Medication consultation interaction system based on artificial intelligence and electronic medical record management method

Abstract

The invention relates to a medication consultation interaction system based on artificial intelligence and an electronic medical record management method, and belongs to the technical field of intelligent medical treatment. The medication consultation interaction system comprises an electronic medical record management module, an intelligent medication decision support engine, an interaction feedback module and a continuous learning module; the intelligent medication decision support engine is in communication connection with the electronic medical record management module and the interactive feedback module, and the continuous learning module is connected with the intelligent medication decision support engine and the interactive feedback module. In the invention, by integrating the natural language processing unit, the system can automatically analyze and standardize the unstructured medical text and convert the unstructured medical text into high-quality structured data, and by deeply integrating the intelligent engine and the electronic medical record, the conversion from passive warning to active decision support is realized; the medication safety, rationality and diagnosis and treatment efficiency are effectively improved, and the system performance is optimized through a continuous learning mechanism.

Description

Technical Field

[0001] This invention relates to the field of smart healthcare technology, and in particular to an artificial intelligence-based medication consultation and interaction system and electronic medical record management method. Background Technology

[0002] Medication consultation and interaction system is an intelligent software tool specifically designed to assist medical staff in making drug-related decisions. It can provide real-time online information query and decision support. Electronic medical records refer to systems that use computer and information technology to create, store, manage, transmit and access patient medical records. They are a digital and intelligent upgrade of paper medical records.

[0003] Electronic medical records (EMRs) serve as the data foundation, providing the medication consultation system with all the patient information needed for review: diagnosis, allergy history, test results, and current medication list. The medication consultation system utilizes the data provided by EMRs, runs a built-in knowledge base and rule base, performs intelligent calculations and judgments, and ultimately feeds back the review results to the doctor.

[0004] Electronic medical record (EMR) systems have become a core infrastructure of modern hospitals, used to store patients' medical information. However, traditional EMRs, as information recorders and repositories, have limited inherent decision support capabilities. Existing medication consultation systems are typically based on simple rule bases (e.g., IF drug A AND drug B THEN warning), which have the following drawbacks: 1. Low accuracy of alerts: The rule engine cannot combine the specific situation of the patient, resulting in a large number of unnecessary and low-risk alerts, which will cause doctors to become fatigued and ignore important warnings. 2. Lack of in-depth analysis: It cannot handle complex medical logic. For example, for cancer patients, the risk of interaction between two drugs may be far less than their therapeutic benefits, but the system will still mechanically pop up a warning. 3. Passive rather than proactive: The system only issues warnings after medication is prescribed, and cannot provide doctors with proactive, personalized medication recommendations before prescription; 4. Inability to continuously evolve: The rule base requires manual maintenance and updates, making it difficult to adapt to the rapid iteration of medical knowledge; Therefore, there is an urgent need for a solution that can deeply understand patients' conditions, provide precise and intelligent medication decision support, and can self-evolve and optimize. To this end, we propose an artificial intelligence-based medication consultation and interaction system and electronic medical record management method. By deeply integrating an intelligent engine and electronic medical records, it realizes the transformation from passive warning to proactive decision support, effectively improving medication safety, rationality and treatment efficiency, while optimizing system performance through a continuous learning mechanism. Summary of the Invention

[0005] This invention provides an artificial intelligence-based medication consultation interaction system and electronic medical record management method, which solves the problems of data silos and low utilization rate, one-sidedness and high false alarm rate in existing electronic medical record systems and medication consultation tools, lack of depth and operability in decision support, static nature of the system and high maintenance cost.

