A microfluidic system and method for glycolipid drug testing

Abstract

The application discloses a kind of blood vessel microfluidic device and method suitable for glycolipid drug test, the device constructs three-dimensional biomimetic microvessel chip using induced pluripotent stem cells, and integrates multiple sensors;Through integrated environmental control module, programmed blood flow loading and drug exposure are realized.Intelligent analysis system includes multimodal intelligent observation module, based on microscope imaging and artificial intelligence speed measurement technology, three-dimensional flow field is inverted;Microcirculation digital twin module reconstructs blood vessels based on medical imaging and constructs calibratable hemodynamic model, for simulation and prediction;And wet-dry interaction module, through two-way interface, realizes "experiment-simulation" closed loop, injects entity data into digital twin to calibrate, and converts the prediction instruction of digital twin into chip control command to drive verification experiment.The system realizes complete technical chain from biomimetic model construction, intelligent observation, digital twin simulation to active experiment optimization.

Description

Technical Field

[0001] This application belongs to the interdisciplinary field of biomedical engineering and microfluidics, specifically relating to a vascular microfluidic system and method for testing glycolipid drugs. Background Technology

[0002] Glucose and lipid metabolism disorders such as diabetes and atherosclerosis are serious global public health problems that threaten human health, and their pathological processes are closely related to microcirculatory dysfunction. As the final site for the exchange of substances between blood and tissues, abnormal hemodynamics in the microcirculation (such as blood flow turbulence and shear stress imbalance) is a key factor leading to tissue ischemia and hypoxia, vascular lesions, and even damage to target organs such as the heart, brain, and kidneys. Clinical studies have confirmed that abnormal glucose metabolism (prediabetes and diabetes) is directly related to microcirculatory disorders in organs such as the skin and subclinical atherosclerosis, and that macrovascular complications of diabetes significantly increase the risk of microvascular complications. Therefore, studying blood rheological properties and evaluating the effects of drugs on improving microcirculatory function in a biomimetic and controllable microenvironment is of great significance for the development of new drugs and personalized treatment of glucose and lipid metabolism disorders.

[0003] Currently, research and drug testing on microcirculatory dysfunction in glucose and lipid metabolism diseases mainly rely on animal models, traditional in vitro models, microfluidic chips, and various observation and analysis techniques. However, existing microcirculation research and drug testing models have certain limitations:

[0004] Animal models are the cornerstone of traditional pathophysiological research and preclinical drug testing, providing a systematic biological environment. However, differences in vascular anatomy, hemodynamics, and metabolic pathways between species often lead to serious biases when translating data into human models. Furthermore, high costs, ethical constraints, and technical limitations preventing real-time, non-destructive, high-resolution dynamic observation of deep microcirculation make it difficult to accurately capture transient blood flow-cell interactions.

[0005] While traditional in vitro models (such as two-dimensional static cell culture) avoid species differences, their limitations are more pronounced. They are completely unable to simulate the three-dimensional topology of blood vessels in vivo and physiologically relevant mechanical stimuli such as blood flow shear forces. The morphology and function of cells on a rigid plane differ greatly from their in vivo state, resulting in poor predictability of drug response data based on such models. Many candidate drugs that are "effective" in two-dimensional culture fail to reproduce their efficacy in the complex in vivo environment.

[0006] Existing microfluidic chip models (Organ-on-a-Chip, OOC) have partially overcome the aforementioned shortcomings, enabling the reconstruction of three-dimensional environments and the application of fluid shear forces at the micrometer scale. However, most current systems are functionally limited: some focus on structural biomimicry but lack multi-parameter real-time sensing capabilities, while others can only perform endpoint destructive analysis and cannot achieve long-term dynamic observation. Most critically, they generally lack the technical means to synchronously correlate physical stimuli (blood flow), cellular functional responses, and tissue morphological changes with high spatiotemporal resolution, resulting in fragmented data acquisition that makes it difficult to reveal the core biological process of "hemodynamic changes driving vascular remodeling."

[0007] The lag in observation and analysis technologies further exacerbates research bottlenecks. Traditional microscopes struggle to achieve high-quality three-dimensional imaging under rapid flow conditions. Standard microparticle image velocimetry (μPIV) typically only provides two-dimensional cross-sectional information, making it difficult to reconstruct the full flow field within complex three-dimensional networks. Simultaneously, data analysis is highly reliant on manual intervention, lacking the ability to automatically and intelligently extract key features from multimodal data. Although artificial intelligence has begun to be applied to microfluidic systems for intelligent detection, existing methods are mostly in the offline, static analysis stage, unable to form a real-time closed loop with the experimental system, and lack mechanisms to deeply embed physical laws to ensure the rationality of predictions. Summary of the Invention

[0008] This application provides a vascular microfluidic system and method for testing glycolipid drugs, in order to address the limitations of traditional in vitro drug screening models.

[0009] To solve the above-mentioned technical problems, one technical solution adopted in this application is: a vascular microfluidic system for glycolipid drug testing, comprising:

[0010] Bionic microcirculation components are used to construct highly realistic three-dimensional vascular organoids and microfluidic culture environments to execute programmed intervention programs.

[0011] The multimodal intelligent observation component, connected to the biomimetic microcirculation component, can simultaneously collect hemodynamic, cellular function, and environmental parameters;

[0012] The digital twin component is connected to the multimodal intelligent observation component, whereby the digital twin component is used to build a virtual simulation platform that is synchronized with the physical entity in real time;

[0013] The dry-wet closed-loop interaction component is connected to the digital twin component. The dry-wet closed-loop interaction component is used to realize two-way data interaction and adaptive optimization between physical experiments and digital simulations.

[0014] Furthermore, the biomimetic microcirculation component includes:

[0015] A dry-wet closed-loop interactive component for preparing three-dimensional vascular organoids by inducing pluripotent stem cells;

[0016] A microfluidic chip module is located in a three-dimensional vascular organoid, which is used to collect data within the three-dimensional vascular organoid.

[0017] The perfusion culture module is used to simulate physiological blood flow shear force;

[0018] The AI ​​analysis module performs real-time acquisition and analysis of hemodynamic data based on selective plane illumination microscopy and microparticle image velocimetry.

[0019] Furthermore, the dry and wet closed-loop interaction components include:

[0020] iPSCs aggregate forming units are used with Matrigel basement membranes and aggregate culture media;

[0021] Mesodermal directional differentiation unit, used with differentiation medium containing CHIR99021 and BMP-4;

[0022] Cellular angiogenesis lineage directed differentiation unit, used in differentiation medium containing VEGF-A, Forskolin and A83-01;

[0023] Vascular sprouting and network forming units are used to enhance simulated blood flow shear forces through mechanical stimulation;

[0024] Vascular organoid forming units are used to extract vascular networks through collagenase treatment.

[0025] Furthermore, the microfluidic chip module includes an upper chip and a lower chip. The upper chip contains microchannels, and the lower chip contains organoid culture chambers. Multiple pairs of ITO transparent electrodes are symmetrically arranged on both sides of the microchannels, and a ring-shaped ITO electrode array is integrated at the bottom of the chamber. It also incorporates a miniature pressure sensor, flow sensor, temperature sensor, and pH sensor.

