A cerebral hemorrhage perioperative medication scheme intelligent recommendation method and system

Abstract

The application discloses a brain hemorrhage perioperative medication scheme intelligent recommendation method and system, acquires physiological parameters and operation time sequence characteristics of a patient, constructs a patient risk image, carries out risk matching with a drug knowledge base to generate a candidate drug set, extracts drug attribute characteristics to determine a drug function grouping, carries out adaptability scoring based on the patient risk image to generate a grouping adaptability matrix, carries out drug interaction analysis to identify synergistic pairs and antagonistic conflict pairs, carries out combination optimization on the synergistic pairs to generate a synergistic scheme, carries out conflict exclusion screening on the antagonistic conflict pairs to generate safety constraints, constructs a medication scheme topology, carries out dose sensitivity anomaly detection and correction to generate a correction factor to generate a time-sharing drug administration scheme, carries out real-time monitoring to carry out perfusion evaluation to determine a safety level and trigger dose adjustment, forms an optimized medication scheme, and realizes intelligent recommendation of perioperative medication.

Description

Technical Field

[0001] This invention relates to the field of medical decision support technology, and in particular to an intelligent recommendation method and system for perioperative medication regimens in cases of cerebral hemorrhage. Background Technology

[0002] Intracerebral hemorrhage is a common critical illness in neurosurgery. Its perioperative medication management involves balancing multiple treatment goals, such as hemostasis, anticoagulation, neuroprotection, and blood pressure control. Drug selection and dosage adjustment need to comprehensively consider various factors, including individual patient differences, surgical risks, and drug interactions. Traditional medication decisions mainly rely on physician experience and clinical guidelines, but in the face of complex patient conditions and diverse drug combinations, it is difficult to achieve precise individualized treatment plans.

[0003] Existing clinical decision support methods have significant shortcomings in recommending medication regimens. On the one hand, they lack a systematic assessment of patients' multidimensional risk characteristics, making it impossible to accurately match applicable drug types and functional groups. On the other hand, drug interaction analysis is not in-depth enough, particularly lacking effective optimization mechanisms for key issues such as hemostasis-anticoagulation balance and synergistic multi-target neuroprotection. Furthermore, the lack of systematic methods for planning the timing of medication and individualizing dosage adjustments at various perioperative stages makes it difficult to guarantee the safety and effectiveness of medication regimens. Therefore, there is an urgent need for an intelligent medication regimen recommendation method to provide comprehensive decision support for perioperative medication in cerebral hemorrhage. Summary of the Invention

[0004] This invention discloses an intelligent recommendation method and system for perioperative medication regimens in patients with cerebral hemorrhage. The system aims to integrate multidimensional risk characteristics of patients with drug knowledge base information, perform functional grouping and suitability assessment of candidate drugs, and deeply analyze the synergistic and antagonistic relationships between drugs. Furthermore, through combination optimization and conflict elimination, a safe and efficient medication regimen topology is constructed. Combined with dose sensitivity correction and real-time perfusion monitoring, dynamic dose adjustment is achieved, ultimately forming an individualized time-segmented dosing regimen covering the entire preoperative, intraoperative, and postoperative process. This provides accurate, safe, and dynamic intelligent recommendation support for perioperative medication decisions in patients with cerebral hemorrhage.

[0005] The first aspect of this invention proposes an intelligent recommendation method for perioperative medication regimens in cases of cerebral hemorrhage, comprising the following steps:

[0006] Obtain patient physiological parameters and surgical timing features, construct a patient risk profile based on the physiological parameters, and perform risk matching between the patient risk profile and a drug knowledge base to generate a candidate drug set;

[0007] Drug attribute features are extracted from the candidate drug set to determine drug function groups. Based on the patient risk profile, the drug function groups are assessed for suitability to generate a group suitability matrix. Drug interaction analysis is performed based on the group suitability matrix to identify synergistic and antagonistic pairs.

[0008] The synergistic pairs are combined and optimized to generate synergistic schemes, and the antagonistic pairs are eliminated and screened to generate safety constraints. The synergistic schemes and the safety constraints are used to construct a medication plan topology.

[0009] A baseline dose is obtained by calculating the dose based on the topology of the medication regimen. An abnormal dose sensitivity is detected and corrected for the baseline dose to generate a correction factor. The correction factor is then combined with the surgical timing features to generate a timed dosing regimen.

[0010] Based on the time-segmented dosing regimen, real-time monitoring is performed to obtain dynamic patient indicators. Based on the dynamic patient indicators, perfusion assessment is performed to determine the safety level. Based on the safety level, dose adjustment is triggered to form an optimized dosing regimen.

[0011] A second aspect of this invention provides an intelligent recommendation system for perioperative medication regimens in cases of cerebral hemorrhage, comprising:

[0012] The data acquisition module is used to acquire the patient's physiological parameters and surgical time sequence characteristics, construct a patient risk profile based on the physiological parameters, and perform risk matching between the patient risk profile and the drug knowledge base to generate a candidate drug set.

[0013] The drug screening module is used to extract drug attribute features from the candidate drug set to determine drug function groups, perform a suitability score on the drug function groups based on the patient risk profile to generate a group suitability matrix, and perform drug interaction analysis based on the group suitability matrix to identify synergistic pairs and antagonistic pairs.

[0014] The scheme construction module is used to combine and optimize the synergistic pairs to generate synergistic schemes, perform conflict elimination screening on the antagonistic pairs to generate safety constraints, and construct a medication scheme topology using the synergistic schemes and the safety constraints.

[0015] The dosage optimization module is used to calculate the baseline dose based on the topology of the medication regimen, perform dose sensitivity anomaly detection and correction on the baseline dose to generate a correction factor, and use the correction factor in combination with the surgical timing features to generate a timed dosing regimen.

[0016] The monitoring and adjustment module is used to perform real-time monitoring to obtain dynamic indicators of the patient based on the time-sharing dosing regimen, perform perfusion assessment based on the dynamic indicators of the patient to determine the safety level, and trigger dose adjustment to form an optimized dosing regimen according to the safety level.

[0017] The beneficial effects of this invention are reflected in the following points: 1. By constructing a multidimensional patient risk profile including bleeding risk, thrombosis risk, organ function, and surgical risk, and combining the anatomical characteristics and pathophysiological features of the bleeding site for site-specific safety assessment, and dynamically adjusting drug adaptation weights for different bleeding sites, precise drug screening based on individual patient characteristics and differences in bleeding sites is achieved. The antagonistic balance window is identified through hemostasis-anticoagulation antagonism detection, and synergistic and antagonistic conflict pairs are simultaneously identified through target complementarity analysis, avoiding the problem of solely considering efficacy while ignoring antagonistic risks, thus improving the safety and rationality of drug combinations. 2. Multi-target signaling pathway classification is performed for neuroprotective drugs. Cross-pathway synergistic effect detection identifies cross-nodes and synergistic factors between different pathways, revealing the synergistic mechanism of multi-drug combination therapy, and achieving synergistic protection covering multiple pathways such as anti-oxidative stress, anti-calcium overload, anti-inflammation, and neurorepair promotion. By extracting pharmacokinetic parameters to divide the preoperative, intraoperative, and postoperative stages, identifying temporal conflicts between drug combinations and arranging them sequentially, a medication timing topology consistent with the perioperative pathophysiological process was constructed. This avoided problems such as mismatched peak drug concentrations and superimposed antagonistic effects, improving the overall coordination of the medication regimen. 3. Dose sensitivity analysis identified high-sensitivity nodes in the dose-response relationship. Safe dose ranges were established for these nodes, and correction factors were output, enabling individualized and precise dose adjustment. This avoided efficacy fluctuations or adverse reactions caused by standard dose regimens in sensitive areas. Continuous time-series co-variance detection of cerebral perfusion pressure parameters was used to assess cerebral perfusion status. A response rate gradient was established to reflect the reserve of cerebral blood flow autoregulation function. Dose adjustment was triggered based on safety levels, achieving dynamic dose optimization based on real-time monitoring and perfusion assessment. This ensured perioperative cerebral perfusion safety and reduced the risk of secondary brain injury. Attached Figure Description

[0018] The accompanying drawings illustrate specific examples of the technical solutions described in this invention and, together with the detailed embodiments, form part of the specification, serving to explain the technical solutions, principles, and effects of this invention.

[0019] Figure 1 This is a flowchart illustrating an intelligent recommendation method for perioperative medication regimens in cerebral hemorrhage according to the present invention.

[0020] Figure 2 This is a structural block diagram of an intelligent recommendation system for perioperative medication regimens in cerebral hemorrhage according to the present invention. Detailed Implementation

[0021] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0022] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0023] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0024] The technical solutions of the embodiments of this application will be described below.

[0025] like Figure 1 As shown, this embodiment of the invention provides an intelligent recommendation method for perioperative medication regimens in cases of cerebral hemorrhage, including the following steps S110-S150:

[0026] Step S110: Obtain the patient's physiological parameters and surgical timing features, construct a patient risk profile based on the physiological parameters, and perform risk matching between the patient risk profile and the drug knowledge base to generate a candidate drug set.

[0027] Specifically, the patient's physiological parameters and surgical timing characteristics were acquired. Patient physiological parameters were collected from preoperative examination records, electronic medical records, and bedside monitoring equipment, encompassing four categories: hematological indicators, vital signs, imaging features, and medical history data. Hematological indicators included coagulation function indicators and complete blood count (CBC) indicators. Coagulation function indicators, such as platelet count, prothrombin time, activated partial thromboplastin time, and international normalized ratio (INR), reflected the patient's coagulation status. CBC indicators, such as hemoglobin concentration and hematocrit, reflected the patient's blood volume status. Vital signs included systolic blood pressure, diastolic blood pressure, heart rate, and body temperature. Systolic and diastolic blood pressure reflected circulatory system function, while heart rate reflected cardiac workload. Imaging features were extracted from head CT scan images, including hemorrhage volume, hemorrhage location, and midline shift. Hemorrhage volume was calculated using multi-level cumulative addition; hemorrhage location identified anatomical regions; and midline shift quantified the mass effect. Medical history data was extracted from electronic medical records, including history of hypertension, atrial fibrillation, prolonged bed rest, and past medication records. Surgical timing features included the planned surgical time, surgical type, and anesthesia method. The planned surgical time recorded the time interval from onset to surgery. The surgical type in the surgical timing features distinguished between craniotomy, minimally invasive surgery, and conservative treatment. The anesthesia method in the surgical timing features distinguished between general anesthesia and local anesthesia. Patient physiological parameters and surgical timing features were stored in structured data tables. Abnormal values ​​in patient physiological parameters were marked using threshold criteria; prothrombin time exceeding the upper limit of normal was marked as coagulation dysfunction, and systolic blood pressure exceeding the target range was marked as poor blood pressure control.