[0006] The solution to the above-mentioned technical problems of the present invention is as follows: a medication consultation interaction system and electronic medical record management method based on artificial intelligence, including an electronic medical record management module, an intelligent medication decision support engine, an interactive feedback module, and a continuous learning module. The intelligent medication decision support engine is communicatively connected to the electronic medical record management module and the interactive feedback module, and the continuous learning module is connected to the intelligent medication decision support engine and the interactive feedback module. The electronic medical record management method includes the following steps: S1 receives and stores the patient's latest medical records through the electronic medical record management module; S2, when a doctor issues or modifies a medical order, the intelligent medication decision support engine is automatically triggered, and real-time calls and analysis of all relevant data of the patient; S3, the intelligent medication decision support engine matches patient data with the knowledge graph and inputs it into the multimodal risk assessment model for comprehensive calculation, generating a review report that includes risk level, specific cause, source of evidence and optimization suggestions; S4 pushes the review report to the physician through the interactive feedback module and records the physician's acceptance, rejection or modification actions; S5, the continuous learning module, uses physician feedback and patient records of subsequent treatment effects and adverse reactions to reinforce the risk assessment model and update the model.

[0007] Based on the above technical solution, the present invention can be further improved as follows.

[0008] Furthermore, the electronic medical record management module is used to centrally store and manage patients' structured and unstructured medical data. This module integrates a data preprocessing unit, which uses natural language processing technology to perform entity recognition and relation extraction on unstructured medical record text and image reports, transforming them into structured data that can be computed by the intelligent medication decision support engine. The integrated data preprocessing unit is specifically responsible for processing unstructured text data (such as physician-written medical records and the conclusions of imaging descriptions) using natural language processing technology. Through the two key technologies of entity recognition and relation extraction, the messy text is transformed into well-organized, semantically labeled structured data, achieving high data availability.

[0009] Furthermore, the intelligent medication decision support engine is used to read and parse patient data in real time. It has a built-in medical knowledge graph and machine learning model, which can identify medication risks and optimize treatment plans, and generate medication review results. The intelligent medication decision support engine includes knowledge graph units and multimodal risk assessment models, realizing a leap from "single rule judgment" to "multi-dimensional comprehensive judgment". Traditional systems can only perform simple "if-A-then-B" rule judgments, while this system combines the association query of the knowledge graph and the predictive ability of the machine learning model to simulate the thinking process of senior pharmacists and clinical experts, and conduct deeper and more individualized risk insights, which greatly improves the depth and breadth of the review.

[0010] Furthermore, the knowledge graph unit stores multidimensional relationships between drugs, diseases, genes, and allergens. The multimodal risk assessment model integrates patients' real-time physiological parameters, test indicators, genetic information, and medical history data to predict individualized risks of adverse drug reactions, insufficient efficacy, and drug interactions. The knowledge graph unit specifically stores the associations between drugs and diseases, genes, allergens, etc., forming a reasonable semantic network. The multimodal risk assessment model is specifically responsible for integrating patients' real-time physiological parameters, test indicators, and other multimodal data to perform individualized risk prediction. By introducing precision medicine dimensions such as genetic information, the system can warn of risks that traditional methods cannot detect (such as drug metabolism abnormalities caused by specific genotypes). By integrating real-time physiological parameters, the system can dynamically assess the risk in the patient's current state, achieving truly "individualized" medication safety monitoring and significantly reducing missed and false alarms.

[0011] Furthermore, the multimodal risk assessment model is an ensemble learning model. Its input features include the embedding vectors of real-time patient data and knowledge graph query results, and its output is risk probability and confidence level. This improves the prediction accuracy and the credibility of the results. Ensemble learning models (such as XGBoost and Random Forest) usually have higher prediction accuracy and stability than single models. The output of "risk probability and confidence level" rather than a simple "yes / no" judgment provides physicians with richer decision-making reference information. The level of confidence can help physicians judge the firmness of the system prompts, thereby rationally allocating attention and prioritizing the handling of high-confidence high-risk warnings.

[0012] Furthermore, the interactive feedback module is used to intuitively display the review results to the user and receive user feedback. The review results provided by the interactive feedback module include high-risk, medium-risk, and advisory levels, and provide evidence-based explanations and alternative medication recommendations for each warning. This effectively alleviates "warning fatigue" and improves the efficiency of doctor-patient communication. Tiered warnings enable physicians to quickly identify truly urgent issues and avoid being overwhelmed by a large number of low-value warnings. Providing explanations and alternative solutions not only enhances the credibility of the system but also transforms the system's role from "error finder" to "decision-making assistant," directly providing physicians with a path to solve problems and greatly improving work efficiency and adherence to prescription modifications.