[0026] Furthermore, the perfusion culture module includes a pneumatic main pump for generating physiological pulsating flow; the perfusion culture module monitors parameters in real time through sensors and dynamically adjusts the flow rate and drug administration regimen.

[0027] Furthermore, the AI ​​analysis module is used to capture blood flow imaging data, sensor data, and Stokes equations to achieve multimodal automatic analysis of two-dimensional image data and infer three-dimensional flow fields, obtaining key hemorheological parameters including velocity field, pressure field, wall shear force, and hematocrit distribution.

[0028] One technical solution adopted in this application is: a method for testing glycolipid drugs using the above-mentioned vascular microfluidic system, comprising:

[0029] S1. First, prepare vascular organoids using the vascular network formed by the aggregation and differentiation of iPSCs;

[0030] S2. Fabrication of a microfluidic chip with a bilayer structure;

[0031] S3. Inject the mixture of vascular organoids and fibrin hydrogel into the microfluidic chip culture chamber, solidify at 37°C for 30 minutes, connect to the perfusion system, and culture at a shear stress pulsating flow of 1-2 dyn / cm² for 72 hours;

[0032] S4. Stepwise increase the shear stress to 4-6 dyn / cm², replace with high-glucose medium containing 25mM glucose, and culture for 7 days to construct a diabetic pathological model;

[0033] S5. The candidate drug was injected in a gradient manner through a multi-channel micropump, with 3 concentration groups and 1 control group, and ≥3 parallel samples in each group;

[0034] S6. Start the SPIM-μPIV fusion imaging system to simultaneously acquire flow field data, multi-channel fluorescence images and sensor data, with an acquisition cycle of 30s / time and continuous monitoring for 72 hours;

[0035] S7. The physical information neural network inverts the three-dimensional flow field, and the digital twin model simulates the drug intervention effect, outputting the drug dose-effect relationship, safety window, microcirculation function improvement score and toxicity risk assessment results.

[0036] Furthermore, the method for preparing vascular organoids in step S1 includes:

[0037] S11. iPSC aggregate formation: Matrigel-coated culture, iPSC inoculum density 2×10⁻⁶ 5 Cells / mL, static culture for 24-72 hours to form homogeneous aggregates;

[0038] S12. Directed differentiation of mesoderm: cultured for 48 hours in a medium containing 3 μM CHIR99021 and 30 ng / ml BMP-4;

[0039] S13. Angiogenesis lineage differentiation: cultured in a medium containing 100 ng / mL VEGF-A, 10 μM Forskolin and 1 μM A83-01 for 72 hours;

[0040] S14. Induction of vascular sprouting: Dynamic culture on a 30rpm track shaker for 72 hours, with mechanical shear force applied;

[0041] S15. Three-dimensional network maturation: Matrigel and type I collagen were mixed in a 3:1 gel and embedded in a gel, then cultured under 3% O2 ​​hypoxia for 5 days;

[0042] S16. Organoid extraction: Treat with 1 mg / mL collagenase IV for 15 minutes, and screen 70 μm cells to obtain functional vascular organoids with a diameter of 100-300 μm.

[0043] Furthermore, the method for fabricating the microfluidic chip module in step S2 includes:

[0044] S21. Design a double-layer chip positive mold structure. Spin-coated silicon wafer with SU-8 photoresist with a thickness of 50μm is obtained by ultraviolet exposure and development.

[0045] S22.PDMS prepolymer and curing agent are mixed at a ratio of 10:1, poured onto the surface of the positive mold, and initially cured at 35℃ for 7 hours and then deeply cured at 60℃ for 4 hours.

[0046] S23. Magnetron sputtering deposits 100nm ITO thin film, photolithography defines electrode patterns, and screen printing creates conductive paths;

[0047] S24. Pressure, flow, temperature, and pH sensors are embedded in the lower PDMS chamber and connected to an external data interface;

[0048] S25. After the upper and lower PDMS layers are cleaned with 100W plasma for 30 seconds, they are aligned and bonded at 60℃ for 48 hours to complete the fabrication of the microfluidic chip.

[0049] Furthermore, the method in step S7 includes: using a physical information neural network that fuses ConvLSTM and Fourier neural operators, embedding Stokes equation physical constraints, inputting two-dimensional μPIV particle image data, and outputting three-dimensional velocity field, pressure field, wall shear stress distribution and hematocrit evolution field, with a prediction error ≤5%.

[0050] The beneficial effects of this application are:

[0051] This application prepares three-dimensional vascular organoids with chamber structures through iPSC-directed differentiation and mechanical stimulation, and combines them with PDMS microfluidic chips that integrate multi-parameter sensing and programmable fluid control to construct an advanced in vitro model that highly simulates the real human microcirculation in terms of cell composition, three-dimensional architecture and hemodynamic environment.

[0052] This application The fusion imaging system supports long-term, three-dimensional, multi-channel simultaneous observation of fluorescence and flow fields; innovative The technology is based on physical information neural networks, which enables intelligent inversion of three-dimensional flow fields and key blood rheological parameters from limited two-dimensional data, solving the problem that traditional methods cannot obtain three-dimensional blood dynamic information in real time.

[0053] This application utilizes digital twin technology to construct a virtual microcirculation that is synchronized with and continuously calibrated in real time with the physical entity. Combined with an AI time-series prediction model, it enables dynamic extrapolation and prospective assessment of drug effects and pathological evolution. The unique dry-wet closed-loop interaction mechanism forms a complete iterative cycle of "experimental observation → digital extrapolation → intelligent decision-making → active verification → model optimization," which greatly improves the predictability, efficiency, and intelligence of drug screening.

[0054] This application can programmatically reconstruct the microenvironment of metabolic diseases such as high blood sugar, high lipids, and abnormal shear stress. Combined with multi-scale simulation of digital twins, it can accurately predict and evaluate the effect of drugs on microcirculation function and their potential vascular toxicity risks. It is particularly suitable for the early evaluation of the efficacy and safety of drugs such as hypoglycemic and lipid-regulating agents at the microcirculation level. Attached Figure Description

[0055] Figure 1 This is a schematic diagram of an embodiment of the vascular microfluidic system for glycolipid drug testing according to this application; Figure 2 yes Figure 1 A structural block diagram of an embodiment of the biomimetic microcirculation component; Figure 3 yes Figure 2 A schematic diagram of the structure of a microfluidic chip module in one embodiment. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to specific embodiments.

[0058] Numerous specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways than those described herein, and therefore the invention is not limited to the specific embodiments disclosed in the following specification.

[0059] See Figure 1 , Figure 1This is a schematic diagram of an embodiment of the vascular microfluidic system for glycolipid drug testing according to this application. The system includes: a biomimetic microcirculation component 1, a multimodal intelligent observation component 2, a digital twin component 3, and a wet-dry closed-loop interaction component 4. The biomimetic microcirculation component 1 is used to construct a highly realistic three-dimensional vascular organoid and microfluidic culture environment to execute programmed intervention protocols. The multimodal intelligent observation component 2 is connected to the biomimetic microcirculation component 1 and synchronously acquires hemodynamic, cellular function, and environmental parameters. The digital twin component 3 is connected to the multimodal intelligent observation component 2, and is used to construct a virtual simulation platform that is synchronized with the physical entity in real time. The wet-dry closed-loop interaction component 4 is connected to the digital twin component 3, and is used to realize bidirectional data interaction and adaptive optimization between physical experiments and digital simulations.

