Intelligent monitoring and early warning method for adverse reactions of antipsychotic drugs based on deep learning
Patent Information
- Application Number
- CN202610831102.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-10
- Publication Date
- 2026-09-22
AI Technical Summary
[0004]本发明的目的在于克服现有技术的不足，适应现实需要，提供基于深度学习的抗精神病药物不良反应智能监测与预警方法，以解决当前现有的抗精神病药物不良反应监测技术在进行风险评估时，往往侧重于患者当前状态数据的分析，对于联合用药产生的相互影响考虑不足，难以准确反映不同药物组合对应的不良反应风险差异，同时，对于患者既往发生的不良反应事件，多数监测方式仅进行简单记录或统计，未能结合不良反应发生时间及严重程度对历史风险进行动态分析，导致历史不良反应信息对当前风险评估的指导作用有限，从而影响不良反应风险识别的准确性及预警结果的可靠性的技术问题
[0024]1.本发明通过建立药物组合风险数据集，对不同抗精神病药物及辅助治疗药物联合使用时产生的不良反应风险进行分析，能够将联合用药因素纳入风险评估过程，提高对复杂用药场景下不良反应风险的识别能力。
Smart Images
Figure CN122800100A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of drug monitoring, specifically to a method for intelligent monitoring and early warning of adverse reactions to antipsychotic drugs based on deep learning. Background Technology
[0002] Antipsychotic drugs are widely used in the clinical treatment of schizophrenia, bipolar disorder, and other mental illnesses, playing an important role in improving patients' mental symptoms and controlling disease progression. With the continuous advancement of medical informatization, medical institutions have gradually established electronic medical record systems, laboratory testing systems, and pharmaceutical management systems, and use technologies such as machine learning, data mining, and deep learning to analyze patient medication data in order to monitor, assess, and warn of adverse drug reactions. Current technologies typically collect data such as patients' medication records, test indicators, vital signs, and medical records, and combine them with risk assessment models to identify and monitor possible adverse reactions during the use of antipsychotic drugs.
[0003] However, existing adverse reaction monitoring technologies for antipsychotic drugs often focus on analyzing the patient's current condition data when conducting risk assessments, failing to adequately consider the interactions resulting from combined medication use. This makes it difficult to accurately reflect the differences in adverse reaction risks associated with different drug combinations. Furthermore, for past adverse reaction events, most monitoring methods only perform simple recording or statistical analysis, neglecting to dynamically analyze historical risks by combining the timing and severity of adverse reactions. This limits the guiding role of historical adverse reaction information in current risk assessments, thus affecting the accuracy of adverse reaction risk identification and the reliability of early warning results. Therefore, new technical solutions are needed to address these issues. Summary of the Invention
[0004] The purpose of this invention is to overcome the shortcomings of existing technologies, adapt to practical needs, and provide a deep learning-based intelligent monitoring and early warning method for adverse reactions to antipsychotic drugs. This addresses the technical problems that current antipsychotic drug adverse reaction monitoring technologies often focus on analyzing the patient's current state data during risk assessment, failing to adequately consider the mutual influence of combined medications, and thus struggling to accurately reflect the differences in adverse reaction risks corresponding to different drug combinations. Furthermore, for patients' past adverse reaction events, most monitoring methods only perform simple recording or statistics, failing to dynamically analyze historical risks by combining the time and severity of adverse reactions. This results in limited guidance of historical adverse reaction information for current risk assessment, thereby affecting the accuracy of adverse reaction risk identification and the reliability of early warning results.
[0005] To achieve the objectives of this invention, the technical solution adopted is as follows: A deep learning-based intelligent monitoring and early warning method for adverse reactions to antipsychotic drugs is designed, comprising the following steps:
[0006] S1. Obtain the medical data of the target patient, wherein the medical data includes structured data and unstructured data;
[0007] S2. Preprocess the medical data and extract patient medication information, health status information, and historical adverse reaction information;
[0008] S3. Establish a drug combination risk dataset based on the combined medication use of the target patients, and determine the risk parameters corresponding to different drug combinations;
[0009] S4. Establish a historical risk dataset based on the occurrence time and severity of historical adverse events in the target patients, and determine the impact coefficient corresponding to each historical adverse event;
[0010] S5. Construct risk analysis data based on the patient's medication information, health status information, risk parameters, and impact coefficients;
[0011] S6. Input the risk analysis data into the pre-trained deep learning prediction model to obtain the adverse reaction risk score corresponding to the target patient;
[0012] S7. Determine the risk level based on the adverse reaction risk score;
[0013] S8. Output the early warning result based on the risk level.
