Risk determination method and device, equipment and storage medium

Abstract

The invention provides a risk determination method and device, equipment and a storage medium, and relates to the technical field of artificial intelligence. In some embodiments of the invention, health state information, medication information, time biological characteristics and drug molecular map structural characteristics are input into a health risk prediction model to obtain a health risk prediction result; under the condition that the health risk prediction result is that the risk exists, determining medication adjustment information according to medication information and a medication dosage rule; sending the health risk prediction result and the medication adjustment information to an early warning terminal, so that the early warning terminal displays the risk prediction result and the medication adjustment information; according to the invention, based on the health state information, the medication information, the time biological characteristics and the drug molecular map structural characteristics of the user, the health risk prediction model is utilized to accurately predict the health risk, the health risk prediction efficiency is improved, and the health risk prediction accuracy is improved.

Description

Technical Field

[0001] This disclosure relates to the field of artificial intelligence technology, and in particular to a risk determination method, apparatus, device, and storage medium. Background Technology

[0002] Currently, the identification and diagnosis of drug-induced hypertension (i.e., hypertension caused by medication) mainly relies on the physician's experience and the patient's medication history. This diagnostic method may be subjective and uncertain, leading to delays or errors in diagnosis. Furthermore, because the causes of drug-induced hypertension are diverse, including the effects of different medications and individual patient differences, personalized treatment plans are necessary.

[0003] Currently, the efficiency and accuracy of predicting the potential health risks associated with drugs are low. Summary of the Invention

[0004] This disclosure provides a risk determination method, apparatus, device, and storage medium to address the problems of low efficiency and low accuracy in existing health risk prediction methods.

[0005] The technical solution disclosed herein is as follows:

[0006] This disclosure provides a risk determination method, including:

[0007] Acquire users' health status information, medication information, chronobiological characteristics, and drug molecular diagram structural characteristics;

[0008] The health status information, medication information, chronobiological characteristics, and drug molecular diagram structural characteristics are input into the health risk prediction model to obtain the health risk prediction results.

[0009] If the health risk prediction result indicates that there is a risk, medication adjustment information is determined based on the medication information and medication dosage rules;

[0010] The health risk prediction results and the medication adjustment information are sent to the early warning terminal so that the early warning terminal can display the risk prediction results and the medication adjustment information.

[0011] Optionally, the health risk prediction model includes a traditional machine learning module, a deep learning module, and an ensemble module. The step of inputting the health status information, the medication information, the chronobiological characteristics, and the drug molecular graph structural characteristics into the health risk prediction model to obtain the health risk prediction result includes:

[0012] Within the health risk prediction model, the health status information, the medication information, the chronobiological characteristics, and the drug molecular diagram structural characteristics are input into the health risk prediction model to obtain at least one first prediction result;

[0013] The health status information, medication information, chronobiological characteristics, and drug molecular graph structural characteristics are input into the deep learning module to obtain a second prediction result;

[0014] The first prediction result and the second prediction result are input into the integrated module to obtain the health risk prediction result.

[0015] Optionally, the traditional machine learning module includes: support vector machine, random forest, and gradient boosting decision tree; the step of inputting the health status information, the medication information, the chronobiological features, and the drug molecule graph structure features into the health risk prediction model to obtain at least one first prediction result includes:

[0016] The health status information, medication information, time biological characteristics, and drug molecule graph structure characteristics are input into the support vector machine to obtain a first prediction result;

[0017] The health status information, medication information, time biological characteristics, and drug molecule graph structure characteristics are input into the random forest to obtain a first prediction result;

[0018] The health status information, medication information, chronobiological features, and drug molecular graph structural features are input into a gradient boosting decision tree to obtain a first prediction result.

[0019] Optionally, the deep learning module includes: a graph neural network, a recurrent neural network, and an attention mechanism network; the step of inputting the health status information, the medication information, the chronobiological features, and the drug molecule graph structure features into the deep learning module to obtain a second prediction result includes:

[0020] The drug molecule graph structure features are input into the graph neural network to obtain the drug feature vector;

[0021] The time-biological features are input into the recurrent neural network to obtain time-series features;

[0022] The health status information, medication information, chronobiological features, drug molecular graph structural features, drug feature vectors, and temporal features are input into the attention mechanism network to obtain the second prediction result.

[0023] Optionally, inputting the first prediction result and the second prediction result into the integration module to obtain the health risk prediction result includes:

[0024] The first prediction result and the second prediction result are input into the integration module, and the first prediction result and the second prediction result are weighted and voted to obtain the health risk prediction result.

[0025] Optionally, the method further includes:

[0026] Construct a drug risk knowledge graph based on drug name, clinical symptoms, mechanism of action, and risk factors;

[0027] Based on the drug risk knowledge graph, risk warning rules and medication dosage rules are determined.

[0028] Optionally, the method further includes:

[0029] Based on the health status information, medication information, and time biological characteristics, determine whether the risk warning rule has been triggered;

[0030] If the aforementioned risk warning rules are met, an alert action will be executed.