[0028] Patient risk profiles are constructed based on patient physiological parameters. A patient risk profile is defined as a multidimensional feature vector quantifying a patient's perioperative risk status. This feature vector includes four dimensions: bleeding risk score, thrombosis risk score, organ function score, and surgical risk level. The bleeding risk score is calculated based on coagulation function indicators among the patient's physiological parameters. The formula is: R_bleeding = α × PT / PT_reference + β × APTT / APTT_reference + γ × (PLT_reference - PLT) / PLT_reference, where R_bleeding is the bleeding risk score, α, β, and γ are weighting coefficients, PT is prothrombin time, PT_reference is the normal reference value for prothrombin time, APTT is activated partial thromboplastin time, APTT_reference is the normal reference value for activated partial thromboplastin time, and PLT is the platelet count, PLT_reference is the normal reference value for platelet count. The thrombosis risk score is calculated based on the patient's medical history and vital signs. A history of hypertension is assigned 0.3 points, a history of atrial fibrillation 0.4 points, and prolonged bed rest 0.3 points. An additional 0.2 points are added for a patient's systolic blood pressure consistently above 160 mmHg. The thrombosis risk score is the sum of the assigned values ​​for each risk factor. The organ function score integrates liver and kidney function indicators from the patient's physiological parameters. Liver function is assessed using alanine aminotransferase (ALT) and total bilirubin, while kidney function is assessed using serum creatinine and blood urea nitrogen. Higher organ function scores indicate more severe functional impairment. Surgical risk level is determined comprehensively based on the type of surgery, patient age, and American College of Anesthesiologists (ACAS) classification. Craniotomy for hematoma evacuation is classified as a high-risk procedure. The patient risk profile also includes bleeding site characteristics, extracted from imaging features of the patient's physiological parameters, including bleeding location, volume, and midline shift, and integrated with vascular distribution patterns and site risk levels in anatomical regions. The patient risk profile includes a list of risk factors, recording the specific factors that lead to elevated scores in each dimension.

[0029] A risk matching process is performed between the patient's risk profile and a drug knowledge base to generate a candidate drug set. The drug knowledge base stores the indications, contraindications, risk characteristics, and efficacy parameters of commonly used drugs for the perioperative period of intracerebral hemorrhage. It includes detailed information on several drugs, including hemostatic agents, anticoagulants, neuroprotective agents, and antihypertensive drugs. The risk matching process queries the drug knowledge base based on the patient's risk profile scores across various dimensions. The matching rules prioritize hemostatic agents when the patient's risk profile shows a high bleeding risk score, uses anticoagulants cautiously when the patient's risk profile shows a high thrombosis risk score, and prioritizes hemostatic agents such as tranexamic acid when the patient's risk profile shows a bleeding risk score exceeding a set threshold. Tranexamic acid, aminocaproic acid, and desmopressin are included in the candidate drug set among the hemostatic agents. The matching of anticoagulants needs to be considered in conjunction with the surgical sequence characteristics and assessed based on the thrombosis risk score in the patient's risk profile during the postoperative stage. The matching of neuroprotective drugs is based on the patient's surgical risk level and organ function score from the risk profile. When the surgical risk level is high and the organ function score is low, indicating good organ function, neuroprotective drugs are recommended from the drug knowledge base. Edaravone, nimodipine, and citicoline are included in the candidate drug set. The matching of antihypertensive drugs is based on the patient's blood pressure value among physiological parameters. When the systolic blood pressure exceeds the target range, antihypertensive drug matching is triggered. Nicardipine and urapidil are included in the candidate drug set from the drug knowledge base. The candidate drug set is output in the form of a drug list, which includes two fields: drug name and drug category.

[0030] Step S120: Extract drug attribute features from the candidate drug set to determine drug function groups; perform fit scoring on drug function groups based on patient risk profiles to generate a group fit matrix; and perform drug interaction analysis based on the group fit matrix to identify synergistic and antagonistic pairs.

[0031] Specifically, drug attribute features are extracted from the candidate drug set to determine drug functional groups. The candidate drug set includes several types of drugs in several categories, such as hemostatic drugs, anticoagulants, neuroprotective drugs, and antihypertensive drugs. Each drug in the drug set is associated with detailed drug attribute features in the drug knowledge base. Drug attribute features include two dimensions: pharmacological mechanism of action and target of action. The pharmacological mechanism of action describes the molecular biological process by which the drug exerts its effect, and the target of action identifies the receptor, enzyme, or ion channel that the drug binds to. Drug functional groups are clustered based on the pharmacological mechanism of action in the drug attribute features. Hemostatic drugs are further subdivided according to their mechanism of action into several functional groups, including antifibrinolytic drugs (tranexamic acid, aminocaproic acid), procoagulant drugs (desmopressin), and vasoconstrictor drugs. The target of antifibrinolytic drugs is plasminogen, and the targets of procoagulant drugs are V2 receptor and coagulation factor VIII. Neuroprotective drugs are further subdivided according to their mechanism of action into several functional groups, including free radical scavenging drugs (edaravone), calcium channel blockers (nimodipine), and neurotrophic drugs (cytidine diphosphate choline). Antihypertensive drugs are subdivided into several functional groups based on their mechanism of action, including calcium channel blockers (nicardipine) and vasodilators (urapidil). All drugs in the candidate drug set are clustered to form a hierarchical structure: the first level is the drug category, the second level is the drug functional group, and the third level is the specific drug.

[0032] In some embodiments, the step of generating a group fit matrix by performing a fit score on the drug function grouping based on the patient risk profile includes: extracting bleeding site features from the patient risk profile; performing a site-specific safety assessment on the drug function grouping based on the bleeding site features to generate site fit weights; performing risk calibration calculations using the site fit weights to generate fit values; and arranging the fit values ​​according to the grouping dimension to form a group fit matrix.

[0033] Bleeding site features are extracted from patient risk profiles. The patient risk profile includes four dimensions: bleeding risk score, thrombosis risk score, organ function score, and surgical risk level. It contains a risk factor list recording specific risk factors and bleeding site features describing the anatomical and pathological characteristics of the bleeding location. These features are extracted from the imaging information of the patient risk profile. The bleeding location identifies anatomical regions, and the anatomical characteristics of this region include vascular distribution patterns and the positional relationships of nerve nuclei. For basal ganglia hemorrhage, site-specific risks include high risk of rebleeding from perforating arteries, limited space for hemostasis, and sensitivity to neurological damage. Bleeding site features include bleeding volume and midline shift; larger bleeding volume indicates massive bleeding, and midline shift indicates a significant mass effect. Bleeding site features are stored as structured data, with data fields including bleeding location, anatomical region, vascular distribution, bleeding volume, midline shift, and site risk level. The vascular distribution information in the bleeding site features describes the number and course of perforating arteries. The basal ganglia region contains multiple perforating vessels, such as the lenticulostriate arteries. These vessels have thin walls and lack collateral circulation, making hemostasis difficult after rupture.

[0034] Site-specific safety assessments of drug functional groups based on hemorrhage site characteristics generate site-adaptation weights. The site-specific safety assessment considers the anatomical and pathophysiological features of different hemorrhage sites within the hemorrhage site characteristics, evaluating the safety and efficacy of each drug functional group at that site. For basal ganglia hemorrhage, the site-specific assessment considers the risk of rebleeding from perforating arteries and drug penetration into deep brain tissues. Antifibrinolytic drug functional groups, by inhibiting plasminogen activation and reducing thrombus dissolution, show significant effects in perforating artery hemorrhage. Procoagulant drug functional groups, by enhancing coagulation factor activity and accelerating thrombus formation, have a rapid onset of action in massive hemorrhages but carry a risk of thromboembolism. Free radical scavenging drug functional groups exhibit good penetration into deep brain tissues, protecting neurons surrounding the hemorrhage site and reducing secondary brain injury. Site-adaptation weights represent the degree of fit of each drug functional group to the specific hemorrhage site identified by the hemorrhage site characteristics. The weight values ​​range from 0 to 1, with higher values ​​indicating better fit. Antifibrinolytic drugs have a high site-fit weight, procoagulant drugs have a medium site-fit weight, free radical scavenging drugs have a high site-fit weight, and calcium channel blockers have a medium-to-high site-fit weight. The site-fit weights consider the bleeding volume and midline shift characteristics of the bleeding site. In cases of massive bleeding, the site-fit weights of procoagulant drugs are increased, and in cases of significant midline shift, the site-fit weights of neuroprotective drugs are increased. The site-fit weights are stored in vector form, with the vector dimension equal to the number of drug function groups.

[0035] Fit values ​​are generated through risk calibration calculations using site-specific weighting. The risk calibration calculation comprehensively considers site-specific weighting and various dimensions of the patient's risk profile. The calibration formula is S = w_site × (1 - α × R_risk adjustment), where S is the fit value, w_site is the site-specific weighting, α is the calibration coefficient determined through clinical data, and R_risk adjustment is the risk adjustment amount corresponding to each drug function group. The risk adjustment amount of hemostatic drug function groups is negatively correlated with the bleeding risk score; the higher the bleeding risk score, the smaller the risk adjustment amount and the higher the fit value, as hemostatic drugs are more needed when the bleeding risk is high. Hemostatic drugs with higher site-specific weighting have further enhanced fit values ​​at high bleeding risk. The risk adjustment amount of procoagulant drugs is positively correlated with the thrombosis risk score; a higher thrombosis risk score increases the risk adjustment amount of procoagulant drugs, leading to a lower fit value, as procoagulant drugs may increase thrombotic events at high thrombosis risk. Lower organ function scores indicate good function, and the risk adjustment amounts of each drug function group are small, with a positive impact on the fit value. The fit value is jointly determined by site-specific weighting and risk adjustment. Antifibrinolytic drugs have higher fit values ​​due to their high site-specific weighting and low bleeding risk adjustment, while procoagulant drugs have moderate fit values, and free radical scavenging drugs have higher fit values. Risk calibration calculations take into account the surgical timeline; the fit value for anticoagulant drugs in the preoperative stage is set to 0 to prohibit the use of these drugs.

