A cardiovascular and cerebrovascular drug virtual clinical trial system and method

Abstract

The application discloses a kind of cardiovascular and cerebrovascular drug virtual clinical test system and method, belong to medical information technology field.System includes: virtual patient construction module, generates blood vessel digital twin based on CTA image data;Drug metabolism simulation module, drug distribution and metabolism virtual simulation are carried out in blood vessel digital twin, and output drug time curve and target organ drug concentration;Efficacy / adverse reaction prediction module, based on preset evaluation index and threshold value prediction efficacy score and adverse reaction risk grade;Closed-loop feedback optimization module, according to the predicted result adjustment parameter generation optimization test scheme.Four modules constitute iterative optimization loop, cover clinical I-IV phase whole process virtual test.The application couples blood vessel digital twin with pharmacokinetic model, realizes the dynamic iteration of virtual clinical test by closed-loop feedback optimization, can shorten cardiovascular and cerebrovascular drug clinical test cycle, reduce research and development cost.

Description

Technical Field

[0001] This invention belongs to the field of medical information technology, specifically relating to a virtual clinical trial system and method for cardiovascular and cerebrovascular drugs based on digital twin technology. Background Technology

[0002] The development of cardiovascular and cerebrovascular drugs faces challenges such as long development cycles (typically 8-12 years), high costs (over $1 billion per drug), and high failure rates. Traditional clinical trials are constrained by ethical restrictions, difficulties in patient recruitment, and individual differences, making it difficult to comprehensively assess the efficacy and safety of drugs. Existing virtual clinical trial technologies (such as CN118350298B and CN117831640A) only cover single stages (such as vascular modeling or drug screening) and have not formed a complete chain from virtual patient construction → drug metabolism simulation → efficacy prediction → closed-loop optimization, especially lacking a systematic solution for the entire process of clinical trials (Phase I-IV) for cardiovascular and cerebrovascular drugs. The new Patent Examination Guidelines, which will come into effect on January 1, 2026, explicitly require that AI-related patent applications must clearly describe the details of the combination of algorithms and specific technical fields. Summary of the Invention

[0003] This invention provides a virtual clinical trial system for cardiovascular and cerebrovascular drugs. It generates digital twins of blood vessels through a virtual patient construction module, simulates pharmacokinetic processes through a drug metabolism simulation module, outputs quantitative evaluations through an efficacy / adverse reaction prediction module, and iteratively adjusts parameters through a closed-loop feedback optimization module, forming a four-module closed-loop circuit to achieve a full-process virtual trial from Phase I to IV clinical trials. This system can shorten the clinical trial cycle for cardiovascular and cerebrovascular drugs, reduce R&D costs, and minimize ethical risks. Attached Figure Description

[0004] Figure 1 This is a schematic diagram of the system module connection relationship and data flow of the present invention;

[0005] Figure 2 This is a timing diagram of the virtual clinical trial process for phases I-IV.

[0006] Figure 3 This is a diagram of a layered system architecture. Detailed Implementation

[0007] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0008] Example 1: System Architecture and Module Connection Relationships

[0009] See Figure 1 The virtual clinical trial system for cardiovascular and cerebrovascular drugs provided by the present invention includes a virtual patient construction module 100, a drug metabolism simulation module 200, an efficacy / adverse reaction prediction module 300, and a closed-loop feedback optimization module 400.

[0010] The virtual patient construction module 100 is configured to acquire the cardiovascular and cerebrovascular CTA image data of the target patient, perform three-dimensional vascular reconstruction based on the CTA image data, and generate a vascular digital twin. Specifically, the CTA image data is a DICOM format coronary CTA image sequence with a slice thickness of 0.5-1.0 mm. The virtual patient construction module 100 further includes: a CTA image segmentation unit, which uses the U-Net++ deep learning segmentation model to segment the cardiovascular and cerebrovascular regions of CTA image data and obtain a vascular segmentation mask; a 3D reconstruction unit, which uses the moving cube algorithm to reconstruct the surface based on the vascular segmentation mask, obtains a triangular mesh model, and then divides the triangular mesh model into tetrahedral meshes to generate a 3D vascular geometric model; a boundary condition setting unit, which sets the inlet flow boundary conditions and outlet pressure boundary conditions of the 3D vascular geometric model based on the physiological parameters (heart rate, blood pressure, cardiac output) of the target patient, forming a complete vascular digital twin; and a virtual sample amplification unit, which uses a cycleGAN generative adversarial network to amplify the CTA image data, generating virtual patient CTA images with different degrees of vascular lesions (stenosis rates of 30%, 50%, 70%, and 90%) and different vascular configurations to construct a virtual patient population. The vascular digital twin includes vascular geometric structural parameters (vessel diameter, stenosis rate, plaque volume, and vascular bifurcation angle) and hemodynamic boundary conditions (inlet flow velocity, outlet pressure, and wall shear stress).

