Medication management method and device based on AI decision engine, medium and equipment

Abstract

The invention discloses a medication management method and device based on an AI decision engine, a medium and equipment, relates to the technical field of health management, can be applied to a medical health business scene, and comprises the following steps: collecting multi-dimensional data of a user; dynamic fusion and intelligent analysis are carried out on the multi-dimensional data of the user through an AI decision engine, a preliminary drug use recommendation scheme is generated, and the preliminary drug use recommendation scheme comprises a target drug use opportunity window matched with the real-time state of the user and a target dose recommendation value; performing risk optimization processing on the preliminary drug use recommendation scheme, submitting the preliminary drug use recommendation scheme to a doctor terminal, and forming a final drug use recommendation scheme after receiving a confirmation instruction of the doctor terminal; and based on the final medication recommendation scheme, the intelligent device is linked to execute medication management operation, and the final medication recommendation scheme is synchronously presented to the user terminal. According to the method, the medication opportunity and dosage can be adjusted according to the real-time change of the individual health indexes of the user, and the core requirements of the user for dynamic, safe and personalized medication management are met.

Description

Technical Field

[0001] This application relates to the field of health management technology, and in particular to a medication management method, device, medium and equipment based on an AI decision engine. Background Technology

[0002] In the healthcare field, standardized treatment of chronic diseases, medication safety for the elderly, and the integration of daily health management with medication have become core areas of focus. With the increasing aging population, my country's chronic disease patient population continues to expand. These patients require long-term, regular medication, and the effectiveness of medication is closely related to real-time health indicators (such as blood pressure and blood sugar). Meanwhile, the elderly frequently experience problems such as missed doses, incorrect dosages, and difficulty finding medications due to declining vision and memory, seriously affecting treatment outcomes and even posing safety risks. Furthermore, the demand for integrated health management is increasingly prominent. Users expect medication management to be integrated with daily health monitoring and medical resources, rather than simply providing isolated medication storage and reminders. Against this backdrop, the integration of IoT and AI technologies with healthcare services is becoming a trend, urgently requiring an intelligent medication management solution that adapts to core healthcare scenarios and balances safety and accuracy to address the pain points of traditional medication management being disconnected from health monitoring and unable to meet personalized medical needs.

[0003] Currently, medication management products and technologies in the industry mainly revolve around drug information management and fixed reminder functions. Mainstream solutions often employ a model of photo recognition of drug instructions combined with manual reminder settings. Static information such as drug name, specifications, and dosage is stored through mobile applications or smart medicine cabinets, and pop-up or voice reminders are triggered based on user-preset fixed times. Existing medication management solutions lack the ability to dynamically perceive and comprehensively analyze users' real-time health status (such as heart rate, blood pressure, and blood sugar fluctuations) and daily habits (such as sleep patterns, meal scenarios, and activity trajectories). This directly results in extremely poor personalization adaptability of existing solutions, rigid and inflexible medication reminder rules, and a lack of targeted medication recommendations. They cannot adjust medication timing and dosage based on real-time changes in individual user health indicators, nor can they predict medication risks. Furthermore, reminders frequently become disconnected from the user's actual situation (e.g., reminders for pre-meal medications are triggered when the user has not eaten, or strong, disruptive reminders are triggered during sleep periods). Ultimately, they fail to meet users' core needs for dynamic, safe, and personalized medication management. Summary of the Invention

[0004] In view of this, this application provides a medication management method, device, medium and equipment based on an AI decision engine, which can adjust the timing and dosage of medication according to the real-time changes of individual user health indicators, thereby meeting the user's core needs for dynamic, safe and personalized medication management.

[0005] According to a first aspect of this application, a medication management method based on an AI decision engine is provided, the method comprising: Collect multi-dimensional user data, including real-time health indicator data, daily behavior data, and static drug information data; The AI ​​decision engine dynamically integrates and intelligently analyzes the user's multi-dimensional data to generate a preliminary medication recommendation plan. The preliminary medication recommendation plan includes a target medication timing window and a target dose recommendation value that are adapted to the user's real-time status. After risk optimization processing, the preliminary medication recommendation plan is submitted to the doctor's terminal, and the final medication recommendation plan is formed upon receiving confirmation from the doctor's terminal. Based on the final medication recommendation plan, the intelligent device is linked to perform medication management operations, and the final medication recommendation plan is presented to the user terminal simultaneously.

[0006] According to a second aspect of this application, a medication management device based on an AI decision engine is provided, the device comprising: The data acquisition module is used to collect multi-dimensional user data, including real-time health indicator data, daily behavior data, and static drug information data. The generation module is used to dynamically fuse and intelligently analyze the user's multi-dimensional data through an AI decision engine to generate a preliminary medication recommendation plan. The preliminary medication recommendation plan includes a target medication timing window and a target dose recommendation value that are adapted to the user's real-time status. The submission module is used to perform risk optimization processing on the preliminary medication recommendation plan and submit it to the doctor's terminal. After receiving the confirmation instruction from the doctor's terminal, the final medication recommendation plan is formed. The execution module is used to coordinate with the smart device to perform medication management operations based on the final medication recommendation plan, and simultaneously present the final medication recommendation plan to the user terminal.

[0007] According to a third aspect of this application, a storage medium is provided that stores a computer program thereon, which, when executed by a processor, implements the above-described medication management method based on an AI decision engine.

[0008] According to a fourth aspect of this application, an electronic device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor executes the program to implement the above-described medication management method based on an AI decision engine.

[0009] By employing the aforementioned technical solutions, the medication management method, device, medium, and equipment based on an AI decision engine provided in this application collect multi-dimensional data on users' real-time health indicators, daily behaviors, and static drug information. The AI ​​decision engine dynamically fuses and intelligently analyzes this multi-dimensional data to generate a preliminary medication plan that adapts to the user's real-time status, including a medication timing window and recommended dosage values. This plan is then optimized for risk and confirmed by the doctor's terminal to form a final plan. Finally, the intelligent device executes the medication management operation and presents the plan to the user's terminal. This effectively solves the problem of existing medication management plans lacking dynamic perception and comprehensive analysis capabilities, breaking the limitations of traditional fixed reminders and static medication suggestions. Medication recommendations can accurately match the user's real-time status, improving the personalization and targeting of medication suggestions. Simultaneously, through risk optimization and doctor confirmation, dual safeguards are provided, reducing medication risks and preventing reminders from becoming disconnected from the user's actual situation. Ultimately, this meets users' core needs for dynamic, safe, and personalized medication management.

[0010] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0011] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 The illustration shows a flowchart of a medication management method based on an AI decision engine provided in an embodiment of this application; Figure 2 A flowchart illustrating a medication management method based on an AI decision engine, according to another embodiment of this application, is shown. Figure 3 This illustration shows a schematic diagram of a medication management device based on an AI decision engine, as provided in an embodiment of this application. Figure 4 A schematic diagram of the structure of a medication management device based on an AI decision engine, provided in another embodiment of this application, is shown. Detailed Implementation

[0012] The present application will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of the present application can be combined with each other.

