Tabo and conflict medicine warning system for AI auxiliary hospital
By using an AI-assisted hospital drug contraindication and conflict warning system, drug conflicts can be monitored and warned in real time, solving the problem of difficulty in judging drug contraindications and conflicts in existing technologies, and improving medication safety and system intelligence.
Patent Information
- Application Number
- CN202511039076.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-28
- Publication Date
- 2025-11-11
AI Technical Summary
In existing hospitals, doctors and nurses have difficulty accurately identifying drug contraindications and conflicts in real time during the drug prescription and administration process. Existing systems are unable to intelligently analyze drug conflicts and toxicity burden, leading to reduced medication safety.
An AI-assisted hospital drug contraindication and conflict warning system is adopted, which includes a data layer, a data processing layer, an AI monitoring and early warning layer, and an interaction layer. It monitors and issues warnings in real time through a drug classification model, and optimizes drug classification rules by combining knowledge graphs and machine learning to provide intuitive drug conflict information and warnings.
It improves the safety and accuracy of drug administration, reduces the probability of incorrect medication, optimizes medication plans, and seamlessly integrates with hospital information systems, ensuring the accuracy and timeliness of AI recognition.
Smart Images
Figure CN120932804A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of medical device technology, and more specifically to an AI-assisted hospital-use contraindication and conflict drug warning system. Background Technology
[0002] Currently, with the rapid development of the medical industry, the complexity of drug types and their interactions is increasing. Existing hospital prescriptions and medication administration mainly rely on doctors' personal experience and the basic database support of medical software systems. However, due to the vast variety of drugs and the continuous updating of pharmacological and toxicological knowledge, it is becoming increasingly difficult for doctors and nurses to accurately determine drug contraindications and conflicts in a short period of time, leading to a higher risk of errors in the prescription and administration process. Existing problems include:
[0003] Doctors' knowledge is outdated, and they cannot keep up with all the latest drug contraindications and interactions in real time. Relying on traditional database queries is time-consuming and not intuitive.
[0004] Existing medical software has limitations; most hospital software can only provide basic drug information and cannot intelligently analyze key issues such as drug interactions and toxicity burden.
[0005] Nurses rely on manual confirmation when administering medications. They typically depend on paper or electronic medical records during the administration process, and there is no automated medication verification mechanism, which increases the possibility of errors.
[0006] Drug conflict identification is lagging behind, and existing systems struggle to dynamically and intelligently identify whether drug conflicts exist when patients are prescribed medications at different times, leading to reduced medication safety.
[0007] The existence of these problems makes it imperative for medical institutions to develop an AI-based decision support system to automatically identify drug conflicts and contraindications, thereby improving the safety of clinical medication use. Summary of the Invention
[0008] In view of this, the present invention provides an AI-assisted hospital contraindication and conflict drug warning system to solve the problems existing in the background art.
[0009] To achieve the above objectives, the present invention adopts the following technical solution:
[0010] An AI-assisted hospital-based contraindication and conflict drug warning system includes: a data layer, a data processing layer, an AI monitoring and early warning layer, and an interaction layer. The data layer provides basic data and transmits it to the data processing layer. The data processing layer trains a drug classification model based on the basic data and performs drug classification. The AI monitoring and early warning layer monitors doctors' prescribing practices in real time, and issues an early warning if drug conflicts are found in the drug classification results, while continuously optimizing the drug classification model. The interaction layer presents the early warning to the doctor and saves the warning record.
[0011] Preferably, the basic data in the data layer includes: drug information integrated from a drug database; patient medication data obtained from patient electronic medical records; historical adverse reaction data obtained from an adverse reaction database; and drug interaction information obtained from a drug interaction knowledge graph.
[0012] Preferably, the data processing layer includes:
[0013] The feature extraction module extracts drug structure, pharmacological, and interaction features from the basic data;
[0014] The AI training module uses the output of the feature extraction module as input data to train drug classification models for various types of drugs.
[0015] The classification prediction module takes the drug information to be calculated as input and obtains the classification result through the drug classification model.
[0016] Preferably, the AI monitoring and early warning layer includes:
[0017] The AI early warning module analyzes the interaction between the drug and the patient's existing medications in real time based on the data from the classification and prediction module when the doctor prescribes the medication. If a drug conflict is found, an early warning will be issued.
[0018] The AI monitoring module continuously optimizes drug classification models for various medications.
[0019] Preferably, the interaction layer includes:
[0020] Doctor's electronic prescription system: Receives warnings from the AI warning module when a doctor prescribes medication and generates alerts accordingly;
[0021] Hospital HIS system: Integrates early warning data from the AI early warning module into the hospital information system for subsequent analysis;
[0022] The intelligent warning feedback module allows doctors to view the details of the warning and manually confirm or adjust it.