[0013] Furthermore, the continuous learning module is used to iteratively optimize the machine learning model based on user feedback and subsequent diagnosis and treatment results. Traditional rule systems require manual maintenance and updates, which are time-consuming, labor-intensive, and outdated. This system, through continuous learning, can automatically adapt to the medication habits of local hospitals, the response characteristics of specific populations, and the latest clinical practices, forming a virtuous cycle of becoming smarter and more accurate with use, thus ensuring the long-term vitality and effectiveness of the system.

[0014] Furthermore, in step S2, the parsing process includes: using a NER model to extract disease entities from the free text diagnosis written by the physician and mapping them to a standardized medical dictionary, thereby achieving a high-precision understanding of the natural language of clinicians. This ensures that no matter how the physician writes the diagnosis, the system can accurately capture its core meaning, providing accurate input for subsequent knowledge graph matching and risk analysis, and fundamentally reducing misjudgments caused by information extraction errors.

[0015] Furthermore, in step S5, the strategy for updating the model includes: if a physician repeatedly ignores a certain type of prompt that the system judges to be low-risk and ultimately does not cause adverse consequences, the system automatically reduces the level or frequency of that type of prompt. This achieves "localization" and "personalization" of the system's behavior, allowing the system to learn and respect the collective wisdom and experience of the hospital's clinical experts. If a certain type of warning is generally considered by experts to be negligible and indeed harmless in the current clinical environment, the system will automatically adjust its sensitivity, thereby reducing unnecessary interference with the hospital's doctors and making the system prompts more in line with actual clinical needs. This is an intelligent behavior that a static system cannot achieve.

[0016] The beneficial effects of this invention are as follows: This invention provides an artificial intelligence-based medication consultation interaction system and electronic medical record management method, which have the following advantages: 1. It solves the problems of data silos and low utilization rates in existing electronic medical record systems and medication consultation tools. Traditional electronic medical record systems can only store data and cannot effectively process and utilize massive amounts of unstructured text data (such as medical records and image reports). This results in a large amount of medical data containing key information being in a "computable" state, unable to provide in-depth support for clinical decision-making. By integrating a natural language processing (NLP) unit, the system can automatically parse and standardize unstructured medical text, transforming it into high-quality structured data, completely releasing the value of medical big data, and laying a solid data foundation for advanced artificial intelligence applications. 2. This system addresses the issues of biased review and high false alarm rates in existing electronic medical record systems and medication consultation tools. Traditional medication alert systems based on simple rule bases lack a deep understanding of the patient's overall condition (such as genetic information and real-time physiological indicators), and can only make isolated, mechanical rule judgments, generating a large number of low-value, non-personalized alerts. This leads to "alert fatigue" among clinicians, who may overlook truly high-risk warnings. By deeply integrating medical knowledge graphs with multimodal machine learning models, the system can simulate expert thinking and conduct multi-dimensional, individualized comprehensive risk assessments. This not only significantly improves the accuracy of risk identification (reducing missed alarms), but more importantly, it significantly improves the precision of alerts (reducing false alarms), effectively alleviating alert fatigue. 3. It solves the problem of insufficient depth and operability in decision support in existing electronic medical record systems and medication consultation tools. Existing systems can usually only point out "problems" but cannot explain "why there are problems" or "what should be done". They lack evidence-based explanations and feasible alternatives, and cannot effectively assist physicians in optimizing clinical decisions. The review report provided by the system includes risk level, detailed reasons, evidence sources and specific optimization suggestions, transforming the system's role from a simple "error finder" to a true "clinical decision support provider". This greatly improves physicians' work efficiency and compliance with system recommendations. 4. This system addresses the static nature and high maintenance costs of existing electronic medical record systems and medication consultation tools. Traditional systems require manual updates and maintenance of their rule bases, failing to adapt automatically to the rapid iteration of medical knowledge, differences in medication habits across hospitals, and the characteristics of specific patient groups. This results in high maintenance costs and delays, making it difficult to consistently guarantee accuracy and effectiveness. The system innovatively introduces a feedback-based continuous learning mechanism. By collecting practical feedback from physicians and real patient outcomes, the system automatically adjusts its risk assessment model and alert strategies, making them increasingly aligned with local clinical practice and patient characteristics. This creates a virtuous cycle of "becoming more accurate with use," ensuring the system's long-term viability and adaptability, significantly reducing subsequent maintenance costs. By integrating multimodal data such as genetic information and real-time physiological parameters, the system achieves true "one-person-one-policy" medication safety monitoring, providing early warnings of individualized risks that traditional methods cannot detect (such as abnormalities caused by drug metabolism gene mutations), significantly improving medical quality and patient safety.