[0036] The fit values ​​are arranged according to grouping dimensions to form a group fit matrix. The group fit matrix organizes the fit values ​​generated by risk calibration calculations in a two-dimensional table format. The rows of the group fit matrix correspond to drug function groups, and the columns correspond to different perioperative stages (preoperative, intraoperative, and postoperative). The elements of the group fit matrix are the fit values ​​of each drug function group at each stage after site fit weight calibration. Antifibrinolytic drugs have moderate fit values ​​in the preoperative stage, the highest fit values ​​in the intraoperative stage due to the highest intraoperative bleeding risk requiring strong hemostasis and the site fit weight being fully reflected in the intraoperative stage, and the fit values ​​drop back to moderate in the postoperative stage due to the reduced postoperative bleeding risk. Procoagulant drugs show a trend of first increasing and then decreasing fit values ​​in the preoperative, intraoperative, and postoperative stages. Intraoperative fit values ​​are relatively high, but are lower than antifibrinolytic drugs due to the risk regulation limitation of the thrombosis risk score. Free radical scavenging drugs showed high fit values ​​across all three stages, with the highest fit value observed during the intraoperative stage due to the most severe cerebral ischemia-reperfusion injury. Calcium channel blockers showed higher fit values ​​postoperatively than preoperatively and intraoperatively, as the increased risk of postoperative cerebral vasospasm led to an upward adjustment in site-specific fit weighting. The group fit matrix has the same number of rows as the total number of drug functional groups and the same number of columns as the number of perioperative stages. This matrix provides a quantitative fit benchmark for each functional group at each stage for subsequent drug interaction analysis.

[0037] In some embodiments, the step of identifying synergistic and antagonistic pairs based on the group fit matrix through drug interaction analysis includes: performing hemostasis-anticoagulation antagonism detection on the group fit matrix to generate antagonistic pair identifiers; extracting antagonistic balance windows from the antagonistic pair identifiers to generate balance constraint parameters; performing target complementarity analysis on the balance constraint parameters to generate synergistic markers; and identifying synergistic and antagonistic pairs based on the synergistic markers.

[0038] A hemostatic-anticoagulation antagonism test was performed on the grouping fit matrix to generate antagonistic pair identifiers. Hemostatic and anticoagulant drugs have a direct antagonistic relationship in their pharmacological effects. The functional antagonism test identifies antagonistic functional group pairs by analyzing the mechanism of action and target sites of each drug functional group in the grouping fit matrix. Antifibrinolytic drugs stabilize thrombi by inhibiting plasminogen activation, while heparin-type anticoagulants inhibit the coagulation cascade by activating antithrombin III. The opposite trends in their fit values ​​in the grouping fit matrix reflect their antagonistic relationship. Procoagulant drugs accelerate the coagulation process by supplementing coagulation factors, while warfarin-type anticoagulants slow coagulation by inhibiting vitamin K-dependent coagulation factor synthesis; similarly, they have an antagonistic relationship. The antagonistic pair identifiers record antagonistic functional group pairs, and the identifiers include three fields: Antagonistic Group 1, Antagonistic Group 2, and Antagonism Intensity. The functional antagonistic pair markers for heparin-type antifibrinolytic drugs are rated as strong antagonism. The functional antagonistic pair markers for warfarin-type procoagulant drugs are also rated as strong antagonism. The functional antagonistic pair markers include antagonism within calcium channel blockers; nimodipine dilates cerebral blood vessels when used for neuroprotection, while nicardipine may decrease cerebral perfusion pressure when used for antihypertensive purposes, indicating a potential conflict between the two in regulating cerebral perfusion.

[0039] The antagonistic balance window is extracted from the functional antagonistic pair identifiers to generate balance constraint parameters. The antagonistic balance window is defined as the parameter range within which the antagonistic drug pair in the functional antagonistic pair identifiers achieves therapeutic equilibrium within a specific dose ratio and time window. The boundaries of the antagonistic balance window are determined by combination drug safety data in the clinical database. The antagonistic balance window for the antifibrinolytic-antagonistic drug antagonistic pair in the functional antagonistic pair identifiers is obtained through a query of the clinical database. When tranexamic acid is used in combination with low molecular weight heparin, the antagonistic effect in the functional antagonistic pair identifiers is controllable when the tranexamic acid dose is controlled within an appropriate range and low molecular weight heparin is initiated at an appropriate time postoperatively. The antagonistic balance window for the procoagulant-antagonistic drug antagonistic pair in the functional antagonistic pair identifiers requires a certain dosing interval. During the interval, the efficacy of the hemostatic drug in the functional antagonistic pair identifiers gradually diminishes before the anticoagulant drug is initiated. The balance constraint parameters are defined as the boundary conditions of the antagonistic balance window corresponding to each functional antagonistic pair identifier, and include two dimensions: maximum dose ratio and minimum time interval. The balance constraint parameters include a maximum dose ratio and a minimum time interval for both antifibrinolytic and anticoagulant drugs. Similarly, the balance constraint parameters for procoagulant-anticoagulant drugs also include a maximum dose ratio and a minimum time interval. These balance constraint parameters take into account the patient's risk profile. When a patient has a high bleeding risk score, the maximum dose ratio is increased to allow for a higher proportion of hemostatic drugs. Conversely, when a patient has a high thrombosis risk score, the minimum time interval is prolonged to ensure adequate hemostasis before anticoagulation is initiated.

[0040] Target complementarity analysis is performed on balance constraint parameters to generate synergistic markers. Target complementarity analysis identifies drug combinations acting on different targets but producing synergistic effects within the safety boundaries defined by the balance constraint parameters. The balance constraint parameters restrict the use conditions of antagonistic pairs, while target complementarity analysis assesses the synergistic potential of non-antagonistic combinations. Antifibrinolytic drugs act on plasminogen activator, and procoagulant drugs act on coagulation factors; both produce synergistic hemostatic effects through different targets within the safety range of the balance constraint parameters. Free radical scavenging drugs act on reactive oxygen species, calcium channel antagonists act on calcium channels, neurotrophic drugs act on nerve growth factor receptors, and several neuroprotective drugs protect neurons through different mechanisms. Target complementarity analysis calculates the target overlap and pathway synergy between different drug functional groups by querying the drug knowledge base for targets and signaling pathways. Synergistic markers represent the synergistic potential of drug combinations; the marker value ranges from 0 to 1, with higher values ​​indicating stronger synergistic effects. The synergistic markers of antifibrinolytic drugs and procoagulants were high; the synergistic markers of free radical scavengers and calcium channel blockers were moderately high; and the synergistic markers of free radical scavengers and neurotrophic drugs were moderate. The synergistic markers considered the fit values ​​in the group fit matrix; the synergistic markers were upregulated when both group fit values ​​were high.

[0041] Synergistic and antagonistic pairs are identified based on synergistic markers. Synergistic pairs are defined as drug combinations with synergistic markers above a set threshold and no functional antagonistic markers. Antagonistic pairs are defined as drug combinations with functional antagonistic markers and where balance constraint parameters are difficult to satisfy. Antifibrinolytic drugs and procoagulants have high synergistic markers exceeding the threshold, and since neither has a functional antagonistic marker, they are identified as synergistic pairs. This combination enhances hemostasis through a dual mechanism of antifibrinolysis and procoagulant action. Free radical scavengers and calcium channel blockers have moderately high synergistic markers exceeding the threshold and are identified as synergistic pairs. This combination protects neurons through both free radical scavenging and blocking calcium overload. Free radical scavengers and neurotrophic drugs have moderately high synergistic markers exceeding the threshold and are identified as synergistic pairs. Antifibrinolytic drugs and anticoagulants have functional antagonistic markers with strong antagonism, and the balance constraint parameters require a long minimum time interval, making simultaneous use difficult in the early postoperative period; these are identified as antagonistic pairs. Procoagulants and warfarin-like drugs are also identified as antagonistic pairs, and their combined use is prohibited during the perioperative period. Synergistic and antagonistic pairs are output as a list of drug combinations, which includes three fields: combination members, combination type, and synergistic marker value or antagonistic strength level.

[0042] Step S130: Combine and optimize the synergistic pairs to generate synergistic schemes, eliminate and screen antagonistic pairs to generate safety constraints, and construct the medication plan topology using the synergistic schemes and safety constraints.

[0043] In some embodiments, the step of combining and optimizing the synergistic pair to generate a synergistic scheme includes: classifying the synergistic pair into multi-target brain protection pathways to generate pathway coverage combinations; detecting cross-pathway synergistic effects through the pathway coverage combinations to generate synergistic features; identifying the optimal treatment window based on the synergistic features to generate synergistic intensity markers; and generating a synergistic scheme based on the synergistic intensity markers.

[0044] This study categorizes synergistic drug pairs into multi-target neuroprotective pathways, generating pathway coverage combinations. The neuroprotective drug combinations within these synergistic pairs involve multiple neuroprotective signaling pathways. The multi-target neuroprotective pathway classification categorizes the pathways involved in the synergistic pairs into four main categories based on their mechanisms of action: anti-oxidative stress pathways, anti-calcium overload pathways, anti-inflammatory response pathways, and pro-neuroprotective pathways. Edaravone activates the anti-oxidative stress pathway by scavenging free radicals, which includes two key steps: superoxide dismutase activation and lipid peroxidation inhibition. Nimodipine activates the anti-calcium overload pathway by blocking L-type calcium channels, which includes two steps: reducing intracellular calcium ion concentration and inhibiting calcium-dependent enzyme activity. Citicoline activates the pro-neuroprotective pathway by supplementing phospholipid precursors, which includes two steps: cell membrane repair and enhanced neurotransmitter synthesis. The multi-target neuroprotective pathway classification identifies the pathway types covered by each drug in the synergistic pairs: edaravone covers the anti-oxidative stress pathway and part of the anti-inflammatory response pathway; nimodipine covers the anti-calcium overload pathway; and citicoline covers the pro-neuroprotective pathway. A pathway coverage combination is defined as a set of drugs that can cover multiple complementary pathways. The edaravone-nimodipine pathway coverage combination covers three pathways: anti-oxidative stress, anti-calcium overload, and anti-inflammation, with a pathway coverage rate of 75%. The edaravone-cytidine-choline pathway coverage combination covers three pathways: anti-oxidative stress, anti-inflammation, and neurorepair, with a pathway coverage rate of 75%. The synergistic three-drug pathway coverage combination (edaravone-nimodipine-cytidine-choline) covers all four pathways, with a pathway coverage rate of 100%.