[0023] As can be seen from the above technical solution, compared with the prior art, the present invention discloses an AI-assisted hospital drug contraindication and conflict warning system. Through the real-time AI early warning system, doctors can intuitively see drug conflict information and optimize medication plans; automatic checks and verifications are performed before medication, reducing the probability of incorrect medication; this system can be seamlessly integrated with the hospital information management system to improve the hospital's intelligence level; it prevents drug conflicts and contraindications, improving the safety of patient medication; and it obtains the latest drug information online daily to ensure the accuracy and timeliness of AI recognition. Attached Figure Description
[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0025] Figure 1 This is a structural schematic diagram provided for the present invention. Detailed Implementation
[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0027] This invention discloses an AI-assisted hospital-use contraindication and conflict drug warning system, such as... Figure 1 As shown, it includes: a data layer, a data processing layer, an AI monitoring and early warning layer, and an interaction layer. The data layer provides basic data and transmits it to the data processing layer; the data processing layer trains a drug classification model based on the basic data and performs drug classification; the AI monitoring and early warning layer monitors doctors' prescriptions in real time, and issues an early warning if there are drug conflicts in the drug classification results, while continuously optimizing the drug classification model; the interaction layer presents the early warning to the doctor as a warning and saves the warning record.
[0028] In one specific embodiment, the basic data in the data layer includes:
[0029] Drug database: Provides detailed information on all drugs on the market.
[0030] Electronic medical record (EHR): contains the patient's historical medication information and disease status.
[0031] Adverse reaction database: Provides data on drug side effects and interaction risks.
[0032] Drug Interaction Knowledge Graph (DTI): Constructing a network of relationships between drugs, targets, and adverse reactions.
[0033] In one specific embodiment, the data processing layer includes:
[0034] The feature extraction module extracts drug structure, pharmacology, and interaction features from basic data;
[0035] The AI training module uses the output of the feature extraction module as input data and trains the ABCD classification model using methods such as GNN and deep learning.
[0036] The classification prediction module takes the drug information to be calculated as input and calculates the ABCD classification result in real time.
[0037] In one specific embodiment, the AI monitoring and early warning layer includes:
[0038] The AI early warning module analyzes the interaction between the drug and the patient's existing medications in real time based on data from the classification and prediction module when the doctor prescribes the medication. If a drug conflict is found, an early warning will be issued.
[0039] The AI monitoring module continuously optimizes drug classification models for various drugs. The data for optimizing drug classification models comes from a daily automatically updated drug database. The AI updates the ABCD drug database daily via the Internet to ensure the real-time nature of label data. It also continuously optimizes drug classification rules by combining data sources from medical regulatory agencies and authoritative pharmacopoeias. Furthermore, it analyzes new drug listing information through machine learning technology and automatically classifies it into the ABCD category library.
[0040] The intelligent rule engine, based on knowledge graphs and machine learning, provides the ability to dynamically adjust rules.
[0041] The intelligent rule engine is a key submodule of the system in this invention. It is primarily used to automatically identify potential drug conflicts and contraindications based on current drug data and clinical context, and dynamically adjust rules to improve the accuracy and timeliness of alerts. This module integrates knowledge graph and machine learning technologies to achieve automatic rule evolution and personalized updates.
[0042] The tools and technologies used consist of:
[0043] Drug Interaction Knowledge Graph (DTI)
[0044] Construct semantic networks for structural relationships, target associations, and adverse reactions among drugs;
[0045] Used to infer potential conflicts, contraindications, and similarities between drugs;
[0046] The knowledge semantic foundation that supports the rule engine.
[0047] Machine learning models
[0048] Including Graph Neural Networks (GNNs), Random Forests, XGBoost, RNNs, Transformers, etc., used for:
[0049] Learn the complex mapping between drugs and organ toxicity;
[0050] Predicting potential cumulative toxicity risks;
[0051] Analyze the risk models after a new drug is launched on the market;
[0052] Quantify the probability of adverse interactions between drugs;
[0053] The learning outcomes are transformed into rule candidates, which serve as a reference for subsequent rule adjustments.
[0054] Dynamic Rule Updating Mechanism
[0055] Data is updated daily from sources such as drug databases, electronic medical records, FAERS, and pharmacopoeias;
[0056] Newly discovered drug interaction features are automatically embedded into the knowledge graph;
[0057] Machine learning is used to evaluate indicators such as conflict intensity and frequency, and to filter and adjust rule weights and trigger thresholds.
[0058] It can be viewed as an expert rule system that "evolves on its own".