[0014] Preferably, the structured data includes at least one of the following: patient basic information, disease diagnosis information, antipsychotic drug use information, biochemical test information, vital sign information, and past adverse reaction records.
[0015] Preferably, the unstructured data includes at least one of electronic medical records, nursing records, follow-up records, medical records, and patient complaint information.
[0016] Preferably, in step S2, unstructured data is subjected to text cleaning, word segmentation, entity recognition, and semantic encoding to extract information on changes in patient symptoms and medication use.
[0017] Preferably, the risk parameter is determined based on the correlation between drug combinations and corresponding adverse reactions in historical cases.
[0018] Preferably, the drug combination includes a combination of two or more antipsychotic drugs used together, or a combination of an antipsychotic drug and an adjuvant therapy drug used together.
[0019] Preferably, the influence coefficient is determined based on the time interval between the occurrence time of historical adverse reaction events and the current monitoring time, as well as the severity of the event.
[0020] Preferably, the risk level includes low risk, medium risk and high risk, and the risk level corresponding to the target patient is determined according to the correspondence between the adverse reaction risk score and the preset risk interval.
[0021] Preferably, the early warning result includes the type of adverse reaction, the risk level, and risk warning information, wherein the risk warning information is used to indicate the current risk of adverse reactions for the target patient.
[0022] Preferably, the deep learning prediction model is constructed using a long short-term memory network, a convolutional neural network, a Transformer network, or a combination of the above networks.
[0023] Compared with the prior art, the beneficial effects of the present invention are as follows:
[0024] 1. This invention establishes a drug combination risk dataset to analyze the adverse reaction risks arising from the combined use of different antipsychotic drugs and adjuvant therapy drugs. It can incorporate combination drug factors into the risk assessment process and improve the ability to identify adverse reaction risks in complex medication scenarios.
[0025] 2. This invention determines the impact coefficient by combining the occurrence time and severity of historical adverse events, enabling historical risk information to participate in current risk assessment and dynamically reflect changes in patient risk, thereby improving the match between risk assessment results and the patient's actual condition.
[0026] 3. This invention uses patient medication information, health status information, drug combination risk parameters, and historical adverse reaction impact coefficients as analytical bases, and employs a deep learning model for comprehensive evaluation. This enables graded early warning of the risk of adverse drug reactions in patients, improving the accuracy and reliability of early warning results and providing auxiliary reference for clinical medication safety management. Attached Figure Description
[0027] Figure 1 This is a schematic diagram of the overall process of the present invention;
[0028] Figure 2 This is a schematic diagram of the data preprocessing and information extraction process of the present invention;
[0029] Figure 3 This is a schematic diagram illustrating the risk analysis data construction and prediction process of the present invention;
[0030] Figure 4 This is a schematic diagram illustrating the risk level classification and early warning result output of the present invention. Detailed Implementation
[0031] The present invention will be further described below with reference to the accompanying drawings and embodiments:
[0032] A deep learning-based intelligent monitoring and early warning method for adverse reactions to antipsychotic drugs, see [link to article]. Figures 1 to 4 This includes the following steps:
[0033] S1. Obtain the target patient's medical data, which includes structured and unstructured data;
[0034] S2. Preprocess medical data and extract patient medication information, health status information, and historical adverse reaction information;
[0035] S3. Establish a drug combination risk dataset based on the combined medication use of the target patients, and determine the risk parameters corresponding to different drug combinations;
[0036] S4. Establish a historical risk dataset based on the occurrence time and severity of historical adverse events in the target patients, and determine the impact coefficient corresponding to each historical adverse event;
[0037] S5. Construct risk analysis data based on patient medication information, health status information, risk parameters, and impact coefficients;
[0038] S6. Input the risk analysis data into the pre-trained deep learning prediction model to obtain the adverse reaction risk score corresponding to the target patient;
[0039] S7. Determine the risk level based on the adverse reaction risk score;
[0040] S8. Output early warning results based on risk level.
[0041] Specifically, during use, the medical data of the target patient is first collected. The medical data includes structured and unstructured data. The structured data includes the patient's basic information, disease diagnosis information, antipsychotic drug use information, biochemical test information, vital signs information, and past adverse reaction records. The unstructured data includes electronic medical records, nursing records, follow-up records, diagnosis and treatment records, and patient complaints. After the data collection is completed, the medical data is preprocessed to extract the patient's medication information, health status information, and historical adverse reaction information.