[0013] Currently, medication management products and technologies in the industry mainly revolve around drug information management and fixed reminder functions. Mainstream solutions often employ a model of photo recognition of drug instructions combined with manual reminder settings. Static information such as drug name, specifications, and dosage is stored through mobile applications or smart medicine cabinets, and pop-up or voice reminders are triggered based on user-preset fixed times. Existing medication management solutions lack the ability to dynamically perceive and comprehensively analyze users' real-time health status (such as heart rate, blood pressure, and blood sugar fluctuations) and daily habits (such as sleep patterns, meal scenarios, and activity trajectories). This directly results in extremely poor personalization adaptability of existing solutions, rigid and inflexible medication reminder rules, and a lack of targeted medication recommendations. They cannot adjust medication timing and dosage based on real-time changes in individual user health indicators, nor can they predict medication risks. Furthermore, reminders frequently become disconnected from the user's actual situation (e.g., reminders for pre-meal medications are triggered when the user has not eaten, or strong, disruptive reminders are triggered during sleep periods). Ultimately, they fail to meet users' core needs for dynamic, safe, and personalized medication management.

[0014] To address the aforementioned technical problems, embodiments of the present invention provide a medication management method based on an AI decision engine, such as... Figure 1 As shown, the method includes: Step 110: Collect multi-dimensional user data, including real-time health indicator data, daily behavior data, and static drug information data.

[0015] Real-time health indicator data refers to dynamic data reflecting the user's immediate physiological state, collected at preset frequencies through wearable health devices (such as smartwatches, blood pressure monitors, and blood glucose meters). This data includes at least heart rate, blood pressure, blood sugar, sleep quality, and activity level, serving as the core physiological basis for judging the user's real-time physical condition and adapting medication plans. Daily behavior data refers to data collected through smart home devices, reflecting the user's daily routine and activity status. This includes voice interaction data collected by smart speakers and action interaction data collected by smart cameras. This data can be used to extract temporal features of behavioral scenarios (such as mealtimes, sleep, and activity periods), providing support for scenario-based adaptation of medication timing. Static drug information data refers to the inherent basic attribute data of drugs, obtained by scanning drug instructions or manually entering information by the user. This data includes at least the drug name, specifications, dosage, expiration date, metabolic cycle, and contraindications. This data serves as the basic reference data for formulating medication plans and is also the core basis for drug lifecycle management.

[0016] In this embodiment of the disclosure, with the user's authorization, an encrypted communication link can be established with the wearable health device through the device SDK, and the user's real-time health indicator data such as heart rate and blood pressure can be synchronized at a preset frequency; at the same time, the user's voice interaction data and action interaction data can be collected by relying on smart home devices to form daily behavior data; and then, by scanning the drug instructions or receiving manual input from the user, static information data of the drug such as drug name, specifications, and expiration date can be obtained, and finally, the three types of data are integrated to form complete multi-dimensional user data.

[0017] By synchronously collecting users' real-time health indicators, daily behavior data, and static drug information, the limitations of existing medication management plans that rely solely on static drug information can be overcome. A multi-dimensional data system integrating physiological state, behavioral scenarios, and drug attributes can be constructed, providing comprehensive data support for the personalized and dynamic adjustment of subsequent medication plans. This effectively avoids the problem of rigid medication recommendations and disconnect from the user's actual situation caused by a single data dimension. At the same time, it can also lay a data foundation for medication risk prediction and full life cycle management of drugs, improving the scientificity and adaptability of medication management plans from the source.

[0018] Step 120: Dynamically fuse and intelligently analyze multi-dimensional user data through an AI decision engine to generate a preliminary medication recommendation plan. The preliminary medication recommendation plan includes a target medication timing window and a target dose recommendation value that are adapted to the user's real-time status.

[0019] The AI ​​decision engine is an intelligent analysis system integrating data structuring, model building, knowledge graph reasoning, and deep learning prediction. It can perform fusion processing and intelligent judgment on multi-source heterogeneous data. Dynamic fusion and intelligent analysis refers to the core processing process in which the AI ​​decision engine constructs a multi-dimensional correlation model after structuring the collected multi-dimensional data, and combines deep learning algorithms to mine the immediate correlation and long-term patterns between data, ultimately realizing the transformation of data value. The preliminary medication recommendation plan is an initial medication suggestion generated by the AI ​​decision engine based on multi-dimensional data analysis, without risk optimization and doctor confirmation. Its core content includes medication timing and dosage recommendations adapted to the user's real-time status. The target medication timing window refers to the most suitable time period for medication selected by combining the user's real-time health index fluctuations and behavioral scenario characteristics, which can maximize drug efficacy and reduce conflicts with the user's life scenarios. The target dosage recommendation value refers to the personalized medication dosage adjusted for the user based on data such as the long-term trend of the user's health indicators and the drug metabolism cycle, which is different from the fixed standard dosage in the instruction manual.

[0020] In this embodiment of the disclosure, an AI decision engine can be used to first integrate and process the collected multi-dimensional user data. Appropriate algorithms are then used to mine the patterns of user health status, behavioral characteristics, and drug compatibility associations contained within the data. This is combined with a built-in medication-related knowledge system to complete a deep fusion analysis of the multi-dimensional data. Based on the analysis results, the target medication timing window and target dosage recommendation value suitable for the user's real-time status are accurately determined, and finally, a preliminary medication recommendation plan is formed. The specific algorithms for data processing, feature mining, and plan generation can employ various implementation methods such as deep learning and knowledge graph reasoning.

[0021] By dynamically fusing and intelligently analyzing multi-dimensional user data through an AI decision engine, the limitations of traditional medication plans relying on single static data can be broken. This enables in-depth correlation analysis of user physiological state, behavioral characteristics, and drug attributes, allowing the generated preliminary medication recommendations to closely match the user's real-time status and accurately match personalized target medication timing windows and target dosage recommendations. This significantly improves the pertinence and suitability of medication advice and effectively solves the problems of rigid reminders and lack of individual differences in dosage recommendations in traditional medication plans. It provides scientific and reasonable initial support for the safety optimization and professional review of subsequent medication plans, ensuring the dynamic and personalized needs of medication management from the core aspects.

[0022] Step 130: After risk optimization of the preliminary medication recommendation plan, submit it to the doctor's terminal, and after receiving the confirmation instruction from the doctor's terminal, form the final medication recommendation plan.