[0059] Overview of operational logic:
[0060] Data input:
[0061] Receive medication instructions from the doctor's electronic prescription system in real time;
[0062] Load the patient's previous medication data and basic health data.
[0063] Rule matching (preliminary judgment):
[0064] Match drug combinations using an existing rule base (static + learned generation);
[0065] The rule base is constructed based on knowledge graph reasoning paths (such as two drugs sharing the same target / organ, co-occurrence of adverse reactions, etc.).
[0066] Machine learning analysis (dynamic scoring):
[0067] Simultaneously run models such as GNN, RNN, and XGBoost to predict risk levels;
[0068] Analyze the actual risks of the current medication combination in the context of the patient's history.
[0069] Rules are dynamically adjusted:
[0070] Update the weights of the current matching rules based on the model's prediction results;
[0071] If the model score is higher than the rule threshold but the current rule is not triggered, the system will automatically adjust the rule to form a new rule item or optimize the old rule.
[0072] If a new drug is launched, the system will automatically establish its linkage relationship with known drugs through a knowledge graph and generate initial rules.
[0073] Output the warning results:
[0074] If a rule is triggered (static or dynamically generated), a medication warning will be sent to the doctor or nurse in real time.
[0075] All rule triggers are recorded and fed back to the system for use in the next model training.
[0076] In one specific embodiment, the AI monitoring module also includes AI monitoring of nurses administering medication. Before administering medication to a patient, the nurse needs to scan the drug's QR code on the bedside table.
[0077] After scanning, the system automatically matches the patient's recent medication history and checks for any conflicts in categories A, B, C, or D.
[0078] If a conflict is detected, the system will prompt the nurse to verify it. The nurse must click to confirm before entering new medications.
[0079] In one specific embodiment, the interaction layer includes:
[0080] Doctor's electronic prescription system: Receives alerts from the AI warning module when a doctor prescribes medication and generates warnings accordingly;
[0081] Hospital HIS system: Integrates early warning data from the AI early warning module into the hospital information system for subsequent analysis;
[0082] The intelligent warning feedback module allows doctors to view the details of warnings and manually confirm or adjust them.
[0083] In one specific embodiment, feature extraction of the data layer includes:
[0084] Chemical structure analysis: Tanimt similarity between drugs is calculated using molecular fingerprint (MF).
[0085] Target similarity: Based on the protein-drug interaction network (DTI), the target similarity of drug action is calculated.
[0086] Adverse reaction analysis: A drug-adverse reaction correlation matrix was constructed based on FAERS (FDA Adverse Event Reprting System) data.
[0087] Pharmacokinetic parameters such as half-life, plasma protein binding rate, and volume of distribution are used to calculate the effect of drug accumulation in vivo.
[0088] A drug classification model is constructed using a multi-level classification model, mapping drugs to four categories: A, B, C, and D.
[0089] (1) Category A (same effects, may lead to additive effects or overdose)
[0090] Methods: Clustering algorithms (such as K-means or Hierarchical Clustering) were used to group drugs based on their mechanism of action (MA, fActin) and target similarity.
[0091] formula:
[0092]
[0093] If Sim MOA (D i D j )>θ A If the threshold is met, the classification is A.
[0094] Let there be two drugs D. i and D j The mechanisms of action are MOA(D) i ) and MOA(D j This paper uses the Jaccard similarity coefficient to measure the mechanistic similarity between two sets. The Jaccard coefficient is defined as the ratio of the size of the intersection to the size of the union of two sets, and therefore the formula is as described above.
[0095] When this value exceeds a certain threshold θA, it is determined that the two drugs have highly similar mechanisms of action, and there may be a risk of overlapping effects.
[0096] MOA(D i ): The set of mechanisms of action of drug Di (such as the pathways it affects, target proteins, etc.).
[0097] ∩ and ∪ are set operations, representing intersection and union respectively.
[0098] |·|: The number of elements in the set.
[0099] θA: The similarity threshold for classifying a class as A, which is set empirically or determined based on training data.
[0100] (2) Category B (obvious conflict, may cause serious adverse reactions)
[0101] Methods: We used knowledge graphs and graph neural networks (GNNs) to identify serious interactions in the drug-target-adverse reaction relationship graph.
[0102] formula
[0103]
[0104] If Risk B >θ B It is then classified as category B, where:
[0105] R represents all adverse reactions.
[0106] w r The weighting of adverse reactions (based on clinical severity grading),
[0107] ·P(D i D j |r) represents the probability r of the adverse reaction caused by the drug combination.
