, and the high-calorie intake node is connected to the weight gain node through the leads_to edge. This graph structure completely expresses the key medical facts and their interrelationships in the pre-diagnosis briefing, providing a structured knowledge basis for subsequent quantitative analysis and report generation.

[0074] 2) Based on the constructed ground truth network, quantitatively evaluate and prioritize the requirement contradictions described in the pre-diagnosis briefing; by calculating the logical conflict intensity between the weight loss target value and the self-reported behavior data, assign a conflict score to each requirement contradiction. The conflict score is composed of the product of the target achievement difficulty coefficient and the behavior deviation degree. The formula is , where is the target difficulty, is the behavior deviation degree, and are adjustment coefficients; sort all the identified requirement contradictions in descending order according to the conflict scores to form a priority sequence. Specifically: First, the system reads the established fact network and extracts key data related to conflicting needs. A conflicting need refers to the logical conflict between the target patient's weight loss goal and their self-reported dietary or exercise behaviors. The system locates the weight loss goal value node in the network and obtains its numerical representation, such as a planned weight loss of 5 kg per month. Simultaneously, the system traverses the behavioral habit entity nodes in the network, extracting relevant self-reported behavioral data, such as daily calorie intake or weekly exercise duration. The system calculates two key parameters for each identified conflicting need. The first parameter is the difficulty coefficient for achieving the goal. This coefficient is calculated based on the difference between the weight loss target and the patient's current physiological state. For example, for a patient with a high body mass index, a monthly weight loss target of 5 kg might be assigned a lower difficulty coefficient. For another patient who is close to the ideal weight, the same goal may be more challenging. The second parameter is the behavioral deviation. This parameter quantifies the degree of discrepancy between self-reported behavioral data and the ideal behavior required to achieve weight loss goals. For example, the patient's self-reported daily calorie intake. Compared with the system's calculated recommended intake The deviation between them can be expressed by the formula The calculations show that the system assigns a conflict score to each conflicting demand using these parameters. The conflict score is the product of the difficulty coefficient for achieving the goal and the degree of behavioral deviation, with an adjustment coefficient used for calibration. The specific formula is as follows: The adjustment coefficient is among them. and Based on clinical experience and presuppositions Used to adjust the influence weight of target difficulty The influence weights used to modulate behavioral deviations. For example, in patients with metabolic obesity, a higher weight might be set. The system emphasizes the accuracy of dietary behavior. It iterates through all conflicting demands and calculates their respective conflict scores. The system sorts all identified conflicting demands in descending order based on their conflict scores. Conflicts with higher scores indicate stronger logical conflicts and require priority attention and handling. The system generates a structured priority sequence, which not only includes the order of the conflicting demands but also records the original description, calculation parameters, and final conflict score for each demand. This priority sequence provides a quantitative basis for subsequent risk assessment and report generation.

[0075] 3) The prioritized list of conflicting needs is mapped to the patient's obesity profile. Potential complication risk factors associated with each high-priority conflict are retrieved from the medical knowledge graph. Based on the severity and correlation strength of these risk factors, targeted clinical warnings are generated and linked to the corresponding conflicts, forming a risk-assessed enhanced conflict list. This list includes both the original conflict description and relevant clinical risk annotations. Specifically: The system first performs a data association operation, connecting the priority-sorted list of conflicting needs with the multidimensional feature profile of the target patient. This process uses a unique patient identifier to link the two types of data. The system reads key dimension information from the obesity feature profile, including etiological classification, physiological metabolic indicators, and historical risk assessment results. For example, the system will determine if the patient in the profile is classified as having metabolic obesity and will read specific values ​​such as fasting blood glucose and blood lipids.

[0076] The system accesses the medical knowledge graph and performs risk factor retrieval for each high-priority conflict in the priority list. The system translates the conflict into a knowledge graph query. For example, for a high-priority conflict between a patient's weight loss goal and a high-fat diet, the system queries the knowledge graph for the high-fat diet node and iterates through its associated potential complication risk factor nodes, such as type 2 diabetes and hyperlipidemia. The retrieval process also retrieves the severity level of these risk factor nodes. And the strength of the association between this factor and contradictory behaviors. .