[0045] Synergistic features are generated by detecting cross-pathway synergistic effects through pathway coverage combinations. Cross-pathway synergistic effects refer to the mutual enhancement between different pathways in a pathway coverage combination. Pathways generate synergistic protection through signal molecule cross-activation or synergistic targeting of common downstream effects. The cross-pathway synergy between the anti-oxidative stress pathway and the anti-calcium overload pathway is manifested in free radical scavenging reducing calcium channel oxidative damage, and calcium overload inhibition reducing mitochondrial free radical generation, forming a positive feedback loop. The cross-pathway synergy between the anti-inflammatory response pathway and the pro-neural repair pathway is manifested in inflammation suppression providing a stable microenvironment for neural repair, and growth factors released during neural repair further inhibiting the inflammatory response. Cross-pathway synergistic effect detection identifies cross-nodes and synergistic factors between pathways in the pathway coverage combination by querying signaling pathway databases. For the edaravone-nimodipine pathway coverage combination, cross-pathway synergistic effect detection identified two key cross-nodes: mitochondrial function and synaptic plasticity. Synergistic features represent the strength and scope of cross-pathway synergistic effects, and the parameters of synergistic features include three dimensions: the number of cross-nodes, synergistic strength, and pathway activation sequence. The synergistic characteristics of the edaravone-nimodipine combination show two crossover nodes, indicating a high synergistic strength because the two pathways form a positive feedback loop. The activation sequence is that edaravone first activates the antioxidant pathway, followed by nimodipine activating the anti-calcium overload pathway. The synergistic characteristics of the edaravone-cytidine diphosphate choline combination show three crossover nodes, indicating a moderate synergistic strength.

[0046] The optimal therapeutic window is identified based on synergistic features to generate synergistic strength markers. The optimal therapeutic window is defined as the time and dose range within which the drug combination produces the maximum synergistic effect. The therapeutic window is influenced by three factors: pharmacokinetic parameters, pathophysiological processes, and individual patient differences. The optimal therapeutic window is determined by calculating the intersection of the overlapping periods of drug peak concentrations and the pathway activation sequence in the synergistic features. The optimal therapeutic window for the edaravone-nimodipine combination is 1.5 to 3 hours after administration, during which the blood concentrations of both drugs are within the therapeutic range and the cross-pathway synergistic effect in the synergistic features is strongest. The pathological process of secondary brain injury after cerebral hemorrhage includes three stages: early oxidative stress injury (0-6 hours), intermediate inflammatory response (6-24 hours), and late neural repair (after 24 hours). The temporal matching degree of the synergistic features is calculated based on the correspondence between the drug combination and the pathological process stage. The edaravone-nimodipine combination has the highest temporal matching degree of synergistic feature activation during the early oxidative stress injury stage, while the edaravone-cytidine choline combination has the highest temporal matching degree of synergistic feature activation during the late neural repair stage. The synergistic strength marker indicates the degree of synergistic effect of drug combinations within the optimal therapeutic window. The marker value comprehensively considers the number of crossover nodes, pathway coverage, and therapeutic window matching. The synergistic strength marker for the edaravone-nimodipine combination is 0.88, for the edaravone-cytidine choline combination it is 0.76, and for the three-drug combination it is 0.92.

[0047] A synergistic regimen was generated based on the synergistic strength markers. The regimen integrated the drug combinations and timing of administration from the synergistic strength markers. The three-drug combination (edaravone-nimodipine-cytidine choline) with the highest synergistic strength marker was selected as the preferred regimen, covering all four neuroprotective pathways and achieving a synergistic strength marker of 0.92. The timing of administration of the three-drug combination was implemented in stages according to the pathological progression. Edaravone and nimodipine were initiated immediately postoperatively to address early oxidative stress and calcium overload, while citicoline was initiated 24 hours postoperatively to promote nerve repair. The specific dosage of each drug was determined through dosage calculation in step S140. The synergistic regimen also included a two-drug combination as an alternative. This two-drug combination was used when the patient was allergic to or had contraindications to a certain drug. The edaravone-nimodipine combination was suitable for the early postoperative stage, and the edaravone-cytidine choline combination was suitable for the late postoperative stage. The synergistic strength markers for the alternatives were 0.88 and 0.76, respectively. The enhancement regimens are provided in a tiered format: Tier 1 is a three-drug combination, Tier 2 is an edaravone-nimodipine combination, and Tier 3 is an edaravone-cytidine choline combination. The enhancement regimens also specify the applicable conditions and monitoring requirements for each regimen, including daily Glasgow Coma Scale scores and neurological function assessments.

[0048] Conflict elimination screening was conducted to generate safety constraints for antagonistic drug pairs. Antagonistic drug pairs included two groups: antifibrinolytic drugs-anticoagulants and procoagulant drugs-warfarin combinations. These pairs exhibited functional antagonistic characteristics and the balance constraint parameters were difficult to meet. Conflict elimination screening assessed the feasibility of using antagonistic drug pairs at various perioperative stages. The screening principle was absolute contraindication of anticoagulants preoperatively and intraoperatively, with selective use postoperatively based on thrombosis risk assessment. For the antifibrinolytic drug-anticoagulant antagonistic pair, the conflict elimination rule was that tranexamic acid should be used at least 12 hours after surgery before low molecular weight heparin could be initiated, with coagulation function indicators monitored during the interval to ensure stable hemostasis. For the procoagulant drug-warfarin antagonistic pair, the conflict elimination rule was that warfarin should be contraindicated during desmopressin use; if the patient had been taking warfarin long-term preoperatively, the medication should be discontinued 3 days preoperatively, and the international normalized ratio (INR) should be monitored to reduce to below 1.5. Conflict elimination screening also identified potential conflicts among calcium channel blockers. When nimodipine and nicardipine are used in combination, cerebral perfusion pressure parameters must be strictly monitored, and the minimum time interval between the two drugs must be no less than 6 hours, as specified in the balance constraint parameters. Safety constraints are defined as limitations to prevent drug antagonism and adverse reactions. These constraints include three types: prohibition constraints, time interval constraints, and dose limit constraints. Prohibition constraints prohibit the use of all anticoagulants before and during surgery. Time interval constraints stipulate a minimum interval of 12 hours between hemostatic drugs and anticoagulants. Dose limit constraints stipulate that a single dose of tranexamic acid should not exceed 1 gram.

[0049] In some embodiments, constructing a medication regimen topology using the synergistic scheme and the safety constraints includes: extracting synergistic drug groups to be combined based on the synergistic scheme; performing conflict detection on the synergistic drug groups and the safety constraints to generate feasible drug combinations; performing phase-adaptive topology reconstruction on the feasible drug combinations to form a temporal topology structure; and selecting the optimal scheme from the temporal topology structure to determine the medication regimen topology.

[0050] The synergistic drug groups to be combined are extracted based on the synergistic protocol. The synergistic protocol includes three levels of drug compatibility recommendations: a primary protocol of three drugs, a secondary protocol of edaravone-nimodipine, and a tertiary protocol of edaravone-cytidine choline. The synergistic drug groups to be combined are extracted from each level of the synergistic protocol. The primary protocol extracts edaravone, nimodipine, and cytidine choline. The secondary protocol extracts edaravone and nimodipine, and the tertiary protocol extracts edaravone and cytidine choline. The synergistic drug groups to be combined also need to integrate hemostatic drug combinations. Tranexamic acid and desmopressin are extracted from the hemostatic portion of the synergistic protocol. Although hemostatic drugs are not members of the neuroprotective synergistic pair, they are a necessary component of the medication protocol topology. The synergistic drug groups are stored as drug sets, with each set element containing the name, category, and route of administration of each drug. The neuroprotective synergistic drug group includes {edaravone intravenous injection, nimodipine oral administration, and citicoline intravenous infusion}, the hemostatic synergistic drug group includes {tranexamic acid intravenous injection and desmopressin intravenous injection}, and nicardipine was retained from the candidate drug pool as an initial member of the antihypertensive synergistic drug group. The proposed synergistic drug groups also include recommendations for the timing of administration of each drug: neuroprotective drugs are recommended to be initiated postoperatively, hemostatic drugs are recommended to be initiated intraoperatively, and antihypertensive drugs are recommended to be initiated preoperatively.

[0051] Feasible drug combinations were generated by performing conflict detection on synergistic drug groups and safety constraints. Safety constraints included three types of restrictions: prohibition constraints, time interval constraints, and dosage upper limit constraints. Conflict detection verified whether each drug and drug pair in the synergistic drug group violated these safety constraints. In the neuroprotective synergistic drug group, nimodipine is a calcium channel blocker, and nicardipine in the antihypertensive drug group is also a calcium channel blocker. Safety constraints stipulated that the two calcium channel blockers could not be used simultaneously to avoid excessive blood pressure reduction. Conflict detection revealed pharmacological antagonism between nimodipine and nicardipine, triggering a drug replacement decision. Nicardipine was replaced with urapidil, an α1-receptor blocker with a different antihypertensive mechanism than calcium channel blockers, and there was no risk of overlapping targets with nimodipine. In the hemostatic synergistic drug group, tranexamic acid and desmopressin are both hemostatic drugs. Safety constraints stipulated that the upper limit of tranexamic acid dosage when used in combination was 1 gram. Conflict detection verified that the dosage regimen complied with the safety constraints. Within the neuroprotective synergistic drug group, the targets of edaravone, nimodipine, and citicoline do not overlap, and there are no prohibition or time interval constraints in the safety constraints; conflict detection passed. A feasible drug combination is defined as a synergistic drug group that passes all safety constraint tests. The feasible neuroprotective drug combination is {edaravone, nimodipine, citicoline}, the feasible hemostatic drug combination is {tranexamic acid, desmopressin}, and the feasible antihypertensive drug combination is {urapidil}.

[0052] For example, the step of performing stage-adaptation topology reconstruction on the feasible drug combinations to form a temporal topology includes: extracting pharmacokinetic parameters corresponding to each drug from the feasible drug combinations; dividing the pharmacokinetic parameters into preoperative, intraoperative, and postoperative stages to generate a stage adaptation scheme; identifying temporal conflicts between drug combinations based on the stage adaptation scheme to generate conflict markers; and arranging the feasible drug combinations temporally according to the conflict markers to form a temporal topology.

[0053] Pharmacokinetic parameters for each drug in the feasible drug combination were extracted. The feasible drug combination includes six drugs: edaravone, nimodipine, citicoline, tranexamic acid, desmopressin, and urapidil. Detailed pharmacokinetic parameters for each drug in the feasible drug combination are associated with the drug knowledge base. The pharmacokinetic parameters include four core parameters: absorption time, volume of distribution, elimination half-life, and clearance rate. For edaravone, the absorption time is immediate after intravenous administration, the volume of distribution is 0.5 L / kg, the elimination half-life is 1.2 hours, and the clearance rate is 400 mL / min. For nimodipine, the absorption time is 30 minutes after oral administration, the volume of distribution is 1.8 L / kg, the elimination half-life is 2 hours, and the clearance rate is 800 mL / min. For tranexamic acid, the absorption time is immediate after intravenous administration, the elimination half-life is 2.5 hours, and it is mainly excreted by the kidneys; the half-life is prolonged in renal insufficiency. Pharmacokinetic parameters also include the peak concentration and time to peak concentration of the plasma concentration-time curve. The peak concentration of edaravone was 30 μg / mL and the time to peak concentration was 1 hour. The peak concentration of nimodipine was 15 ng / mL and the time to peak concentration was 2 hours.