[0060] Among them, the biomimetic microcirculation component 1 includes: a vascular organoid preparation module 11, a microfluidic chip module 12, a perfusion culture module 13, and an AI analysis module 14.

[0061] The vascular organoid preparation module 11 forms a microvascular network with a chamber structure in a biomimetic hydrogel through multi-stage chemical induction and mechanical stimulation. Specifically, it includes an iPSC aggregate forming unit 111, a mesodermal-directed differentiation unit 112, a cell-vascular lineage-directed differentiation unit 113, a vessel sprouting and network formation unit 114, a three-dimensional vascular network maturation unit 115, and a vascular organoid forming unit 116. The iPSC aggregate forming unit 111 uses a Matrigel basement membrane and aggregation medium. The mesodermal-directed differentiation unit 112 uses a differentiation medium containing CHIR99021 and BMP-4; the cell-vascular lineage-directed differentiation unit 113 uses a differentiation medium containing VEGF-A, Forskolin, and A83-01; the vessel sprouting and network formation unit 114 enhances the simulation of blood flow shear force through mechanical stimulation; and the vascular organoid forming unit 116 extracts the vascular network through collagenase treatment.

[0062] The microfluidic chip module 12 is located in the three-dimensional vascular organoid, whereby the microfluidic chip module 12 is used to collect data within the three-dimensional vascular organoid; the microfluidic chip module 12 includes an upper chip 121 and a lower chip 122, the upper chip 121 includes a microchannel 1211, and the lower chip 122 includes an organoid culture chamber 1221; multiple pairs of ITO transparent electrodes 123 are symmetrically arranged on both sides of the microchannel 1211, and a ring ITO electrode array 124 is integrated at the bottom of the chamber 1221, and a micro pressure sensor 125, a flow sensor 126, a temperature sensor 127, and a pH sensor 128 are also built in.

[0063] The perfusion culture module 13 is used to simulate physiological blood flow shear force; the perfusion culture module 13 includes a pneumatic main pump for generating physiological pulsating flow; the perfusion culture module monitors parameters in real time through sensors and dynamically adjusts the flow rate and drug administration regimen.

[0064] The AI ​​analysis module 14 performs real-time acquisition and analysis of hemodynamic data based on selective planar illumination microscopy and microparticle image velocimetry. The AI ​​analysis module 14 uses the captured blood flow imaging data, sensor data, and Stokes equations to achieve multimodal automatic analysis of two-dimensional image data and inference of the three-dimensional flow field, obtaining key hemorheological parameters including velocity field, pressure field, wall shear force, and hematocrit distribution.

[0065] This application also provides a method for testing glycolipid drugs using the above-mentioned vascular microfluidic system, including:

[0066] S1. First, prepare vascular organoids using the vascular network formed by the aggregation and differentiation of iPSCs.

[0067] Specifically, in the protocol for the formation of induced pluripotent stem cell (iPSC) aggregates, s1 includes the following steps:

[0068] S11. Inducing the formation of aggregates of pluripotent stem cells (IPCs):

[0069] Matrigel basement membrane preparation:

[0070] Thawing and dilution: Transfer Matrigel from -20°C to a 4°C ice box to thaw for 3-4 hours, ensuring the entire process is performed on ice to prevent premature gelation of Matrigel during operation. Use a pre-chilled P1000 pipette to dilute Matrigel with Duchenne Modified Eagle Medium and F-12 Nutrient Mixture (1:1) (DMEM / F-12) at a volume ratio of 1:50, avoiding the generation of air bubbles.

[0071] Spreading and curing: The diluted Matrigel was spread evenly in a six-well plate at a rate of 1.5 μL / cm², shaken horizontally for 30 seconds, and then placed in a 37°C, 5% CO2 incubator to cure for 15 minutes to form a base film of uniform thickness.

[0072] Preparation of S11.iPSCs single-cell suspension:

[0073] Dissociation and resuspension: iPSCs (NC8 strain) were pretreated with 0.5 mM ethylenediaminetetraacetic acid (EDTA) for 2 minutes, followed by resuspension with 0.1%... Digest at 37°C for 3 minutes, then mechanically pipette to form a single-cell suspension;

[0074] Aggregate formation: Cells / mL resuspended in a solution containing Cell suspensions were obtained in DMEM / F-12:20% KOSR aggregation medium and seeded in Matrigel-coated 6-well plates at 37°C. Incubate in CO2 statically for 24–72 hours to allow the cells to spontaneously form aggregates of uniform size.

[0075] S12. Directed differentiation of the mesoderm;

[0076] Differentiation medium 1 preparation: Use N2B27 basal medium supplemented with 3μM CHIR99021 and 30ng / ml BMP-4, adjusting the osmotic pressure to... pH 7.4;

[0077] Differentiation culture: Use P1000 pipettes to transfer cell aggregates to differentiation medium 1, 3 mL per well, and incubate at 37°C, 5% CO2 for 48 hours. Change the medium every 12 hours to maintain the activity of differentiation factors.

[0078] S13. Directed differentiation of cellular angiogenesis lineages:

[0079] Differentiation medium 2 preparation: N2B27 medium containing 100 ng / mL vascular endothelial growth factor A (VEGF-A), 10 μM Forskolin, and 1 μM A83-01 (TGF-β inhibitor) was used to activate the differentiation medium. and Signaling pathways promote endothelial precursor cell differentiation.

[0080] Differentiation culture: Cell aggregates were collected using the above method, and the cell aggregates were resuspended in differentiation medium 2 using a pipette to obtain a cell suspension. 3 ml of the cell suspension was added to each well of a six-well plate and cultured for 72 hours.

[0081] S14. Blood vessel sprouting and network formation:

[0082] Differentiation medium 3 preparation: A medium containing bone morphogenetic protein 4 (BMP-4) (30 ng / ml), VEGF-A (30 ng / ml) and fibroblast growth factor 2 (FGF-2) (50 ng / ml) was used to stimulate angiogenesis and branching.

[0083] Mechanical stimulation enhancement: 3 ml of cell suspension was added to each well of a six-well plate, and the six-well plate was placed on a track shaker (30 rpm, 1 mm amplitude) for dynamic culture to simulate the shear force of blood flow in vivo. The medium was changed every 24 hours.

[0084] S15. Maturation of the three-dimensional vascular network:

[0085] Composite gel embedding:

[0086] Mix Matrigel and Collagen I at a 3:1 volume ratio on ice, then add 10 mM 4-(2-hydroxyethyl)piperazine-1-ethanesulfonic acid (HEPES) (pH 7.4) to adjust the osmotic pressure. To avoid bubble formation, after embedding the cell aggregates into the gel, pre-curing was performed at 25°C for 30 minutes, followed by final curing at 37°C for 1 hour, forming a layer with a thickness of [missing information]. 3D support frame.

