Dynamic electronic prescription generation method and system based on artificial intelligence

Abstract

The invention relates to the technical field of electronic prescriptions, in particular to a dynamic electronic prescription generation method and system based on artificial intelligence. The method comprises the following steps: collecting the latest physiological state parameter flow of a patient, carrying out real-time health state evaluation, and generating a personalized patient state map; historical medical records of a patient are extracted, time sequence pathological evolution tracking is carried out, and a pathological evolution trajectory is generated; performing multi-parameter time sequence difference comparison calculation and drug curative effect quantitative evaluation based on the personalized patient state map and the pathological evolution trajectory to obtain a curative effect evaluation report; performing allergic drug identification on the patient based on the historical medical record of the patient, and performing secondary drug screening to obtain a drug candidate set; and performing intelligent matching calculation on the drug candidate set according to the curative effect evaluation report, and performing combinatorial optimization analysis to generate a final effective electronic prescription. The electronic prescription is automatically updated and adjusted based on the state change of the patient, the risk of allergic prescriptions is reduced, and the safety of the prescriptions is improved.

Description

Technical Field

[0001] This invention relates to the field of electronic prescription technology, and in particular to a method and system for generating dynamic electronic prescriptions based on artificial intelligence. Background Technology

[0002] With the continuous development of medical informatization and artificial intelligence technologies, medical services are gradually evolving from the traditional static diagnosis and treatment model towards personalization, intelligence, and dynamism. Electronic prescriptions, as a crucial bridge connecting doctors' diagnoses and drug treatment, directly impact the scientific rigor of clinical decision-making and the precision of patient treatment through their level of intelligence. Especially in emerging medical scenarios such as chronic disease management, telemedicine, and elderly health care, traditional static electronic prescriptions are no longer sufficient to meet the dynamic treatment needs arising from real-time changes in patients' health status.

[0003] Compared to traditional prescription systems, dynamic electronic prescription systems not only rely on the doctor's initial diagnosis but also need to automatically adjust drug dosage, administration time, and compatibility based on real-time physiological changes in the patient during treatment, such as fluctuations in multiple parameters like heart rate, blood pressure, body temperature, blood sugar, and respiratory rate. This new prescription generation model requires the system to have efficient physiological data processing capabilities, accurate health status assessment models, and flexible prescription recommendation mechanisms. However, traditional electronic prescription systems are mostly based on fixed templates and rule bases, lacking the ability to perceive and respond to real-time changes in the patient's health status, making it difficult to meet the needs of individualized dynamic adjustments. Furthermore, modern clinical practice faces the challenge of accessing and processing massive amounts of heterogeneous health data, such as data from wearable devices, remote monitoring systems, and electronic health records (EHRs), with inconsistent formats and frequent updates, placing higher demands on the system's data fusion, anomaly detection, and trend analysis capabilities. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes a dynamic electronic prescription generation method and system based on artificial intelligence, thereby resolving at least one of the aforementioned technical problems.

[0005] To achieve the above objectives, the present invention provides a dynamic electronic prescription generation method based on artificial intelligence, comprising the following steps: Step S1: Collect the patient's latest physiological status parameter stream, conduct real-time health status assessment, and generate a personalized patient status map; Step S2: Extract the patient's historical medical records, perform time-series pathological evolution tracking, and generate a pathological evolution trajectory; Step S3: Based on the personalized patient status map and pathological evolution trajectory, perform multi-parameter temporal difference comparison calculation and quantitative evaluation of drug efficacy to obtain an efficacy evaluation report; Step S4: Identify the patient's allergic drugs based on the patient's medical history and perform secondary drug screening to obtain a drug candidate set; Step S5: Based on the efficacy evaluation report, perform intelligent matching calculations on the drug candidate set and conduct combination optimization analysis to generate the final effective electronic prescription.

[0006] This specification provides an artificial intelligence-based dynamic electronic prescription generation system for executing the artificial intelligence-based dynamic electronic prescription generation method described above, including: The health status assessment module is used to collect the latest physiological status parameter streams of patients, perform real-time health status assessments, and generate personalized patient status maps. The pathological evolution module is used to extract patients' historical medical records, track the temporal pathological evolution, and generate pathological evolution trajectories. The efficacy assessment module is used to perform multi-parameter time-series difference comparison calculations based on personalized patient status atlases and pathological evolution trajectories, and to conduct quantitative assessments of drug efficacy to obtain efficacy assessment reports. The drug screening module is used to identify patients' allergic drug information based on their historical medical records and to perform secondary drug screening to obtain a drug candidate set. The drug combination module is used to intelligently match and calculate the drug candidate set based on the efficacy evaluation report, perform combination optimization analysis, and generate the final effective electronic prescription.

[0007] The beneficial effects of this invention are specifically as follows: It enables dynamic real-time monitoring of patients' health status, avoiding reliance solely on static physical examination results. By using multi-source physiological parameters (such as heart rate, blood pressure, blood oxygen, blood glucose, and respiratory rate) to form a personalized health status map, it improves adaptability to individual differences. It can promptly detect potential abnormal changes in patients, providing accurate real-time health baseline data for subsequent prescription decisions. By comprehensively utilizing historical electronic medical records, previous examinations, and treatment records, it forms a temporal evolution model of the patient's disease. Through trajectory-based display, it can reveal the potential development trend and long-term risk points of the disease. It helps to combine real-time health data with medical history evolution, enabling prescription recommendations to consider not only the current state but also the dynamics of the disease course and long-term health management. By introducing multi-parameter difference analysis, it compares real-time health parameters with pathological trajectories, quantifying the actual efficacy and sensitivity of patients to different drugs. It avoids the "one-size-fits-all" approach of traditional prescriptions based solely on group clinical experience, achieving accurate assessment of individual patient efficacy. It generates objective and quantifiable efficacy reports, providing reliable reference for doctors and AI decision-making modules. By automatically identifying patients' allergy history and adverse drug reactions through medical record analysis, the risk of allergic prescriptions is reduced. A secondary screening of the drug library eliminates potentially dangerous drugs, ensuring the safety and individual suitability of recommended medications. A candidate drug set is generated, providing a high-safety benchmark for subsequent optimization and reducing medical risks. Combining efficacy assessment with the candidate drug set, AI algorithms are used for personalized drug matching and combination optimization to improve treatment outcomes. Optimization can be performed based on different disease states, combined drug mechanisms, and drug interactions, avoiding conflicts and duplications found in traditional manual prescriptions. Finally, a dynamic electronic prescription is generated, supporting automatic updates and adjustments as the patient's condition changes, achieving true precision medicine and intelligent healthcare. Attached Figure Description

[0008] Figure 1 This is a schematic diagram of the steps of a dynamic electronic prescription generation method based on artificial intelligence according to the present invention; Figure 2 This is a detailed flowchart illustrating the implementation steps of step S1. Figure 3 This is a flowchart illustrating the detailed implementation steps of step S2. Detailed Implementation

[0009] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0010] This application provides a method and system for generating dynamic electronic prescriptions based on artificial intelligence. The executing entities of the method and system include, but are not limited to, mechanical equipment, data processing platforms, cloud server nodes, and network upload devices that can be considered general computing nodes in this application. The data processing platform includes, but is not limited to, at least one of an audio-visual management system, an information management system, and a cloud-based data management system.