[0023] Risk optimization refers to the process of systematically identifying, assessing, and adjusting the initial medication recommendation plan. By analyzing potential medication safety hazards in the plan, the plan is revised and improved to enhance the safety of the medication recommendations. The doctor terminal is a dedicated terminal for medical staff to view and review medication recommendation plans. It allows medical staff to evaluate the plans and provide feedback based on their professional knowledge and clinical experience. The confirmation instruction is the instruction issued by the doctor terminal to the system based on the review results of medical staff, approving the optimized initial medication recommendation plan. It is the key basis for transforming the initial plan into the final plan. The final medication recommendation plan is a formal medication recommendation that has undergone risk optimization and has been professionally reviewed and confirmed by doctors. It has the effect of execution and is the core basis for subsequent medication management operations performed by intelligent devices.

[0024] In this embodiment of the disclosure, a comprehensive risk optimization process can be carried out on the preliminary medication recommendation plan generated by the AI ​​decision engine using preset risk assessment logic and methods. After optimization, the plan is pushed to the doctor's terminal for professional review by medical staff. The medical staff evaluates the plan based on their own professional judgment. If they approve the plan, they send a confirmation instruction to the system through the doctor's terminal. After receiving the confirmation instruction, the system officially confirms the optimized preliminary medication recommendation plan as the final medication recommendation plan. The risk optimization process can be implemented using various methods such as algorithm model evaluation and rule verification.

[0025] By optimizing the initial medication recommendations for risk assessment, potential medication safety hazards can be identified and mitigated in advance, improving the safety and rationality of the medication plan. The process of submitting the recommendations to the doctor's terminal and obtaining confirmation instructions can incorporate clinical review by professional medical personnel, compensating for the limitations of pure algorithm analysis. This ensures that the final medication recommendations are both personalized to the individual user's condition and compliant with professional medical standards, forming a dual guarantee of algorithm optimization and professional review. This effectively solves the problem of traditional medication plans lacking safety verification and professional endorsement, providing users with medication guidance that is both personalized and safe.

[0026] Step 140: Based on the final medication recommendation plan, link the smart device to perform medication management operations and simultaneously present the final medication recommendation plan to the user terminal.

[0027] Medication management operations refer to various intelligent management behaviors carried out throughout the entire medication process, including medication reminders, medication dispensing supervision, and drug lifecycle management. User terminals are terminal devices, such as mobile phones and tablets, that allow users to view medication plans and receive reminder information. They are the main carriers for users to obtain medication-related information.

[0028] In this embodiment of the disclosure, based on the determined final medication recommendation plan, collaborative management instructions adapted to various smart devices can be generated, sent to the corresponding smart devices and driven to perform corresponding medication management operations. At the same time, the final medication recommendation plan is pushed to the user terminal so that the user can view the core information in the plan at any time. The specific type of smart device and the specific method of medication management operation can be flexibly selected according to the actual application scenario.

[0029] Based on the final medication recommendation plan and linked with intelligent devices to execute medication management operations, a closed-loop management of the entire process of medication recommendations from plan generation to implementation can be achieved. With the automation and intelligence of intelligent devices, the accuracy and timeliness of medication management can be effectively guaranteed. At the same time, the final plan is presented to the user's terminal, allowing the user to clearly understand their own medication arrangement, improve medication adherence, and solve the problems of disconnect between plan and execution and untimely access to user information in traditional medication management. It can provide users with convenient, efficient and comprehensive medication management services.

[0030] In summary, the medication management method based on an AI decision engine provided in this application collects multi-dimensional data on users' real-time health indicators, daily behaviors, and static drug information. Using an AI decision engine, it dynamically integrates and intelligently analyzes this multi-dimensional data to generate a preliminary medication plan that adapts to the user's real-time status, including medication timing windows and recommended dosage values. This plan is then optimized for risk and confirmed by the doctor's terminal to form a final plan. Finally, it is linked to intelligent devices to execute medication management operations and present the plan to the user's terminal. This effectively solves the problem of existing medication management plans lacking dynamic perception and comprehensive analysis capabilities, breaking through the limitations of traditional fixed reminders and static medication suggestions. Medication recommendations can accurately match the user's real-time status, improving the personalization and targeting of medication advice. Simultaneously, through risk optimization and doctor confirmation, dual safeguards are provided, reducing medication risks and preventing reminders from becoming disconnected from the user's actual situation. Ultimately, it meets users' core needs for dynamic, safe, and personalized medication management.

[0031] Furthermore, as a refinement and extension of the specific implementation methods of the above embodiments, and to fully illustrate the implementation methods of this embodiment, this embodiment also provides another medication management method based on an AI decision engine, such as... Figure 2 As shown, the method includes: Step 210: Collect multi-dimensional user data, including real-time health indicator data, daily behavior data, and static drug information data.

[0032] For embodiments of this disclosure, the steps may include: after obtaining user authorization, establishing an encrypted communication link with the wearable health device through the device SDK, synchronizing the user's real-time health indicator data at a preset frequency, the real-time health indicator data including at least heart rate, blood pressure, blood sugar, sleep quality, and activity level; collecting the user's daily behavior data through smart home devices, the daily behavior data including voice interaction data and motion interaction data; and collecting static drug information data by scanning the drug instructions or receiving manual input from the user, the static drug information data including at least the drug name, drug specifications, drug dosage, drug expiration date, metabolic cycle, and contraindications.

[0033] Step 220: Dynamically fuse and intelligently analyze multi-dimensional user data through an AI decision engine to generate a preliminary medication recommendation plan.

[0034] For embodiments of this disclosure, step 220 may include the following steps: Step 220-1: Perform structured processing on the multi-dimensional data to extract core parameters of health indicators, temporal features of behavioral scenarios, and key attributes of drugs.

[0035] Structured processing refers to the process of cleaning, transforming, integrating, and classifying multi-dimensional data from multiple sources with different formats according to unified standards and rules, transforming it into a data form with clear logical relationships that can be directly recognized and analyzed by algorithms. Core parameters of health indicators are core data extracted from real-time health indicator data that have a key impact on medication plan formulation. They are core indicators reflecting the user's physiological state, such as peak and trough blood pressure, blood sugar fluctuation range, and average heart rate. Temporal characteristics of behavioral scenarios are user behavior patterns extracted from daily behavior data and marked with a time dimension. They can reflect the user's behavioral state at different times, such as the distribution of meal times, sleep duration and rhythm, and activity level change cycles. Key drug attributes are core attributes closely related to medication safety and efficacy, screened from static drug information data, such as drug dosage range, metabolic cycle duration, contraindications, and expiration date threshold.

[0036] In this embodiment of the disclosure, the collected multi-dimensional data, including real-time health indicator data, daily behavior data, and drug static information data, can first be cleaned to remove invalid and abnormal data. Then, the data can be converted and integrated according to preset data classification rules and standard formats to complete the structured processing of the multi-dimensional data. Subsequently, based on the core requirements for medication plan formulation, the core parameters of health indicators can be extracted from the structured real-time health indicator data through feature extraction algorithms, the temporal features of behavioral scenarios with time dimension can be mined from daily behavior data, and the key attributes of drugs can be screened from drug static information data, providing accurate feature data support for subsequent model building and medication plan generation.