[0087] Functional maturity development:

[0088] Differentiation medium 4 was prepared using a mature medium containing 15% fetal bovine serum (FBS), 100 ng / mL VEGF-A and 100 ng / mL FGF-2, supplemented with SB43152 (5 μM, a TGF-β inhibitor) to inhibit mesenchymal transition; cultured in a 3% O2 ​​hypoxic incubator for 5 days to promote metabolic adaptation and lumen formation of vascular endothelial cells, with the medium being changed every 24 hours.

[0089] S16. Induces the formation of vascular organoids;

[0090] On day 18, the cells were treated with 1 mg / mL collagenase IV at 37°C for 15 minutes, and the vascular network was released by gentle pipetting. The cells were washed three times with phosphate-buffered saline solution to remove residual gel. Complete vascular organoids (100-300 μm in diameter) were screened through a 70 μm cell sieve and transferred to a neuron-endothelial co-culture system (NEI) containing 10% FBS for pre-adaptation for 12 hours. The vascular network was established by extracting the vascular network of single cell aggregates from the gel and further culturing the vascular network in well plates to form vascular organoids through self-assembly.

[0091] S2. Fabrication of microfluidic cores

[0092] Specifically, in this embodiment, a microvascular microfluidic chip adapted for vascular organoid culture and detection is simultaneously prepared. This chip integrates ITO electrodes and microsensors, enabling fluid control and real-time monitoring. The specific steps are as follows:

[0093] S21. Positive mold preparation: The chip structure was designed using AutoCAD. The upper microchannel was 1 mm wide and 50 μm deep, and the lower organoid culture chamber was 3 mm in diameter and 200 μm deep. Three pairs of 100 μm wide ITO electrodes were symmetrically designed on both sides of the microchannel, and a 2 mm diameter ring-shaped ITO electrode array was designed at the bottom of the chamber. The designed layout was submitted for photolithography. SU-8 2050 photoresist was spin-coated on a 4-inch silicon wafer with a thickness of 50 μm. After UV exposure (dose 180 mJ / cm²), development (SU-8 developer, 3 minutes), and baking (150℃, 30 minutes), the photoresist positive mold was obtained.

[0094] S22. PDMS Casting: Mix PDMS prepolymer and curing agent (Dow Corning, Sylgard 184) thoroughly at a mass ratio of 10:1, vacuum for 30 minutes to remove air bubbles, and slowly cast onto the surface of the silicon male mold. Preliminary curing is carried out in an oven at 35°C for 7 hours, followed by deep curing at 60°C for 4 hours. After complete curing, the PDMS substrate is peeled off from the male mold, and fluid inlet / outlet and electrode lead holes are punched out using a biopsy punch.

[0095] S23. Electrode fabrication: The PDMS substrate is placed in a magnetron sputtering instrument, and a 10nm chromium adhesion layer and a 100nm ITO functional layer are deposited sequentially, with the sheet resistance controlled at 15-20Ω / □; the electrode pattern is defined by photolithography, excess ITO is removed by wet etching, and silver paste conductive path is prepared by screen printing to connect the electrode to the external lead.

[0096] S24. Sensor Integration: Embed a miniature pressure sensor (Measurement Specialties, MS5837), a miniature flow sensor (Sensirion, SLG075), a platinum resistance temperature sensor, and a pH sensor (Senova, pH-FET) in the reserved positions in the chamber of the lower PDMS substrate, and solder them to the external data acquisition circuit through conductive silver paste.

[0097] S25. Chip bonding: Place the upper and lower PDMS substrates into a plasma cleaner, process with 100W power for 30 seconds, remove and quickly align and bond, place in a 60℃ oven for 48 hours for bonding, and perform a leakage test after bonding: pass deionized water with a pressure of 200kPa, and hold the pressure for 30 minutes if there is no leakage, which is considered qualified.

[0098] The chip prepared in this embodiment can realize real-time monitoring of pressure (0-10kPa, accuracy ±2%), flow rate (0-100μL / min, accuracy ±5%), temperature (35-40℃, accuracy ±0.2℃), and pH (6.5-8.0, accuracy ±0.1). The ITO electrode can be used for impedance sensing and electrical stimulation experiments to meet the requirements of long-term dynamic culture.

[0099] S3. Inject the mixture of vascular organoids and fibrin hydrogel into the chip culture chamber of the microfluidic chip module, cure at 37°C for 30 minutes, connect to the perfusion system, and culture at a shear stress pulsating flow of 1-2 dyn / cm² for 72 hours.

[0100] Specifically, this step involves implanting the prepared vascular organoids into a microvascular microfluidic chip for cell seeding, multi-stage perfusion culture, and dynamic monitoring through staged fluorescent labeling, ultimately constructing a functional microvascular network. The specific method is as follows:

[0101] S31. Preparation of composite hydrogels and cell suspension formulation

[0102] Material pretreatment and vascular organoid transfer: Thrombin and fibrin hydrogel were stored at 4°C to prevent premature coagulation. Using a pre-chilled P1000 pipette, vascular organoids were gently aspirated from the six-well plate and transferred to EP tubes containing 1 mL of pre-chilled DMEM / F-12 medium. The tubes were centrifuged for 5 minutes to completely remove residual matrix gel and were ready for use.

[0103] Preparation of composite hydrogel: Centrifuged vascular organoids and thrombin are mixed evenly at a volume ratio of 1:10, followed by the addition of fibrin solution. The mixture is then gently stirred until a homogeneous composite hydrogel is formed, avoiding the formation of air bubbles that could damage the structure of the vascular organoids.

[0104] Preparation of cell suspension: Human umbilical vein endothelial cells (HUVECs) at a concentration of 6 × 10⁻⁶ were used. 6 (cells / ml) and fibroblasts (concentration 2×10⁻⁶) 6 The cells were mixed at a ratio of 3:1 (cells / ml) and resuspended in endothelial cell medium (ECM). The ECM medium was supplemented with 20% fetal bovine serum, 1% penicillin-streptomycin and 50 ng / mL VEGF-A. After preparation, the medium was placed in the reservoir of the microfluidic system for later use, and a sterile environment was maintained throughout the process.

[0105] S32. Chip preprocessing and electrode functionalization

[0106] Chip washing and passivation: The microvascular microfluidic chip prepared in step S27 was sequentially washed with 75% ethanol, sterile deionized water, and PBS buffer, ensuring no air bubbles remained in the channels at each step. The final wash used Minimum Essential Medium α-Modification (α-MEM) medium containing 15% FBS, filling the entire channel and allowing it to stand for 30 minutes to form a protein passivation layer on the inner wall of the channel, reducing cell adhesion loss.

[0107] Electrode functionalization: 1) Immerse the chip electrode area in ethanol for 30 minutes to activate the electrode surface; 2) Drop-coat a 1.5wt% poly(3,4-ethylenedioxythiophene)-polystyrene sulfonate (PEDOT:PSS) conductive polymer to uniformly cover the electrode surface; 3) Dry in a 45℃ vacuum drying oven to form a stable ion permeation layer; 4) Inject ECM medium containing 0.1mg / mL collagenase IV into the chip channel to passivate the electrode and improve its biocompatibility and detection stability.