[0042] After obtaining relevant information, a drug combination risk dataset is established based on the patient's current combined medication situation. The corresponding risk parameters are determined by combining the correlation between different drug combinations and adverse reactions in historical cases. When a patient is using two or more antipsychotic drugs in combination, or when an antipsychotic drug is used in combination with an adjuvant therapy drug, the corresponding risk parameters reflect the impact of different drug combinations on the risk of adverse reactions, so that the combined medication factor can participate in the subsequent risk assessment.
[0043] At the same time, the adverse reaction events that patients have experienced in the past are compiled, and a historical risk dataset is established based on the occurrence time and severity of the adverse reaction events. The corresponding impact coefficient is determined. The closer the historical adverse reaction event is to the current monitoring time, or the higher the severity of the adverse reaction, the greater its corresponding impact. Conversely, the corresponding impact decreases. By incorporating historical adverse reaction information into the analysis process, the risk assessment can be comprehensively judged in combination with the patient's past medication history.
[0044] Subsequently, patient medication information, health status information, risk parameters corresponding to drug combinations, and impact coefficients corresponding to historical adverse reaction events are integrated to form risk analysis data. The risk analysis data is then input into a pre-trained deep learning prediction model for analysis and processing to obtain the adverse reaction risk score for the target patient.
[0045] The risk level is determined based on the correspondence between the obtained risk score and the preset risk range. The risk level includes low risk, medium risk and high risk. Finally, the warning result is output, which includes the type of adverse reaction, the risk level and risk warning information, so that medical staff can understand the possible medication risks of patients in a timely manner and provide a reference for subsequent monitoring and management.
[0046] For details, see Figure 2 Structured data includes at least one of the following: basic patient information, disease diagnosis information, antipsychotic drug use information, biochemical test information, vital signs information, and past adverse reaction records.
[0047] For more details, see Figure 2 Unstructured data includes at least one of the following: electronic medical records, nursing records, follow-up records, medical records, and patient complaints.
[0048] Further, see Figure 2 In step S2, unstructured data is cleaned, segmented, identified as entities, and semantically encoded to extract information on changes in patient symptoms and medication use.
[0049] Further, see Figure 3 Risk parameters are determined based on the association between drug combinations and corresponding adverse reactions in historical cases.
[0050] It is worth noting that, see Figure 1 Drug combinations include combinations of two or more antipsychotic drugs used together, or combinations of antipsychotic drugs used together with adjuvant therapy drugs.
[0051] It is worth noting that, see Figure 3The impact coefficient is determined based on the time interval between the occurrence time of historical adverse reaction events and the current monitoring time, as well as the severity of the event.
[0052] It is worth mentioning that, see Figure 4 The risk levels include low risk, medium risk, and high risk. The risk level of the target patient is determined based on the correspondence between the adverse reaction risk score and the preset risk range.
[0053] It is worth noting that, see Figure 4 The warning results include the type of adverse reaction, the risk level, and risk warning information, among which the risk warning information is used to inform the target patient of the current risk of adverse reaction.
[0054] It is worth emphasizing that, see Figure 3 Deep learning prediction models are constructed using long short-term memory networks, convolutional neural networks, Transformer networks, or combinations of these networks.
[0055] Example 1
[0056] The present invention will be described using the monitoring of adverse reactions to antipsychotic drugs in hospitalized patients in a psychiatric hospital as an example.
[0057] First, acquire the medical data of the target patients. The structured data includes the patient's age, gender, disease diagnosis information, medication records, biochemical test results, vital signs information, and past adverse reaction records. The unstructured data includes electronic medical records, nursing records, ward round records, and patient chief complaints.
[0058] The patient is a 42-year-old male diagnosed with schizophrenia. He has been receiving antipsychotic medication for a long time, and his current medication regimen is a combination of olanzapine and sodium valproate. After data collection, the structured data was organized, and the electronic medical records, nursing records and patient complaints were processed to extract information such as the patient's recent weight changes, sleep patterns, diet, and medication adjustments.
[0059] Subsequently, a drug combination risk dataset was established based on the patient's current combined medication use, and the risk parameters corresponding to the drug combination were determined based on the occurrence of adverse reactions corresponding to the combined use of olanzapine and sodium valproate in historical cases.
[0060] Meanwhile, the patient's past adverse reaction records were analyzed. The patient had experienced weight gain six months ago and elevated blood glucose levels two months ago. The corresponding impact coefficients were determined based on the time interval between the occurrence time of each adverse reaction event and the current monitoring time, as well as the severity of the adverse reaction, and a historical risk dataset was established.
[0061] Subsequently, patient medication information, health status information, risk parameters corresponding to drug combinations, and impact coefficients corresponding to historical adverse events were integrated to form risk analysis data. The risk analysis data was then input into a pre-trained deep learning prediction model for analysis, resulting in a risk score of 0.82 for the patient to develop metabolic-related adverse reactions.