[0077] Then, the system determines the severity of the retrieved risk factors. and correlation strength This generates targeted clinical alerts. The system uses a weighted calculation model to assess the overall alert level for each risk factor. The calculation formula is: ,in and These are predefined weighting coefficients. According to... Given a given numerical range, the system selects a matching alert text from a pre-set alert template library. For example, for a high severity level... And strong correlation These risk factors may trigger warnings. Maintaining a high-fat diet for a long period will significantly increase your risk of developing type 2 diabetes. It is recommended to adjust your diet immediately.

[0078] Finally, the system binds these generated clinical alerts to their corresponding conflicting needs. Each high-priority conflict no longer only has its original description in the list, but is expanded into a structured data object containing the original conflict description, the calculated conflict score, one or more associated risk factors, and the corresponding clinical alert. This information-bound list is the risk assessment-enhanced conflict list, which provides depth and clinical context for structured pre-diagnosis reports.

[0079] 4) Based on the preset structured pre-diagnosis report template, the system automatically populates and formats the basic fact network, the list of contradictions enhanced by risk assessment, and relevant quantitative indicators. The template includes three main chapters: patient summary, contradiction analysis, and risk assessment recommendations. The system generates the final structured pre-diagnosis report document according to the data structure defined in the template. This document can be directly imported into the multidisciplinary collaborative diagnosis and treatment subsystem for doctors to view and use. Specifically: The system first loads a pre-defined structured pre-diagnosis report template, which defines the complete structure and data fields of the report in XML or JSON format. The template explicitly includes three main sections: Patient Summary, Conflict Analysis, and Risk Assessment Recommendations. Each section is further subdivided into multiple fields; for example, the Patient Summary section includes basic information and key health indicators, the Conflict Analysis section includes a list of conflict points and a conflict score, and the Risk Assessment Recommendations section includes clinical warnings and priority interventions. The system then performs data extraction and mapping operations. It extracts the patient's core medical facts from the basic fact network, including standardized chief complaint entities and behavioral habit entities and their relationships. It extracts complete information for each conflict point from the risk assessment-enhanced list of conflict points, including the original conflict description and conflict score. In addition, the system will collect relevant quantitative indicators, such as the difficulty coefficient of achieving the target. and behavioral deviation The system automatically populates data according to the template-defined data structure. For the patient summary section, the system maps basic patient information from the basic fact network to the corresponding fields. For the conflict analysis section, the system populates the conflict list field with the risk assessment enhancement conflict points in priority order and creates a standardized display format for each entry. For the risk assessment recommendations section, the system categorizes and organizes the clinical warnings associated with each conflict point and fills them into the corresponding fields. All numerical data, such as conflict scores... All data is formatted according to the precision and units required by the template. The system generates the final structured pre-diagnosis report document. The system serializes the fully populated data structure into two document formats: one is a machine-readable JSON document, preserving the complete data structure and numerical precision; the other is a formatted document, such as PDF or HTML, that is easy for doctors to read directly. Both document formats can be directly imported into the multidisciplinary collaborative diagnosis and treatment subsystem via application programming interface, ensuring that doctors can promptly access and use these structured pre-diagnosis analysis results.