[0037] By structuring multi-dimensional data and extracting core features, redundant and interfering information in the original data can be effectively eliminated, transforming heterogeneous multi-source data into a standardized data form that can be analyzed by algorithms. At the same time, the extracted core parameters of health indicators, temporal features of behavioral scenarios, and key attributes of drugs can accurately focus on the core needs of medication plan formulation. This lays a high-quality data foundation for subsequent construction of three-dimensional dynamic models and generation of personalized medication plans, avoiding the bias in analysis results caused by data clutter and unclear features, and improving the efficiency and accuracy of subsequent intelligent analysis.

[0038] Step 220-2: Construct a three-dimensional dynamic model based on the core parameters of health indicators, the temporal characteristics of behavioral scenarios, and the key attributes of drugs, and dynamically generate a user medication profile based on the three-dimensional dynamic model.

[0039] Among them, the three-dimensional dynamic model is a data analysis model constructed with three core dimensions: core parameters of health indicators, temporal characteristics of behavioral scenarios, and key attributes of drugs. It can depict the relationship between the three and update dynamically over time, and can intuitively reflect the interaction between users' physiological state, behavioral habits, and drug attributes. The user medication profile is a set of individualized medication-related characteristics extracted from the analysis results of the three-dimensional dynamic model. It covers key information such as the user's health status patterns, behavioral scenario preferences, and drug compatibility characteristics, and is the core basis for generating personalized medication plans.

[0040] In this embodiment of the disclosure, the extracted core parameters of health indicators, temporal features of behavioral scenarios, and key attributes of drugs can be used as three core dimensions. Through a preset association algorithm and modeling logic, a dynamic mapping relationship between the three can be established to form a three-dimensional dynamic model that can reflect the user's physiological, behavioral, and drug interaction status in real time. Based on this model, the drug adaptation rules of users under different health states and different behavioral scenarios can be continuously mined and integrated to dynamically generate a user medication profile that comprehensively depicts the user's individualized medication needs.

[0041] By constructing a three-dimensional dynamic model and generating user medication profiles, the limitations of isolated analysis of health indicators, behavioral scenarios, and drug attributes can be overcome. This enables dynamic quantification and visualization of the relationships among the three, accurately capturing individualized health status patterns, behavioral scenario preferences, and drug suitability characteristics. This provides clear user characteristic references for subsequent medication knowledge graph matching and personalized adjustments to medication timing and dosage, significantly improving the targeting and suitability of subsequent medication recommendations and effectively solving the problem of "one-size-fits-all" traditional medication plans.

[0042] Step 220-3: Call the preset medication knowledge graph and use multi-hop reasoning technology to match basic medication information that matches the user's medication profile.

[0043] Among them, multi-hop reasoning technology refers to intelligent algorithm technology that performs logical deduction in a knowledge graph by crossing links of multiple entities and relationships. It can mine indirect but highly relevant medication information based on user characteristics and association rules in the knowledge graph. Basic medication information refers to medication-related content matched from the medication knowledge graph that is initially adapted to the user's medication profile. It includes core information such as the range of recommended drugs, basic medication dosage, and the timing of regular medication, and is the basis for subsequent personalized adjustments.

[0044] In this embodiment of the disclosure, a pre-constructed medication knowledge graph can be retrieved, and the core feature tags in the dynamically generated user medication profile can be used as the starting point for retrieval and reasoning. Multi-hop reasoning technology is used to perform multi-dimensional inferences along the association links such as drug-indication-health status-behavioral scenarios in the knowledge graph, to explore the matching relationship between the user medication profile and various medication information in the knowledge graph, and finally to select basic medication information that is highly compatible with the user's individual characteristics.

[0045] By invoking a pre-defined medication knowledge graph and combining it with multi-hop reasoning technology to match basic medication information, the professional medical knowledge system of the knowledge graph can ensure the scientific and standardized nature of medication recommendations. At the same time, the multi-hop reasoning technology can break through the limitations of single-dimensional matching, and can uncover the deep correlation between user characteristics and medication information. This ensures that the matched basic medication information closely matches the individualized needs of the user's medication profile, avoiding the limitations of traditional keyword-based matching. This provides a scientific and appropriate basis for the precise adjustment of subsequent medication timing and dosage.

[0046] Step 220-4: Extract the corresponding derived time-series data from the three dimensions of the three-dimensional dynamic model. The derived time-series data includes the fluctuation trend of health indicators, the time-series characteristics of behavioral scenarios, and the data related to the drug metabolism cycle.

[0047] Among them, the derived time-series data is a collection of derived data with time dimension characteristics, obtained by processing, analyzing and refining the original data of the three dimensions of the three-dimensional dynamic model through time-series processing, analysis and refinement. It can reflect the regularity and correlation of the data changes over time in each dimension. The health indicator fluctuation trend is extracted from the time-series data of the core parameters of health indicators. It reflects the data characteristics of the changes in the user's physiological indicators in different time periods, such as the daily fluctuation range of blood pressure and the post-meal blood sugar change curve. The behavioral scenario time-series characteristics are behavioral pattern characteristics with time stamps extracted from the user's daily behavior-related data. They can reflect the periodicity and regularity of the user's behavioral scenarios, such as the distribution of daily meal times and changes in sleep rhythm. The drug metabolism cycle correlation data is time-series data that combines the key attributes of drugs with the user's health indicators and behavioral scenario time-series data to mine the correlation between the drug metabolism process and the user's physiological state and behavioral habits. For example, the matching relationship between the peak of drug metabolism and the user's meal time, and the correlation of changes in the user's blood pressure within the metabolic cycle.

[0048] In this embodiment of the disclosure, data from three dimensions—core parameters of health indicators, temporal characteristics of behavioral scenarios, and key attributes of drugs—in a three-dimensional dynamic model can be used as data sources. Based on a preset temporal analysis algorithm, the data in each dimension are processed in terms of time sequence, pattern mining, and correlation analysis. The fluctuation trends of health indicators that reflect the patterns of user physiological changes, the temporal characteristics of behavioral scenarios that reflect user behavior patterns, and the drug metabolism cycle correlation data that characterizes the relationship between drug metabolism and user physiology and behavior are extracted respectively. Finally, they are integrated to form a complete set of derived temporal data.

[0049] Extracting corresponding derived time-series data from the three dimensions of a three-dimensional dynamic model can transform discrete raw dimensional data into correlated data with time patterns. This allows for the discovery of deep correlations between health indicators, behavioral scenarios, and drug metabolism over time. This not only compensates for the inability of static data to reflect change patterns but also provides precise time-series feature inputs for subsequent time-series deep learning model analysis. This ensures the scientific validity and accuracy of adjusting medication timing windows and dosage recommendations, further enhancing the personalized adaptation of medication recommendation plans.