[0017] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it according to the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Specific embodiments of the present invention are given in detail below with reference to the accompanying drawings. Attached Figure Description

[0018] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 A system flowchart of an artificial intelligence-based medication consultation interaction system and electronic medical record management method provided in an embodiment of the present invention; Figure 2 This is a flowchart illustrating an artificial intelligence-based medication consultation interaction system and electronic medical record management method, as provided in an embodiment of the present invention. Detailed Implementation

[0019] The following is in conjunction with the appendix Figure 1-2 The principles and features of the present invention are described below. The examples given are for illustrative purposes only and are not intended to limit the scope of the invention. The invention is described more specifically in the following paragraphs by way of example with reference to the accompanying drawings. The advantages and features of the invention will become clearer from the following description. It should be noted that the drawings are in a very simplified form and use non-precise proportions, and are only used to facilitate and clarify the illustration of the embodiments of the invention.

[0020] It should be noted that when a component is described as "fixed to" another component, it can be directly on the other component or may have a component in between. When a component is considered "connected to" another component, it can be directly connected to the other component or may have a component in between. When a component is considered "set on" another component, it can be directly set on the other component or may have a component in between. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0022] like Figure 1-2 As shown, the present invention provides an artificial intelligence-based medication consultation interaction system and electronic medical record management method, including an electronic medical record management module, an intelligent medication decision support engine, an interactive feedback module, and a continuous learning module. The intelligent medication decision support engine is communicatively connected to the electronic medical record management module and the interactive feedback module, and the continuous learning module is connected to the intelligent medication decision support engine and the interactive feedback module.

[0023] Preferably, the electronic medical record management module is used to centrally store and manage patients' structured and unstructured medical data. The electronic medical record management module integrates a data preprocessing unit, which uses natural language processing technology to perform entity recognition and relation extraction on unstructured medical record text and image reports, and transforms them into structured data that can be calculated by the intelligent medication decision support engine. The integrated data preprocessing unit is specifically responsible for processing unstructured text data (such as physician's handwritten medical records and the conclusions of imaging descriptions) using natural language processing technology. Through the two key technologies of entity recognition and relation extraction, the messy text is transformed into regular, semantically labeled structured data, achieving high data availability.

[0024] Preferably, the intelligent medication decision support engine is used to read and analyze patient data in real time. It has a built-in medical knowledge graph and machine learning model, which can identify medication risks and optimize treatment plans, and generate medication review results. The intelligent medication decision support engine includes knowledge graph units and multimodal risk assessment models, realizing a leap from "single rule judgment" to "multi-dimensional comprehensive judgment". Traditional systems can only perform simple "if-A-then-B" rule judgments, while this system combines the association query of the knowledge graph and the predictive ability of the machine learning model to simulate the thinking process of senior pharmacists and clinical experts, and conduct deeper and more individualized risk insights, which greatly improves the depth and breadth of the review.