[0011] Please see Figures 1 to 3 This invention provides a dynamic electronic prescription generation method based on artificial intelligence, comprising the following steps: Step S1: Collect the patient's latest physiological status parameter stream, conduct real-time health status assessment, and generate a personalized patient status map; Step S2: Extract the patient's historical medical records, perform time-series pathological evolution tracking, and generate a pathological evolution trajectory; Step S3: Based on the personalized patient status map and pathological evolution trajectory, perform multi-parameter temporal difference comparison calculation and quantitative evaluation of drug efficacy to obtain an efficacy evaluation report; Step S4: Identify the patient's allergic drugs based on the patient's medical history and perform secondary drug screening to obtain a drug candidate set; Step S5: Based on the efficacy evaluation report, perform intelligent matching calculations on the drug candidate set and conduct combination optimization analysis to generate the final effective electronic prescription.

[0012] In the embodiments of the present invention, see Figure 1 The diagram below illustrates the steps of a dynamic electronic prescription generation method based on artificial intelligence according to the present invention. In this example, the steps of the dynamic electronic prescription generation method based on artificial intelligence include: Step S1: Collect the patient's latest physiological status parameter stream, conduct real-time health status assessment, and generate a personalized patient status map; In this embodiment, the latest physiological parameters of the patient are collected, including electrocardiogram signals (sampling frequency 500Hz), blood oxygen saturation (accuracy ±2%), core body temperature (accuracy ±0.1°C), blood pressure fluctuation (accuracy ±3mmHg), respiratory rate (range 5-40 breaths / minute), and skin conductance (measurement range 0-100μS), etc., as multidimensional continuous physiological monitoring data. The collected physiological parameter stream is preprocessed, and the Kalman filter algorithm is used to remove signal noise. The filtering window is set to 30 seconds, the smoothing coefficient is 0.85, and timestamp calibration is performed to ensure that the data synchronization accuracy is within 10 milliseconds. The multidimensional physiological parameters are normalized using the Z-score normalization method to construct a 15-dimensional normalization model. A standardized physiological data matrix of ×1440 is used, where 15 represents the types of monitoring parameters and 1440 corresponds to minute-level sampling over 24 hours. Features are extracted from the standardized data matrix using a Long Short-Term Memory Neural Network (LSTM), with 128 hidden layer nodes, a learning rate of 0.001, and 200 training epochs to identify patient physiological state patterns and health risk indicators. Combining static features such as patient age, gender, BMI (normal range 18.5-24.9), and basal metabolic rate, a multimodal fusion algorithm is used to construct a personalized patient status map that includes physiological activity scores (0-100 points), health stability index (0-1 range), and disease risk warning levels (1-5 levels).

[0013] Step S2: Extract the patient's historical medical records, perform time-series pathological evolution tracking, and generate a pathological evolution trajectory; In this embodiment, complete historical medical record data of patients is extracted from the Hospital Information System (HIS), including structured and unstructured medical data such as outpatient records, inpatient medical records, laboratory reports, and imaging examinations from the past 5 years. The total amount of data is typically in the range of 50-500MB. Medical text semantic analysis technology based on a BERT pre-trained model is used to perform deep semantic parsing of the medical record text. The word vector dimension is set to 768, the maximum sequence length is 512, and the accuracy of disease entity identification reaches 94.2%. Key medical information such as symptom descriptions, diagnostic conclusions, and treatment plans are extracted. Using medical knowledge graph and ontology mapping technology, the extracted medical information is standardized into ICD-10 disease codes. A unified medical data representation format was established using disease coding and ATC drug classification coding. Based on time series analysis, a time window of 30 days and a step size of 7 days were set to track the temporal changes in the patient's disease development. The fluctuation trends of biochemical indicators (such as blood glucose change rate ±15mg / dL, liver function enzyme value fluctuation ±20U / L) and changes in symptom severity scores (using a 0-10 level scoring standard) were analyzed. A multidimensional temporal pathological evolution trajectory was constructed, including key parameters such as the disease progression rate index (dP / dt, where P is the pathological severity), treatment response time constant (usually 3-14 days), and duration of the recovery stabilization period, forming an individualized pathological evolution trajectory map for each patient.

[0014] Step S3: Based on the personalized patient status map and pathological evolution trajectory, perform multi-parameter temporal difference comparison calculation and quantitative evaluation of drug efficacy to obtain an efficacy evaluation report; In this embodiment, a multi-dimensional time-series comparative analysis is performed based on the current personalized patient status profile and the historical pathological evolution trajectory. The magnitude and rate of change of key physiological indicators are calculated, and a significance threshold of p < 0.05 is set. Paired t-tests are used to analyze the statistical significance of parameter differences. A efficacy evaluation model is constructed, defining the efficacy improvement rate. The IR (Intensity Reduction) calculation formula is IR = (S0 - St) / S0 × 100%, where S0 is the baseline symptom score, St is the post-treatment score, and the IR value ranges from 0 to 100%, with a value greater than 30% considered a significant improvement. A pharmacokinetic-pharmacodynamic (PK-PD) joint model was used to analyze the blood drug concentration-time curves of previous medications, calculating key parameters such as drug half-life (t1 / 2), clearance rate (CL), and volume of distribution (Vd). t1 / 2 typically ranges from 1 to 24 hours, and CL is 0.5-2.0 L / h / kg. A machine learning regression algorithm was used to establish a drug efficacy prediction model. A random forest algorithm was used, with 100 decision trees and a maximum depth of 10. The training data included over 1000 patient samples, and the model's prediction accuracy reached 87.3%. A comprehensive analysis was conducted on drug efficacy indicators (efficacy score: 0-1), safety index (0-10), and patient tolerability (tolerance rate). (0-100%), generating a comprehensive efficacy assessment report that includes quantitative efficacy evaluation, side effect risk rating, and pharmacoeconomic analysis.