[0031] This disclosure also provides a risk determination apparatus, including:

[0032] The acquisition module is used to acquire users' health status information, medication information, chronobiological characteristics, and drug molecular diagram structural characteristics.

[0033] The prediction module is used to input the health status information, the medication information, the time biological characteristics, and the drug molecular diagram structure characteristics into the health risk prediction model to obtain the health risk prediction result;

[0034] The determination module is used to determine medication adjustment information based on the medication information and medication dosage rules when the health risk prediction result indicates that there is a risk.

[0035] The display module is used to send the health risk prediction results and the medication adjustment information to the early warning terminal, so that the early warning terminal can display the risk prediction results and the medication adjustment information.

[0036] This disclosure also provides an electronic device, including:

[0037] processor;

[0038] Memory used to store processor-executable instructions;

[0039] The processor is configured to execute instructions to implement the steps in the above method.

[0040] This disclosure also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0041] The technical solutions provided by the embodiments of this disclosure have at least the following beneficial effects:

[0042] In some embodiments of this disclosure, user health status information, medication information, chronobiological characteristics, and drug molecular graph structure characteristics are obtained; these information are input into a health risk prediction model to obtain a health risk prediction result; if the health risk prediction result indicates the presence of risk, medication adjustment information is determined based on medication information and dosage rules; the health risk prediction result and medication adjustment information are sent to an early warning terminal for the terminal to display. This disclosure, based on user health status information, medication information, chronobiological characteristics, and drug molecular graph structure characteristics, utilizes a health risk prediction model to accurately predict health risks, improving the efficiency and accuracy of health risk prediction.

[0043] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0044] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure, and are not intended to unduly limit this disclosure.

[0045] Figure 1 A flowchart illustrating a risk determination method provided for an exemplary embodiment of this disclosure;

[0046] Figure 2 A schematic diagram of a risk determination device provided for an exemplary embodiment of this disclosure;

[0047] Figure 3 A schematic diagram of the structure of an electronic device provided for an exemplary embodiment of this disclosure. Detailed Implementation

[0048] To enable those skilled in the art to better understand the technical solutions of this disclosure, the technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings.

[0049] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure.

[0050] It should be noted that the user information involved in this disclosure includes, but is not limited to, user device information and user personal information; the collection, storage, use, processing, transmission, provision and disclosure of user information in this disclosure all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0051] To address the aforementioned technical issues, some embodiments of this disclosure involve acquiring a user's health status information, medication information, chronobiological characteristics, and drug molecular graph structural characteristics; inputting these information into a health risk prediction model to obtain a health risk prediction result; determining medication adjustment information based on medication information and dosage rules when the health risk prediction result indicates a risk; and sending the health risk prediction result and medication adjustment information to an early warning terminal for display. This disclosure, based on a user's health status information, medication information, chronobiological characteristics, and drug molecular graph structural characteristics, utilizes a health risk prediction model to accurately predict health risks, thereby improving the efficiency and accuracy of health risk prediction.

[0052] The technical solutions provided by the embodiments of this disclosure are described in detail below with reference to the accompanying drawings.

[0053] Figure 1 This is a flowchart illustrating a risk determination method provided as an exemplary embodiment of this disclosure. Figure 1 As shown, the method includes:

[0054] S101: Obtain user's health status information, medication information, chronobiological characteristics, and drug molecular diagram structural characteristics;

[0055] S102: Input health status information, medication information, chronobiological characteristics, and drug molecular diagram structural characteristics into the health risk prediction model to obtain health risk prediction results;

[0056] S103: When the health risk prediction result indicates that there is a risk, medication adjustment information shall be determined based on medication information and medication dosage rules;

[0057] S104: Send the health risk prediction results and medication adjustment information to the early warning terminal so that the early warning terminal can display the risk prediction results and medication adjustment information.

[0058] In this embodiment, the entity executing the above method can be a terminal device or a server.

[0059] The terminal device includes, but is not limited to, mobile stations (MS), mobile terminals, mobile phones, handsets, and portable equipment. This terminal device can communicate with one or more core networks via a radio access network (RAN). For example, the terminal device can be a mobile phone (or "cellular" phone), a computer with wireless communication capabilities, a computer with wireless transceiver capabilities, a virtual reality (VR) terminal device, an AR terminal device, a wireless terminal in industrial control, a wireless terminal in self-driving, a wireless terminal in remote medical care, a wireless terminal in a smart grid, a wireless terminal in transportation safety, a wireless terminal in a smart city, a wireless terminal in a smart home, etc. The operating systems installed on the terminal device include, but are not limited to, iOS, Android, Windows, Linux, and Mac OS. In different networks, terminals may be called by different names, such as: user equipment, mobile station, user unit, station, cellular phone, personal digital assistant, wireless modem, wireless communication device, handheld device, laptop, cordless phone, wireless local loop station, television, etc. For ease of description, this embodiment will simply refer to it as terminal device.