[0054] Pharmacokinetic parameters were divided into preoperative, intraoperative, and postoperative phases to generate phase-appropriate regimens. The phase division determined the optimal dosing phase for each drug based on the onset and duration of action of the pharmacokinetic parameters, with the principle of matching the peak efficacy period of the pharmacokinetic parameters with the peak therapeutic demand period. Urapidil's pharmacokinetic parameters show an onset time of 5 minutes and a duration of action of 4-6 hours. The phase-appropriate regimen was applied to the preoperative phase to stabilize blood pressure before surgery. Dosing 2 hours preoperatively ensured that urapidil maintained an effective antihypertensive concentration during surgery. Tranexamic acid and desmopressin both have onset times within 15 minutes and a duration of action of 4-6 hours. The phase-appropriate regimen was applied to the intraoperative phase to achieve rapid hemostasis during the surgical procedure. Edaravone and nimodipine have durations of action of 8 hours and 6 hours, respectively. The phase-appropriate regimen was applied to the postoperative phase to continuously protect brain tissue. Edaravone's elimination half-life of 1.2 hours requires repeated dosing every 12 hours to maintain an effective concentration. Citicoline has a slow onset of action but a duration of action lasting up to 24 hours. A phased adaptation protocol applies it to the late postoperative period to promote nerve repair. The phased adaptation protocol defines the timing and frequency of administration for each drug at each stage. Dosage frequency is determined based on the elimination half-life in pharmacokinetic parameters; drugs with shorter half-lives are administered more frequently, and drugs with longer half-lives are administered less frequently. The phased adaptation protocol includes urapidil monotherapy in the preoperative stage, tranexamic acid and desmopressin combination therapy in the intraoperative stage, and edaravone, nimodipine, and citicoline sequentially in the postoperative stage.

[0055] Conflict markers are generated based on the phased adaptation protocol to identify temporal conflicts between drug combinations. A temporal conflict is defined as an overlap in the dosing times of different drugs, coupled with pharmacological interactions or cumulative adverse reactions. Conflict identification is achieved by comparing the dosing times of each drug in the phased adaptation protocol with a drug interaction database. Both nimodipine and urapidil have antihypertensive effects; the overlap in their dosing times in the phased adaptation protocol may lead to excessive blood pressure reduction, thus identifying a potential temporal conflict. Conflict markers categorize conflicts into three types: pharmacodynamic additive, pharmacokinetic interference, and adverse reaction additive. The nimodipine-urapidil conflict is classified as pharmacodynamic additive, with a moderate severity rating, and a recommended dosing interval of at least 2 hours. The tranexamic acid-low molecular weight heparin conflict is classified as pharmacodynamic antagonistic, with a severe severity rating, and a recommended dosing interval of at least 12 hours, consistent with the time difference between intraoperative tranexamic acid administration and postoperative low molecular weight heparin administration in the phased adaptation protocol. Edaravone showed no significant timing conflict with other neuroprotective drugs, and the conflict was marked as non-conflict. The conflict markers also recorded conflict resolution strategies. The resolution strategy for the nimodipine-urapidil conflict was staggered dosing, with urapidil administered preoperatively and nimodipine administered postoperatively, with an interval of approximately 8 hours between the two administrations.

[0056] A temporal topology is formed by sequentially arranging feasible drug combinations based on conflict markers. The temporal arrangement adjusts the dosing order and time intervals of each drug according to the conflict marker resolution strategy, with the principle of avoiding temporal conflicts marked as moderate to severe. Urapidil is placed at the beginning of the preoperative phase, administered 2 hours before surgery. Tranexamic acid and desmopressin are placed in the intraoperative phase, administered at the skin incision and intraoperative bleeding periods, respectively; the conflict markers do not identify any temporal conflict between the two drugs. Edaravone and nimodipine are placed at the beginning of the postoperative phase, administered 30 minutes after surgery, more than 6 hours after urapidil administration to avoid antihypertensive drug overlap. Citicoline is placed in a subsequent position in the postoperative phase, administered 24 hours after surgery. The temporal topology organizes the arranged drug dosing events in the form of a time axis, with zero representing the start of surgery, negative times representing preoperative times, and positive times representing postoperative times. The nodes of the temporal topology represent each drug dosing event, with node attributes including dosing time, drug name, and route of administration. The temporal topology is stored in a directed graph data structure. The topological sequence of the graph reflects the temporal order of drug administration. The time intervals between adjacent nodes in the temporal topology all satisfy the constraint requirements of the conflict marker.

[0057] The optimal dosing regimen was selected from the temporal topology to form the dosing regimen topology. The temporal topology contained multiple feasible drug dosing sequences, each differing in efficacy and safety. The optimal regimen was selected by evaluating the overall efficacy of each sequence in the temporal topology using a multi-objective decision-making algorithm. The efficacy score formula was E = w1 × efficacy score + w2 × safety score. Weighting coefficients w1 and w2 were set to 0.6 and 0.4 respectively based on clinical treatment priorities. The efficacy score was calculated based on a combination of synergistic strength markers and pathway coverage, while the safety score was calculated inversely based on the number and severity of conflicts indicated by conflict markers. The dosing sequence containing the three-drug neuroprotective combination (edaravone-nimodipine-cytidine choline) had the highest efficacy score (0.92) and safety score (0.85) in the temporal topology, indicating a high overall efficacy score. The dosing sequence containing the two-drug neuroprotective combination (edaravone-nimodipine) had an efficacy score of 0.88 and a safety score of 0.90, with a slightly lower overall efficacy score than the three-drug combination. The optimal treatment regimen is a three-drug neuroprotective combination sequence. The topology of this regimen includes five key nodes: preoperative urapidil for blood pressure reduction, intraoperative tranexamic acid and desmopressin for hemostasis, postoperative edaravone and nimodipine for immediate neuroprotection, and cytidine diphosphate choline for nerve repair 24 hours postoperatively. The topology is defined as the temporal topology of the optimal regimen, with each node labeled with the administration time, drug name, and route of administration. The topology is output in two forms: a visual timeline and a structured data table. The timeline visually displays the temporal relationship of drug administration, while the data table records detailed parameter information for each node.

[0058] Step S140: Calculate the baseline dose based on the medication regimen topology, perform dose sensitivity anomaly detection and correction on the baseline dose to generate a correction factor, and use the correction factor in combination with the surgical timing features to generate a time-sharing dosing regimen.

[0059] Specifically, the baseline dose is calculated based on the medication regimen topology. The medication regimen topology includes five key nodes: preoperative urapidil for blood pressure reduction, intraoperative tranexamic acid and desmopressin for hemostasis, postoperative edaravone and nimodipine for immediate brain protection, and cytidine diphosphate choline for neurological repair 24 hours postoperatively. Each node in the medication regimen topology is associated with a drug name and administration time. The baseline dose is defined as the recommended dose of each drug in a standard adult patient. The baseline dose is calculated based on the dose range in the drug's instructions and individualized adjustments to the patient's weight. The formula for calculating the baseline dose of urapidil is D=k×W, where D is the baseline dose, k is a dose coefficient per unit body weight (1 mg / kg / hour), and W is the patient's weight. The baseline dose of urapidil is 70 mg / hour via continuous intravenous infusion. The baseline dose of tranexamic acid consists of a single loading dose and a maintenance dose. The single loading dose is a fixed dose of 1 gram intravenously injected, and the maintenance dose is calculated based on the surgical duration in the surgical sequence characteristics and is 1 gram via continuous intravenous infusion until the end of the surgery. The baseline dose of desmopressin is calculated using the formula D = k × W, where k is the dose coefficient per unit body weight (0.3 μg / kg), and W is the patient's weight. The baseline dose is 21 μg diluted and administered slowly intravenously. The baseline dose of edaravone is a fixed dose of 30 mg every 12 hours for 7 days post-surgery. The baseline dose of nimodipine is 60 mg orally every 4 hours for 21 days. The baseline dose of citicoline is 500 mg intravenously daily for 14 days. Baseline doses are also associated with dose adjustment rules: the baseline dose of tranexamic acid is halved in patients with renal insufficiency due to increased risk of drug accumulation; the baseline dose of edaravone is reduced to 20 mg in patients with abnormal liver function due to decreased clearance.

[0060] In some embodiments, the step of detecting and correcting dose sensitivity anomalies in the reference dose to generate a correction factor includes: performing sensitivity analysis on the reference dose to obtain a dose sensitivity sequence; performing anomaly detection on the dose sensitivity sequence to identify high-sensitivity nodes; identifying dose boundaries based on the high-sensitivity nodes to establish a safe dose range; and combining the safe dose range with the dose sensitivity sequence to output a correction factor.

[0061] A dose-sensitivity sequence was obtained by performing a sensitivity analysis on the baseline dose. The dose-sensitivity analysis assessed the changes in efficacy and the incidence of adverse reactions at different single-dose levels of the baseline dose. The analytical method involved querying clinical databases to identify efficacy and safety indicators for each drug at different doses. For tranexamic acid, the dose-sensitivity analysis used a single loading dose as the analysis object, sampling at a fixed step size within the range above and below the baseline dose. The hemostasis success rate and thrombotic event incidence corresponding to each dose level constituted a dose-response curve. The hemostasis success rate was 65% at a tranexamic acid dose of 0.5 g, 88% at a dose of 1 g, 92% at a dose of 1.5 g, and 93% at a dose of 2 g. The thrombotic event incidence was 2% at doses below 1 g, 5% at a dose of 1.5 g, and 12% at a dose of 2 g. Dose sensitivity is defined as the ratio of the rate of change in efficacy to the rate of change in dose. The sensitivity of tranexamic acid is 46% / g in the 0.5-1 gram dose range and 8% / g in the 1-1.5 gram dose range. The decrease in sensitivity from 46% / g to 8% / g indicates that the efficacy plateau has been reached near the baseline dose of 1 gram. Dose sensitivity analysis of edaravone was sampled within the range of the baseline dose of 30 mg. At a dose of 30 mg, the improvement rate of neurological function was 72%, and at a dose of 45 mg, the improvement rate was 75%, with a sensitivity of 0.2% / mg. Low sensitivity indicates that the efficacy of edaravone changes slowly within this dose range. Dose sensitivity sequences record the sensitivity values ​​of each drug in different dose ranges. The dose sensitivity sequences are grouped by drug, with each group containing sensitivity values ​​for multiple dose ranges. The dose sensitivity sequence for tranexamic acid is [46% / g, 8% / g, 2% / g] corresponding to three dose ranges, and the dose sensitivity sequence shows a decreasing trend in sensitivity.