[0015] Step S4: Identify the patient's allergic drugs based on the patient's medical history and perform secondary drug screening to obtain a drug candidate set; In this embodiment, drug allergy information is extracted from the patient's medical history. Named Entity Recognition (NER) technology is used to identify the names of allergenic drugs, achieving an accuracy rate of 96.8%. Simultaneously, allergic reaction types (such as rash, dyspnea, anaphylactic shock, etc.) and severity levels (1-4) are extracted. A personalized drug contraindication database is established for each patient, including drug chemical structure similarity analysis. A molecular fingerprint similarity threshold of 0.85 is set to identify structurally similar potential sensitizing drugs, expanding the scope of safety screening. Drug metabolism capacity is assessed based on the patient's liver and kidney function test results, with liver function indicators including ALT (normal value <50 U / L), AST (normal value <40 U / L), and total bilirubin (normal value <17.1 μmol / L). The renal function indicators include serum creatinine (normal range 44-133 μmol / L), blood urea nitrogen (normal range 3.6-9.5 mmol / L), and glomerular filtration rate (normal range >90 mL / min / 1.73 m²). Safety screening is conducted using a drug interaction database covering the interactions of over 1500 commonly used drugs to identify drug combinations that may produce synergistic toxicity, antagonistic effects, or metabolic competition. A multi-level screening strategy is employed: first, drugs with known allergies are excluded (exclusion rate approximately 8-12%); second, drugs with metabolic pathway conflicts are screened out (exclusion rate approximately 5-8%); and finally, positive screening is performed based on indication matching (matching threshold >0.7), ultimately resulting in a safe drug candidate set containing 15-25 candidate drugs.

[0016] Step S5: Based on the efficacy evaluation report, perform intelligent matching calculations on the drug candidate set and conduct combination optimization analysis to generate the final effective electronic prescription.

[0017] In this embodiment, based on key indicators in the efficacy evaluation report, including expected efficacy improvement (target value > 50%), side effect risk rating (acceptable range 1-3), and treatment adherence prediction (target value > 80%), intelligent matching calculation is performed on the drug candidate set. A multi-objective optimization function is constructed, using a weighted summation method to integrate efficacy weight (w1=0.4), safety weight (w2=0.3), cost-effectiveness weight (w3=0.2), and convenience weight (w4=0.1). The objective function is F=w1×E+w2×S+w3×C+w4×U, where E is the efficacy score, S is the safety score, C is the cost-effectiveness score, and U is the ease of use score. A genetic algorithm is used for drug combination optimization, with a population size of 50, a crossover probability of 0.8, a mutation probability of 0.1, and an evolutionary generation of 1. In the 2000s, the optimal drug combination is sought; the optimal dosage is calculated based on pharmacokinetic modeling, taking into account individual differences such as patient weight (usually 50-100kg), age (affecting metabolic rate by 10-30%), and liver and kidney function (affecting clearance by 20-50%) to determine a personalized dosage range; the dosing schedule is optimized using the principles of chronopharmacology, analyzing factors such as the time to peak drug absorption (Tmax usually 0.5-4 hours), the influence of circadian rhythms (cortisol peak time 6-8 am), and the patient's sleep patterns to develop a dosing schedule accurate to the hour; finally, a detailed personalized electronic prescription is generated, including drug name, strength and dosage (accurate to milligrams), dosing frequency (1-4 times daily), treatment duration (usually 7-28 days), and medication precautions, with the prescription validity period set at 30 days.

[0018] In this embodiment, see Figure 2 The diagram below illustrates the detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include: The latest physiological state parameter stream of the patient is collected, including real-time electrocardiogram signal, blood oxygen saturation, body temperature change, blood pressure fluctuation, respiratory rate and skin conductance signal; The physiological state parameter stream is identified by parameter source timestamps and then time-series aligned to obtain a time-series synchronized parameter stream. Anomaly detection and adaptive noise filtering are performed on the timing synchronization parameter stream to generate a filtered and optimized parameter stream. Perform multi-parameter variation trend analysis on the filter optimization parameter flow to generate multiple parameter trend curves; Real-time health status assessment is performed on multiple parameter trend curves to generate personalized patient status maps.

[0019] In this embodiment, a real-time acquisition channel for multi-dimensional vital signs is established. This relies on the collaborative operation of multiple types of physiological sensors and monitoring devices, including an electrocardiogram (ECG) monitoring module (for real-time acquisition of ECG signals) and a photoplethysmography (PPG) sensor (for monitoring blood oxygen saturation). The system includes a high-precision temperature sensor (for acquiring body temperature change curves), a continuous non-invasive blood pressure monitoring device (based on pulse wave conduction time or cuff measurement methods), a respiratory rate sensor (achieved through thoracic impedance measurement or airflow detection), and a skin conductance sensor (to reflect the intensity of sympathetic nerve activity). During data acquisition, a reasonable sampling rate needs to be set according to the characteristics of different signals. For example, the sampling frequency for electrocardiogram (ECG) signals is typically set at 250–500 Hz to ensure the capture of key features such as the QRS complex; a blood oxygen saturation update frequency of around 1 Hz is sufficient for clinical needs; body temperature change signals can be recorded at intervals of 30 seconds to 1 minute; continuous blood pressure monitoring can achieve non-invasive measurements at a frequency of once per minute or higher; respiratory rate signals are recommended to be sampled at 10 Hz for frequency domain analysis; and skin conductance signals are often sampled at 5–10 Hz to capture rapid changes in emotional stress. All sensor data is transmitted to a unified data access module via wearable devices or bedside monitoring, with precise time stamps to ensure the integrity, continuity, and traceability of the original physiological data. Due to differences in physical operating principles, sampling frequencies, and signal transmission delays among various sensors, raw data streams often cannot be perfectly matched on the timeline. Direct use can lead to distortions in the comparison and fusion of different parameters. Therefore, a time stamp identification and timing alignment method is needed to incorporate various parameters into a unified time reference. A unified time synchronization mechanism should be established for all acquisition devices, using a high-precision clock to timestamp each set of data at the millisecond level (global synchronization can be achieved through NTP or PTP protocols). Subsequently, data at different sampling rates should be interpolated and resampled. For example, ECG signals can be directly used as a high-frequency reference, while low-sampling-rate blood pressure or body temperature data require methods such as linear interpolation, spline interpolation, or Kalman filtering prediction to fill in missing points and maintain a time correspondence with the high-frequency signal. To avoid alignment deviations caused by short-term data loss, a time sliding window mechanism can be used, such as a sliding window in 5-second increments, to ensure that all parameters are sampled at least once within that time period.