[0060] In this embodiment, the implementation form of the server is not limited. For example, the server can be a conventional server, a cloud server, a cloud host, a virtual center, or other server devices. The server mainly consists of a processor, hard disk, memory, system bus, and other common computer architecture types.

[0061] In this embodiment, the user's health status information, medication information, chronobiological characteristics, and drug molecular graph structure characteristics are acquired; these information are then input into a health risk prediction model to obtain a health risk prediction result; if the health risk prediction result indicates the presence of risk, medication adjustment information is determined based on the medication information and dosage rules; the health risk prediction result and medication adjustment information are sent to an early warning terminal for display. This disclosure, based on the user's health status information, medication information, chronobiological characteristics, and drug molecular graph structure characteristics, utilizes a health risk prediction model to accurately predict health risks, improving the efficiency and accuracy of health risk prediction.

[0062] In some embodiments of this disclosure, a comprehensive and multi-dimensional knowledge base for drug-induced hypertension needs to be constructed first. Adverse reaction information related to blood pressure, including detailed data such as dosage, method of use, and contraindications, is extracted and analyzed from various drug instruction manual databases. The system also systematically searches relevant domestic and international literature, extracts clinical research results and evidence-based medicine, and establishes a complete literature database. Furthermore, by integrating data from the adverse reaction reporting system, a comprehensive adverse reaction database is constructed, enabling comprehensive monitoring of adverse drug reactions. Finally, by collecting typical case data from within the hospital, a rich clinical case database is established; this real-world data provides crucial support for the practical application of the system. The integration of these databases lays a solid data foundation for subsequent knowledge graph construction and health risk prediction model development. This disclosure constructs a comprehensive and multi-dimensional knowledge base for drug-induced hypertension by integrating multi-source knowledge such as instruction manuals, expert consensus / guidelines, internet case mining, typical cases, and literature, comprehensively covering relevant information on drug-induced hypertension and improving the accuracy and practicality of the knowledge base. In constructing the knowledge base, this disclosure particularly emphasizes combining drug information with the characteristics of drug-induced hypertension. This method helps to more accurately identify and correlate the relationship between medications and hypertension, thereby improving the accuracy of diagnosis and early warning.

[0063] In the above embodiments, data cleaning and standardization are crucial steps in constructing the knowledge base for drug-induced hypertension. First, generic drug names are uniformly converted and standardized to ensure consistent naming of the same drug across different data sources, which is essential for accurately identifying drugs that may cause hypertension. Second, since drug-induced hypertension is primarily related to oral and injectable administration routes, data from other routes of administration (such as topical and inhalation) need to be excluded to improve data specificity. Strict deduplication and verification mechanisms are employed to ensure data consistency, removing duplicate records and maintaining consistency across all data fields. Considering that the active pharmaceutical ingredient (API) is a key factor leading to elevated blood pressure, information on the API is primarily retained, while information on excipients such as salts and metals that may interfere with the analysis is removed. Furthermore, because the interactions of multiple components in compound preparations can affect the accurate assessment of blood pressure changes, data from single-component preparations are prioritized, while data from compound preparations are excluded to ensure the accuracy and reliability of the data analysis.

[0064] In some embodiments of this disclosure, a drug risk knowledge graph is constructed based on drug name, clinical symptoms, mechanism of action, and risk factors; risk warning rules and medication dosage rules are determined based on the drug risk knowledge graph.

[0065] In one exemplary embodiment, a multi-level knowledge graph model is constructed using an ontology approach. At the drug level, the system records detailed information on drugs that may cause hypertension, including key information such as the drug's chemical classification (e.g., nonsteroidal anti-inflammatory drugs, glucocorticoids), molecular structural characteristics, recommended dosage range, and dosing regimen. This information helps physicians accurately identify potential hypertensive drugs and make appropriate adjustments to their medication regimens. At the symptom level, the system comprehensively describes the clinical manifestations of drug-induced hypertension, including the temporal pattern of blood pressure elevation (e.g., immediate or gradual increase after medication), the magnitude of blood pressure changes, and whether heart rate changes are present. It also includes common accompanying symptoms such as headache, dizziness, and blurred vision, which are important indicators for early identification of drug-induced hypertension. At the mechanism level, the system analyzes in depth the mechanisms of action of different drugs that cause hypertension, such as sodium and water retention, sympathetic nervous system activation, and vascular endothelial dysfunction. It also integrates information on various risk factors, including factors that may affect the risk of drug-induced hypertension, such as the patient's age, underlying diseases, and renal function. This mechanistic knowledge provides a theoretical basis for the development of personalized medication regimens. These three levels of knowledge are closely linked through ontological relationships, forming a complete knowledge network that can support the system in conducting intelligent risk assessment and early warning.