[0108] S33. Composite hydrogel injection and curing

[0109] The pretreated chip was connected to the micropump control system via sterile tubing. The system control software was used to set a low-speed, stable, non-pulsating flow program for the main pneumatic pump, with a flow rate controlled at 50 μL / min, for 5 minutes. This flow field was used to smoothly inject the vascular organoid-fibrin hydrogel composite into the target culture chamber of the chip, with precise flow rate control throughout to avoid air bubbles or shear damage. Immediately after injection, a "zero flow" program was executed to allow the composite hydrogel to fully solidify in a static environment, ensuring stable positioning of the vascular organoid within the chamber.

[0110] S34. Multi-stage perfusion culture and parameter monitoring

[0111] Establish perfusion connection: After the chip has been cured, it is connected to the micropump control system, independent drug reservoir and culture medium reservoir through sterile fluid pipeline to form a closed-loop perfusion system, ensuring that the pipeline connection is leak-free.

[0112] Perform a multi-stage perfusion culture procedure:

[0113] Phase 1 (Physiological State Simulation): Physiological pulsating flow is generated through the main pneumatic pump, and the average shear stress is set to... The pulsation frequency is (Simulating the human resting heart rate), continuous perfusion for 72 hours promotes the adaptation of vascular organoids to the chip microenvironment and the initial formation of a network structure.

[0114] Phase Two (Pathological State Simulation and Programmed Drug Administration): Stepwise increase of mean shear stress to It simulates the high shear stress pathological state commonly seen in diabetic patients; simultaneously, at specific time windows every 6 hours, it applies a continuous oscillating flow or intermittent reverse flow for 1 hour to simulate the low oscillating shear stress environment in areas prone to atherosclerosis. Through the system's integrated independent drug reservoir and control software, it achieves programmed drug delivery, precisely controlling the dosing time, dosage, and flow rate.

[0115] Parameter control and maintenance during perfusion: Shear stress, pressure, flow rate, pH, temperature gradient, and local oxygen partial pressure at the microvascular wall are monitored in real time using chip-integrated microsensors. Flow rate, pulsation frequency, or drug administration regimen are dynamically adjusted based on the monitoring data. Fresh culture medium containing 50 ng / mL VEGF-A and 10 ng / mL FGF-2 is replaced daily. Cell metabolic activity is monitored using a pH sensor, and the pH of the culture medium is maintained between 7.2 and 7.4. The dynamic culture period is 7 days, which can be appropriately extended according to experimental needs until a functionalized self-assembled microvascular network is formed.

[0116] Sample pretreatment: After perfusion culture, the perfusion medium was replaced with heparinized whole blood or blood simulant, and perfusion continued for 1-2 hours at physiologically relevant shear stress to stabilize the flow field environment and prepare samples for subsequent flow field data acquisition. Electron impedance spectroscopy was used to detect tissue electrical impedance in the organoid culture chamber to assess the integrity of the microvascular network.

[0117] S35. Staged Fluorescent Labeling and Dynamic Monitoring: To avoid fluorescence signal interference and achieve accurate monitoring at different stages, a staged fluorescent labeling strategy is adopted. The specific operation is as follows:

[0118] Reference markers (24 hours before the experiment): Culture medium containing Hoechst 33342 dye was injected into the chip at a flow rate of 10 μL / min using a perfusion system. After incubation for 30 minutes, the chip was continuously perfused with fresh dye-free culture medium for 30 minutes to remove residual dye. This dye was used to mark the location of cell nuclei, serving as a spatial positioning reference for subsequent imaging.

[0119] Dynamic monitoring markers (before and after drug intervention): First, Fluo-4 AM, used for monitoring calcium signaling, was loaded and incubated in dye-containing medium for 45 minutes, followed by perfusion with dye-free medium for 30 minutes; then, markers for detecting cell membrane morphology were loaded. Incubate for 15-20 minutes, rinse briefly, and begin imaging immediately; this stage allows for simultaneous monitoring of calcium signal dynamics and cell morphology changes.

[0120] Endpoint analysis markers (after the experiment): Mark the endpoints containing... Inject the mixed culture medium with PI into the chip, incubate for 20-30 minutes, and perform endpoint imaging directly without rinsing. The signal of live / dead cells is distinguished by spectral separation to complete the final assessment of microvascular network function and cell viability.

[0121] S4. Gradually increase the shear stress to 4-6 dyn / cm², replace with a high-glucose culture medium containing 25mM glucose, and culture for 7 days to construct a diabetic pathological model.

[0122] S5. The candidate drug was injected in a gradient manner through a multi-channel micropump, with 3 concentration groups and 1 control group, and ≥3 parallel samples in each group;

[0123] S6. Start the SPIM-μPIV fusion imaging system to simultaneously acquire flow field data, multi-channel fluorescence images and sensor data, with an acquisition cycle of 30s / time and continuous monitoring for 72 hours;

[0124] Specifically, this step involves building a microcirculation observation system based on selective plane illumination microscopy (SPIM) and microparticle image velocimetry (μPIV) technology, and simultaneously integrating artificial intelligence velocimetry (AIV) technology to achieve spatiotemporal coupled acquisition, intelligent analysis, and model optimization of hemodynamic and cellular function data. The specific methods are as follows:

[0125] S61. Construction of the SPIM-μPIV Fusion Microcirculation Observation System

[0126] The microcirculation system consists of an optical imaging module, a microcirculation simulation module, a data acquisition and analysis module, and a support device. These modules work together to achieve accurate acquisition and real-time control of multimodal data. The specific setup and calibration process is as follows:

[0127] Optical Imaging Module: Multiwavelength Selective Plane Illumination Microscope (SPIM) with High Sensitivity camera

[0128] Microcirculation simulation module: microvascular microfluidic chip, peristaltic pump, temperature control device

[0129] Data acquisition and analysis module: image acquisition card (used to transmit image data captured by the camera to the computer in real time), pressure sensor, flow sensor, synchronization control device, image processing software.

[0130] Supporting components: water tank, storage tank, water pipes, brackets and fixing devices

[0131] S62. Construction and Calibration of a Multi-Wavelength SPIM-μPIV Fusion Imaging System

[0132] System optical architecture: A selective plane illumination microscope system with a 45° incident angle is designed, constructing an orthogonal optical path structure, with the excitation and detection optical paths forming a 90° angle on the sample plane. The microvascular microfluidic chip is placed horizontally, and the excitation light penetrates the chip's thin-walled optical window (150μm polypropylene film) at a 45° angle. The detection objective lens collects optical signals vertically upwards, avoiding disturbance to the fluid environment inside the chip and ensuring flow field stability.