[0062] Risk levels are determined based on a pre-defined risk range: a risk score below 0.40 corresponds to low risk, a risk score between 0.40 and 0.70 corresponds to medium risk, and a risk score above 0.70 corresponds to high risk. Since this patient's risk score is 0.82, their risk level is determined to be high risk.
[0063] Finally, an early warning result is generated based on the determined risk level. The early warning result shows that the patient has a high risk of metabolic-related adverse reactions, and also provides information on the risk of continuous weight gain and abnormal blood sugar, so that medical staff can keep abreast of the patient's medication risk and strengthen the monitoring of relevant indicators.
[0064] Through the above implementation process, factors related to combined medication and historical adverse reactions are both involved in risk assessment, enabling the obtained adverse reaction risk score to more accurately reflect the patient's current risk status and improve the accuracy and reliability of the early warning results.
[0065] In addition, all components designed in this invention are general standard parts or components known to those skilled in the art. Their structure and principle can be known to those skilled in the art through technical manuals or conventional experimental methods. Those skilled in the art can fully implement them, so there is no need to elaborate. The content protected by this invention does not involve improvements to the internal structure and method.
Claims
1. A method for intelligent monitoring and early warning of adverse reactions to antipsychotic drugs based on deep learning, characterized in that, Includes the following steps: S1. Obtain the medical data of the target patient, wherein the medical data includes structured data and unstructured data; S2. Preprocess the medical data and extract patient medication information, health status information, and historical adverse reaction information; S3. Establish a drug combination risk dataset based on the combined medication use of the target patients, and determine the risk parameters corresponding to different drug combinations; S4. Establish a historical risk dataset based on the occurrence time and severity of historical adverse events in the target patients, and determine the impact coefficient corresponding to each historical adverse event; S5. Construct risk analysis data based on the patient's medication information, health status information, risk parameters, and impact coefficients; S6. Input the risk analysis data into the pre-trained deep learning prediction model to obtain the adverse reaction risk score corresponding to the target patient; S7. Determine the risk level based on the adverse reaction risk score; S8. Output the early warning result based on the risk level.
2. The method for intelligent monitoring and early warning of adverse reactions to antipsychotic drugs based on deep learning according to claim 1, characterized in that, The structured data includes at least one of the following: basic patient information, disease diagnosis information, antipsychotic drug use information, biochemical test information, vital signs information, and past adverse reaction records.
3. The method for intelligent monitoring and early warning of adverse reactions to antipsychotic drugs based on deep learning according to claim 1, characterized in that, The unstructured data includes at least one of electronic medical records, nursing records, follow-up records, medical records, and patient complaints.
4. The method for intelligent monitoring and early warning of adverse reactions to antipsychotic drugs based on deep learning according to claim 1, characterized in that, In step S2, unstructured data is cleaned, segmented, identified as entities, and semantically encoded to extract information on changes in patient symptoms and medication use.
5. The method for intelligent monitoring and early warning of adverse reactions to antipsychotic drugs based on deep learning according to claim 1, characterized in that, The risk parameters are determined based on the correlation between drug combinations and corresponding adverse reactions in historical cases.
6. The method for intelligent monitoring and early warning of adverse reactions to antipsychotic drugs based on deep learning according to claim 1, characterized in that, The drug combination includes combinations of two or more antipsychotic drugs used together, or combinations of antipsychotic drugs used together with adjuvant therapy drugs.
7. The method for intelligent monitoring and early warning of adverse reactions to antipsychotic drugs based on deep learning according to claim 1, characterized in that, The impact coefficient is determined based on the time interval between the occurrence time of historical adverse reaction events and the current monitoring time, as well as the severity of the event.
8. The method for intelligent monitoring and early warning of adverse reactions to antipsychotic drugs based on deep learning according to claim 1, characterized in that, The risk levels include low risk, medium risk, and high risk. The risk level of the target patient is determined based on the correspondence between the adverse reaction risk score and the preset risk interval.
9. The method for intelligent monitoring and early warning of adverse reactions to antipsychotic drugs based on deep learning according to claim 1, characterized in that, The early warning results include the type of adverse reaction, the risk level, and risk warning information, wherein the risk warning information is used to indicate the current risk of adverse reactions for the target patient.
10. The method for intelligent monitoring and early warning of adverse reactions to antipsychotic drugs based on deep learning according to claim 1, characterized in that, The deep learning prediction model is constructed using a long short-term memory network, a convolutional neural network, a Transformer network, or a combination of the above networks.