[0080] In another embodiment, the present invention provides an obesity full-cycle management system based on an artificial intelligence large model, comprising a multimodal data fusion and preprocessing module, an obesity-specific large model core module, and a full-cycle management application module. The multimodal data fusion and preprocessing module integrates and structures data from different sources, including electronic medical record data, wearable device data, and patient-reported data. Electronic medical record data covers patient complaints, medical history, laboratory test results, imaging reports, and other information. Wearable device data includes real-time collected physiological parameters such as heart rate, steps, sleep, weight, and body fat. Patient-reported data includes dietary logs, exercise experiences, and self-reported attributions input via voice or text / image interaction. This module converts multi-source data into structured multimodal patient data in a unified format through data cleaning, entity extraction, and standardization, providing a foundation for subsequent analysis. The obesity-specific large model core module is built upon a large language model fine-tuned with high-quality obesity domain knowledge, integrating a medical inference engine capable of professional clinical logical reasoning, effectively suppressing AI illusions, and ensuring the scientific validity and safety of output recommendations. This module possesses the following core capabilities: intelligent pre-diagnosis and demand conflict identification, automatically collecting information through natural language interaction in the pre-diagnosis stage to identify logical contradictions between the patient's weight loss goals and current behaviors, generating a structured pre-diagnosis brief; multi-dimensional etiological classification and risk assessment, comprehensively analyzing structured multimodal patient data to classify obesity etiologically and assess individualized health risks; and personalized plan generation and dynamic optimization, based on the patient's multi-dimensional feature profile, combining the logic of nutrition, exercise, psychology, and other disciplines to generate an initial personalized intervention plan including dietary recommendations and exercise plans, and dynamically adjusting plan parameters based on post-diagnosis monitoring data. The full-cycle management application module includes a pre-diagnosis intelligent pre-diagnosis subsystem, an in-diagnosis multi-disciplinary collaborative diagnosis and treatment subsystem, and a post-diagnosis dynamic tracking and follow-up subsystem. The pre-diagnosis intelligent pre-diagnosis subsystem automates the collection and preliminary analysis of patient information, improving physician efficiency. The in-diagnosis multi-disciplinary collaborative diagnosis and treatment subsystem provides physicians with visualized multi-dimensional patient feature profiles and evidence-based intervention plan suggestions, supporting online collaboration among multidisciplinary experts to develop plans. The post-diagnosis dynamic tracking and follow-up subsystem automatically integrates data from wearable devices to generate dynamic health records, providing continuous health education and reminders to consolidate long-term, remote intervention effects and prevent rebound. This module generates a customized management plan based on the initial personalized intervention plan and the target patient's multidimensional profile, and dynamically adjusts the customized management plan based on the latest monitored physiological and behavioral data to generate the final personalized intervention plan. The specific workflow is as follows: In the pre-diagnosis stage, intelligent pre-consultation is conducted through voice or text-based interaction to collect patients' health information and weight loss needs. The core module of the obesity-specific disease model identifies conflicting needs, pre-loads clinical decisions, and generates a pre-diagnosis briefing containing key issues. Conflicting needs refer to the logical conflict between the target patient's weight loss goals and their self-reported dietary or exercise behaviors. During the diagnosis stage: Pre-diagnosis information and clinical examination data are integrated to construct a multi-dimensional profile of the patient's obesity characteristics. The core module of the obesity-specific disease model is invoked to conduct nutritional and exercise capacity assessments, generating personalized dietary recommendations and exercise plans as initial personalized intervention programs. Doctors refer to the model's recommendations to formulate the final treatment plan, and the plan information is synchronized to the management platform. In the post-diagnosis stage: Data on weight, body fat, and dietary and exercise adherence reported by patients are continuously received from wearable devices. The core module of the obesity-specific disease model evaluates the intervention effect in real time based on the data stream. If indicators deviate from the expected trajectory, automatic warnings are issued and suggestions for plan adjustments are provided, such as adjusting exercise intensity or optimizing the diet. Personalized health knowledge is pushed to patients, and regular follow-ups are conducted to form a closed-loop management system. The full-cycle management application module dynamically adjusts the customized management plan based on the latest monitoring data, generating the final personalized intervention plan.