[0133] Multi-wavelength laser excitation system integration: The system is equipped with four independent lasers, which are spatially combined using a dichroic mirror to achieve switching between multi-channel fluorescence excitation and μPIV particle illumination; the parameters of each laser are as follows:

[0134] 1. 405 nm diode laser (power 10 ), used to stimulate Dye-labeled cell nuclei; 2. 488 nm solid-state laser (power 10 ), used to excite the calcium signal labeled with Fluo-4 AM dye and the scattered light from μPIV tracer particles; 3. 561 nm DPSS laser (power 20 ), as a backup channel, compatible with various orange / red fluorescent dyes; 4. 640 nm diode laser (power 15 ), used to stimulate Dye-labeled cell membranes. Each laser is equipped with a TTL modulation interface, via... The card enables independent on / off control, with a channel switching time of ≤5 seconds. This ensures synchronized timing of multimodal data acquisition.

[0135] Debugging of the light sheet generation module: The multi-wavelength laser beam after beam combining first passes through the beam expander system ( Diverging lens Converging lenses extend the beam diameter to Then through A focal length cylindrical lens is compressed to form a static light sheet. At the excitation wavelength, the thickness (full width at half maximum) of the optical sheet is controlled as follows: Effective lighting height The lighting width fully covers The chip observation area ensures overall imaging coverage of the sample.

[0136] Imaging module assembly: 1. The objective lens is a 20× ultra-long working distance objective lens (Mitutoyo M Plan ApoSL, numerical aperture NA 0.28, working distance WD 30.5 mm), paired with a 100 mm focal length tube lens to achieve a total magnification of 10×; 2. The imaging device adopts high sensitivity. The camera (Hamamatsu ORCA Flash 4.0) balances high quantum efficiency (>80%) and high frame rate performance (>100 fps at full resolution) to meet the needs of high-speed flow fields and dynamic fluorescence signal acquisition; 3. The filter system is equipped with a six-position automatic filter wheel and corresponding wavelength emission filter groups (such as 525 / 50 nm, 690 / 50 nm, etc.) to achieve precise separation of multi-channel fluorescence signals.

[0137] 3D scanning and synchronous control configuration: 1. The scanning system adopts a 3D precision translation stage ( (Series), with a positioning accuracy of 100 nm and a Z-axis scanning step size of 1-10. Adjustable, enabling three-dimensional imaging of samples; 2. Synchronous control is programmed based on the Laboratory Virtual Instrument Engineering Workbench (LabVIEW) platform, and controls the laser TTL switch, camera exposure, filter wheel switching, translation stage movement, and peristaltic pump operation through the NI-DAQmx card, with a trigger delay error ≤1. This ensures that all devices are synchronized in a timely manner.

[0138] System performance calibration: 1. Optical sheet performance calibration: 500 nm fluorescent microspheres ( The system is calibrated by embedding a hydrogel into a chip, measuring the point diffusion function via Z-axis scanning, and setting the system resolution to lateral. axial 2. Multi-channel spatial registration: Using multi-color fluorescent microsphere samples, images were acquired sequentially using lasers of different wavelengths. The spatial offsets of the 405 nm, 561 nm, and 640 nm channels relative to the 488 nm reference channel were calculated using a cross-correlation algorithm. 1. Establish a registration lookup table, and the software automatically corrects to ensure multi-channel alignment accuracy <1 pixel; 2. Noise suppression and distortion correction: Use multi-band filters to reduce background noise and ensure the signal-to-noise ratio of the particle image. ;pass Spatial calibration is performed using a standard grid plate to calculate pixel resolution. Combined with a dynamic distortion correction algorithm, optical aberrations are compensated to improve imaging accuracy.

[0139] Microcirculation simulation module flow control: The microfluidic chip inlet is connected to a high-precision peristaltic pump, and embedded micro pressure sensors and flow sensors are deployed at the inlet and sidewalls of the flow channel to collect local pressure and flow signals in real time, providing data support for flow field regulation; a temperature control module is used to maintain the fluid temperature inside the chip at 37.0±0.2℃, and the fluid temperature gradient is eliminated by preheating the reservoir, simulating the physiological temperature environment in vivo, and avoiding the impact of temperature fluctuations on cell activity and hemorheological properties.

[0140] Real-time control closed-loop establishment: The set control parameters (flow rate, pressure waveform) of the micro-pump system are used as the precise inlet boundary conditions for the digital twin fluid simulation. The digital twin model is based on... The technology monitors vascular network morphology (number of branches, length, and diameter) and flow field data in real time, compares and analyzes them with preset target parameters, and generates optimization and control suggestions. Control commands are automatically sent to the micropump control system via an API interface, dynamically adjusting its control program to form a closed-loop control system of "acquisition-analysis-optimization-control," ensuring that the flow field environment accurately matches experimental requirements.

[0141] S64. Multimodal Data Acquisition and Preprocessing

[0142] Time-division multiplexing acquisition timing design: Establish a periodic acquisition protocol, with a single acquisition cycle of 30 seconds, divided into three stages: 1. model( ): Enable Laser (full power) Camera with 1. High-speed continuous capture of 4 frames of particle scattering light images for flow field calculation; 2. Multicolor fluorescence mode ( ): Open in sequence (exposure (Collecting cell nuclear signals) (exposure (Collecting calcium signals) (exposure 1. Acquiring cell membrane signals) Laser to complete multi-channel fluorescence image acquisition; 2. Data processing and storage ( Real-time processing Image-based flow field parameters are calculated, and the flow field data is aligned with the fluorescence image by timestamp to ensure spatiotemporal coupling consistency.

[0143] High-speed response acquisition mode trigger: When the system detects a sudden change in the flow field or a drug intervention signal, it automatically switches to high-speed acquisition mode, increasing the acquisition frequency to [missing information]. Collect continuously for 3-5 minutes, prioritizing capture. Calcium signaling and Particle imaging accurately records instantaneous biophysical response processes.

[0144] Experimental endpoint analysis: After the long-term dynamic observation ended, through... Perform rapid 3D scanning and data acquisition (Living cells) and (Dead cell) fluorescence images were obtained, and the ratio of the red fluorescence (dead cell) area to the total red + green fluorescence area (total cells) was calculated to quantitatively assess cell mortality under different drug conditions, providing a quantitative basis for the analysis of experimental results.

[0145] Image acquisition parameter optimization and preprocessing: 1. Parameter optimization: Set the camera exposure time to the laser pulse interval. To avoid blurring of particle motion; adjust the laser pulse interval. Matching the fluid flow rate to ensure particle displacement remains within the cross-correlation window; controlling the tracer particle concentration to 0.1% volume fraction to maintain a particle density of 5-15 particles / cross-correlation window (window size 32×32 pixels) in a single frame image, ensuring the accuracy of flow field calculation; 2. Image preprocessing: Non-Local Means (NLM) filtering is used to remove high-frequency noise while preserving particle edge information; histogram equalization is used to enhance the contrast between particles and background, ensuring particle recognition rate. Based on particle diameter distribution statistics (target diameter) 1) Remove abnormally clustered or broken particle images; 2) Data integration and storage: Integrate the preprocessed image sequences, pressure signals, and flow data according to timestamps. The formatted database is labeled with experimental conditions such as flow rate, temperature, and shear rate, providing standardized data for subsequent AIV model training and validation.