[0066] In the above embodiments, a visual drug risk knowledge graph is constructed; where nodes represent entities, such as drug name, symptoms (e.g., headache), mechanisms (e.g., sodium and water retention), and patient characteristics (e.g., age > 65 years). Edges represent relationships between entities, such as "Drug A → leads to → symptom B", "Mechanism C → associated with → risk factor D".

[0067] This disclosure utilizes a constructed knowledge graph and a multi-level rule generation method to form a complete early warning rule system for drug-induced hypertension. The rule generation process mainly includes three stages: rule extraction, rule optimization, and rule verification.

[0068] During the rule extraction phase, the system first identifies key drug-related features from the knowledge graph and constructs risk warning rules. For nonsteroidal anti-inflammatory drugs (NSAIDs), the focus is on features such as COX-2 selectivity, dosage, and treatment duration; for glucocorticoids, the focus is on features such as receptor affinity and cumulative dosage. Simultaneously, the system also extracts rules related to patient characteristics, such as age over 65 years, renal insufficiency, and concurrent use of multiple vasopressors—all high-risk factors. These rules are expressed in the IF-THEN format, for example: "IF: Patient age > 65 years AND using selective COX-2 inhibitors AND treatment duration > 4 weeks, THEN: Increased risk of drug-induced hypertension."

[0069] In one exemplary embodiment, it is determined whether a risk warning rule has been triggered based on health status information, medication information, and chronobiological characteristics; if the risk warning rule has been triggered, an alert operation is performed.

[0070] At the medication regimen level, the system generates a series of dose-related dosage rules. For example, for certain nonsteroidal anti-inflammatory drugs (NSAIDs), the system generates dosage adjustment suggestions when the daily dose exceeds 75% of the recommended range. For medications requiring long-term use, the system also generates rules for step-down dosing regimens, gradually adjusting the dose to reduce the risk of elevated blood pressure. These rules particularly emphasize the principle of individualized medication, taking into account the differences in drug sensitivity among different patients.

[0071] Regarding monitoring rules, the system generates corresponding blood pressure monitoring protocols based on the temporal characteristics of different drugs that induce hypertension. For example, for drugs that may cause acute hypertension, the system generates more frequent blood pressure monitoring rules; while for drugs that may cause chronic hypertension, it generates rules for long-term follow-up monitoring. The monitoring rules also include warning thresholds for blood pressure changes, such as specific indicators like "a single systolic blood pressure increase of 20 mmHg from baseline" or "the average of three consecutive measurements increase of 15 mmHg from baseline."

[0072] The rule optimization phase primarily focuses on the accuracy and usability of the rules. The system employs an evidence level assessment method, assigning weights to each rule based on the reliability of its source. For example, evidence from large-scale randomized controlled trials has the highest weight, while evidence from expert experience has a relatively lower weight. Simultaneously, the system establishes a rule conflict resolution mechanism, automatically selecting the most suitable rule based on its priority and evidence level when multiple rules are triggered simultaneously.

[0073] During the rule validation phase, the system validates the generated rules using real case data. The validation process includes sensitivity analysis (whether the rules can identify real cases of drug-induced hypertension) and specificity analysis (whether they generate too many false positive alerts). For rules that perform poorly, the system makes corresponding adjustments and optimizations. For example, if a rule causes too many false positive alerts, the system will improve its specificity by adjusting the rule's trigger conditions or adding limiting conditions.

[0074] Furthermore, the system has established a dynamic rule update mechanism. As new medical evidence emerges and clinical practice accumulates, the system can update and improve its rule base in a timely manner. This dynamic update mechanism ensures that the rule system can continuously maintain its scientific rigor and timeliness. For example, when a new drug is found to potentially cause hypertension, the system can quickly generate corresponding warning rules; when the risk assessment criteria for a certain type of drug change, the system can also adjust the parameters of the relevant rules in a timely manner.

[0075] It should be noted that the health risk prediction model includes traditional machine learning modules, deep learning modules, and ensemble modules.

[0076] Before using a health risk prediction model, it is necessary to first construct the model. During the model construction process, a systematic dataset was built. Considering the characteristics of drug-induced hypertension, training data was collected and organized from multiple dimensions. Positive samples mainly included two types of data: first, clinically diagnosed cases of drug-induced hypertension, all of which underwent rigorous review by specialists to ensure data accuracy; second, cases of hypertension clearly caused by medication, screened from the adverse reaction monitoring database. For negative samples, cases using the same class of drugs but without elevated blood pressure, as well as data from normotensive patients using other medications, were selected.

[0077] To ensure data quality and representativeness, detailed annotations were performed on each case. Annotations included patient health status, medication information, chronobiological characteristics, and drug molecular map structural features. Health status information included age, sex, renal function, and baseline blood pressure levels; medication information included drug type, dosage, frequency of administration, duration of treatment, and cumulative exposure time; chronobiological characteristics included blood pressure trends and changes in circadian rhythms. Annotation was completed by a team of physicians with extensive clinical experience, and a two-person cross-review process was used to ensure annotation quality. Considering the relatively low incidence of drug-induced hypertension, a combination of stratified sampling and oversampling was used to balance the dataset, ensuring the model could fully learn less common but important feature patterns. The selection of these features was based on expert experience and literature evidence, and can effectively reflect the risk of drug-induced hypertension.