[0025] Preferably, the knowledge graph unit stores multidimensional relationships between drugs, diseases, genes, and allergens. The multimodal risk assessment model is used to integrate patients' real-time physiological parameters, test indicators, genetic information, and medical history data to predict individualized risks of adverse drug reactions, insufficient efficacy, and drug interactions. The knowledge graph unit specifically stores the relationships between drugs and diseases, genes, allergens, etc., forming a reasonable semantic network. The multimodal risk assessment model is specifically responsible for integrating patients' real-time physiological parameters, test indicators, and other multimodal data to perform individualized risk prediction. By introducing precision medicine dimensions such as genetic information, the system can warn of risks that traditional methods cannot detect (such as drug metabolism abnormalities caused by specific genotypes). By integrating real-time physiological parameters, the system can dynamically assess the risk in the patient's current state, achieving truly "individualized" medication safety monitoring and significantly reducing underreporting and false alarms.

[0026] Preferably, the multimodal risk assessment model is an ensemble learning model. Its input features include the embedding vectors of real-time patient data and knowledge graph query results, and the output is risk probability and confidence level. This improves the prediction accuracy and the credibility of the results. Ensemble learning models (such as XGBoost and Random Forest) usually have higher prediction accuracy and stability than single models. The output of "risk probability and confidence level" rather than a simple "yes / no" judgment provides physicians with richer decision-making reference information. The level of confidence can help physicians judge the firmness of the system prompts, thereby rationally allocating attention and prioritizing the handling of high-confidence high-risk warnings.

[0027] Preferably, the interactive feedback module is used to intuitively display the review results to the user and receive user feedback. The review results provided by the interactive feedback module include high-risk, medium-risk, and advisory levels, and provide evidence-based explanations and alternative medication recommendations for each warning. This effectively alleviates "warning fatigue" and improves the efficiency of doctor-patient communication. Tiered warnings enable physicians to quickly identify truly urgent issues and avoid being overwhelmed by a large number of low-value warnings. Providing explanations and alternative solutions not only enhances the credibility of the system but also transforms the system's role from "error finder" to "decision-making assistant," directly providing physicians with a path to solve problems and greatly improving work efficiency and compliance with prescription modifications.

[0028] Preferably, the continuous learning module is used to iteratively optimize the machine learning model based on user feedback and subsequent diagnosis and treatment results. Traditional rule systems require manual maintenance and updates, which are time-consuming, labor-intensive, and lagging. This system, through continuous learning, can automatically adapt to the medication habits of local hospitals, the response characteristics of specific populations, and the latest clinical practices, forming a virtuous cycle of becoming smarter and more accurate with use, ensuring the long-term vitality and effectiveness of the system.