[0081] The beneficial effects of this invention are as follows: Its technological innovation lies in the creation of a large-scale medical model focused on obesity, solving the professionalism and reliability issues of general-purpose models in medical scenarios through a technical approach combining authoritative knowledge, clinical data, and a dedicated inference engine. The innovative treatment model shifts from passive treatment to proactive, end-to-end management. Through a three-tiered prevention network design, it moves the management focus forward, covering the entire life cycle and addressing the pain points of late intervention and management gaps. It improves efficiency and homogenization by enhancing physician efficiency through intelligent pre-consultation and decision support functions, empowering primary care physicians to provide obesity management services of the same quality as those offered by top medical institutions. It enhances patient compliance by extending management to the home environment through humanized interaction, continuous dynamic feedback, and personalized science education, improving the problems of poor patient compliance and easy relapse in traditional management models. Specifically: This marks the first time a large-scale medical model specifically focused on obesity has been created. The core module of this obesity-specific model is implemented through a technical approach combining authoritative knowledge, clinical data, and a dedicated inference engine. Specifically, the model is trained based on authoritative medical guidelines and expert consensus, deeply integrating structured multimodal patient data provided by a multimodal data fusion and preprocessing module, and incorporating a dedicated medical inference engine for clinical logic reasoning. This technical approach ensures the scientific validity and safety of the output recommendations, effectively addressing issues such as inaccurate information generation and insufficient professional knowledge in general-purpose large-scale models within medical scenarios. It achieves a shift from passive treatment to proactive, end-to-end management. By constructing a three-tiered prevention network design covering risk identification, etiological classification, and precise management, the management focus is shifted forward. The system achieves early risk identification through a pre-diagnosis intelligent pre-consultation subsystem, completes precise etiological classification through a multidisciplinary collaborative diagnosis and treatment subsystem during diagnosis, and implements long-term precise management through a post-diagnosis dynamic tracking and follow-up subsystem. This design extends the service scope to the entire life cycle, effectively solving the problems of late intervention and management gaps in traditional models. Intelligent pre-consultation and decision support functions improve doctors' work efficiency. The pre-diagnosis intelligent consultation subsystem automatically generates structured pre-diagnosis reports, while the in-diagnosis multidisciplinary collaborative treatment subsystem provides intervention plan suggestions based on multidimensional patient profiles. These functions enable primary care physicians to receive decision support at the same level as top medical institutions, gaining professional guidance in developing initial personalized intervention plans and generating customized management plans, thereby promoting the balanced allocation of medical resources. The system extends management to the home environment through user-friendly natural language interaction, continuous dynamic feedback, and personalized health knowledge delivery. The post-diagnosis dynamic tracking and follow-up subsystem continuously receives physiological parameter data from wearable devices and behavioral data proactively reported by patients. Based on this latest monitoring data, the customized management plan is dynamically adjusted to generate the final personalized intervention plan. This continuous management and adjustment effectively improves the problems of poor patient compliance and easy relapse after intervention in traditional management models.

[0082] like Figure 2 As shown in the figure, an embodiment of the present invention provides a method for full-cycle obesity management based on an artificial intelligence large model, comprising: S1. Integrate and structure the obesity-related data of the target patients to generate structured multimodal patient data; S2. The large language model obtained by training with knowledge in the obesity domain processes the structured multimodal patient data and outputs a pre-diagnosis briefing containing identified conflicting needs and an initial personalized intervention plan containing dietary recommendations and exercise plans. Conflicting needs refer to the logical conflict between the target patient's weight loss goal and the target patient's self-reported diet or exercise behavior. S3. Generate a customized management plan based on the initial personalized intervention plan and the multidimensional feature profile of the target patient.

[0083] Optionally, the above technical solution also includes: dynamically adjusting the customized management plan based on the latest monitored physiological parameter data and behavioral data of the target patient to generate a final personalized intervention plan.

[0084] Optionally, the above technical solution also includes: performing pre-diagnosis analysis based on the pre-diagnosis briefing to generate a structured pre-diagnosis report.