[0062] For example, the step of detecting anomalies and identifying high-sensitivity nodes in the dose sensitivity sequence includes: performing statistical distribution analysis on the dose sensitivity sequence to generate sensitivity distribution features; performing outlier detection on the sensitivity distribution features to generate abnormal sensitive intervals; establishing a node candidate set based on the abnormal sensitive intervals to trace the source of drug interaction sensitivity; and selecting the point with the highest sensitivity from the node candidate set to determine it as a high-sensitivity node.

[0063] Statistical distribution analysis was performed on the dose sensitivity sequences to generate sensitivity distribution characteristics. The statistical distribution analysis assessed the numerical distribution pattern of the dose sensitivity sequences, and the analysis method included two steps: skewness calculation and kurtosis calculation. The skewness of the tranexamic acid dose sensitivity sequence [46% / g, 8% / g, 2% / g] was 1.65, indicating a right-skewed distribution, and the kurtosis was 2.85, indicating a peaked distribution. The right-skewed distribution indicates low sensitivity in most dose intervals but extremely high sensitivity in a few intervals, suggesting a narrow dose window risk. The sensitivity distribution characteristics also included parameter estimation of the probability density function. The probability density function of the tranexamic acid sensitivity distribution approximately follows an exponential distribution, and the parameter λ was estimated to be 0.053. The edaravone dose sensitivity sequence [0.2% / mg, 0.15% / mg, 0.1% / mg] had a skewness of 0.32, approaching a symmetrical distribution, and a kurtosis of -0.85, indicating a relatively flat distribution. This type of sensitivity distribution suggests a wide therapeutic window and sufficient dose adjustment margin. The dose sensitivity sequence of desmopressin has a skewness of 1.42, also exhibiting a right-skewed distribution, and its sensitivity distribution characteristics are similar to those of tranexamic acid. The skewness value of the sensitivity distribution characteristics is used to set the threshold for outlier detection. When the skewness is greater than 1, an exponential distribution model is used to detect outliers, and when the skewness is close to 0, a normal distribution model is used.

[0064] Outlier detection is performed based on sensitivity distribution characteristics to generate abnormal sensitive intervals. Outlier detection identifies low probability density regions based on the probability density function of the sensitivity distribution characteristics. Drugs with a skewness greater than 1 in the sensitivity distribution characteristics are modeled using an exponential distribution model, while drugs with a skewness close to 0 are modeled using a normal distribution model. The probability density threshold for the sensitivity distribution characteristics of tranexamic acid is set to 10% of the probability density at the distribution mean, corresponding to a sensitivity value of 35% / g. Dose intervals with sensitivity exceeding 35% / g in the dose sensitivity sequence are marked as abnormal sensitive intervals. For example, the sensitivity of tranexamic acid in the dose range of 0.5 to 1 gram exceeds the threshold at 46% / g and is marked as an abnormal sensitive interval. The probability density threshold for desmopressin is 15% of the probability density at the distribution mean, corresponding to a sensitivity threshold of 40% / (μg / kg). The sensitivity of the dose range of 0.2 to 0.3 μg / kg is marked as an abnormal sensitive interval at 55% / (μg / kg). Abnormally sensitive intervals were further validated using a box plot method. Intervals in the dose sensitivity sequence with sensitivity exceeding the upper quartile plus 1.5 times the interquartile range were also marked as abnormally sensitive intervals. This dual detection method reduced the false positive rate for abnormally sensitive intervals. The sensitivity distribution characteristic skewness of edaravone was close to 0, and the sensitivity of each interval was below the threshold; no abnormally sensitive intervals were detected, indicating that edaravone dose adjustment is highly safe.

[0065] Based on abnormally sensitive regions, a candidate set of nodes was established for tracing drug interaction sensitivity. The source tracing analysis analyzed the causes of abnormally sensitive regions, and the source tracing method used drug interaction databases to identify interaction factors affecting the dose-response relationship. The abnormally sensitive region of tranexamic acid (0.5 to 1 gram) corresponds to the rapid improvement phase of hemostatic effect in clinical observation. Source tracing analysis revealed that within this abnormally sensitive region, the binding of tranexamic acid to endogenous plasminogen reached saturation; with further dose increases, the plasminogen binding sites were already filled, and the synergistic effect slowed down. The abnormally sensitive region of desmopressin (0.2 to 0.3 μg / kg) showed that the interaction between desmopressin and V2 receptors was in a dose-dependent activation phase within this range; the receptor activation rate reached 90% at a dose of 0.3 μg / kg, and the receptor approached saturation beyond this dose. Drug interaction sensitivity analysis also identified the impact of combined drug use on sensitivity. When tranexamic acid was used in combination with desmopressin, the abnormal sensitivity range shifted towards lower doses, suggesting that the synergistic effect of the two drugs allows tranexamic acid to achieve an effective plasminogen inhibitory concentration at a lower dose. The candidate dose set included dose points corresponding to all abnormal sensitivity ranges and dose points affecting interactions. For tranexamic acid, the candidate dose set included three candidate dose points: {0.5g monotherapy point, 1g monotherapy point, 0.7g combination therapy point}. The 0.7g combination therapy point is the critical dose point after the shift in the abnormal sensitivity range when tranexamic acid was used in combination with desmopressin. For edaravone, the candidate dose set included two candidate dose points: {30mg standard point, 20mg dose reduction point}. The dose reduction point corresponds to dose adjustments in patients with abnormal liver function.

[0066] The nodes with the highest sensitivity were selected from the candidate set to be high-sensitivity nodes. The selection of the highest sensitivity nodes involved comparing the sensitivity values ​​and clinical efficacy indicators of each node in the candidate set, with the selection principle being the node with the highest sensitivity value and satisfactory clinical efficacy. In the tranexamic acid candidate set, the 0.5g monotherapy node had the highest sensitivity at 46% / g, but the hemostasis success rate at this dose was only 65%, lower than the clinically required 80%. The 1g monotherapy node in the candidate set had a sensitivity of 8% / g, but a hemostasis success rate of 88%, meeting clinical needs. The 0.7g combination therapy node had a sensitivity of 25% / g and a hemostasis success rate of 82%, exceeding clinical requirements. Considering both sensitivity and efficacy, the 0.7g combination therapy node in the candidate set was selected as a high-sensitivity node. This high-sensitivity node requires strict dosage control in combination therapy regimens; a dosage deviation of 0.1g can cause a fluctuation of approximately 8 percentage points in the hemostasis success rate. In the desmopressin candidate dose set, the highest sensitivity was 55% at 0.25 μg / kg, which was selected as a high-sensitivity dose, corresponding to a V2 receptor activation rate of approximately 75%, located in the steep segment of the dose-response curve. In the edaravone candidate dose set, the sensitivity was 0.2% / mg at the 30 mg standard dose set and 0.25% / mg at the 20 mg reduced dose set; both doses showed low sensitivity and flat dose-response curves. No doses meeting the high-sensitivity criteria were found in the candidate set. The results of the high-sensitivity dose selection indicate that tranexamic acid and desmopressin are dose-sensitive drugs, while edaravone is a dose-tolerant drug.

[0067] A safe dose range is established based on dose boundary identification at high-sensitivity nodes. Dose boundary identification determines the safe dose fluctuation range around high-sensitivity nodes. The boundary identification method calculates the upper and lower bounds of the dose based on dual constraints of efficacy and safety. For the tranexamic acid high-sensitivity node at 0.7g, the dose boundary identification considers two constraints: a hemostasis success rate of no less than 80% and a thrombotic event incidence rate of no more than 5%. The lower bound of the safe dose range is determined by efficacy constraints as 0.65g, corresponding to a hemostasis success rate of 80%, and the upper bound of the safe dose range is determined by safety constraints as 0.85g, corresponding to a thrombotic event incidence rate of 5%. This safe dose range is within the single-dose upper limit of 1g specified by the S130 safety constraints. For the desmopressin high-sensitivity node at 0.25μg / kg, the safe dose range is calculated based on the constraint of a receptor activation rate of 60%-90%. The lower bound of the safe dose range at 0.22μg / kg corresponds to a receptor activation rate of 60%, which is the minimum effective activation level. The upper bound at 0.28μg / kg corresponds to a receptor activation rate of 85%, which is close to saturation but has not entered the high-incidence area of ​​adverse reactions. The safe dose range for edaravone at the standard dose of 30 mg is relatively wide, [25 mg, 35 mg]. Because this drug has low sensitivity and a wide therapeutic window, dose fluctuations within this range have an impact of no more than 2 percentage points on the improvement rate of neurological function. The safe dose range also considers individual patient differences. For patients with renal insufficiency, the safe dose range for tranexamic acid is narrowed to [0.35 g, 0.50 g] due to drug accumulation leading to a prolonged duration of effective concentration. For elderly patients, the safe dose range for urapidil is reduced to [35, 56] mg / hour because elderly patients have increased sensitivity to antihypertensive drugs.

[0068] A correction factor is output by combining the safe dose range and dose sensitivity sequence. The correction factor is defined as an adjustment coefficient of the baseline dose, used to correct the baseline dose to the optimal dose within the safe dose range. The calculation formula is F = (D_optimal - D_baseline) / D_baseline, where F is the correction factor, D_optimal is the recommended dose within the safe dose range, and D_baseline is the baseline dose. The baseline dose of tranexamic acid of 1 gram is outside the safe dose range [0.65 g, 0.85 g]. The dose sensitivity sequence shows that tranexamic acid has a sensitivity as high as 46% / gram in the 0.5 to 1 gram range, indicating that dose changes in this range have a significant impact on efficacy. D_optimal is selected at the upper limit of the safe dose range of 0.85 g to maximize hemostatic effect, and the correction factor is -0.15. The baseline dose of desmopressin at 0.3 μg / kg is outside the safe dose range [0.22, 0.28] μg / kg. Dose sensitivity sequences show a sensitivity of 55% / (μg / kg) in the 0.2 to 0.3 μg / kg range, placing it in the high-sensitivity region. The optimal D_selection is 0.25 μg / kg, corresponding to the best receptor activation rate, with a correction factor of -0.17. The baseline dose of edaravone at 30 mg is in the middle of the safe dose range [25 mg, 35 mg]. Dose sensitivity sequences show sensitivity below 0.25% / mg in all intervals, with a correction factor of 0. The baseline dose of urapidil at 70 mg / hour has a safe dose range of [35, 56] mg / hour. Dose sensitivity sequences show moderate sensitivity, with the optimal D_selection being 49 mg / hour, with a correction factor of -0.3. The dose sensitivity sequences for nimodipine and citicoline show sensitivity below the clinical significance threshold in all intervals, and their safe dose ranges cover the baseline dose, with correction factors of 0 for both.