[0146] S65. Intelligent Analysis of Blood Rheological Processes Based on AIV Technology: A multimodal data analysis model is constructed by fusing μPIV imaging data, sensor detection data, and Stokes equations using Physical Information Neural Networks (PINNs) to achieve automatic analysis of two-dimensional image data, three-dimensional flow field inference, and quantification of blood rheological parameters. The specific process is as follows:

[0147] Multimodal data preprocessing and feature extraction: Initial hematocrit distribution was extracted using the microscopic image analysis software ImageJ. The computation area is divided into a 32×32 grid to facilitate feature extraction and model calculation.

[0148] Spatiotemporal alignment and coordinate normalization: 1. Spatiotemporal alignment: Based on hardware-synchronized timestamps, ... Image sequence (Time resolution) Multichannel fluorescence image sequences , , Sensor data stream , , , Sampling rate and initial hematocrit distribution 1. Perform precise alignment to ensure data spatiotemporal consistency; 2. Coordinate normalization: normalize physical coordinates. Mapped to The interval, in which Corresponding to the bottom surface of the microchannel, The top surface simplifies the computational complexity of the model.

[0149] S66. Three-dimensional physical information neural network modeling

[0150] Network architecture design: adopting and The fusion architecture, coupled with a decoder, enables physical field output. The functions of each layer are as follows:

[0151] 1. Input layer: Receives normalized four-dimensional spacetime coordinates. , where x and y are planar spatial positions, z is the normalized depth coordinate, and t is the time coordinate;

[0152] 2. Hidden layer: Contains parallel spatiotemporal feature extraction paths and spatial feature modeling paths, ultimately achieving feature fusion;

[0153] 3. Output Layer: Maps the fused features to physical space, outputting a three-dimensional velocity field. Pressure field and hematocrit distribution .

[0154] Hidden layer design details: 1. Spatiotemporal feature extraction path ( Module): adopts multi-layer The structure captures spatiotemporal correlation features through a gating mechanism. The mathematical expression for the gating machine is:

[0155] ;

[0156] ;

[0157] ;

[0158] ;

[0159] ;

[0160] ;

[0161] In the formula, Indicates memory state, Candidate memory states This represents the input information. Indicates based on The obtained hidden state, This represents the activation function. These represent the weight matrices for the input gate, forget gate, and output gate, respectively. These represent the bias parameters for the input gate, forget gate, and output gate, respectively.

[0162] Spatial feature modeling path ( (Module): Using Fourier neural operators, a parameterized integral kernel is constructed in Fourier space:

[0163] ;

[0164] in and These represent the Fourier transform and its inverse transform, respectively. This is a frequency domain parameterized matrix.

[0165] Feature fusion: Output "temporal characteristics" Spatial characteristics compared to FNO output The features are concatenated, and the concatenated features are then passed through a 1×1 convolutional layer. The features are then fused to obtain a unified feature representation. .

[0166] Output layer: fused features Through a decoder consisting of two-dimensional convolutional layers, it is gradually mapped to physical space; three-dimensional velocity field. Pressure field and hematocrit distribution

[0167] Physical constraint embedding:

[0168] 1. Hard-coded boundary conditions: Physical boundary conditions are applied directly to the output layer to ensure that the model output conforms to the actual flow field.

[0169] 2. No-slip boundary conditions: ;

[0170] 3. Conditions for hematocrit fixation at the point of entry: ;

[0171] Soft constraints on physical equations: through loss functions The forced flow field satisfies the governing equations, including: 1. Momentum conservation equation ,in, As a deviatoric stress tensor, its calculation incorporates a dynamic viscosity model to describe the non-Newtonian rheological properties of blood: 2. Mass conservation equation 3. Blood cell migration equation .

[0172] Dynamic viscosity model: A dynamic viscosity model is constructed by combining hematocrit and shear rate.

[0173] ;

[0174] ;

[0175] in, It's plasma viscosity. It is the maximum hematocrit (usually taken as 0.6). These are model parameters (usually set to 2.5). This model accurately describes the relationship between blood viscosity and shear rate and blood cell concentration.

[0176] S67. Adaptive Loss Function Optimization: The total loss function is composed of a weighted average of data loss, physical loss, initial condition loss, and boundary condition loss. It is used to balance the model's data fitting ability and physical rationality. Its expression is as follows:

[0177] ;

[0178] in, and These are dynamically adjusted weighting coefficients.

[0179] The specific definitions of each loss term are as follows:

[0180] 1. Data loss The mean squared error (MSE) calculation is used to ensure the consistency between the model output and the measured data. The expression is:

[0181] ;

[0182] In the formula, For the sample size, To predict the velocity field for the model, For the actual measured velocity field, To predict the pressure field for the model, The measured pressure value from the sensor is used to constrain the model's fitting accuracy to the real flow field using this loss term.

[0183] 2. Physical loss By calculating the residuals of the momentum conservation equation, mass conservation equation, and blood cell migration equation, the predicted results are forced to conform to the laws of fluid mechanics and biophysics. The expression is as follows:

[0184] ;

[0185] In the formula, The number of sampling points is a physical constraint. Blood density, For blood dynamic viscosity, The hematocrit distribution is used as a loss term to avoid physically meaningless predictions from the model.

[0186] 3. Initial Condition Loss The constraint ensures that the model's initial output matches the measured initial data, guaranteeing the accuracy of the training starting point. The expression is:

[0187] ;

[0188] 4. Boundary condition loss The constraint model's output at the blood vessel boundary conforms to preset physical conditions (such as no slip boundary), expressed as:

[0189] ;

[0190] In the formula, , These represent the number of sampling points for the initial conditions and the boundary conditions, respectively. The initial velocity, Both are boundary velocities, determined by measured data or prior physiological knowledge.

[0191] To achieve balanced optimization of each loss term, a dynamic weight adjustment strategy is adopted: every 1000 iterations, the gradient magnitude of each loss term is calculated. ( (Representing the corresponding loss term), the weight coefficients are updated according to the following formula:

[0192] ;

[0193] By dynamically adjusting the weights, a balanced optimization between data fitting and physical constraints is ensured, thereby improving the model's accuracy and stability.

[0194] S68. Intelligent Data Interaction Interface Definition: Design a two-way data interaction interface to achieve efficient linkage between the physical system and the digital twin system, as detailed below:

[0195] Interface A (Data Assimilation Interface, Wet → Dry): Standardizes, timestamps, and encapsulates the multimodal real-time data generated by the microfluidic chip system. It integrates a Kalman filter data assimilation algorithm and uses real-time observation data to automatically and continuously calibrate key parameters of the digital twin (metabolic flux, vascular wall elasticity coefficient, pharmacokinetic parameters), eliminates state deviations between the digital twin and the physical entity, ensures consistency between the two states, and provides support for accurate simulation and prediction.

[0196] Interface B (Active Experiment Interface, Dry → Wet): Receives control commands from the digital twin system and translates them into executable operation commands for the microfluidic control system. These commands include adjusting the peristaltic pump flow rate and pulsation mode, switching perfusion fluids (e.g., switching between standard culture medium and high glucose culture medium), controlling the drug concentration and timing in the drug reservoir, and triggering specific imaging modes (e.g., high-speed acquisition mode). This enables the digital twin model to actively control physical experiments, improving the level of automation and intelligence in experiments.