[0078] The model construction employs a multi-model fusion strategy, comprehensively leveraging the advantages of traditional machine learning and deep learning. In traditional machine learning, Support Vector Machines (SVMs) are used to handle classification problems in high-dimensional feature spaces, with their kernel functions optimized to adapt to the distribution characteristics of the feature space. Random Forest models provide robust predictive capabilities by integrating multiple decision trees, making them particularly suitable for handling clinical data with missing values. XGBoost (Gradient Boosting Decision Tree) models continuously optimize predictive performance through gradient boosting, with the depth and number of tree structures carefully tuned to balance model complexity and generalization ability. In deep learning, a specialized neural network architecture was designed. Graph Neural Networks (GNNs) are used to directly learn the structural features of drug molecule graphs, and their message passing mechanism effectively captures interactions between atoms. Recurrent Neural Networks (RNNs) handle time-series data such as blood pressure monitoring, maintaining long-term dependencies through LSTM units. Attention mechanisms integrate features from different sources, highlighting the contributions of important features through dynamic weight allocation.

[0079] A systematic approach was employed in the model optimization process. For hyperparameter optimization, grid search combined with cross-validation was used to comprehensively explore the parameter space. To improve model interpretability, SHAP value analysis was used to quantify the contribution of different features, which is of significant value for understanding the model's predictive basis and guiding clinical practice. The final ensemble model adopted a weighted voting method, with weights dynamically adjusted based on the performance of each base model on the validation set, ensuring the stability and reliability of the prediction results.

[0080] A multi-layered validation strategy was employed during the model validation and evaluation phase to comprehensively assess the performance and practical value of the health risk prediction model. In the internal validation stage, a ten-fold cross-validation method was used for comprehensive model evaluation. Specifically, the entire dataset was randomly divided into 10 equal-sized subsets. Nine subsets were selected as training data each time, and the remaining subset was used for validation. This process was repeated 10 times to fully evaluate the model's stability and generalization ability. This validation method is particularly suitable for evaluating prediction tasks with imbalanced clinical data, such as drug-induced hypertension, because it can make full use of limited positive samples.

[0081] For external validation, independent test datasets from different time periods were first used to evaluate the model's performance on new data. Simultaneously, Y-randomization validation was performed, randomly shuffling the labels to verify the model's reliability. Of particular note, group validation was conducted for different categories of hypertensive drugs (such as nonsteroidal anti-inflammatory drugs and glucocorticoids) to assess the model's predictive accuracy for different drug categories.

[0082] In the clinical application validation phase, the focus is on the model's performance in real-world clinical environments. Real-world medication and blood pressure monitoring data are collected to evaluate the model's actual warning effectiveness. Simultaneously, clinical experts are invited to evaluate and provide feedback on the warning results, and the model parameters and warning rules are continuously optimized based on expert opinions. This dynamic optimization mechanism ensures that the model can adapt to various complex situations in clinical practice.

[0083] To quantify and evaluate model performance, a series of scientific evaluation metrics were employed. In the early warning of drug-induced hypertension, sensitivity (SE) reflects the model's ability to correctly identify patients who may experience adverse hypertension reactions, while specificity (SP) indicates the model's ability to correctly exclude low-risk patients. Accuracy (ACC) comprehensively reflects the model's overall predictive accuracy, while the Matthews correlation coefficient (MCC) provides a more objective performance assessment in cases of imbalanced samples. These metrics were calculated based on true positive (TP), true negative (TN), false positive (FP), and false negative (FN) data from the confusion matrix.

[0084] The specific calculation method is as follows:

[0085] Sensitivity (SE) = TP / (TP + FN);

[0086] Specificity (SP) = TN / (TN+FP);

[0087] Accuracy (ACC) = (TP + TN) / (TP + TN + FP + FN);

[0088] Matthews correlation coefficient (MCC) = (TP×TN-FP×FN) / √((TP+FP)(TP+FN)(TN+FP)(TN+FN));

[0089] Among them, TP is a true positive, TN is a true negative, FP is a false positive, and FN is a false negative.

[0090] After rigorous validation and evaluation, this disclosure achieved an accuracy exceeding 0.85 on the test set, demonstrating the effectiveness and reliability of the method in early warning of drug-induced hypertension. More importantly, the system can provide clinicians with timely early warning information on the risk of drug-induced hypertension, facilitating the development of safer and more personalized medication regimens and significantly improving patient medication safety. This early warning capability has significant clinical value in preventing drug-induced hypertension, especially for patients using high-risk medications.