[0020] First, abnormal parameters are detected, and then the signal is optimized using adaptive noise filtering techniques. Anomaly detection can be divided into two categories: threshold-based detection and pattern recognition-based detection. The former directly marks anomalies when the heart rate exceeds 200 beats / min or falls below 30 beats / min, and considers oxygen saturation below 85% or above 100% as invalid values. The latter can use statistical modeling methods (such as Gaussian distribution models) or machine learning anomaly detection methods (such as One-Class SVM or Isolation Forest) to identify atypical waveforms and anomalies. For noise processing, adaptive filters (such as adaptive filters based on LMS or RLS algorithms) can dynamically adjust filtering parameters according to the characteristics of the input signal, effectively suppressing time-varying noise. In ECG signal processing, wavelet decomposition and reconstruction techniques can be combined to remove high-frequency interference while preserving key waveform features. For low-frequency signals such as skin conductance and respiratory rate, low-pass filters with a cutoff frequency of 0.5–1 Hz can be used to smooth jitter. After this step, noise and artifacts are significantly reduced, resulting in a higher signal-to-noise ratio in the "filtered optimized parameter stream." Time series signals are smoothed, for example using moving averages or exponentially weighted moving averages (EWMA), to reduce noise interference from short-term fluctuations. Subsequently, time series forecasting and modeling methods are applied to estimate trends in key parameters. For instance, ARIMA models are suitable for linear trend prediction, while deep learning methods based on recurrent neural networks (LSTM) are better suited for capturing nonlinear dynamic patterns. For example, by modeling the RR interval sequence of electrocardiogram signals, heart rate variability (HRV) can be extracted to assess autonomic nervous system regulation; in the joint analysis of blood oxygenation and respiratory rate, the potential association between hypoxic events and respiratory instability can be identified; and in the interactive analysis of skin conductance and body temperature trends, the relationship between stress response and changes in metabolic levels can be discovered. The time window for trend curves is typically set to 30 seconds to 5 minutes to balance real-time performance and trend stability. Furthermore, correlation analysis and mutual information calculations can identify potential coupling effects between different parameters. For example, sharp fluctuations in blood pressure accompanied by a decrease in heart rate variability often indicate an abnormal increase in sympathetic nervous system activity.

[0021] This typically relies on multi-parameter fusion algorithms derived from artificial intelligence. Common methods include Bayesian networks based on probabilistic graphical models to handle causal relationships between multiple parameters; fuzzy logic-based reasoning to transform continuous physiological data into qualitative health levels; and deep learning-based multimodal fusion models (such as Transformer structures) to handle complex nonlinear interactions. When determining a patient's condition, it's not enough to rely on a single indicator; multiple trends must be considered for a comprehensive assessment. For example, a persistently elevated heart rate accompanied by decreased blood oxygen saturation and increased skin conductance may indicate anxiety or hypoxia risk; drastic blood pressure fluctuations and abnormal ECG waveforms may suggest a cardiovascular emergency. The resulting "personalized patient condition atlas" is a visualized, multi-dimensional health status description tool, usually presented as curves, color blocks, and risk level markers, integrating dynamic information from parameters such as ECG, blood pressure, blood oxygen saturation, respiration, body temperature, and skin conductance into a panoramic health profile.

[0022] In this embodiment, see Figure 3 The diagram below illustrates the detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include: Identify the patient's electronic identity information and retrieve the patient's historical medical records; Deep semantic analysis is performed on the patient's medical history to extract the patient's pathological information; the patient's pathological information includes past disease records, diagnosis results, examination reports, disease type, symptom description and treatment plan; The patient's historical medical records are extracted in a structured manner to obtain the historical consultation records; The pathological evolution of patients' pathological information and historical consultation records is tracked over time to generate a pathological evolution trajectory.

[0023] In this embodiment, patient electronic identity information is typically stored in a Medical Information System (HIS), Electronic Health Record (EHR), or a regional medical big data platform. The identity verification process generally relies on a multi-factor authentication mechanism, such as cross-verification using unique identifiers like ID card numbers, medical insurance card numbers, and hospital visit card numbers combined with facial recognition or fingerprint recognition. To ensure information security and privacy compliance, an access protocol based on encrypted transmission and a permission control system are required to ensure the identity verification process complies with the Personal Information Protection Law and medical data security regulations. After successful identity matching, the patient's historical medical records are automatically retrieved from the EHR database, including outpatient records, inpatient records, examination and test results, imaging data, and previous prescription data. Due to the diverse sources of medical records, data formats may differ between structured (e.g., diagnostic codes), semi-structured (e.g., examination report forms), and unstructured (e.g., free text written by doctors). Therefore, a unified data interface standard needs to be established at this stage, such as using the HL7 FHIR protocol for medical record exchange, to ensure the consistency and integrity of data extraction. This process relies on Natural Language Processing (NLP) technology, combined with medical knowledge graphs and clinical terminology databases (such as ICD-10 and SNOMED-CT). The semantic parsing process involves multiple stages: First, medical text preprocessing is performed, including word segmentation, stop word removal, and spell correction. Then, a Named Entity Recognition (NER) model is used to identify medical entities in the text, such as disease names, symptom descriptions, examination items, and medication information. Next, a relation extraction model is used to establish relationships between different entities, such as "disease-symptom," "examination-result," and "treatment-response." Further, Semantic Role Labeling (SRL) technology is used to extract the semantic structure of sentences, thereby clarifying the pathological event of "a patient being diagnosed with a certain disease and receiving a certain treatment at a certain time." Through this process, the extracted pathological information includes past medical records (such as hypertension and diabetes), definitive diagnoses (such as cerebral infarction), examination report results (such as elevated blood pressure and abnormal blood sugar), disease types (chronic and acute), symptom descriptions (such as chest tightness and shortness of breath), and historical treatment plans (such as drug therapy and surgical intervention).