[0069] A timed dosing regimen was generated using correction factors combined with surgical timing characteristics. The correction factors adjusted the baseline dose to the optimal dose: tranexamic acid was adjusted to 0.85 g by correction factor -0.15, desmopressin to 17.5 μg by correction factor -0.17, and urapidil to 49 mg / hour by correction factor -0.3. Edaravone, nimodipine, and citicoline were maintained at baseline doses with correction factors of 0. Surgical timing characteristics included a planned surgical time of 8 hours after symptom onset, surgical type of craniotomy for hematoma evacuation, estimated surgical duration of 4 hours, and general anesthesia. These characteristics defined the time points for each perioperative stage. The timed dosing regimen allocated the optimal dose adjusted by the correction factors to the specific time points corresponding to the surgical timing characteristics: the preoperative stage began 2 hours before surgery, the intraoperative stage began at the skin incision time, and the postoperative stage began at the end of surgery. In the timed dosing regimen, urapidil was administered intravenously at a rate of 49 mg / hour preoperatively until the end of surgery. The total dosing time was approximately 6 hours, calculated by adding the 4-hour surgery duration to the 2-hour preoperative period. During the intraoperative phase, tranexamic acid was administered intravenously at the time of skin incision, starting with a loading dose of 0.85 g, followed by a continuous infusion of 0.85 g until the end of surgery. Desmopressin was administered intravenously at a rate of 17.5 μg diluted and slowly injected over 30 minutes during the intraoperative bleeding period. Postoperatively, edaravone was administered intravenously at a rate of 30 mg 30 minutes postoperatively, followed by every 12 hours. Nimodipine was started orally at a rate of 60 mg every 4 hours, starting 6 hours postoperatively, with a delayed initiation to avoid overlapping the antihypertensive effects with urapidil. Citicoline was administered intravenously at a rate of 500 mg once daily for 14 days, starting 24 hours postoperatively, to match the pathological progression of the late neurological repair phase.

[0070] Step S150: Real-time monitoring is performed based on the time-sharing dosing regimen to obtain dynamic patient indicators. Perfusion assessment is performed based on the dynamic patient indicators to determine the safety level. Dosage adjustment is triggered based on the safety level to form an optimized dosing regimen.

[0071] Specifically, real-time monitoring of patient dynamic indicators was performed based on a timed dosing regimen. The timed dosing regimen included multiple administration points, such as preoperative urapidil for blood pressure reduction, intraoperative tranexamic acid and desmopressin for hemostasis, and postoperative edaravone and nimodipine for brain protection. Each time point in the timed dosing regimen was associated with corresponding monitoring requirements. Real-time monitoring collected patient dynamic indicators through bedside monitoring equipment and invasive monitoring devices. The monitoring frequency of patient dynamic indicators was dynamically adjusted according to the medication stage and drug characteristics of the timed dosing regimen. During the preoperative urapidil administration period in the timed dosing regimen, blood pressure was monitored every 30 minutes, including systolic blood pressure, diastolic blood pressure, and mean arterial pressure. Two hours before surgery, the systolic blood pressure was 165 mmHg at the start of administration; 30 minutes after administration, the systolic blood pressure decreased to 152 mmHg; and one hour after administration, the systolic blood pressure stabilized at 148 mmHg. In the timed dosing regimen, coagulation function was monitored every 2 hours after intraoperative tranexamic acid administration. Prothrombin time and activated partial thromboplastin time (APPT) were among the dynamic indicators reflecting hemostatic effect. After administration of 0.85 g at skin incision, prothrombin time decreased from 13.2 seconds to 11.8 seconds 2 hours later. Liver function was monitored daily during postoperative edaravone administration. ALT (alanine aminotransferase) reflected the liver safety of edaravone; it was 35 U / L on postoperative day 1 and rose to 48 U / L on postoperative day 3, remaining within the normal range. Other dynamic indicators included intracranial pressure monitoring and neurological function assessment. Intracranial pressure was measured via a lateral ventricle drain. Immediately postoperative intracranial pressure was 18 mmHg, decreasing to 14 mmHg 6 hours postoperatively. Neurological function was assessed using the Glasgow Coma Scale; a score of 10 immediately postoperatively and 12 at 24 hours postoperatively indicated improved neurological function.

[0072] In some embodiments, the step of determining the safety level based on the patient's dynamic indicators perfusion assessment includes: extracting intracranial pressure features and blood pressure features from the patient's dynamic indicators to generate cerebral perfusion pressure parameters; performing continuous time-series co-variance detection on the cerebral perfusion pressure parameters to obtain perfusion pressure fluctuation features; establishing a response rate gradient based on the perfusion pressure fluctuation features to generate a risk level; and combining the cerebral perfusion pressure parameters with the risk level to determine the safety level.

[0073] Intracranial pressure (CPP) and blood pressure characteristics were extracted from patient dynamic indicators to generate cerebral perfusion pressure (CPP) parameters. Patient dynamic indicators included continuous 24-hour postoperative CPP and blood pressure monitoring records, monitored hourly. CPP characteristics were extracted from the CPP monitoring data; a coefficient of variation of 0.25 indicated high CPP fluctuations. Blood pressure characteristics were extracted from the mean arterial pressure (MAP) monitoring data; the mean MAP from 0 to 6 hours postoperatively was 92 mmHg. The CPP parameter was defined as the difference between mean MAP and CPP, reflecting the actual blood perfusion status of brain tissue. The calculation formula was CPP = MAP - ICP, where CPP is CPP, MAP is mean MAP, and ICP is CPP. Two hours post-surgery, the mean arterial pressure was 90 mmHg, intracranial pressure was 20 mmHg, and cerebral perfusion pressure was 70 mmHg. Four hours post-surgery, the mean arterial pressure was 93 mmHg, intracranial pressure was 18 mmHg, and cerebral perfusion pressure was 75 mmHg. Six hours post-surgery, the mean arterial pressure was 95 mmHg, intracranial pressure was 16 mmHg, and cerebral perfusion pressure was 79 mmHg. The cerebral perfusion pressure parameters include a temporal sequence and statistical characteristics. The mean cerebral perfusion pressure from 0 to 6 hours post-surgery was 73 mmHg, and the trend of the cerebral perfusion pressure parameters was a fluctuating upward trend.

[0074] Cooperative deviation detection of cerebral perfusion pressure parameters over a continuous time period was used to obtain perfusion pressure fluctuation characteristics. The continuous time period was defined as a 6-hour monitoring window post-surgery. Cooperative deviation detection identified deviation patterns between cerebral perfusion pressure parameters and the target range within this time period. The target range for cerebral perfusion pressure parameters after intracerebral hemorrhage was set at 60 to 70 mmHg. This range was determined based on the cerebral blood flow autoregulation curve and the metabolic needs of brain tissue. A cerebral perfusion pressure below 60 mmHg indicates insufficient brain perfusion, while a value above 70 mmHg may increase the risk of rebleeding. The time-series sequence of cerebral perfusion pressure parameters was compared with the target range hour by hour. At 2 hours post-surgery, a cerebral perfusion pressure of 70 mmHg was within the target range boundary; at 4 hours post-surgery, a cerebral perfusion pressure of 75 mmHg exceeded the upper limit of the target; and at 6 hours post-surgery, a cerebral perfusion pressure of 79 mmHg exceeded the upper limit of the target. Co-existing deviation is defined as a simultaneous deviation of cerebral perfusion pressure and mean arterial pressure from their target ranges. In the 4-6 hours post-surgery, cerebral perfusion pressure consistently exceeded the target upper limit by 70 mmHg, and mean arterial pressure consistently exceeded the recommended upper limit by 85 mmHg, indicating a co-existing upward deviation. The deviation for cerebral perfusion pressure was 9 mmHg, and for mean arterial pressure, it was 8 mmHg. Perfusion pressure fluctuation characteristics describe the deviation properties of the time-series cerebral perfusion pressure parameters. These characteristics include both the magnitude and direction of deviation. In the 0-6 hours post-surgery, the deviation in perfusion pressure fluctuation characteristics was 9 mmHg, with an upward deviation. The magnitude and direction of deviation in these characteristics are used for risk level assessment.

[0075] Risk levels were generated by establishing a response rate gradient based on perfusion pressure fluctuation characteristics. The response rate was defined as the speed at which cerebral perfusion pressure parameters recovered to the target range after drug dosage adjustment. The response rate gradient reflects the reserve capacity of cerebral autoregulation. After detecting an upward bias in perfusion pressure fluctuation characteristics 6 hours post-surgery, the urapidil infusion rate was reduced. One hour later, the mean arterial pressure decreased from 95 mmHg to 89 mmHg, and the cerebral perfusion pressure parameter decreased from 79 mmHg to 73 mmHg, with a response rate of 6 mmHg / hour. The response rate gradient was defined as the trend of response rate change over time. The response rate was 6 mmHg / hour at 6-7 hours post-surgery and 3 mmHg / hour at 7-8 hours. A negative response rate gradient indicates a decreasing response rate, suggesting a gradual recovery of cerebral autoregulation. The risk level is determined comprehensively based on the deviation of the response rate gradient and the perfusion pressure fluctuation characteristics. The risk level determination rules are as follows: if the response rate gradient is less than a set threshold and the deviation in the perfusion pressure fluctuation characteristics is less than 10 mmHg, the risk level is low; if the response rate gradient is moderate and the deviation is 10 to 20 mmHg, the risk level is medium; and if the response rate gradient is greater than the threshold or the deviation is greater than 20 mmHg, the risk level is high. Currently, the response rate gradient is negative, and the deviation in the perfusion pressure fluctuation characteristics is 9 mmHg, close to 10 mmHg, but the rate of decrease of the response rate gradient is slow. Therefore, the overall risk level is determined to be medium.

[0076] The safety level is determined by combining cerebral perfusion pressure parameters with risk level. A mean cerebral perfusion pressure of 73 mmHg exceeds the upper limit of the target range of 70 mmHg. A fluctuating upward trend in the cerebral perfusion pressure parameters indicates persistently high perfusion pressure, and the risk level is medium, indicating limited autoregulation. The safety level comprehensively assesses three dimensions: the absolute value of the cerebral perfusion pressure parameters, their trend, and the risk level. The safety level also considers the deviation direction of perfusion pressure fluctuations as an auxiliary parameter for subsequent dose adjustment strategies. The safety level is divided into three levels: safe, alert, and dangerous. The criteria for determining the safe level are a mean cerebral perfusion pressure of 60 to 70 mmHg with a stable or upward trend and a low risk level. The criteria for determining the alert level are a mean cerebral perfusion pressure of 50 to 60 mmHg or 70 to 80 mmHg with a fluctuating trend and a medium risk level. The alert level indicates that the cerebral perfusion pressure parameters deviate from the target, but the autoregulation function is still present. The criteria for determining the hazard level within the safety level are: a mean cerebral perfusion pressure parameter below 50 mmHg or above 80 mmHg with a rapidly changing trend, indicating a high risk level. Currently, the mean cerebral perfusion pressure parameter is 73 mmHg, falling within the 70-80 mmHg range, with a fluctuating upward trend, classifying it as a medium risk level, and the safety level as alert. The safety level also includes recommended actions: a safety level recommendation is to maintain the current medication regimen; an alert level recommendation is to adjust the antihypertensive drug dosage; and a hazard level recommendation is to initiate emergency intervention and a specialist consultation.