[0091] In some embodiments of this disclosure, health status information, medication information, chronobiological characteristics, and drug molecular graph structure features are input into a health risk prediction model to obtain a health risk prediction result. One possible approach is to input the health status information, medication information, chronobiological characteristics, and drug molecular graph structure features into the health risk prediction model to obtain at least one first prediction result; input the health status information, medication information, chronobiological characteristics, and drug molecular graph structure features into a deep learning module to obtain a second prediction result; and input the first and second prediction results into an integration module to obtain the final health risk prediction result. It should be noted that the health risk prediction result may indicate a high risk of drug-induced hypertension, but does not necessarily mean that the user will develop drug-induced hypertension. The health risk prediction result may also be a health risk level, such as low risk, medium risk, and high risk of drug-induced hypertension.

[0092] In the above embodiments, the traditional machine learning module includes: Support Vector Machine (SVM), Random Forest, and Gradient Boosting Decision Tree; health status information, medication information, chronobiological features, and drug molecular graph structure features are input into the health risk prediction model to obtain at least one first prediction result. One possible implementation is to input health status information, medication information, chronobiological features, and drug molecular graph structure features into a Support Vector Machine to obtain a first prediction result; input health status information, medication information, chronobiological features, and drug molecular graph structure features into a Random Forest to obtain a first prediction result; and input health status information, medication information, chronobiological features, and drug molecular graph structure features into a Gradient Boosting Decision Tree to obtain a first prediction result. The first prediction result can be a probability value or a category label.

[0093] In the above embodiments, the deep learning module includes a graph neural network, a recurrent neural network, and an attention mechanism network. Health status information, medication information, chronobiological features, and drug molecular graph structure features are input into the deep learning module to obtain a second prediction result. One possible implementation is to input the drug molecular graph structure features into the graph neural network to obtain a drug feature vector; input the chronobiological features into the recurrent neural network to obtain temporal features; and input the health status information, medication information, chronobiological features, drug molecular graph structure features, drug feature vector, and temporal features into the attention mechanism network to obtain the second prediction result. The input to the graph neural network can be the drug molecular graph structure features and other drug-related features that may be combined; the output of the graph neural network can be the drug feature vector or a direct risk prediction contribution. The output of the recurrent neural network can be a risk score or a temporal feature useful to other models. The input to the attention mechanism network integrates features from different sources, such as features from traditional machine learning modules and the output features of the graph neural network and the recurrent neural network. The output of the attention mechanism network can be weights assigned to different features, or a weighted fusion of feature representations or risk predictions.

[0094] In the above embodiments, the first and second prediction results are input into the integration module to obtain the health risk prediction result. One possible approach is to input the first and second prediction results into the integration module, and then perform a weighted vote on the first and second prediction results to obtain the health risk prediction result. The integration module integrates the predictions from each model through weighted voting to arrive at a final, more stable, and accurate early warning information for drug-induced hypertension risk, such as a final risk classification or risk score. The weights may be dynamically adjusted based on the performance of each base model on the validation set.

[0095] In some embodiments of this disclosure, when the health risk prediction result indicates the presence of risk, medication adjustment information is determined based on medication information and dosage rules. Medication adjustment information can be medication warning information, for example, generating a warning message "Patient is using a certain medication at a dosage exceeding the recommended dose by 75%, dosage adjustment is recommended" when a patient is using a certain medication at a dosage exceeding the recommended dose; medication adjustment information can also be medication recommendations, such as a step-down dosing regimen.

[0096] Figure 2 This is a schematic diagram of the structure of a risk determination device 20 provided for an exemplary embodiment of this disclosure. (See diagram below.) Figure 2 As shown, the risk determination device 20 includes: an acquisition module 21, a prediction module 22, a determination module 23, and a display module 24.

[0097] Among them, the acquisition module 21 is used to acquire the user's health status information, medication information, chronobiological characteristics and drug molecular diagram structural characteristics;

[0098] Prediction module 22 is used to input health status information, medication information, time biological characteristics and drug molecular diagram structural characteristics into the health risk prediction model to obtain health risk prediction results;

[0099] Module 23 is used to determine medication adjustment information based on medication information and medication dosage rules when the health risk prediction result indicates that there is a risk.

[0100] The display module 24 is used to send the health risk prediction results and medication adjustment information to the early warning terminal so that the early warning terminal can display the risk prediction results and medication adjustment information.

[0101] Optionally, the health risk prediction model includes a traditional machine learning module, a deep learning module, and an ensemble module. The prediction module 22, when inputting health status information, medication information, chronobiological characteristics, and drug molecular diagram structural characteristics into the health risk prediction model to obtain health risk prediction results, is used for:

[0102] Within the health risk prediction model, health status information, medication information, chronobiological characteristics, and drug molecular diagram structural characteristics are input into the health risk prediction model to obtain at least one first prediction result.

[0103] Health status information, medication information, chronobiological characteristics, and drug molecular diagram structural characteristics are input into the deep learning module to obtain the second prediction result;

[0104] The first and second prediction results are input into the integrated module to obtain the health risk prediction results.