[0085] Optionally, in the above technical solution, obesity-related data of the target patient is integrated and structured to generate structured multimodal patient data, including: The system integrates and structures electronic medical record data of target patients, physiological parameter data collected by wearable devices, and behavioral data actively reported by target patients to generate structured multimodal patient data.

[0086] It should be noted that the beneficial effects of the obesity full-cycle management method based on an artificial intelligence large model provided in the above embodiments are the same as the beneficial effects of the obesity full-cycle management system based on an artificial intelligence large model described above, and will not be repeated here. Furthermore, the system and method embodiments provided in the above embodiments belong to the same concept, and their specific implementation processes are detailed in the method embodiments, and will not be repeated here.

[0087] The obesity full-cycle management system based on artificial intelligence big data model of the present invention can be a computer program (including program code) running on a computer device. For example, the obesity full-cycle management system based on artificial intelligence big data model of the present invention is an application software that can be used to execute the corresponding steps in the obesity full-cycle management method based on artificial intelligence big data model of the present invention.

[0088] In some embodiments, the obesity full-cycle management system based on an artificial intelligence big model of the present invention can be implemented in a combination of hardware and software. As an example, the obesity full-cycle management system based on an artificial intelligence big model of the present invention can be a processor in the form of a hardware decoding processor, which is programmed to execute the obesity full-cycle management method based on an artificial intelligence big model of the present invention. For example, the processor in the form of a hardware decoding processor can be one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components.

[0089] The modules described in the embodiments of this invention can be implemented in software or hardware. The names of the modules are not, in some cases, limiting the scope of the module itself.

[0090] An electronic device according to an embodiment of the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any of the above-mentioned methods for full-cycle obesity management based on a large artificial intelligence model. That is, an electronic device according to an embodiment of the present invention may include, but is not limited to: a processor and a memory; the memory is used to store the computer program; the processor is used to execute the method for full-cycle obesity management based on a large artificial intelligence model shown in any embodiment of the present invention by calling the computer program.

[0091] In one alternative embodiment, an electronic device is provided, such as Figure 3 As shown, Figure 3 The illustrated electronic device 4000 includes a processor 4001 and a memory 4003. The processor 4001 and the memory 4003 are connected, for example, via a bus 4002. Optionally, the electronic device 4000 may further include a transceiver 4004, which can be used for data interaction between the electronic device and other electronic devices, such as sending and / or receiving data. It should be noted that in practical applications, the transceiver 4004 is not limited to one type, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of the present invention.

[0092] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this invention. Processor 4001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0093] Bus 4002 may include a path for transmitting information between the aforementioned components. Bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 4002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 3 The bus 4002 is represented by only one thick line, but this does not mean that there is only one bus or one type of bus.

[0094] The memory 4003 may be ROM (Read Only Memory) or other types of static storage devices capable of storing static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices capable of storing information and instructions, or EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto.

[0095] The memory 4003 stores application code (computer program) for executing the present invention, and its execution is controlled by the processor 4001. The processor 4001 executes the application code stored in the memory 4003 to implement the content shown in the foregoing method embodiments.

[0096] Among them, electronic devices can also be terminal devices, which can be any device that can install applications, including at least one of smartphones, tablets, laptops, desktop computers, smart speakers, smartwatches, smart TVs, and smart in-vehicle devices.

[0097] It should be noted that, Figure 3 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0098] An embodiment of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements any of the above-mentioned methods for full-cycle obesity management based on a large artificial intelligence model.

[0099] Alternatively, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), magnetic tape, a floppy disk, and an optical data storage device, etc.

[0100] In an exemplary embodiment, a computer program product or computer program is also provided, which includes computer instructions stored in a computer-readable storage medium. The processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the aforementioned obesity full-cycle management method based on a large-scale artificial intelligence model.

[0101] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0102] It should be understood that the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of methods and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0103] The computer-readable storage medium provided in this invention can be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EEPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0104] The aforementioned computer-readable storage medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the method shown in the above embodiments.

[0105] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this invention is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this invention.