[0050] Step 220-5: Based on the derived time series data, use a time series deep learning model to adjust the medication timing window and recommended dosage values ​​for the basic medication information to form a preliminary medication recommendation plan.

[0051] Among them, the temporal deep learning model is a deep learning model that excels at processing temporal data and mining temporal correlation patterns. It can adopt a CNN-LSTM hybrid architecture. The CNN-LSTM hybrid architecture is a composite model architecture that integrates convolutional neural network (CNN) layers and long short-term memory network (LSTM) layers. The CNN layer is responsible for extracting local features, and the LSTM layer is responsible for learning global temporal dependencies. The two work together to achieve accurate analysis of temporal data. Local features are features extracted from derived temporal data by the CNN layer that reflect changes in the user's real-time state. They can capture real-time information such as short-term fluctuations in health indicators and instantaneous switching of behavioral scenarios, supporting real-time adaptation of medication plans. Global dependencies are correlations that reflect long-term patterns learned from derived temporal data by the LSTM layer. They cover long-term trends of health indicators, cyclical patterns of behavioral scenarios, and temporal correlations between drug metabolism and health indicators, providing a basis for long-term adaptation of medication plans.

[0052] Accordingly, for embodiments of this disclosure, the steps may include: extracting local features of derived time-series data through a CNN layer, whereby local features are used to capture the instantaneous changes in the user's real-time physiological state and behavioral scenarios, including short-term fluctuations in health indicators and instantaneous switching features of behavioral scenarios; learning global dependencies of derived time-series data through an LSTM layer, whereby global dependencies are used to mine the long-term correlation between the user's physiological patterns and drug metabolism, including long-term trends in health indicators, periodic patterns of behavioral scenarios, and the temporal correlation between drug metabolism and health indicators; predicting the efficacy suitability and dosage safety under different medication timings based on local features and global dependencies, and selecting target medication timing windows and target dose recommendation values ​​that are suitable for the current health state and current behavioral scenarios; adjusting the medication timing windows and dose recommendation values ​​in the basic medication information to the target medication timing windows and target dose recommendation values ​​to form a preliminary medication recommendation scheme.

[0053] Based on derived time-series data, a time-series deep learning model with a hybrid CNN-LSTM architecture is used to adjust basic medication information. This approach can accurately capture users' real-time physiological and behavioral changes through the CNN layer, ensuring the real-time adaptability of medication plans. At the same time, the LSTM layer can explore the correlation between long-term physiological patterns and drug metabolism, ensuring the long-term rationality of medication plans. This dual precision adjustment of real-time state adaptation and long-term pattern alignment can effectively solve the problem that traditional medication plans cannot take into account both real-time changes and long-term patterns. It can significantly improve the personalization and scientific accuracy of the initial medication recommendation plan, providing a high-quality initial basis for subsequent risk optimization and professional review.

[0054] Step 230: Conduct a comprehensive risk assessment of the preliminary medication recommendation plan to determine the type and level of medication risk.

[0055] Among them, the full-dimensional risk assessment refers to the systematic risk identification and judgment process carried out around the initial medication recommendation plan from multiple dimensions such as the user's individual health status, behavioral scenario adaptability, and drug attribute matching degree. It aims to comprehensively investigate potential safety hazards during medication. Medication risk type is a category of various safety hazards that may occur during medication based on the full-dimensional assessment results, such as health indicator adaptability risk, behavioral scenario conflict risk, and drug metabolism contraindication risk. Risk level is a severity level of the identified medication risk type based on indicators such as the probability of occurrence and the degree of harm of the medication risk. It can usually be divided into three levels: high, medium, and low, providing a basis for decision-making for subsequent risk optimization.

[0056] In this embodiment of the disclosure, the generated preliminary medication recommendation plan can be compared and analyzed in multiple dimensions with the user's core health indicators, behavioral scenario time sequence characteristics, and key drug attributes. Combined with preset risk assessment rules and pharmaceutical safety standards, various potential medication safety hazards in the plan can be comprehensively identified, the corresponding medication risk types can be clarified, and then the various risks can be classified and judged according to indicators such as the probability of occurrence of the risk and the potential harm to the user's health, so as to finally determine the medication risk type and the corresponding risk level.

[0057] Conducting a comprehensive risk assessment of the initial medication recommendation plan and determining the type and level of medication risks can comprehensively identify potential safety hazards in the medication plan from multiple dimensions, accurately locate the specific category and severity of risks, provide a clear basis for subsequent targeted risk optimization and treatment, effectively avoid medication safety issues caused by loopholes in the plan, improve the safety and reliability of the medication plan, and provide advance protection for users' medication health.

[0058] Step 240: Optimize the initial medication recommendation plan based on the type and level of medication risk.

[0059] Risk optimization refers to the process of revising and improving the initial medication recommendation plan based on the medication risk type and risk level determined by the assessment, using differentiated adjustment strategies. The core purpose is to eliminate or reduce potential medication safety hazards.

[0060] In this embodiment of the disclosure, based on the types and corresponding risk levels of medication risks obtained from the comprehensive risk assessment, a preset differentiated risk handling strategy can be matched. The initial medication recommendation plan can be adjusted in a targeted manner for different types and levels of risks. For example, for high-risk contraindication risks, medication recommendations can be directly removed; for medium-risk suitability risks, the timing or dosage of medication can be adjusted; and for low-risk warning risks, safe medication instructions can be added. Finally, the risk optimization of the initial medication recommendation plan is completed, forming a safer and more reasonable optimized plan.

[0061] Targeted risk optimization based on medication risk type and risk level can accurately target various safety hazards in the initial medication recommendation plan. Through differentiated strategies, it can achieve precise control and elimination of risks, effectively reduce the possibility of irrational medication, and significantly improve the safety and reliability of medication plan.

[0062] Step 250: After risk optimization of the preliminary medication recommendation plan, submit it to the doctor's terminal, and after receiving the confirmation instruction from the doctor's terminal, form the final medication recommendation plan.

[0063] In this embodiment of the disclosure, the initial medication recommendation plan can be optimized for risk first, and then the optimized plan can be pushed to the doctor's terminal for medical staff to conduct a comprehensive review in conjunction with the user's health indicators, behavioral characteristics and clinical diagnosis and treatment guidelines. After the medical staff approves the evaluation, they can send a confirmation instruction to the system through the doctor's terminal. After receiving the instruction, the system will officially determine the optimized initial medication recommendation plan as the final medication recommendation plan.

[0064] Step 260: Based on the final medication recommendation plan, link the smart device to perform medication management operations and simultaneously present the final medication recommendation plan to the user terminal.