[0024] In addition to pathological information, medical records contain a large amount of consultation records, usually in the form of question-and-answer sessions between doctors and patients, such as "Patient's chief complaint: headache for three days" or "Present illness: accompanied by fever, no vomiting." This type of data is of great value for disease tracking and prescription recommendation, but its original form is mostly free text, making it difficult to use directly. Therefore, structured extraction of consultation records is necessary. Using an intent recognition model in natural language processing, different consultation modules are distinguished, such as chief complaint, present illness, past medical history, allergy history, and family history. Then, information extraction methods are used to structure and label the key information in these modules. For example, the chief complaint can be extracted as "symptoms + duration"; the present illness can be transformed into "time series + symptom changes"; and the allergy history is directly extracted as "allergenic drugs + reaction type." To ensure the medical accuracy of the structured results, normalization is often performed using a medical terminology standard library, for example, mapping "dizziness" and "vertigo" to the same standard symptom entry. The final generated "historical consultation records" exist in the form of structured tables or knowledge units, with corresponding relationships between timestamps, symptoms, signs, doctor's questions, and patient's answers. The extracted pathological information is sorted and merged with the consultation records according to timestamps to form a continuous sequence of pathological events. For example, a patient was diagnosed with hypertension five years ago, developed heart failure three years ago, and underwent coronary stent implantation one year ago; these key milestones need to be arranged sequentially along a timeline. Simultaneously, a correlation is established between pathological information and consultation records, such as an evolutionary chain of "symptom-diagnosis-treatment." To better describe the dynamics of the disease course, time series modeling methods can be used for pattern recognition of the evolutionary trajectory, such as using Hidden Markov Models (HMMs) to identify implicit transition patterns in the disease course, or using Recurrent Neural Networks (RNNs, LSTMs) to predict the possible future trajectory of the disease. Furthermore, visualization techniques can be used to display the pathological trajectory in the form of a timeline or atlas, allowing doctors to clearly see the complete process of the disease from initial symptoms to multiple follow-up visits and treatment adjustments.

[0025] In this embodiment, step S3 includes the following steps: Multi-parameter temporal difference comparison calculations were performed based on personalized patient status atlases and pathological evolution trajectories to extract the difference values ​​and variation amplitudes of different physiological parameters; Based on the differences and the magnitude of change, the direction of parameter change is analyzed to extract physiological improvement indicators and physiological deterioration indicators. Collect patient feedback information, perform subjective state analysis, and generate patient feedback state characteristics; Based on patient feedback status characteristics, physiological improvement indicators, and physiological deterioration indicators, a recovery trend analysis was conducted, and the efficacy of the drugs was quantitatively evaluated to obtain an efficacy evaluation report.

[0026] In this embodiment, physiological parameters at different time periods are compared, such as electrocardiogram characteristics (heart rate, HRV), dynamic blood pressure levels (systolic blood pressure, diastolic blood pressure, and pulse pressure), and blood oxygen saturation (…). The parameters considered included average values ​​and the number of hypoxic events, respiratory rate stability, body temperature variation, and skin conductance. The calculation method employed time-series differencing and statistical comparison to extract the differences in each parameter at different time points. Further, a sliding window mechanism (e.g., 5-minute, 30-minute, 24-hour windows) was used to calculate the variation range of the parameters. To ensure accuracy, normalization was used to make parameters with different dimensions comparable. For example, blood pressure variation was expressed in mmHg, while blood oxygen variation was expressed as a percentage; after standardization, these could be uniformly used for difference analysis. Directional judgment rules were set for each parameter. For example, if the heart rate gradually returned to the normal range (60–100 beats / min) and the HRV value increased, it indicated improved autonomic nervous function, which could be used as an indicator of improvement. If blood pressure decreased to the target range (e.g., from 160 / 100 mmHg to 130 / 80 mmHg in a hypertensive patient), it was also considered an improvement. Similarly, a sustained increase in blood oxygen saturation and a decrease in hypoxic events were also indicators of improvement. Conversely, if there are repeated increases in blood pressure, a drop in blood oxygen saturation below 90%, a persistently rapid respiratory rate, or worsening abnormalities in the electrocardiogram waveform, these should be marked as indicators of deterioration. To improve reliability, a multi-parameter joint judgment method can be used, that is, simultaneously referring to the changes in two or more indicators to confirm the trend. For example, "elevated blood pressure + increased heart rate" together indicate an increased cardiovascular burden.

[0027] Subjective feedback from patients is incorporated. This feedback can be collected through various methods, including self-reporting modules on wearable devices, symptom scoring questionnaires on mobile applications, and voice input via bedside interaction. Content typically covers pain levels, sleep quality, fatigue, emotional state, and perceived drug side effects. To ensure the calculability of subjective information, natural language processing and quantification are required. For example, a patient's input "Recently, my dizziness has significantly improved" can be interpreted as symptom relief and assigned a corresponding improvement weight; while "I feel nauseous after taking medication" is transformed into a side effect characteristic of adverse drug reactions. Sentiment analysis and semantic classification methods can be used to categorize feedback into positive, neutral, or negative states, while simultaneously quantifying pain or discomfort using a Likert scale (e.g., 0–10 score). The resulting "patient feedback status characteristics" include the direction of subjective symptom change (improvement or worsening), the intensity of adverse reactions, and an emotional state index. By fusing objective physiological improvement / deterioration indicators with subjective feedback status characteristics, a comprehensive assessment of the patient's recovery trend and a quantitative evaluation of drug efficacy can be achieved. The physiological parameters showing improvement and deterioration are weighted and calculated, with weights assigned according to the importance of different indicators. For example, blood pressure and electrocardiogram indicators have higher weights in cardiovascular disease prescriptions, while blood oxygen and respiratory rate are more critical in respiratory disease prescriptions. Subsequently, patient feedback characteristics are incorporated into the evaluation system, and a multidimensional scoring model is used to integrate subjective and objective indicators. For instance, an evaluation matrix is ​​constructed using the Analytic Hierarchy Process (AHP) to quantify physiological indicators and patient subjective feelings into a unified health improvement score. Furthermore, time series analysis is used to assess the overall recovery trend, determining whether the patient exhibits a pattern of gradual improvement, fluctuating recovery, or continuous deterioration. Regarding drug efficacy evaluation, the differences in parameters before and after drug use are compared with feedback characteristics to quantify the comprehensive effects of the drug in symptom relief, sign improvement, and side effects. The final output, the "Efficacy Evaluation Report," is presented in the form of visual charts and text descriptions, including key improvement indicators, key deterioration indicators, a summary of the patient's subjective state, drug efficacy scores, and trend predictions.