[0029] The specific working principle and usage method of this invention are as follows: S1, Patient Data Integration and Preprocessing: Data from the hospital information system is continuously extracted into the electronic medical record management module of this system through the ETL process. The data preprocessing unit (NLP server) of this module automatically processes unstructured text (such as physician-written medical records and imaging reports). Algorithm representation: Let T be the input clinical text. Let S be a sequence of tokens S = (w1, w2, ..., wn) from T. The NER model M_ner predicts a sequence of labels L = (l1, l2, ...,ln) for S, where l_i ∈ {B-Drug,I-Drug, B-Disease, I-Disease, B-LabTest, I-LabTest, ..., O}. M_ner can be a deep learning model such as BiLSTM-CRF or BERT-CRF: L = CRF(BiLSTM(Embedding(S))) or L = CRF(BERT(S)). The extracted entities are then mapped to standard medical concepts (eg, UMLS CUI, RxNorm codes) using a dictionary D_map; S2, diagnosis and treatment behavior triggers and real-time data synchronization. When a doctor completes diagnosis entry or prescribes medication on the electronic medical record workstation and clicks the "Save" or "Submit" button, this operation will trigger an event. The intelligent medication decision support engine will capture this event and immediately send a data query request to the electronic medical record management module to obtain all relevant data of the patient. algorithm: / / Event Listener onEvent(PhysicianSavesOrder(patient_id, order_list)) { / / 1. Asynchronously send messages to the message queue message_queue.push({event_type: "order_review", data: {patient_id,order_list}}); / / 2. Return a response immediately, allowing the doctor's interface to continue operating without waiting. return Response("Order submitted. ai review in progress...");} / / Message Consumer consumer = subscribe(message_queue); For message in consumer: if message.event_type == "order_review": / / Obtain comprehensive patient data patient_data = EMR_Module.retrieve_patient_data(message.patient_id); / / Call the intelligent engine review_result=ai_Engine.review_orders(patient_data,message.order_list); / / Push results to the interaction module UI_Module.push_review_result(message.patient_id, review_result); S3, the core reasoning and risk assessment engine for artificial intelligence, receives patient data and medical orders to be reviewed, and executes two sub-tasks in parallel: knowledge graph query and multimodal machine learning model reasoning. Finally, the two results are merged to generate the final risk assessment and recommendations. The parsing process includes: using the NER model to extract disease entities from free text diagnoses written by physicians and mapping them to a standardized medical dictionary; 1. Knowledge graph query algorithm: Query knowledge related to the current drug and patient conditions; Let KG = (V, E) be the medical knowledge graph, where V is a setofentities and E is a set of relations. For a given drug d_i in order_list and a patient condition c_j (eg, a disease orgenotype) from patient_data: Execute a query Q to find paths between d_i and c_j with relations r∈ {interacts_with, contraindicates, metabolized_by, ...}. The result R_kg is a set of tuples (d_i, r, c_j, evidence_score). 2. Multimodal risk assessment model algorithm: Integrates all information to make probability predictions; Let P represent the patient's feature vector. It is a concatenation of: P_demographic: demographic features (age, weight), P_labs: laboratory values (eGFR, INR), P_genetic: genetic markers (one-hot encoded), P_conditions: binary vector of existing conditions, P_medications: binary vector of current medications. Let D be the feature vector of the newly ordered drug(s). The model M_risk is a function that maps the combined feature vectorX = concatenate(P, D) to a risk probability y_hat ∈ [0,1]. y_hat = M_risk(X), where M_risk can be a Gradient Boosting Machine(e.g., XGBoost) or a DeepNeural Network. The loss function for training M_risk is typically binary cross-entropy: L = - [y * log(y_hat)+ (1 - y) * log(1 - y_hat)], where y is thegroundtruth label (1 if adverse event occurred, 0 otherwise). 3. Result integration: Integrate the atlas results and model predictions; The final risk score S_final is a weighted combination: S_final = α * evidence_score + (1 - α) * y_hat where α is a tunable parameter. The system then maps S_final to arisk level (eg, High if S_final>0.7); S4, Interactive Decision Support and Clinical Feedback: The results generated by the intelligent medication decision support engine are immediately sent back to the interactive interface of the physician's workstation. Warning information is presented in a hierarchical, structured, and operable manner. Physicians can choose to adopt, modify, or ignore the suggestion based on their clinical judgment. algorithm: function displayAlert(alert_level, message, evidence, suggestions): color = getColor(alert_level) / / eg, Red for High displayInSidebar(color, message, evidence, buttons(suggestions)) onPhysicianAction(alert_id, action, alternative_drug): logAction(alert_id, action, alternative_drug) / / Record the decision; S5, closed-loop feedback and continuous self-optimization The system anonymizes the physician's decisions (outcomes) and the patient's subsequent efficacy and adverse reaction records (outcomes) and stores them in the learning database. It uses this new data regularly (e.g., weekly) to incrementally train or fine-tune the multimodal risk assessment model, thereby optimizing its future predictive accuracy. Algorithm: Reinforcement learning / online learning algorithm: The system optimizes based on feedback signals from physicians; The physician's action A (eg, "accepted") and the patient outcome O (eg, "no ADE") are used to create a reward signal R. For example: R = +1 if (A == "accept" and O == "no ADE") or (A == "ignore" and O == "ADE"), else R = -1. The goal of the continuous learning module is to update the model parameters θ of M_risk to maximize the expected future reward. This can be framed as a contextual bandit problem and optimized using algorithms like Policy Gradient or Thompson Sampling. Alternatively, the new data (X, y) can be added to the training set for periodic batch re-training of M_risk.