[0065] Medication management operations may include, but are not limited to, medication dispensing monitoring operations and drug lifecycle management operations. Medication dispensing monitoring operations refer to the real-time monitoring and verification of the user's medication dispensing process to ensure that the medication dispensing behavior complies with the requirements of the plan, such as medication time period and drug dosage. Drug lifecycle management operations refer to the management operations of drugs from warehousing to the end of use, including monitoring the remaining dosage of drugs and verifying the expiration date, to ensure medication safety.

[0066] In this embodiment of the disclosure, the final medication recommendation plan can be used as the core basis to generate and issue collaborative management instructions adapted to various related smart devices, drive the smart devices to perform corresponding medication management operations such as medication dispensing action monitoring and drug life cycle management, and simultaneously push the complete content of the final medication recommendation plan to the user terminal, so that the user can view key information such as medication timing and dosage at any time, so as to realize the simultaneous implementation of medication plan execution and information presentation.

[0067] As one possible implementation, when medication management operations include medication retrieval monitoring operations, step 260 of the embodiment may include: parsing the medication period, target drug, and drug dosage in the final medication recommendation plan, generating a medication retrieval monitoring instruction for the smart medicine box, which instructs the smart medicine box to monitor in real time whether the weight change of the target drug compartment during the user's medication retrieval process conforms to the drug dosage, and to verify whether the user's medication retrieval behavior is completed within the medication period by combining the medication retrieval action detected by the touch sensor; if the monitored weight change of the target drug compartment does not conform to the drug dosage, and / or no user medication retrieval behavior is detected within the medication period, then a medication retrieval reminder message is automatically sent to the user through at least one of the smart medicine box, wearable health device, and user terminal.

[0068] The system provides a medication monitoring instruction, generated after parsing the final medication recommendation plan. This instruction guides the smart medicine box in performing medication monitoring tasks and includes core execution parameters such as medication time period, target medication, and dosage. A weight sensor, built into the smart medicine box, detects real-time weight changes in the medication compartments to determine if the user's dosage meets the plan's requirements. A touch sensor, also built into the smart medicine box, detects the user's actions of opening the compartments and removing medication, verifying that the action was completed within the specified time period based on time information. Medication reminders are notifications sent to the user via the smart device when abnormal dosage or failure to remove medication on time is detected, urging the user to use medication correctly.

[0069] In another possible implementation of this disclosure, when the medication management operation includes a drug lifecycle management operation, step 260 may include: parsing the target drug identifier, single-dose weight, and preset expiration date threshold in the final medication recommendation plan, generating a lifecycle management instruction for the smart medicine box. The lifecycle management instruction is used to instruct the smart medicine box to calculate the remaining dose of the target drug in real time based on the target drug identifier and single-dose weight using a weight sensor, and to verify whether the actual expiration date of the target drug is within the preset expiration date threshold. If the remaining dose of the target drug is lower than the set threshold, and / or the actual expiration date of the target drug exceeds the preset expiration date threshold, a drug replacement reminder message is automatically sent to the user through at least one of the smart medicine box, wearable health device, and user terminal.

[0070] Among them, the target drug identifier is a unique identifier used to distinguish different drugs, which can accurately locate specific drugs in the smart medicine box that require lifecycle management; the single dose weight refers to the standard weight of the drug required for a single dose, which is the core basis for the smart medicine box to calculate the number of times the remaining drug can be used and the remaining dose; the preset expiration date threshold is a pre-set warning threshold for drug expiration, and the system will trigger an alert mechanism when the actual expiration date of the drug approaches or exceeds this threshold; the lifecycle management instruction is generated by the system after parsing the relevant parameters of the final medication recommendation plan, and is used to guide the smart medicine box to perform tasks such as calculating the remaining dose of the drug and verifying the expiration date; the drug replacement reminder message is a reminder message pushed by the system to the user when the remaining dose of the drug is lower than the set threshold or the expiration date exceeds the preset threshold, which is used to remind the user to replace the drug in time.

[0071] In summary, the technical solution in this application collects real-time health indicators, daily behaviors, and static drug information from multiple dimensions. After structured processing, a three-dimensional dynamic model is constructed to generate a user medication profile. This is combined with a medication knowledge graph to match basic medication information. Then, a CNN-LSTM hybrid architecture temporal deep learning model is used to extract local features and global dependencies from the derived temporal data, enabling precise adjustment of medication timing and dosage to generate a preliminary medication plan. Subsequently, a comprehensive risk assessment and optimization, and confirmation by the doctor's terminal, form the final plan. Finally, intelligent devices are linked to monitor the medication dispensing action and the drug's life cycle. By integrating operations such as cycle management and presenting solutions simultaneously, this system effectively addresses the pain points of existing medication management solutions, such as a lack of dynamic perception and comprehensive analysis capabilities, poor personalization, rigid reminders, and insufficient risk prediction. It not only ensures that medication recommendations are accurately adapted to the user's real-time health status and behavioral scenarios, significantly improving the personalization and targeting of medication advice, but also strengthens medication safety through dual safeguards of risk control and doctor review. Furthermore, by leveraging multi-device linkage to achieve intelligent monitoring of the entire medication process, it avoids issues of scenario disconnect and inappropriate medication use, fully meeting users' core needs for dynamic, safe, and personalized medication management.

[0072] Furthermore, as Figure 1 and Figure 2 The specific implementation of the method shown in this embodiment provides a medication management device based on an AI decision engine, such as... Figure 3 As shown, the device includes: a data acquisition module 31, a data generation module 32, a data submission module 33, and an execution module 34.

[0073] The data acquisition module 31 can be used to collect multi-dimensional user data, including real-time health indicator data, daily behavior data, and static drug information data. The generation module 32 can be used to dynamically fuse and intelligently analyze multi-dimensional user data through an AI decision engine to generate a preliminary medication recommendation plan. The preliminary medication recommendation plan includes a target medication timing window and a target dose recommendation value that are adapted to the user's real-time status. The submission module 33 can be used to submit the preliminary medication recommendation plan to the doctor's terminal after risk optimization, and form the final medication recommendation plan after receiving the confirmation instruction from the doctor's terminal. The execution module 34 can be used to link smart devices to perform medication management operations based on the final medication recommendation plan, and simultaneously present the final medication recommendation plan to the user terminal.

[0074] In some embodiments of this application, the data acquisition module 31 can be specifically used to establish an encrypted communication link with the wearable health device through the device SDK after obtaining user authorization, and synchronize the user's real-time health indicator data at a preset frequency. The real-time health indicator data includes at least heart rate, blood pressure, blood sugar, sleep quality, and activity level. It can also collect the user's daily behavior data through smart home devices, including voice interaction data and motion interaction data. Furthermore, it can collect static drug information data by scanning the drug instructions or receiving manual input from the user. The static drug information data includes at least the drug name, drug specifications, drug dosage, drug expiration date, metabolic cycle, and contraindications.