[0028] In this embodiment, step S4 includes the following steps: Identify patients' drug allergies based on their medical history and construct a drug contraindication list. Extract the most recent prescription drug information from the patient's medical history, including drug type, dosage, frequency of use, and course of treatment; Based on the prescription drug information and pathological evolution trajectory, drug metabolism capacity is quantified to generate a drug metabolism capacity value. Dynamic drug adaptability analysis is performed based on drug metabolism capacity values ​​and drug contraindication lists, followed by secondary drug screening to obtain a drug candidate set.

[0029] In this embodiment, allergy information is extracted from the patient's medical history, and a personalized drug contraindication list is constructed based on this information. The patient's medical history typically contains explicit records of allergies, such as "penicillin allergy," "rash reaction to cephalosporins," or "asthma attack with aspirin." This information may be distributed in structured fields (such as allergy history forms), semi-structured documents (such as medical record summaries), or unstructured free text (such as physician progress notes). The identification methods include natural language processing technology and medical terminology database matching. First, named entity recognition (NER) is used to identify entities involving "drug name + allergic reaction." Then, standard medical lexicons (such as ATC drug classification) are used for drug normalization, unifying drug names expressed in different ways, for example, unifying "amoxicillin" and "penicillin-like drugs." Next, the corresponding allergy reaction types (such as rash, anaphylactic shock, gastrointestinal reactions) are extracted and mapped to severity levels. Finally, a "drug contraindication list" is automatically generated, including the names of prohibited drugs, associated reaction categories, and risk levels. The most recent prescription information is extracted from the patient's medical history. Prescriptions typically contain four core elements: drug type, dosage, frequency of use, and duration of treatment. Since prescription data in medical records may be stored in structured fields (such as electronic prescriptions) or exist as scanned copies or text descriptions, different extraction methods are required. For structured data, the prescription form can be read directly; for unstructured text, key information must be extracted using text parsing methods. For example, "Atorvastatin 20mg orally, once nightly, for 30 days" can be parsed as: drug type = atorvastatin, dosage = 20mg, frequency of use = once daily, duration of treatment = 30 days. If the prescription involves multiple drugs, parameters for each drug must be extracted separately, maintaining the corresponding relationships. After extraction, a "most recent prescription drug information set" is generated. This information set is not only input data for analyzing drug efficacy and metabolism but also an important reference for assessing patient medication adherence and subsequent prescription adjustments.

[0030] This approach matches a patient's pathological evolution trajectory (disease progression, organ function changes, treatment history, etc.) with the metabolic pathways of currently prescribed drugs. First, based on the drug's pharmacokinetic characteristics (ADME: absorption, distribution, metabolism, excretion), key metabolic pathways and enzymes are identified, such as CYP450 family enzymes and UGT enzymes. Then, combined with liver and kidney function indicators from the pathological evolution trajectory (such as serum creatinine and liver function enzyme levels), comorbidities (such as chronic kidney disease and cirrhosis), and past drug responses, the patient's metabolic capacity deviation is estimated. For example, in patients with impaired liver function, their CYP3A4 metabolic pathway may be impaired, thereby reducing the clearance rate of certain drugs. Through mathematical modeling or machine learning prediction models, these factors can be integrated into a "drug metabolism capacity value," which represents an individual's metabolic efficiency as a quantitative value (such as a metabolic coefficient between 0 and 1). A higher value indicates normal or enhanced metabolic capacity, while a lower value suggests weakened metabolism or even potential accumulation risk. Contraindicated drugs are first excluded, and then drugs suitable for the patient's metabolic characteristics are screened based on the metabolic capacity value. A drug suitability scoring model was constructed. The model's inputs included drug pharmacokinetic parameters (half-life, metabolic enzyme dependence, clearance pathway), the patient's metabolic capacity, a contraindication list, and treatment guidelines for the target disease. The model output was a "drug suitability index," used to assess the suitability of a particular drug for the current patient. For example, if a patient had a low CYP2D6 metabolic capacity, the suitability score for drugs dependent on that pathway would be lowered, and alternative drugs with lower metabolic dependence would be recommended. Simultaneously, drugs on the contraindication list were directly excluded, even if they had high suitability scores, and would not be included in the candidate set. Finally, the "drug candidate set," obtained after secondary screening, was a dynamically optimized list of medications, including not only drug types but also recommended dosage ranges and usage precautions.

[0031] In this embodiment, the specific steps of step S5 are as follows: Define multi-dimensional solution indices, including efficacy improvement, side effect control, patient satisfaction, and treatment adherence; Based on the multi-dimensional solution index and efficacy evaluation report, the drug candidate set is intelligently matched and calculated, and combined optimization analysis is performed to generate multiple drug combinations. Multiple drug combinations are simulated and validated to generate drug efficacy simulation data; Based on drug efficacy simulation data, drug parameters are optimized to generate personalized electronic prescriptions; The drug parameter optimization specifically involves: calculating the optimal combination of drug types, the optimal dosage ratio, and the ideal dosing time interval, and optimizing the drug dosage distribution and dosing sequence. Personalized electronic prescriptions are uploaded to the cloud for review. Once the doctor's review and approval are detected, a final valid electronic prescription with a digital signature is generated.

[0032] In this embodiment, an evaluation system is established, namely, a multi-dimensional solution index is defined. This index is used to comprehensively evaluate the merits of drug combinations, thereby providing an evaluation basis for subsequent intelligent matching and optimization. The solution index generally includes four dimensions: First, efficacy improvement, that is, the effect of the drug on improving symptoms and reversing the course of the target disease, which can be measured by the improvement of key clinical indicators (such as the degree of blood pressure reduction, blood glucose control range, and cardiac function improvement level); second, side effect control, that is, the degree of control of adverse reactions of the drug on the body, which needs to be quantified by the probability of past side effects, the severity level of adverse events, and individual patient risk assessment; third, patient satisfaction, which is derived from the patient's subjective feedback and quality of life assessment, such as symptom relief, drug tolerance, and the degree of impact on daily life; fourth, treatment adherence, that is, the likelihood of the patient taking the medication on time and in the correct dosage, which is closely related to the method of administration (oral, injection), frequency of administration (once or multiple times daily), and the complexity of the drug. For ease of calculation, these dimensions are quantified into standardized scores, such as a range of 0–100 points, and integrated through weighted coefficients. Different disease types or patient needs will determine the different weights of each dimension. These indicators are used to intelligently match and optimize the drug candidate set. The drug candidate set, derived from previous metabolic adaptability analysis and contraindication screening, already possesses high safety and adaptability. At this point, intelligent computational methods are needed to evaluate the comprehensive performance of candidate drugs in terms of efficacy, side effects, satisfaction, and compliance, and to generate multiple combination schemes. Specifically, a multi-objective optimization model is established, using four solution indices as objective functions, and heuristic or evolutionary algorithms are employed for optimization. For example, genetic algorithms (GA), particle swarm optimization (PSO), or simulated annealing algorithms can be used to generate drug combinations that meet the balance of multi-dimensional indicators. During the combination optimization process, drug interactions (such as synergistic and antagonistic effects), dose-addition effects, and treatment guideline constraints are considered. For example, antihypertensive treatment may require the combination of diuretics, calcium channel blockers, and ACE inhibitors, but functional duplication or metabolic conflicts must be avoided. The final result is not a single scheme, but multiple feasible combinations, each with a comprehensive score and optimization path.