[0030] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Content not described in detail in this specification is prior art known to those skilled in the art.

[0031] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Those skilled in the art can readily implement the present invention based on the accompanying drawings and the above description. However, any modifications, alterations, or variations made by those skilled in the art without departing from the scope of the present invention, utilizing the disclosed technical content, are equivalent embodiments of the present invention. Furthermore, any modifications, alterations, or variations made to the above embodiments based on the essential technology of the present invention are still within the protection scope of the present invention.

Claims

1. An artificial intelligence-based medication consultation and interaction system, comprising an electronic medical record management module, an intelligent medication decision support engine, an interactive feedback module, and a continuous learning module, characterized in that, The intelligent medication decision support engine is communicatively connected to the electronic medical record management module and the interactive feedback module, and the continuous learning module is connected to the intelligent medication decision support engine and the interactive feedback module. The electronic medical record management method includes the following steps: S1 receives and stores the patient's latest medical records through the electronic medical record management module; S2, when a doctor issues or modifies a medical order, the intelligent medication decision support engine is automatically triggered, and real-time calls and analysis of all relevant data of the patient; S3, the intelligent medication decision support engine matches patient data with the knowledge graph and inputs it into the multimodal risk assessment model for comprehensive calculation, generating a review report that includes risk level, specific cause, source of evidence and optimization suggestions; S4 pushes the review report to the physician through the interactive feedback module and records the physician's acceptance, rejection or modification actions; S5, the continuous learning module, uses physician feedback and patient records of subsequent treatment effects and adverse reactions to reinforce the risk assessment model and update the model.

2. The medication consultation and interaction system based on artificial intelligence according to claim 1, characterized in that, The electronic medical record management module is used to centrally store and manage patients' structured and unstructured medical data. The electronic medical record management module integrates a data preprocessing unit, which uses natural language processing technology to perform entity recognition and relationship extraction on unstructured medical record text and image reports, and transforms them into structured data that can be calculated by the intelligent medication decision support engine.

3. The medication consultation and interaction system based on artificial intelligence according to claim 2, characterized in that, The intelligent medication decision support engine is used to read and parse patient data in real time. It has a built-in medical knowledge graph and machine learning model, which can identify medication risks and optimize treatment plans, and generate medication review results. The intelligent medication decision support engine includes knowledge graph units and multimodal risk assessment models.

4. The medication consultation and interaction system based on artificial intelligence according to claim 3, characterized in that, The knowledge graph unit stores multidimensional relationships between drugs, diseases, genes, and allergens. The multimodal risk assessment model is used to integrate patients' real-time physiological parameters, test indicators, genetic information, and medical history data to predict individualized risks of adverse drug reactions, insufficient efficacy, and drug interactions.

5. The medication consultation and interaction system based on artificial intelligence according to claim 4, characterized in that, The multimodal risk assessment model is an ensemble learning model. Its input features include the embedding vectors of real-time patient data and knowledge graph query results, and its output is the risk probability and confidence level.

6. The medication consultation and interaction system based on artificial intelligence according to claim 1, characterized in that, The interactive feedback module is used to intuitively display the review results to the user and receive user feedback. The review results provided by the interactive feedback module include high risk, medium risk, and advisory prompts, and provide evidence-based explanations and alternative drug recommendations for each warning.

7. The medication consultation and interaction system based on artificial intelligence according to claim 1, characterized in that, The continuous learning module is used to iteratively optimize the machine learning model based on user feedback and subsequent diagnosis and treatment results.

8. The medication consultation and interaction system based on artificial intelligence according to claim 1, characterized in that, In step S2, the parsing process includes: extracting disease entities from free text diagnoses written by physicians using a NER model and mapping them to a standardized medical dictionary.

9. The medication consultation and interaction system based on artificial intelligence according to claim 1, characterized in that, In step S5, the strategy for updating the model includes: if a physician repeatedly ignores a certain type of prompt that the system judges as low risk and which ultimately does not cause adverse consequences, the system automatically reduces the level or frequency of that type of prompt.

Images