[0011] The drug metabolism simulation module 200 is connected to the virtual patient construction module 100 via an API interface. It is configured to acquire a vascular digital twin and perform virtual simulation of drug distribution and metabolism processes in the vascular digital twin based on the physicochemical parameters and pharmacokinetic model of the target cardiovascular drug. Specifically, the drug metabolism simulation module 200 includes: a drug parameter input unit configured to receive the molecular weight, lipid solubility (LogP value), protein binding rate, and metabolic enzyme substrate information (such as CYP3A4 and CYP2C9 substrate information) of the target cardiovascular drug; a hemodynamic coupling unit that couples the blood flow velocity field and pressure field of the vascular digital twin with the drug convection-diffusion equation (∂C / ∂t + ∇·(vC) = D∇²C) to calculate the convection and diffusion transport of the drug in the vascular lumen; a tissue distribution calculation unit that calculates the transwall permeation of the drug from the vascular lumen to the vascular wall and surrounding tissues based on the vascular wall permeability parameters of the vascular digital twin (calculated through the Kedem-Katchalsky equation); and a pharmacokinetic fitting unit that fits the drug concentration-time data based on a two-compartment or three-compartment pharmacokinetic model to generate drug-time curves and target organ drug concentrations.

[0012] The efficacy / adverse reaction prediction module 300 is connected to the drug metabolism simulation module 200 via an API interface. It is configured to predict efficacy scores and adverse reaction risk levels based on the drug-time curve and target organ drug concentration, combined with preset cardiovascular drug efficacy evaluation indicators and adverse reaction thresholds. Efficacy evaluation indicators include: the rate of change in target vessel diameter (the ratio of post-treatment vessel diameter to baseline diameter) based on a vascular digital twin, the change in fractional flow reserve (FFR), and the vascular endothelial function index (calculated through shear stress). Adverse reaction thresholds include: the comparison between target organ drug concentration and a preset safe upper limit, and the comparison between the peak blood drug concentration (Cmax) calculated based on the drug-time curve and a preset toxic concentration. The efficacy score is a continuous value from 0 to 100, and the adverse reaction risk levels are divided into low risk (0-30), medium risk (31-70), and high risk (71-100).

[0013] The closed-loop feedback optimization module 400 is connected to the virtual patient construction module 100, the drug metabolism simulation module 200, and the efficacy / adverse reaction prediction module 300 via a data bus. It is configured to adjust the vascular geometry parameters (such as vessel diameter and stenosis rate) of the virtual patient construction module and / or the pharmacokinetic model parameters (such as clearance rate and distribution volume) of the drug metabolism simulation module based on the efficacy score and adverse reaction risk level, thereby generating an optimized virtual trial protocol. The closed-loop feedback optimization module 400 includes: a parameter sensitivity analysis unit, which uses the Sobol index method to determine the most sensitive parameters among vascular geometry parameters and pharmacokinetic model parameters that have the greatest impact on efficacy scores and adverse reaction risk levels; an optimization algorithm unit, which uses Bayesian optimization to iteratively update the values ​​of sensitive parameters with the optimization objectives of maximizing efficacy scores and minimizing adverse reaction risk levels, generating new parameter combinations in each iteration and re-executing drug metabolism simulation and efficacy prediction; and a trial protocol generation unit, which generates an optimized virtual trial protocol based on the optimized sensitive parameters. The virtual trial protocol includes at least: virtual patient inclusion criteria (age range, degree of stenosis, comorbidities), dosing regimen (dosage, dosing frequency, route of administration), and observation endpoints (primary efficacy endpoint, safety endpoint).

[0014] The virtual patient construction module 100, drug metabolism simulation module 200, efficacy / adverse reaction prediction module 300, and closed-loop feedback optimization module 400 constitute an iterative optimization loop, used to simulate virtual trials of target cardiovascular drugs at each stage from clinical phase I to clinical phase IV.