[0106] It should be noted that the terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and represent a limitation on a specific order or sequence. Where appropriate, the order of use for similar objects can be interchanged so that the embodiments of this application described herein can be implemented in an order other than that shown or described.

[0107] Those skilled in the art will recognize that this invention can be implemented as a system, method, or computer program product. Therefore, this invention can be specifically implemented in the following forms: it can be entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, generally referred to herein as a "circuit," "module," or "system." Furthermore, in some embodiments, this invention can also be implemented as a computer program product contained in one or more computer-readable media, which includes computer-readable program code.

[0108] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A comprehensive obesity management system based on a large-scale artificial intelligence model, characterized in that, It includes a multimodal data fusion and preprocessing module, a core module for a large-scale obesity disease model, and a full-cycle management application module; The multimodal data fusion and preprocessing module is used to: integrate and structure the obesity-related data of the target patients to generate structured multimodal patient data; The core module of the obesity-specific big model is used to: process the structured multimodal patient data using a big language model trained with knowledge of the obesity domain, and output a pre-diagnosis brief containing identified conflicting needs and an initial personalized intervention plan containing dietary recommendations and exercise plans. Conflicting needs refer to the logical conflict between the target patient's weight loss goal and the target patient's self-reported diet or exercise behavior. The full-cycle management application module is used to generate a customized management plan based on the initial personalized intervention plan and the multi-dimensional feature profile of the target patient.

2. The obesity full-cycle management system based on an artificial intelligence large model according to claim 1, characterized in that, The full-cycle management application module is also used to: dynamically adjust the customized management plan based on the latest monitored physiological parameter data and behavioral data of the target patient, and generate the final personalized intervention plan.

3. The obesity full-cycle management system based on an artificial intelligence large model according to claim 1, characterized in that, The full-cycle management application module is also used to: perform pre-diagnosis analysis based on the pre-diagnosis briefing and generate a structured pre-diagnosis report.

4. A full-cycle obesity management system based on an artificial intelligence large model according to any one of claims 1 to 3, characterized in that, The multimodal data fusion and preprocessing module is specifically used to: integrate and structure the electronic medical record data of the target patient, the physiological parameter data collected by the wearable device, and the behavioral data actively reported by the target patient to generate structured multimodal patient data.

5. A method for full-cycle obesity management based on a large artificial intelligence model, characterized in that, include: Integrate and structure obesity-related data of target patients to generate structured multimodal patient data; The large language model obtained by training with knowledge in the obesity domain processes the structured multimodal patient data and outputs a pre-diagnosis briefing containing identified conflicting needs and an initial personalized intervention plan containing dietary recommendations and exercise plans. Conflicting needs refer to the logical conflict between the target patient's weight loss goal and the target patient's self-reported diet or exercise behavior. A customized management plan is generated based on the initial personalized intervention plan and the multidimensional feature profile of the target patient.

6. The method for full-cycle obesity management based on a large artificial intelligence model according to claim 5, characterized in that, Also includes: The customized management plan is dynamically adjusted based on the latest physiological and behavioral data of the target patients to generate the final personalized intervention plan.

7. The method for full-cycle obesity management based on a large artificial intelligence model according to claim 5, characterized in that, Also includes: Based on the aforementioned pre-diagnosis briefing, a pre-diagnosis analysis is performed to generate a structured pre-diagnosis report.

8. A method for full-cycle obesity management based on a large artificial intelligence model according to any one of claims 5 to 7, characterized in that, Integrate and structure obesity-related data from target patients to generate structured, multimodal patient data, including: The system integrates and structures the electronic medical record data of the target patient, the physiological parameter data collected by wearable devices, and the behavioral data actively reported by the target patient to generate structured multimodal patient data.

9. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the obesity full-cycle management method based on an artificial intelligence large model as described in any one of claims 5 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the obesity full-cycle management method based on an artificial intelligence large model as described in any one of claims 5 to 8.

Images