[0075] In some embodiments of this application, the generation module 32 can be specifically used to perform structured processing on multi-dimensional data, extract core parameters of health indicators, temporal features of behavioral scenarios, and key attributes of drugs; construct a three-dimensional dynamic model based on the core parameters of health indicators, temporal features of behavioral scenarios, and key attributes of drugs, and dynamically generate a user medication profile based on the three-dimensional dynamic model; call a preset medication knowledge graph, and match basic medication information that matches the user medication profile through multi-hop inference technology; extract corresponding derived temporal data from the three dimensions of the three-dimensional dynamic model, including health indicator fluctuation trends, temporal features of behavioral scenarios, and drug metabolism cycle related data; based on the derived temporal data, use a temporal deep learning model to adjust the medication timing window and dosage recommendation value of the basic medication information to form a preliminary medication recommendation plan.

[0076] In some embodiments of this application, the temporal deep learning model adopts a CNN-LSTM hybrid architecture, including CNN layers and LSTM layers. When adjusting the medication timing window and dosage recommendation suggestions of basic medication information based on derived temporal data using the temporal deep learning model to form a preliminary medication recommendation scheme, the generation module 32 can be used to extract local features of the derived temporal data through the CNN layer. The local features are used to capture the instantaneous changes in the user's real-time physiological state and behavioral scenarios. The local features include short-term fluctuations in health indicators and instantaneous switching features of behavioral scenarios. The global dependencies of the derived temporal data are learned through the LSTM layer. The global dependencies are used to mine the long-term correlation between the user's physiological patterns and drug metabolism. The global dependencies include long-term trends of health indicators, periodic patterns of behavioral scenarios, and temporal correlation between drug metabolism and health indicators. Based on the local features and global dependencies, the efficacy suitability and dosage safety under different medication timing are predicted, and the target medication timing window and target dosage recommendation value that are suitable for the current health state and current behavioral scenario are selected. The medication timing window and dosage recommendation value in the basic medication information are adjusted to the target medication timing window and target dosage recommendation value to form a preliminary medication recommendation scheme.

[0077] In some embodiments of this application, such as Figure 4 As shown, the device also includes: a processing module 35; When performing risk optimization on the initial medication recommendation plan, the processing module 35 can be used to conduct a full-dimensional risk assessment of the initial medication recommendation plan, determine the type and level of medication risk, and perform risk optimization on the initial medication recommendation plan based on the type and level of medication risk.

[0078] In some embodiments of this application, the medication management operation includes a medication retrieval monitoring operation. Execution module 34 is specifically used to parse the medication period, target drug, and dosage in the final medication recommendation plan, and generate a medication retrieval monitoring instruction for the smart medicine box. The medication retrieval monitoring instruction is used to instruct the smart medicine box to monitor in real time whether the weight change of the target drug compartment during the user's medication retrieval process conforms to the dosage, and to verify whether the user's medication retrieval behavior is completed within the medication period by combining the medication retrieval action detected by the touch sensor. If the monitored weight change of the target drug compartment does not conform to the dosage, and / or no user medication retrieval behavior is detected within the medication period, then a medication retrieval reminder message is automatically sent to the user through at least one of the smart medicine box, wearable health device, and user terminal.

[0079] In some embodiments of this application, the medication management operation includes a drug lifecycle management operation. Execution module 34 is specifically used to parse the target drug identifier, single-dose weight, and preset expiration date threshold in the final medication recommendation plan, and generate a lifecycle management instruction for the smart medicine box. The lifecycle management instruction instructs the smart medicine box to calculate the remaining dose of the target drug in real time using a weight sensor based on the target drug identifier and single-dose weight, and to verify whether the actual expiration date of the target drug is within the preset expiration date threshold. If the remaining dose of the target drug is lower than the set threshold, and / or the actual expiration date of the target drug exceeds the preset expiration date threshold, a drug replacement reminder message is automatically sent to the user through at least one of the smart medicine box, wearable health device, and user terminal.

[0080] It should be noted that other corresponding descriptions of the functional units involved in the medication management device based on an AI decision engine provided in this embodiment can be found in [reference needed]. Figure 1 and Figure 2 The corresponding descriptions in [the document] will not be repeated here.

[0081] Based on the above, Figure 1 and Figure 2 Accordingly, this embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the above-described method. Figure 1 and Figure 2 The example shown is a medication management method based on an AI decision engine.

[0082] Based on this understanding, the technical solution of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as CD-ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause an electronic device (such as personal computer, server, or network device, etc.) to execute the methods of various implementation scenarios of this application.

[0083] Based on the above, Figure 1 and Figure 2 The method shown, and Figure 3 , 4 To achieve the above objectives, the present application also provides an electronic device, specifically a personal computer, tablet computer, server, or other network device, as shown in the virtual device embodiment. This device includes a storage medium and a processor; the storage medium stores a computer program; the processor executes the computer program to achieve the above-described objectives. Figure 1 and Figure 2 The example shown is a medication management method based on an AI decision engine.

[0084] Optionally, the aforementioned physical devices may also include a user interface, a network interface, a camera, radio frequency (RF) circuitry, sensors, audio circuitry, a Wi-Fi module, etc. The user interface may include a display screen, input units such as a keyboard, etc., and optional user interfaces may also include USB interfaces, card reader interfaces, etc. The network interface may optionally include standard wired interfaces, wireless interfaces (such as Wi-Fi interfaces), etc.

[0085] Those skilled in the art will understand that the physical device structure provided in this embodiment does not constitute a limitation on the physical device, and may include more or fewer components, or combine certain components, or have different component arrangements.

[0086] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the aforementioned physical device, supporting the operation of information processing programs and other software and / or programs. The network communication module is used to enable communication between the various components within the storage medium, as well as communication with other hardware and software in the information processing physical device.

[0087] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platform, or it can be implemented by hardware.

[0088] This invention collects multi-dimensional data on users' real-time health indicators, daily behaviors, and static drug information. Using an AI decision engine, it dynamically fuses and intelligently analyzes this data to generate a preliminary medication plan that adapts to the user's real-time status, including a medication timing window and recommended dosage. This plan is then optimized for risk and confirmed by the doctor's terminal to form a final plan. Finally, it is linked to a smart device to execute medication management operations and present the plan to the user's terminal. This effectively solves the problem of existing medication management plans lacking dynamic perception and comprehensive analysis capabilities. It breaks through the limitations of traditional fixed reminders and static medication suggestions, allowing medication recommendations to accurately match the user's real-time status, improving the personalization and targeting of medication advice. Simultaneously, through risk optimization and doctor confirmation, it reduces medication risks and avoids situations where reminders are out of sync with the user's actual scenario. Ultimately, it meets users' core needs for dynamic, safe, and personalized medication management.

[0089] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of a preferred embodiment, and the modules or processes shown in the drawings are not necessarily essential for implementing this application. Those skilled in the art will understand that the modules in the apparatus of the embodiment can be distributed within the apparatus of the embodiment as described, or can be modified to be located in one or more apparatuses different from this embodiment. The modules of the above-described embodiment can be combined into one module, or further divided into multiple sub-modules.