[0015] Example 2: Specific Implementation of Virtual Trials at Each Stage of Clinical Phase I-IV

[0016] This system can sequentially execute virtual clinical trials from Phase I to Phase IV. The simulation methods and parameter settings for each phase are as follows:

[0017] Phase I Clinical Simulation: The Virtual Patient Construction Module 100 constructs a digital twin of the blood vessels of a healthy virtual patient based on CTA imaging data of healthy volunteers (no significant vascular stenosis, stenosis rate <20%). The Drug Metabolism Simulation Module 200 simulates a single-dose escalation tolerance trial of the target cardiovascular drug, with doses of 1, 2, 4, and 8 times the preset dose, each dose group containing 6 virtual patients. The Efficacy / Adverse Reaction Prediction Module 300 outputs pharmacokinetic parameters such as pharmacokinetic curves, Cmax, AUC, and elimination half-life at each dose, as well as the adverse reaction risk level. The Closed-Loop Feedback Optimization Module 400 determines the maximum tolerated dose and the recommended starting dose for Phase II.

[0018] Phase II Clinical Simulation: The Virtual Patient Construction Module 100 constructs a virtual patient population of 15 patients per group based on CTA imaging data (stenosis rate 30%-70%) of virtual patients with the target indication. The Drug Metabolism Simulation Module 200 simulates a multiple-dose exploratory trial of the target cardiovascular drug, setting up three dose groups (low, medium, and high). The Efficacy / Adverse Reaction Prediction Module 300 outputs the dose-response relationship and preliminary efficacy data for each dose group. The Closed-Loop Feedback Optimization Module 400 determines the optimal dosage for the Phase III trial.

[0019] Phase III Clinical Simulation: The virtual sample expansion unit expands the virtual patient population to over 200 cases per group. The drug metabolism simulation module 200 simulates a randomized controlled trial of the target cardiovascular drug against a placebo or positive control drug, using a double-blind design. The efficacy / adverse reaction prediction module 300 outputs confirmatory efficacy (incidence of the primary endpoint event) and safety data. The closed-loop feedback optimization module 400 adjusts the trial protocol parameters based on interim analysis results.

[0020] Phase IV Clinical Simulation: A virtual patient population (including complex situations such as concomitant medications and comorbidities) constructed based on real-world data, with no fewer than 500 cases in each group. The drug metabolism simulation module 200 simulates the efficacy and adverse reaction monitoring of a large-scale population after the target cardiovascular drug's market launch, with a simulation period of 1-3 years. The efficacy / adverse reaction prediction module 300 outputs real-world efficacy and rare adverse reaction signals.

[0021] Example 3: Specific numerical examples of drug metabolism simulation

[0022] Taking a certain anticoagulant drug (molecular weight 500-600 Da, lipid solubility LogP=2.5, protein binding rate 90%) as an example:

[0023] The hemodynamic coupling unit performs the following calculations: It acquires the blood flow velocity field v(x,y,z) and pressure field p(x,y,z) of the vascular digital twin, and solves the coupled equations of the Navier-Stokes equations and the convection-diffusion equations using the finite volume method. Boundary conditions are set as follows: inlet velocity is 0.3-0.5 m / s (calculated based on a cardiac output of 5 L / min and the vessel cross-sectional area), and outlet pressure is set to 80% of the mean arterial pressure. The wall shear stress τ is calculated using τ = μ·(∂u / ∂y), where μ is the blood viscosity (3.5 cP), and ∂u / ∂y is the velocity gradient.

[0024] The tissue distribution calculation unit calculates the transwall permeability using the Kedem-Katchalsky equation: Jv = Lp·(ΔP - σ·Δπ), where Lp is the hydraulic conductivity coefficient (1.2 × 10⁻⁻⁶). 7(cm / s / mmHg), σ is the reflection coefficient (0.8), ΔP is the transwall pressure difference, and Δπ is the colloidal osmotic pressure difference. Based on the above calculations, the pharmacokinetic fitting unit uses a two-compartment model to perform nonlinear fitting on the concentration-time data. The fitting equation is: C(t)=A·e^(-αt)+B·e^(-βt), where A and B are the intercepts of the distribution phase and the elimination phase, and α and β are the rate constants of the distribution phase and the elimination phase, respectively. The target parameters obtained from the fitting include: distribution half-life t1 / 2α=0.5h, elimination half-life t1 / 2β=12h, apparent volume of distribution Vd=0.2L / kg, and clearance rate CL=0.8L / h.