[0033] The core of simulation validation is to construct a virtual medication environment based on pharmacokinetic (PK) and pharmacodynamic (PD) models to simulate drug concentration changes, metabolic processes, and pharmacodynamic effects in the body. Specifically, the absorption, distribution, metabolism, and excretion parameters of the drug are first input into the model, such as half-life, oral bioavailability, protein binding rate, and hepatic and renal clearance. Then, combined with individual patient parameters (such as age, weight, and liver and kidney function) and pathological evolution trajectories, the concentration-time curve of the drug in the patient's body is calculated. Next, the PD model is used to establish a functional relationship between drug concentration and pharmacodynamic response, such as the dose-response curve of blood pressure reduction versus drug concentration. During the simulation, drug interactions are also considered, such as metabolic competition or synergistic enhancement that may occur during combined drug use. The final output is a set of "pharmacodynamic usage simulation data," including drug blood concentration curves, pharmacodynamic response curves, side effect risk curves, and overall efficacy predictions.

[0034] In drug combination optimization, the drug regimen with the highest overall score, balancing efficacy and safety, is selected based on simulation results. Secondly, regarding dosage optimization, iterative calculations of blood drug concentrations and pharmacodynamic responses at different doses are used to find the optimal dosage range with the lowest side effects. For example, if a drug shows limited efficacy improvement at a 20mg dose but significantly increased side effects at 40mg, then 30mg might be recommended as the optimal dose. Thirdly, in terms of dosing timing, the ideal dosing interval is calculated based on the drug's half-life and duration of effect to ensure blood drug concentrations remain within the therapeutic window while avoiding excessive peak-to-trough effects. For example, adjusting a certain type of drug to be taken once daily at night can improve compliance and reduce side effects. The generated personalized electronic prescription not only includes the drug name and dosage but also specifies the dosing time, treatment duration, and precautions.

[0035] In this embodiment, the specific steps for uploading the personalized electronic prescription to the cloud for review, and generating a final valid electronic prescription with a digital signature when the doctor's review and confirmation are detected, are as follows: Upload personalized electronic prescriptions to the cloud for review; Doctors conduct a full prescription security check via cloud review and digitally authenticate and sign the prescription. Once the doctor's review and confirmation are detected, a final valid electronic prescription with a digital signature is generated. The final valid electronic prescription is stored and transmitted using blockchain encryption.

[0036] In this embodiment, during the upload process, the prescription content is structured and organized, including basic patient information (a unique identifier after anonymization), a drug list (type, dosage, frequency, and course of treatment), medication precautions, and generated personalized parameter descriptions. To ensure data transmission security, an encrypted transmission protocol (such as TLS 1.3) is used to encrypt the prescription file throughout the process, combined with an identity authentication mechanism (such as OAuth 2.0 or certificate-based two-way authentication) to verify the legitimacy of the upload source. The uploaded prescription is stored in a cloud-based review database, which typically employs a highly available architecture (such as distributed storage and multi-replica mechanisms) to ensure that data is not lost during transmission and storage. Simultaneously, the cloud performs standardized verification of the prescription data, such as checking whether drug codes conform to the ATC classification standard and whether dosages and units comply with national pharmacopoeia specifications. Personalized prescriptions uploaded to the cloud require manual review by doctors to ensure medical safety and legal compliance. When doctors access the prescription through cloud review, they are first provided with an automated review report, including drug contraindication matching results, drug interaction alerts, dosage range compliance assessments, and treatment course rationality analysis. When reviewing prescriptions, doctors can refer to the prompts to make a final confirmation of the safety and suitability of the medication. If any issues are found, the doctor can directly modify the prescription; otherwise, it passes review and is digitally signed. Digital signatures are typically implemented using Public Key Infrastructure (PKI) technology. The doctor's private key is used to generate the prescription signature, and cloud-based review and pharmacies can verify the signature using the doctor's public key. The signature not only ensures the integrity and immutability of the prescription during transmission but also confirms that the prescription indeed comes from a legitimate doctor. Only when the doctor clicks "review and confirm" will a "final valid electronic prescription" with a digital signature be generated. After the doctor reviews and completes the digital signature, the prescription enters the blockchain storage and transmission stage to further enhance its security and traceability. The characteristics of blockchain—decentralization, tamper-proofing, and traceability—make it ideal for storing sensitive medical data. Specifically, the information digest of the final valid electronic prescription (generated using a hash function such as SHA-256) is written to the blockchain, while the complete plaintext data of the prescription is stored in a distributed database or in the cloud to ensure a balance between performance and privacy. Each prescription upload generates a unique block record containing the prescription hash, doctor's digital signature, public key authentication information, and a timestamp. This ensures that any modification to the prescription will result in a change in the hash value, allowing for immediate detection of tampering. During transmission, a peer-to-peer transmission protocol of the blockchain network ensures data replication and synchronization across multiple nodes, achieving network-wide consistency. Pharmacies, health insurance institutions, or regulatory authorities can verify the authenticity and integrity of prescriptions by querying the blockchain upon receipt.This not only ensures the security of electronic prescriptions during cross-institutional circulation, but also enables full-process tracking and accountability confirmation, ultimately forming a reliable electronic prescription circulation and management system.