[0077] Dosage adjustment is triggered based on safety level to optimize the medication regimen. When the safety level is deemed safe, maintenance medication is recommended, and the dosing schedule does not need adjustment. When the safety level is deemed alarming, a dosage adjustment process is triggered, with the adjustment strategy determined by the deviation direction of perfusion pressure fluctuations. An upward deviation in cerebral perfusion pressure indicates excessively high mean arterial pressure or excessively low intracranial pressure. When the safety level is alarming, the adjustment strategy is to reduce the dose of antihypertensive drugs or slow the rate of blood pressure reduction; for example, the urapidil infusion rate is reduced from 49 mg / hour to 35 mg / hour. A downward deviation in cerebral perfusion pressure indicates excessively low mean arterial pressure or excessively high intracranial pressure. When the safety level is alarming, the adjustment strategy is to discontinue or reduce the dose of antihypertensive drugs or use dehydrating agents to lower intracranial pressure. If the mean arterial pressure is below 75 mmHg, antihypertensive drugs are discontinued and vasoactive drugs are administered. Dosage adjustment also considers drug interactions and safety constraints. The antihypertensive effects of urapidil and nimodipine may be additive; when the two drugs are used together, the urapidil dose needs to be additionally reduced to avoid excessive blood pressure drop. When the safety level is determined to be dangerous, the emergency intervention procedure is triggered. Emergency intervention includes discontinuing the suspected medication, initiating vasopressor or antihypertensive therapy, adjusting the dosage of dehydration medications, and notifying the attending physician for bedside assessment. The optimized medication regimen integrates all drug dosing plans after dosage adjustments. The optimized urapidil dosing regimen is a continuous infusion of 35 mg / hour, with monitoring frequency adjusted from every 30 minutes to every 15 minutes to closely observe blood pressure changes. The optimized dosing regimens for edaravone, nimodipine, and citicoline remain unchanged from their original timed dosing schedules. Tranexamic acid, administered intraoperatively, was discontinued after surgery. The optimized medication regimen also includes expected goals: cerebral perfusion pressure is expected to remain stable at 60-70 mmHg 12 hours postoperatively; the Glasgow Coma Scale score is expected to improve to above 13 points 24 hours postoperatively; and no thromboembolic or rebleeding complications are expected 7 days postoperatively.

[0078] To implement the intelligent recommendation method for perioperative medication regimens for cerebral hemorrhage corresponding to the above method embodiments, and to achieve the corresponding functions and technical effects. See also Figure 2 , Figure 2 This diagram illustrates a structural block diagram of an intelligent recommendation system 200 for perioperative medication regimens in cases of cerebral hemorrhage, as provided in an embodiment of this application. For ease of explanation, only the parts relevant to this embodiment are shown. The intelligent recommendation system 200 for perioperative medication regimens in cases of cerebral hemorrhage, as provided in this embodiment of the application, includes:

[0079] Data acquisition module 201 is used to acquire patient physiological parameters and surgical time sequence characteristics, construct patient risk profile based on the physiological parameters, and perform risk matching between the patient risk profile and drug knowledge base to generate candidate drug set;

[0080] Drug screening module 202 is used to extract drug attribute features from the candidate drug set to determine drug function groups, perform a suitability score on the drug function groups based on the patient risk profile to generate a group suitability matrix, and perform drug interaction analysis based on the group suitability matrix to identify synergistic pairs and antagonistic pairs.

[0081] The scheme construction module 203 is used to combine and optimize the synergistic pairs to generate synergistic schemes, eliminate and screen the antagonistic pairs to generate safety constraints, and construct a medication scheme topology using the synergistic schemes and the safety constraints.

[0082] The dosage optimization module 204 is used to calculate the baseline dose based on the topology of the medication regimen, perform dose sensitivity anomaly detection and correction on the baseline dose to generate a correction factor, and use the correction factor in combination with the surgical timing features to generate a timed dosing regimen.

[0083] The monitoring and adjustment module 205 is used to perform real-time monitoring to obtain dynamic indicators of the patient based on the time-sharing dosing regimen, perform perfusion assessment based on the dynamic indicators of the patient to determine the safety level, and trigger dose adjustment to form an optimized dosing regimen according to the safety level.

[0084] The aforementioned intelligent recommendation system 200 for perioperative medication regimens in cerebral hemorrhage can implement the intelligent recommendation method for perioperative medication regimens in cerebral hemorrhage as described in the above-described method embodiments. The options in the above method embodiments are also applicable to this embodiment and will not be detailed here. The remaining content of this application embodiment can be referred to the content of the above method embodiments, and will not be repeated in this embodiment.

[0085] The purpose of the above embodiments is to reproduce and derive the technical solution of the present invention by way of example, and to fully describe the technical solution, purpose and effect of the present invention. The purpose is to enable the public to have a more thorough and comprehensive understanding of the disclosure of the present invention, and not to limit the scope of protection of the present invention.

[0086] The above embodiments are not an exhaustive list based on the present invention, and there may be many other embodiments not listed. Any substitutions and improvements made without departing from the concept of the present invention are within the protection scope of the present invention.

Claims

1. An intelligent recommendation system for perioperative medication regimens in patients with cerebral hemorrhage, characterized in that, include: The data acquisition module is used to acquire the patient's physiological parameters and surgical time sequence characteristics, construct a patient risk profile based on the physiological parameters, and perform risk matching between the patient risk profile and the drug knowledge base to generate a candidate drug set. The drug screening module is used to extract drug attribute features from the candidate drug set to determine drug functional groups, perform a suitability score on the drug functional groups based on the patient risk profile to generate a group suitability matrix, and perform drug interaction analysis based on the group suitability matrix to identify synergistic pairs and antagonistic pairs. This includes: performing hemostasis-anticoagulation antagonism detection on the group suitability matrix to generate antagonistic pair identifiers; extracting antagonistic balance windows from the antagonistic pair identifiers to generate balance constraint parameters; performing target complementarity analysis on the balance constraint parameters to generate synergistic markers; and identifying synergistic pairs and antagonistic pairs based on the synergistic markers. The scheme construction module is used to combine and optimize the synergistic pairs to generate synergistic schemes, perform conflict elimination screening on the antagonistic pairs to generate safety constraints, and construct a medication scheme topology using the synergistic schemes and the safety constraints. The dosage optimization module is used to calculate the baseline dose based on the topology of the medication regimen, perform dose sensitivity anomaly detection and correction on the baseline dose to generate a correction factor, and use the correction factor in combination with the surgical timing features to generate a timed dosing regimen. The monitoring and adjustment module is used to perform real-time monitoring to acquire dynamic patient indicators based on the time-sharing dosing regimen, and to perform perfusion assessment to determine the safety level based on the dynamic patient indicators. This includes: extracting intracranial pressure and blood pressure characteristics from the dynamic patient indicators to generate cerebral perfusion pressure parameters; performing continuous-time-period collaborative deviation detection on the cerebral perfusion pressure parameters to obtain perfusion pressure fluctuation characteristics; establishing a response rate gradient based on the perfusion pressure fluctuation characteristics to generate a risk level; combining the cerebral perfusion pressure parameters and the risk level to determine the safety level; and triggering dose adjustment based on the safety level to form an optimized dosing regimen.

2. The system according to claim 1, characterized in that, The process of generating a grouping fit matrix by performing a fit score on the drug function grouping based on the patient risk profile includes: Extract bleeding site features from the patient risk profile; Based on the characteristics of the bleeding site, site-specific safety assessments are performed on the drug function groups to generate site-adaptive weights. The adaptation value is generated by performing risk calibration calculation based on the adaptation weight of the aforementioned parts. The fitness values ​​are arranged according to the grouping dimension to form a group fitness matrix.

3. The system according to claim 1, characterized in that, The step of combining and optimizing the synergistic pairs to generate synergistic solutions includes: To achieve the synergistic effect, multi-target brain protection pathways are classified and pathway coverage combinations are generated. Synergistic features are generated by detecting cross-pathway synergistic effects through the aforementioned pathway coverage combinations; Based on the aforementioned synergistic features, the optimal treatment window is identified to generate an enhancement intensity marker; An enhancement scheme is generated based on the enhancement intensity marker.

4. The system according to claim 1, characterized in that, The method of constructing a medication regimen topology using the enhancement scheme and the safety constraints includes: Extract the synergistic drug group to be combined according to the aforementioned enhancement scheme; Conflict detection is performed between the synergistic drug group and the safety constraints to generate feasible drug combinations; The feasible drug combinations are subjected to stage-adaptive topology reconstruction to form a temporal topology; The optimal solution is selected from the time-series topology and determined as the medication plan topology.

5. The system according to claim 1, characterized in that, The step of detecting and correcting dose sensitivity anomalies in the baseline dose to generate a correction factor includes: A dose sensitivity sequence was obtained by performing a sensitivity analysis on the baseline dose; Anomaly detection was performed on the dose sensitivity sequence to identify high-sensitivity nodes; A safe dose range is established based on dose boundary identification using the aforementioned highly sensitive nodes; A correction factor is output by combining the safe dose range with the dose sensitivity sequence.

6. The system according to claim 4, characterized in that, The step of performing phased adaptation and topological reconstruction of the feasible drug combinations to form a temporal topology includes: Extract the pharmacokinetic parameters of each drug from the feasible drug combinations; The pharmacokinetic parameters are divided into preoperative, intraoperative, and postoperative stages to generate a stage adaptation scheme. Based on the stage adaptation scheme, temporal conflicts between drug combinations are identified and conflict markers are generated. The feasible drug combinations are sequentially arranged according to the conflict markers to form a temporal topology.

7. The system according to claim 5, characterized in that, The step of detecting and identifying high-sensitivity nodes in the dose-sensitivity sequence includes: Statistical distribution analysis is performed on the dose sensitivity sequence to generate sensitivity distribution characteristics; Outlier detection is performed based on the sensitivity distribution characteristics to generate abnormal sensitivity intervals; Based on the aforementioned abnormally sensitive intervals, a candidate set of nodes is established for tracing the source of drug interaction sensitivity. The node with the highest sensitivity is selected from the candidate node set and identified as a high-sensitivity node.

Images