[0105] Optionally, the traditional machine learning module includes: support vector machine, random forest, and gradient boosting decision tree; the prediction module 22, when inputting health status information, medication information, chronobiological characteristics, and drug molecule graph structure characteristics into the health risk prediction model and obtaining at least one first prediction result, is used for:

[0106] Health status information, medication information, chronobiological characteristics, and drug molecular graph structure features are input into a support vector machine to obtain a first prediction result;

[0107] By inputting health status information, medication information, chronobiological characteristics, and drug molecular graph structure characteristics into a random forest, a first prediction result is obtained.

[0108] Health status information, medication information, chronobiological characteristics, and drug molecular graph structural characteristics are input into a gradient boosting decision tree to obtain a first prediction result.

[0109] Optionally, the deep learning module includes: a graph neural network, a recurrent neural network, and an attention mechanism network; the prediction module 22, when inputting health status information, medication information, chronobiological characteristics, and drug molecule graph structure characteristics into the deep learning module to obtain a second prediction result, is used for:

[0110] The structural features of the drug molecule graph are input into a graph neural network to obtain the drug feature vector;

[0111] By inputting chronobiological features into a recurrent neural network, temporal features are obtained;

[0112] Health status information, medication information, chronobiological characteristics, drug molecular graph structural characteristics, drug feature vectors, and temporal characteristics are input into the attention mechanism network to obtain the second prediction result.

[0113] Optionally, when the prediction module 22 inputs the first prediction result and the second prediction result into the integration module to obtain the health risk prediction result, it is used to:

[0114] The first and second prediction results are input into the integration module, and a weighted vote is performed on the first and second prediction results to obtain the health risk prediction result.

[0115] Optionally, the prediction module 22 can also be used for:

[0116] Construct a drug risk knowledge graph based on drug name, clinical symptoms, mechanism of action, and risk factors;

[0117] Based on the drug risk knowledge graph, risk warning rules and medication dosage rules are determined.

[0118] Optionally, the prediction module 22 can also be used for:

[0119] Based on health status information, medication information, and chronobiological characteristics, determine whether the risk warning rules have been triggered;

[0120] If the risk warning rules are met, an alert action will be taken.

[0121] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0122] Figure 3 This is a schematic diagram of the structure of an electronic device provided as an exemplary embodiment of the present disclosure. For example... Figure 3 As shown, the electronic device includes a memory 31 and a processor 32. Additionally, the electronic device also includes a power supply component 33 and a communication component 34.

[0123] Memory 31 is used to store computer programs and can be configured to store various other data to support operation on the electronic device. Examples of this data include instructions for any application or method used to operate on the electronic device.

[0124] The memory 31 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0125] Communication component 34 is used for data transmission with other devices.

[0126] The processor 32 can execute computer instructions stored in the memory 31 to: acquire the user's health status information, medication information, chronobiological characteristics, and drug molecular diagram structural characteristics; input the health status information, medication information, chronobiological characteristics, and drug molecular diagram structural characteristics into a health risk prediction model to obtain a health risk prediction result; if the health risk prediction result indicates the existence of risk, determine medication adjustment information based on medication information and medication dosage rules; and send the health risk prediction result and medication adjustment information to an early warning terminal for the early warning terminal to display the risk prediction result and medication adjustment information.

[0127] Accordingly, embodiments of this disclosure also provide a computer-readable storage medium storing a computer program. When the computer-readable storage medium stores a computer program, and the computer program is executed by one or more processors, it causes one or more processors to perform... Figure 1 Each step in the method embodiment.

[0128] Accordingly, embodiments of this disclosure also provide a computer program product, which includes a computer program / instructions that are executed by a processor. Figure 1 Each step in the method embodiment.

[0129] The above Figure 3The communication component is configured to facilitate wired or wireless communication between the device containing the communication component and other devices. The device containing the communication component can access wireless networks based on communication standards, such as WiFi, 2G, 3G, 4G / LTE, 5G, or combinations thereof. In one exemplary embodiment, the communication component receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, the communication component also includes a Near Field Communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on Radio Frequency Identification (RFID), Infrared Data Association (IrDA) technology, Ultra-Wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0130] The above Figure 3 The power supply component provides power to the various components of the device in which it resides. The power supply component may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device in which it resides.

[0131] The aforementioned electronic devices also include a display screen and audio components.

[0132] The display includes a screen, which may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen can be implemented as a touchscreen to receive input signals from a user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors can sense not only the boundaries of touch or swipe actions but also the duration and pressure associated with the touch or swipe operation.

[0133] An audio component may be configured to output and / or input audio signals. For example, the audio component includes a microphone (MIC) configured to receive external audio signals when the device containing the audio component is in an operating mode, such as call mode, recording mode, or voice recognition mode. The received audio signals may be further stored in memory or transmitted via a communication component. In some embodiments, the audio component also includes a speaker for outputting audio signals.