[0025] The efficacy / adverse reaction prediction module 300 calculates the efficacy score based on the above parameters: target vessel diameter change rate ΔD = (D_post - D_pre) / D_pre. If ΔD ≥ 15%, the score is excellent (80-100 points); if ΔD is between 5% and 15%, the score is good (50-79 points); and if ΔD < 5%, the score is poor (0-49 points). Adverse reaction risk level: The ratio of the target organ drug concentration C_tissue to the preset safe upper limit C_safe is calculated as R = C_tissue / C_safe. If R < 0.5, it is low risk; 0.5 ≤ R ≤ 0.8, it is medium risk; and R > 0.8, it is high risk. Simultaneously, the ratio of the peak blood drug concentration Cmax to the preset toxic concentration C_toxic is calculated. If Cmax / C_toxic ≥ 0.8, it is directly judged as high risk.

[0026] Example 4: Iterative process of closed-loop feedback optimization

[0027] The parameter sensitivity analysis unit of the closed-loop feedback optimization module 400 performs the following steps: It sets vessel diameter (-20% to +20%), stenosis rate (30%-90%), drug clearance rate (0.5-1.5 times the range), and distribution volume (0.5-1.5 times the range) as candidate parameters. It generates 200 parameter combinations using Latin hypercube sampling. Each parameter combination is input into the drug metabolism simulation module 200 and the efficacy / adverse reaction prediction module 300 to calculate the corresponding efficacy score and adverse reaction risk level. The Sobol index method is used to calculate the first-order sensitivity and total-order sensitivity of each parameter, identifying stenosis rate and drug clearance rate as the most sensitive parameters (sensitivity > 0.3) that have the greatest impact on the efficacy score.

[0028] The optimization algorithm unit employs Bayesian optimization to perform the following iterations: Initially, 20 parameter combinations are generated through random sampling as training data for the Gaussian process regression model, with the Matern 5 / 2 kernel selected as the Gaussian process kernel function. The expectation improvement (EI) function is used for data acquisition, recommending new parameter combinations in each iteration. The multi-objective optimization function is to maximize the efficacy score (target value > 85 points) and minimize the adverse reaction risk level (target value < 30 points), iteratively updating the values ​​of sensitive parameters. Typically, after 30-50 iterations, the Pareto front of the optimization objective function converges. The trial protocol generation unit generates an optimized virtual trial protocol based on the optimized parameters, including: inclusion criteria adjusted to a stenosis rate of 50%-70%, age 45-75 years, and no severe hepatic or renal insufficiency; the dosing regimen adjusted to a loading dose of 300 mg and a maintenance dose of 75 mg / day; and the observation endpoint being the composite endpoint of major adverse cardiovascular events (MACE).

[0029] Example 5: Specific Implementation of Model Validation

[0030] Before step S1, a model validation step is performed: historical clinical trial data for marketed anticoagulants (such as warfarin and rivaroxaban) are obtained, including patient baseline characteristics, dosing regimens, pharmacokinetic data, efficacy endpoints, and safety data for each phase of clinical trials. This data is input into the system for retrospective simulation, calculating the Pearson correlation coefficient (r) between the system-predicted efficacy score and the actual results of historical clinical trials, as well as the Kappa coefficient (Kappa) between the adverse reaction risk level and the actual incidence of adverse events in historical clinical trials. When the Kappa coefficient is lower than a preset threshold (r < 0.80 or Kappa < 0.75), validation is re-executed by adjusting the vascular parameter generation strategy of the virtual patient construction module (e.g., adjusting the noise level of virtual sample amplification) and / or the pharmacokinetic model parameters of the drug metabolism simulation module (e.g., adjusting the clearance model coefficient) until the threshold requirements are met. In a specific validation example, the system-predicted incidence of the primary efficacy endpoint in the rivaroxaban phase III trial deviated from the actual trial results by 6.8%, and the predicted incidence of bleeding events deviated from the actual results by 9.2%, both lower than the preset 15% deviation threshold.

[0031] Example 6: System Layered Architecture

[0032] This system adopts a four-tier layered architecture (see...). Figure 3The physical layer is configured to provide GPU cluster computing resources (no less than 4 NVIDIA A100 GPUs) and distributed data storage resources (no less than 100TB); the data layer is configured to store CTA image data (DICOM format), a database of physicochemical parameters of target cardiovascular drugs (including molecular weight, LogP, pKa, protein binding rate, and metabolic enzyme information), and historical clinical trial data (including trial design schemes and results for each phase); the digital twin layer is configured to deploy a virtual patient construction module and a drug metabolism simulation module, using Docker containerization and Kubernetes for elastic scaling management; the application layer is configured to deploy an efficacy / adverse reaction prediction module and a closed-loop feedback optimization module, and provide a web-based virtual trial scheme configuration interface and a results visualization dashboard (including drug-time curves, dose-response relationship diagrams, adverse reaction risk distribution diagrams, etc.).