[0090] The serial numbers in this application are for descriptive purposes only and do not represent the superiority or inferiority of any particular implementation scenario. The above disclosures are merely a few specific implementation scenarios of this application; however, this application is not limited thereto, and any variations conceived by those skilled in the art should fall within the protection scope of this application.

Claims

1. A medication management method based on an AI decision engine, characterized in that, The method includes: Collect multi-dimensional user data, including real-time health indicator data, daily behavior data, and static drug information data; The AI ​​decision engine dynamically integrates and intelligently analyzes the user's multi-dimensional data to generate a preliminary medication recommendation plan. The preliminary medication recommendation plan includes a target medication timing window and a target dose recommendation value that are adapted to the user's real-time status. After risk optimization processing, the preliminary medication recommendation plan is submitted to the doctor's terminal, and the final medication recommendation plan is formed upon receiving confirmation from the doctor's terminal. Based on the final medication recommendation plan, the intelligent device is linked to perform medication management operations, and the final medication recommendation plan is presented to the user terminal simultaneously.

2. The method according to claim 1, characterized in that, The collection of multi-dimensional user data includes: After obtaining user authorization, an encrypted communication link is established with the wearable health device through the device SDK, and the user's real-time health indicator data is synchronized at a preset frequency. The real-time health indicator data includes at least heart rate, blood pressure, blood sugar, sleep quality and activity level. The user's daily behavior data is collected through smart home devices, including voice interaction data and motion interaction data. Static drug information data is collected by scanning the drug instructions or by receiving manual input from the user. The static drug information data includes at least the drug name, drug specifications, drug dosage, drug expiration date, metabolic cycle, and contraindications.

3. The method according to claim 1, characterized in that, The process of dynamically fusing and intelligently analyzing the user's multi-dimensional data using an AI decision engine to generate a preliminary medication recommendation plan includes: The multi-dimensional data is structured to extract core parameters of health indicators, temporal features of behavioral scenarios, and key attributes of drugs. A three-dimensional dynamic model is constructed based on the core parameters of the health indicators, the temporal characteristics of the behavioral scenarios, and the key attributes of the drugs. A user medication profile is dynamically generated based on the three-dimensional dynamic model. The system invokes a pre-defined medication knowledge graph and uses multi-hop reasoning technology to match basic medication information that is compatible with the user's medication profile. The corresponding derived time-series data are extracted from the three dimensions of the three-dimensional dynamic model. The derived time-series data includes the fluctuation trend of health indicators, the time-series characteristics of behavioral scenarios, and the data related to the drug metabolism cycle. Based on the derived time-series data, a time-series deep learning model is used to adjust the medication timing window and recommended dosage values ​​of the basic medication information to form a preliminary medication recommendation plan.

4. The method according to claim 3, characterized in that, The temporal deep learning model adopts a CNN-LSTM hybrid architecture, including CNN layers and LSTM layers; Based on the derived time-series data, a time-series deep learning model is used to adjust the medication timing window and dosage recommendations for the basic medication information, forming a preliminary medication recommendation scheme, including: The CNN layer extracts local features from the derived time-series data. These local features are used to capture the real-time changes in the user's physiological state and behavioral scenarios. The local features include short-term fluctuations in health indicators and instantaneous switching features of behavioral scenarios. The global dependencies of the derived time-series data are learned through the LSTM layer. These global dependencies are used to mine the long-term correlation between user physiological patterns and drug metabolism. The global dependencies include long-term trends of health indicators, periodic patterns of behavioral scenarios, and the time-series correlation between drug metabolism and health indicators. Based on the local features and the global dependencies, the efficacy and dosage safety of medication at different times are predicted, and the target medication timing window and target dose recommendation value that are suitable for the current health status and current behavior scenario are selected. The medication timing window and recommended dosage value in the basic medication information are adjusted to the target medication timing window and the target recommended dosage value to form a preliminary medication recommendation plan.

5. The method according to claim 1, characterized in that, Risk optimization of the preliminary medication recommendation plan includes: A comprehensive risk assessment was conducted on the preliminary medication recommendation to determine the type and level of medication risk. The initial medication recommendation plan is optimized based on the medication risk type and the risk level.

6. The method according to claim 1, characterized in that, The medication management operation includes medication dispensing monitoring operations, and the medication management operation based on the final medication recommendation plan, which is performed by linking intelligent devices, includes: The medication time period, target drug, and drug dosage in the final medication recommendation plan are analyzed to generate a medication retrieval monitoring instruction for the smart medicine box. The medication retrieval monitoring instruction is used to instruct the smart medicine box to monitor in real time whether the weight change of the target drug compartment is consistent with the drug dosage during the user's medication retrieval process through a weight sensor, and to verify whether the user's medication retrieval behavior is completed within the medication time period by combining the medication retrieval action detected by the touch sensor. If the change in the weight of the target medicine compartment does not conform to the dosage of the medicine, and / or if no medication retrieval behavior of the user is detected during the medication period, a medication retrieval reminder message will be automatically sent to the user through at least one of the following: smart medicine box, wearable health device, and user terminal.

7. The method according to claim 1, characterized in that, The medication management operation includes drug lifecycle management operations. The step of linking intelligent devices to execute medication management operations based on the final medication recommendation plan includes: The system analyzes the target drug identifier, single dose weight, and preset expiration date threshold in the final medication recommendation scheme to generate a lifecycle management instruction for the smart medicine box. The lifecycle management instruction is used to instruct the smart medicine box to calculate the remaining dose of the target drug in real time based on the target drug identifier and the single dose weight through a weight sensor, and to verify whether the actual expiration date of the target drug is within the preset expiration date threshold. If the remaining dose of the target drug is lower than a set threshold, and / or the actual expiration date of the target drug exceeds the preset expiration date threshold, a drug replacement reminder will be automatically sent to the user via at least one of a smart medicine box, a wearable health device, and a user terminal.

8. A medication management device based on an AI decision engine, characterized in that, The device includes: The data acquisition module is used to collect multi-dimensional user data, including real-time health indicator data, daily behavior data, and static drug information data. The generation module is used to dynamically fuse and intelligently analyze the user's multi-dimensional data through an AI decision engine to generate a preliminary medication recommendation plan. The preliminary medication recommendation plan includes a target medication timing window and a target dose recommendation value that are adapted to the user's real-time status. The submission module is used to perform risk optimization processing on the preliminary medication recommendation plan and submit it to the doctor's terminal. After receiving the confirmation instruction from the doctor's terminal, the final medication recommendation plan is formed. The execution module is used to coordinate with the smart device to perform medication management operations based on the final medication recommendation plan, and simultaneously present the final medication recommendation plan to the user terminal.

9. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

10. An electronic device comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.

Images