[0037] In this embodiment, an artificial intelligence-based dynamic electronic prescription generation system is provided for executing the artificial intelligence-based dynamic electronic prescription generation method described above, including: The health status assessment module is used to collect the latest physiological status parameter streams of patients, perform real-time health status assessments, and generate personalized patient status maps. The pathological evolution module is used to extract patients' historical medical records, track the temporal pathological evolution, and generate pathological evolution trajectories. The efficacy assessment module is used to perform multi-parameter time-series difference comparison calculations based on personalized patient status atlases and pathological evolution trajectories, and to conduct quantitative assessments of drug efficacy to obtain efficacy assessment reports. The drug screening module is used to identify patients' allergic drug information based on their historical medical records and to perform secondary drug screening to obtain a drug candidate set. The drug combination module is used to intelligently match and calculate the drug candidate set based on the efficacy evaluation report, perform combination optimization analysis, and generate the final effective electronic prescription.

[0038] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0039] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein are implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present 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 of the invention herein.

Claims

1. An artificial intelligence-based dynamic electronic prescription generation method, characterized by, Comprise the following steps: Step S1: Collect the latest physiological state parameter flow of the patient, perform real-time health status assessment, and generate a personalized patient state atlas; Step S2: Extract the patient's medical history, track the time sequence of pathological evolution, and generate a pathological evolution track; Step S3: Based on the personalized patient state atlas and the pathological evolution track, perform multi-parameter time sequence difference comparison calculation and drug efficacy quantitative evaluation, and obtain the efficacy evaluation report; Step S4: Based on the patient's medical history, identify the patient's allergic drugs, and perform secondary drug screening to obtain a drug candidate set; Step S5: According to the efficacy evaluation report, the drug candidate set is intelligently matched and calculated, and combined optimization analysis is performed to generate the final effective electronic prescription. 2.The dynamic electronic prescription generation method based on artificial intelligence according to claim 1, wherein, The specific steps of step S1 are: Collect the latest physiological state parameter flow of the patient, which includes real-time electrocardiogram signal, blood oxygen saturation, body temperature change, blood pressure fluctuation, respiratory rate and skin conductance signal; Perform parameter source timestamp identification on the physiological state parameter flow and perform time sequence alignment processing to obtain a time sequence synchronized parameter flow; Perform abnormal parameter detection and adaptive noise filtering on the time sequence synchronized parameter flow to generate a filtered optimized parameter flow; Perform multi-parameter change trend analysis on the filtered optimized parameter flow to generate multiple parameter trend curves; Perform real-time health status assessment on the multiple parameter trend curves to generate a personalized patient state atlas. 3.The dynamic electronic prescription generation method based on artificial intelligence according to claim 1, wherein, The specific steps of step S2 are: Identify the patient's electronic identity information and extract the patient's medical history; Perform deep semantic analysis on the patient's medical history to extract the patient's pathological information; The patient's pathological information includes past disease records, diagnosis results, examination reports, disease types, symptom descriptions, and treatment plans; Structure the patient's medical history to obtain historical inquiry records; Perform time sequence pathological evolution tracking on the patient's pathological information and historical inquiry records to generate a pathological evolution track. 4.The dynamic electronic prescription generation method based on artificial intelligence according to claim 1, wherein, The specific steps of step S3 are: Based on the personalized patient state atlas and the pathological evolution track, perform multi-parameter time sequence difference comparison calculation, extract the difference value and change amplitude of different physiological parameters; Perform parameter change direction analysis based on the difference value and change amplitude, extract physiological improvement indicators and physiological deterioration indicators; Collect patient feedback information, perform subjective state analysis, and generate patient feedback state features; Based on the patient feedback state features, physiological improvement indicators and physiological deterioration indicators, perform recovery trend analysis and drug efficacy quantitative evaluation to obtain the efficacy evaluation report. 5.The dynamic electronic prescription generation method based on artificial intelligence according to claim 1, wherein, The specific steps of step S4 are: Based on the patient's medical history, identify the patient's allergic drugs, and construct a drug contraindication list; Extract the latest prescription drug information from the patient's medical history, including drug type, dosage specification, medication frequency and treatment course; Based on the prescription drug information and the pathological evolution track, perform drug metabolism capacity quantification to generate a drug metabolism capacity value; Based on the drug metabolism capacity value and the drug contraindication list, perform dynamic drug adaptability analysis and secondary drug screening to obtain a drug candidate set. 6.The dynamic electronic prescription generation method based on artificial intelligence according to claim 1, wherein, The specific steps of step S5 are: Define a multi-dimensional solution index, including efficacy improvement, side effect control, patient satisfaction and treatment compliance; According to the multi-dimensional solution index and the efficacy evaluation report, the intelligent matching calculation is performed on the drug candidate set, and combination optimization analysis is performed to generate multiple drug combinations; The drug administration scheme simulation verification is performed on the multiple drug combinations to generate pharmacodynamic use simulation data; According to the pharmacodynamic use simulation data, the drug parameter optimization is performed to generate an individualized electronic prescription; The drug parameter optimization specifically includes: calculating the optimal drug species combination, the optimal dose ratio, the ideal drug administration time interval, optimizing the drug dose distribution and the drug administration time sequence arrangement; The individualized electronic prescription is uploaded to a cloud audit system, and when it is detected that the doctor's audit is confirmed, a final effective electronic prescription with a digital signature is generated. 7.The dynamic electronic prescription generation method based on artificial intelligence according to claim 6, wherein, The specific steps of uploading the individualized electronic prescription to the cloud audit system and generating the final effective electronic prescription with the digital signature when the doctor's audit is confirmed are as follows: The individualized electronic prescription is uploaded to the cloud audit system; The doctor performs a full prescription safety check through the cloud audit system and performs digital authentication signature, and when it is detected that the doctor's audit is confirmed, a final effective electronic prescription with a digital signature is generated; The final effective electronic prescription is stored and transmitted by block chain encryption.

8. An artificial intelligence based dynamic electronic prescription generation system characterized in that, The method for generating an electronic prescription based on artificial intelligence is used to perform the method according to claim 1, comprising: a health status evaluation module for collecting the latest physiological state parameter flow of a patient, performing real-time health status evaluation, and generating an individualized patient state atlas; a pathological evolution module for extracting a patient's historical medical record, performing time-series pathological evolution tracking, and generating a pathological evolution track; an efficacy evaluation module for performing multi-parameter time-series difference comparison calculation and drug efficacy quantitative evaluation based on the individualized patient state atlas and the pathological evolution track to obtain an efficacy evaluation report; a drug screening module for identifying patient allergic drugs based on a patient's historical medical record and performing secondary drug screening to obtain a drug candidate set; a drug combination module for performing intelligent matching calculation on the drug candidate set according to the efficacy evaluation report and performing combination optimization analysis to generate a final effective electronic prescription.

Images