[0033] Example 7: Material and Data Source Description

[0034] CTA imaging data were sourced from the coronary CTA examination database of partner hospitals, and data acquisition complied with the Declaration of Helsinki and relevant ethical guidelines. All CTA imaging data underwent anonymization before importation into the system: direct identifiers such as patient name, hospital number, and ID number were removed, retaining only baseline characteristics such as age, gender, height, weight, and blood pressure. Physicochemical parameters of the target cardiovascular drugs were obtained from publicly available drug databases (such as DrugBank and PubChem). Initial parameters of the pharmacokinetic model were referenced from literature values ​​and calibrated to the measured data for the specific drug through model validation procedures.

Claims

1. A virtual clinical trial system for cardiovascular and cerebrovascular drugs, characterized in that, include: The virtual patient construction module is configured to: acquire cardiovascular and cerebrovascular CTA image data of the target patient, perform three-dimensional vascular reconstruction based on the CTA image data, and generate a vascular digital twin; the vascular digital twin includes vascular geometric parameters and hemodynamic boundary conditions; The drug metabolism simulation module is connected to the virtual patient construction module and is configured to: acquire the vascular digital twin, perform virtual simulation of drug distribution and metabolism in the vascular digital twin based on the physicochemical parameters and pharmacokinetic model of the target cardiovascular drug, and output the drug-time curve and drug concentration in the target organ. The efficacy / adverse reaction prediction module is connected to the drug metabolism simulation module and is configured to predict efficacy scores and adverse reaction risk levels based on the drug-time curve and target organ drug concentration, combined with preset cardiovascular and cerebrovascular drug efficacy evaluation indicators and adverse reaction thresholds. The closed-loop feedback optimization module is connected to the virtual patient construction module, the drug metabolism simulation module, and the efficacy / adverse reaction prediction module, respectively, and is configured to: adjust the vascular geometry parameters of the virtual patient construction module and / or the pharmacokinetic model parameters of the drug metabolism simulation module according to the efficacy score and the adverse reaction risk level, and generate an optimized virtual trial plan. The virtual patient construction module, drug metabolism simulation module, efficacy / adverse reaction prediction module, and closed-loop feedback optimization module constitute an iterative optimization loop, which is used to simulate the virtual trials of the target cardiovascular drug at each stage from clinical phase I to clinical phase IV.

2. The virtual clinical trial system for cardiovascular and cerebrovascular drugs according to claim 1, characterized in that, The virtual patient construction module includes: The CTA image segmentation unit uses a deep learning segmentation model to segment the cardiovascular and cerebrovascular regions of the CTA image data and obtain a vascular segmentation mask. The three-dimensional reconstruction unit performs surface reconstruction and mesh generation based on the blood vessel segmentation mask to generate a three-dimensional blood vessel geometric model. The boundary condition setting unit sets the inlet flow boundary conditions and outlet pressure boundary conditions of the three-dimensional vascular geometry model based on the physiological parameters of the target patient. The virtual sample amplification unit uses a generative adversarial network to amplify the CTA image data, generating virtual patient CTA images with different degrees of vascular lesions and different vascular configurations, in order to construct a virtual patient population.

3. The virtual clinical trial system for cardiovascular and cerebrovascular drugs according to claim 1, characterized in that, The drug metabolism simulation module includes: The drug parameter input unit is configured to receive information on the molecular weight, lipid solubility, protein binding rate, and metabolic enzyme substrate of the target cardiovascular drug. The hemodynamic coupling unit couples the blood flow velocity field and pressure field of the digital twin of the blood vessel with the drug convection and diffusion equation to calculate the convection and diffusion transport of the drug in the blood vessel lumen. The tissue distribution calculation unit calculates the trans-wall permeation amount of drug from the blood vessel lumen to the blood vessel wall and surrounding tissues based on the blood vessel wall permeability parameters of the blood vessel digital twin. The pharmacokinetic fitting unit fits the drug concentration-time data based on a multi-compartment pharmacokinetic model to generate the drug-time curve.

4. The virtual clinical trial system for cardiovascular and cerebrovascular drugs according to claim 1, characterized in that, The efficacy evaluation indicators include at least one of the following: the target vessel diameter change rate, the change in blood flow reserve fraction, and the vascular endothelial function index based on the vascular digital twin; the adverse reaction thresholds include: the comparison result of the drug concentration in the target organ with the preset safe upper limit, and the comparison result of the peak blood drug concentration calculated based on the drug-time curve with the preset toxic concentration.