[0197] S7. The physical information neural network inverts the three-dimensional flow field, and the digital twin model simulates the drug intervention effect, outputting the drug dose-effect relationship, safety window, microcirculation function improvement score and toxicity risk assessment results.

[0198] This application, through the aforementioned closely linked implementation steps, constructs a complete closed-loop system encompassing biomimetic model building, intelligent dynamic observation, digital twin simulation, and proactive experimental optimization. This ensures that the drug testing process not only highly simulates the physiological and pathological environment of the human microcirculation but also achieves truly personalized and predictable assessments based on drug characteristics and experimental objectives. Specifically addressing the complexity of efficacy and safety evaluation at the microcirculation scale, the system deeply integrates physical experiments with artificial intelligence simulations, significantly improving the intelligence and foresight of the screening process while maintaining biological relevance.

[0199] The above description is merely an embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A vascular microfluidic system for testing glycolipid drugs, characterized in that, include: Bionic microcirculation components are used to construct highly realistic three-dimensional vascular organoids and microfluidic culture environments to execute programmed intervention programs. The multimodal intelligent observation component, connected to the biomimetic microcirculation component, can simultaneously collect hemodynamic, cell function, and environmental parameters; A digital twin component is connected to the multimodal intelligent observation component, wherein the digital twin component is used to construct a virtual simulation platform that is synchronized with the physical entity in real time; A dry-wet closed-loop interaction component is connected to the digital twin component, wherein the dry-wet closed-loop interaction component is used to realize bidirectional data interaction and adaptive optimization between physical experiments and digital simulations.

2. The system according to claim 1, characterized in that, The biomimetic microcirculation component includes: A module for preparing vascular organoids using induced pluripotent stem cells; A microfluidic chip module is located in the three-dimensional vascular organoid, wherein the microfluidic chip module is used to collect data within the three-dimensional vascular organoid; The perfusion culture module is used to simulate physiological blood flow shear force; The AI ​​analysis module performs real-time acquisition and analysis of hemodynamic data based on selective plane illumination microscopy and microparticle image velocimetry.

3. The system according to claim 1, characterized in that, The vascular organoid preparation module includes: iPSCs aggregate forming units are used with Matrigel basement membranes and aggregate culture media; Mesodermal directional differentiation unit, used with differentiation medium containing CHIR99021 and BMP-4; Cellular angiogenesis lineage directed differentiation unit, used in differentiation medium containing VEGF-A, Forskolin and A83-01; Vascular sprouting and network forming units are used to enhance simulated blood flow shear forces through mechanical stimulation; Vascular organoid forming units are used to extract vascular networks through collagenase treatment.

4. The system according to claim 1, characterized in that, The microfluidic chip module consists of an upper chip and a lower chip. The upper chip includes microchannels, and the lower chip includes an organoid culture chamber. Multiple pairs of ITO transparent electrodes are symmetrically arranged on both sides of the microchannels. A ring-shaped ITO electrode array is integrated at the bottom of the chamber, and a miniature pressure sensor, flow sensor, temperature sensor, and pH sensor are also built in.

5. The system according to claim 1, characterized in that, The perfusion culture module includes a pneumatic main pump for generating physiological pulsating flow; the perfusion culture module monitors parameters in real time through sensors and dynamically adjusts the flow rate and drug administration regimen.

6. The system according to claim 2, characterized in that, The AI ​​analysis module is used to capture blood flow imaging data, sensor data, and Stokes equations to achieve multimodal automatic analysis of two-dimensional image data and infer three-dimensional flow fields, obtaining key hemorheological parameters including velocity field, pressure field, wall shear force, and hematocrit distribution.

7. A method for testing glycolipid drugs using the system described in any one of claims 1-6, characterized in that, include: S1. Preparation of vascular organoids; S2. Fabrication of a microfluidic chip with a dual-sided structure; S3. The vascular organoids and fibrin hydrogel were mixed and injected into the chip culture chamber of the microfluidic chip module, cured at 37°C for 30 minutes, connected to the perfusion system, and cultured with 1-2 dyn / cm² shear stress pulsating flow for 72 hours. S4. Stepwise increase the shear stress to 4-6 dyn / cm², replace with high-glucose medium containing 25mM glucose, and culture for 7 days to construct a diabetic pathological model; S5. The candidate drug was injected in a gradient manner through a multi-channel micropump, with 3 concentration groups and 1 control group, and ≥3 parallel samples in each group; S6. Start the SPIM-μPIV fusion imaging system to simultaneously acquire flow field data, multi-channel fluorescence images and sensor data, with an acquisition cycle of 30s / time and continuous monitoring for 72 hours; S7. The physical information neural network inverts the three-dimensional flow field, and the microcirculation digital twin model simulates the drug intervention effect, outputting the drug dose-effect relationship, safety window, microcirculation function improvement score and toxicity risk assessment results.

8. The method according to claim 7, characterized in that, The method for preparing the vascular organoid in step S1 includes: S11. iPSC aggregate formation: Matrigel-coated culture, iPSC inoculum density 2×10⁻⁶ 5 Cells / mL, static culture for 24-72 hours to form homogeneous aggregates; S12. Directed differentiation of mesoderm: cultured for 48 hours in a medium containing 3 μM CHIR99021 and 30 ng / ml BMP-4; S13. Angiogenesis lineage differentiation: cultured in a medium containing 100 ng / mL VEGF-A, 10 μM Forskolin and 1 μM A83-01 for 72 hours; S14. Induction of vascular sprouting: Dynamic culture on a 30rpm track shaker for 72 hours, with mechanical shear force applied; S15. Three-dimensional network maturation: Matrigel and type I collagen were mixed in a 3:1 gel and embedded in a gel, then cultured under 3% O2 ​​hypoxia for 5 days; S16. Organoid extraction: Treat with 1 mg / mL collagenase IV for 15 minutes, and screen 70 μm cells to obtain functional vascular organoids with a diameter of 100-300 μm.

9. The method according to claim 7, characterized in that, The method for preparing the microfluidic chip module in step S2 includes: S21. Design a double-layer chip positive mold structure. Spin-coated silicon wafer with SU-8 photoresist with a thickness of 50μm is obtained by ultraviolet exposure and development. S22.PDMS prepolymer and curing agent are mixed at a ratio of 10:1, poured onto the surface of the positive mold, and initially cured at 35℃ for 7 hours and then deeply cured at 60℃ for 4 hours. S23. Magnetron sputtering deposits 100nm ITO thin film, photolithography defines electrode patterns, and screen printing creates conductive paths; S24. Pressure, flow, temperature, and pH sensors are embedded in the lower PDMS chamber and connected to an external data interface; S25. After the upper and lower PDMS layers are cleaned with 100W plasma for 30 seconds, they are aligned and bonded at 60℃ for 48 hours to complete the fabrication of the microfluidic chip.

10. The method according to claim 7, characterized in that, Step S7 includes the following method: using a physical information neural network that fuses ConvLSTM and Fourier neural operators, embedding Stokes equation physical constraints, inputting two-dimensional μPIV particle image data, and outputting three-dimensional velocity field, pressure field, wall shear stress distribution and hematocrit evolution field, with a prediction error ≤5%.

Images