[0134] In the embodiments of the apparatus, device, storage medium, and computer program products disclosed herein, user health status information, medication information, chronobiological characteristics, and drug molecular graph structure characteristics are acquired; these information are input into a health risk prediction model to obtain a health risk prediction result; if the health risk prediction result indicates the presence of risk, medication adjustment information is determined based on medication information and dosage rules; the health risk prediction result and medication adjustment information are sent to an early warning terminal for display. This disclosure, based on user health status information, medication information, chronobiological characteristics, and drug molecular graph structure characteristics, utilizes a health risk prediction model to accurately predict health risks, improving the efficiency and accuracy of health risk prediction.

[0135] Those skilled in the art will understand that embodiments of this disclosure can be provided as methods, systems, or computer program products. Therefore, this disclosure can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this disclosure can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0136] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0137] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0138] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0139] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0140] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0141] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0142] 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. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0143] The above are merely specific embodiments of this disclosure, enabling those skilled in the art to understand or implement this disclosure. 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 this disclosure. Therefore, this disclosure is not to be limited to these embodiments, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A risk determination method, characterized in that, include: Acquire users' health status information, medication information, chronobiological characteristics, and drug molecular diagram structural characteristics; The health status information, medication information, chronobiological characteristics, and drug molecular diagram structural characteristics are input into the health risk prediction model to obtain the health risk prediction results. If the health risk prediction result indicates that there is a risk, medication adjustment information is determined based on the medication information and medication dosage rules; The health risk prediction results and the medication adjustment information are sent to the early warning terminal so that the early warning terminal can display the risk prediction results and the medication adjustment information.

2. The method according to claim 1, characterized in that, The health risk prediction model includes a traditional machine learning module, a deep learning module, and an ensemble module. The health status information, medication information, chronobiological characteristics, and drug molecular graph structural characteristics are input into the health risk prediction model to obtain health risk prediction results, including: Within the health risk prediction model, the health status information, the medication information, the chronobiological characteristics, and the drug molecular diagram structural characteristics are input into the health risk prediction model to obtain at least one first prediction result; The health status information, medication information, chronobiological characteristics, and drug molecular graph structural characteristics are input into the deep learning module to obtain a second prediction result; The first prediction result and the second prediction result are input into the integrated module to obtain the health risk prediction result.

3. The method according to claim 2, characterized in that, The traditional machine learning module includes: support vector machine, random forest, and gradient boosting decision tree; the step of inputting the health status information, the medication information, the chronobiological characteristics, and the drug molecule graph structure characteristics into the health risk prediction model to obtain at least one first prediction result includes: The health status information, medication information, time biological characteristics, and drug molecule graph structure characteristics are input into the support vector machine to obtain a first prediction result; The health status information, medication information, time biological characteristics, and drug molecule graph structure characteristics are input into the random forest to obtain a first prediction result; The health status information, medication information, chronobiological features, and drug molecular graph structural features are input into a gradient boosting decision tree to obtain a first prediction result.

4. The method according to claim 2, characterized in that, The deep learning module includes: a graph neural network, a recurrent neural network, and an attention mechanism network; the step of inputting the health status information, the medication information, the chronobiological characteristics, and the drug molecule graph structure characteristics into the deep learning module to obtain a second prediction result includes: The drug molecule graph structure features are input into the graph neural network to obtain the drug feature vector; The time-biological features are input into the recurrent neural network to obtain time-series features; The health status information, medication information, time biological characteristics, drug molecular graph structural characteristics, drug feature vector, and temporal characteristics are input into the attention mechanism network to obtain the second prediction result.

5. The method according to claim 2, characterized in that, The step of inputting the first prediction result and the second prediction result into the integration module to obtain the health risk prediction result includes: The first prediction result and the second prediction result are input into the integration module, and the first prediction result and the second prediction result are weighted and voted to obtain the health risk prediction result.

6. The method according to claim 1, characterized in that, The method further includes: Construct a drug risk knowledge graph based on drug name, clinical symptoms, mechanism of action, and risk factors; Based on the drug risk knowledge graph, risk warning rules and medication dosage rules are determined.

7. The method according to claim 6, characterized in that, The method further includes: Based on the health status information, medication information, and time biological characteristics, determine whether the risk warning rule has been triggered; If the aforementioned risk warning rules are met, an alert action will be executed.

8. A risk determination device, characterized in that, include: The acquisition module is used to acquire users' health status information, medication information, chronobiological characteristics, and drug molecular diagram structural characteristics. The prediction module is used to input the health status information, the medication information, the time biological characteristics, and the drug molecular diagram structure characteristics into the health risk prediction model to obtain the health risk prediction result; The determination module is used to determine medication adjustment information based on the medication information and medication dosage rules when the health risk prediction result indicates that there is a risk. The display module is used to send the health risk prediction results and the medication adjustment information to the early warning terminal, so that the early warning terminal can display the risk prediction results and the medication adjustment information.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to execute instructions to implement the steps of the method as described in any one of claims 1-7.

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

Images