5. The virtual clinical trial system for cardiovascular and cerebrovascular drugs according to claim 1, characterized in that, The closed-loop feedback optimization module includes: The parameter sensitivity analysis unit is used to determine the most sensitive parameters among the vascular geometry parameters and pharmacokinetic model parameters that have the greatest impact on the efficacy score and adverse reaction risk level. The optimization algorithm unit uses Bayesian optimization or multi-objective genetic algorithm to iteratively update the values ​​of the sensitive parameters with the optimization objectives of maximizing the efficacy score and minimizing the adverse reaction risk level. The trial protocol generation unit generates the optimized virtual trial protocol based on the optimized sensitivity parameters. The virtual trial protocol includes at least: virtual patient inclusion criteria, dosing regimen, and observation endpoint.

6. The virtual clinical trial system for cardiovascular and cerebrovascular drugs according to claim 1, characterized in that, The virtual trials for each stage from clinical phase I to clinical phase IV include: Phase I clinical simulation: Based on a digital twin of the blood vessels of a healthy virtual patient, the single-dose escalation tolerance test and pharmacokinetic characteristics of the target cardiovascular drug were simulated. Phase II clinical simulation: Based on the digital twin of the blood vessels of a virtual patient with the target indication, a multi-dose exploratory trial of the target cardiovascular drug was simulated to obtain dose-response relationship and preliminary efficacy. Phase III clinical simulation: Based on an expanded virtual patient population, simulate randomized controlled trials of the target cardiovascular drug against placebo or positive control drugs to obtain confirmatory efficacy and safety data; Phase IV clinical simulation: A virtual patient population constructed based on real-world data to simulate the efficacy and adverse reaction monitoring of the target cardiovascular drug in a large-scale population after its market launch.

7. The virtual clinical trial system for cardiovascular and cerebrovascular drugs according to claim 1, characterized in that, The system adopts a layered architecture, including: The physical layer is configured to provide GPU cluster computing and data storage resources; The data layer is configured to store the CTA image data, the physicochemical parameter database of the target cardiovascular and cerebrovascular drugs, and historical clinical trial data; A digital twin layer is configured to deploy the virtual patient building module and the drug metabolism simulation module; The application layer is configured to deploy the efficacy / adverse reaction prediction module and the closed-loop feedback optimization module, and to provide a virtual trial protocol configuration and result visualization interface.

8. A method for virtual clinical trials of cardiovascular and cerebrovascular drugs, characterized in that, The system described in any one of claims 1 to 7 is used to perform the following steps: S1: Acquire cardiovascular and cerebrovascular CTA image data of the target patient, and construct a digital twin of the blood vessels based on the CTA image data; S2: Based on the physicochemical parameters and pharmacokinetic model of the target cardiovascular drug, a virtual simulation of drug distribution and metabolism is performed in the digital twin of the blood vessels to generate drug-time curves and drug concentrations in target organs; S3: Based on the drug-time curve and target organ drug concentration, combined with preset cardiovascular and cerebrovascular drug efficacy evaluation indicators and adverse reaction thresholds, predict efficacy scores and adverse reaction risk levels. S4: Based on the efficacy score and adverse reaction risk level, adjust the parameters of the vascular digital twin and / or the pharmacokinetic model parameters to generate an optimized virtual trial protocol; S5: Repeat steps S1 to S4 until the preset optimization termination conditions are met, and output the final virtual clinical trial report.

9. The virtual clinical trial method for cardiovascular and cerebrovascular drugs according to claim 8, characterized in that, Before step S1, a model validation step is also included: retrospectively validating the system using historical clinical trial data of marketed cardiovascular and cerebrovascular drugs, calculating the consistency coefficient between the efficacy score and adverse reaction risk level predicted by the system and the actual results of historical clinical trials; when the consistency coefficient is lower than a preset threshold, adjusting the parameters of the virtual patient construction module and / or drug metabolism simulation module.

10. The virtual clinical trial method for cardiovascular and cerebrovascular drugs according to claim 8, characterized in that, The target cardiovascular and cerebrovascular drugs are selected from one of the following: anticoagulants, antiplatelet drugs, lipid-lowering drugs, and antihypertensive drugs; the cardiovascular and cerebrovascular CTA imaging data of the virtual patient group include at least one of the following: coronary artery CTA images, head and neck CTA images, or aortic CTA images.

Images