[0108] We will have all of this related to drug D i D j The set of adverse reactions is represented as a set of events R, where each event r has a corresponding weight wr and a probability of occurrence P(D). i D j |r). The final conflict risk value RiskB(D) i D j ) is the weighted sum of the risks of each event, as shown in the formula above.
[0109] When the total risk value exceeds the threshold θ B At that time, both drugs were classified as Category B.
[0110] R: The known set of serious adverse reactions related to the drug.
[0111] w r The severity weight of response r is usually set according to the clinical classification (such as CTCAE classification).
[0112] P(D i D j |r): Under given reaction r, drug Di With D j Simultaneously, the probability of causing this reaction is used.
[0113] θ B Risk threshold for category B
[0114] (3) Category C (imposes a burden on the same organ and may exacerbate organ damage)
[0115] Methods: Based on biomarker association analysis and organ toxicity databases (such as SIDER), random forests or XGBst were trained to predict the organ damage risk of drugs.
[0116] formula:
[0117]
[0118] If Sim Tox (D i D j )>θ C If so, it is classified as category C.
[0119] The Jaccard similarity coefficient, similar to class A, is used to measure the overlap of organ toxicity between two drugs. The set of organs affected by toxicity is defined as Tox(D). i If the similarity is calculated as shown above, then the similarity formula is as follows:
[0120] When the similarity is higher than the threshold θ C At that time, the drug was classified as Category C.
[0121] Tox(D i Drug D i A group of organs affected (such as the liver, kidneys, and heart).
[0122] θ C : Threshold for organ toxicity similarity classified as C.
[0123] (4) Category D (same toxicology, may lead to cumulative toxic effects)
[0124] Methods: RecurrentNeuralNetwrk (RNN) and Transfrmer were used to process drug structure data to predict toxicity accumulation patterns.
[0125] formula:
[0126]
[0127] If Tox acc >θ D If so, it is classified as category D.
[0128] Let drug D i With Dj The toxicity score of the combined use of the drugs at T consecutive time points is the ToxicityScore (D). i D j ,t), introduce a time decay factor α t (e.g., decreasing weights) to reflect the accumulation of toxicity within the time window. The calculation formula is shown in the formula above.
[0129] When the total toxicity score exceeds the threshold θ D It is classified as Category D.
[0130] T: Length of the forecast time window (e.g., in days).
[0131] α t Time decay factor, usually α t =γ T-t (0<γ<1).
[0132] ToxicityScore(D i D j, t): The toxicity score predicted by the model on day t.
[0133] θ D : Cumulative toxicity threshold for category D.
[0134] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.
[0135] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. An AI-assisted hospital-based system for warning of contraindications and conflicts with medications, characterized in that, include: The system comprises a data layer, a data processing layer, an AI monitoring and early warning layer, and an interaction layer. The data layer provides basic data and transmits it to the data processing layer. The data processing layer trains a drug classification model based on the basic data and performs drug classification; the AI monitoring and early warning layer monitors doctors' prescriptions in real time, and issues an early warning if there are drug conflicts in the drug classification results, while continuously optimizing the drug classification model; the interaction layer presents the early warning to the doctor as a warning, and saves the warning record.
2. The AI-assisted hospital-use contraindication and conflict drug warning system according to claim 1, characterized in that, The basic data in the data layer includes: drug information integrated from drug databases; patient medication data obtained from patients' electronic medical records; historical adverse reaction data obtained from adverse reaction databases; and drug interaction information obtained from drug interaction knowledge graphs.
3. The AI-assisted hospital-use contraindication and conflict drug warning system according to claim 1, characterized in that, The data processing layer includes: The feature extraction module extracts drug structure, pharmacological, and interaction features from the basic data; The AI training module uses the output of the feature extraction module as input data to train drug classification models for various types of drugs. The classification prediction module takes the drug information to be calculated as input and obtains the classification result through the drug classification model.
4. The AI-assisted hospital-use contraindication and conflict drug warning system according to claim 3, characterized in that, The AI monitoring and early warning layer includes: The AI early warning module analyzes the interaction between the drug and the patient's existing medications in real time based on the data from the classification and prediction module when the doctor prescribes the medication. If a drug conflict is found, an early warning will be issued. The AI monitoring module continuously optimizes drug classification models for various medications.
5. The AI-assisted hospital-use contraindication and conflict drug warning system according to claim 4, characterized in that, The interaction layer includes: Doctor's electronic prescription system: Receives warnings from the AI warning module when a doctor prescribes medication and generates alerts accordingly; Hospital HIS system: Integrates early warning data from the AI early warning module into the hospital information system for subsequent analysis; The intelligent warning feedback module allows doctors to view the details of the warning and manually confirm or adjust it.