An artificial intelligence-based poultry intestinal organoid auxiliary culture analysis system

Abstract

The application relates to the technical field of organoid culture analysis, and discloses an artificial intelligence-based poultry intestinal tract organoid auxiliary culture analysis system, which comprises microfluidic culture, boundary perception, flow field reconstruction, topology analysis, fluid control and drug analysis modules. The system collects flow velocity vector and pressure drop data of the visual boundary of a microfluidic chip, inverses a three-dimensional flow field inside the organoid by using a deep learning model with embedded physical constraints, constructs a geometric network based on the flow field to generate a topological penetration efficiency index, adjusts a fluid focusing ratio adaptively to optimize penetration depth according to the index, and corrects drug screening data by decoupling causality. The application solves the problems of invisible microenvironment inside the organoid and low fluid control precision, effectively overcomes the limitations that traditional optical detection is difficult to penetrate into the inside of the tissue and that an implanted sensor is easy to interfere with the microenvironment.

Description

Technical Field

[0001] This invention relates to the field of organoid culture and analysis technology, specifically to an artificial intelligence-based assisted culture and analysis system for poultry intestinal organoids. Background Technology

[0002] In the poultry farming industry, gut health is directly related to the growth performance and economic benefits of poultry. Frequent outbreaks of intestinal diseases caused by viruses, coccidia, and bacteria have always been a thorny issue affecting the industry's development. To deeply explore the mechanisms of intestinal pathology and screen for highly effective veterinary drugs, constructing high-fidelity in vitro models is particularly crucial.

[0003] In existing organoid microfluidic culture applications, syringe pumps are typically used to drive the fluid to maintain a basic metabolic environment. Experimental procedures usually involve immobilizing the organoids in specific chambers of a microarray or in a matrix gel, setting a constant perfusion rate, and continuously introducing culture medium or drug-containing media into the channels. Monitoring and analysis primarily rely on periodic observation of the organoids' morphological growth using optical microscopy, or on staining with biochemical reagents at the experimental endpoint to determine cell viability and metabolic levels by detecting fluorescence intensity.

[0004] However, existing organoids typically grow quite densely and have optical thickness, making it difficult for conventional optical imaging to penetrate their surface tissues. The internal fluid flow is essentially in an invisible blind zone, and numerical simulations lacking real-time boundary data cannot accurately recreate the true microenvironment. Fluid control strategies are also relatively crude, and finding a balance between penetration depth and cellular shear damage at a constant flow rate often results in excessive external erosion while the internal tissue remains nutrient-deficient. Therefore, this invention provides an artificial intelligence-based assisted culture and analysis system for poultry intestinal organoids to address the shortcomings of existing technologies. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an artificial intelligence-based assisted culture and analysis system for poultry intestinal organoids, which solves the problems of the invisible internal microenvironment of organoids, the difficulty in balancing fluid penetration depth and cell shear damage, and the low accuracy of drug screening results due to physical delivery barriers in existing technologies.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] The first aspect of this invention provides an artificial intelligence-based assisted culture and analysis system for poultry intestinal organoids, comprising:

[0008] A microfluidic culture module for immobilizing poultry intestinal organoids and providing a fluid culture environment with central flow channels and lateral sheath flow channels;

[0009] The boundary sensing module is used to collect the flow velocity vector and the total pressure drop at the inlet and outlet of the visible boundary region of the microfluidic culture module.

[0010] The flow field reconstruction module is used to invert and calculate the three-dimensional flow field distribution data of the porous media region inside the organoid based on the flow velocity vector and the total pressure drop value using a deep learning model constrained by embedded physical equations.

[0011] The topology analysis module is used to extract the effective flow region from the three-dimensional flow field distribution data, construct a geometric network model that reflects microscopic connectivity, and generate a topology penetration efficiency index accordingly.

[0012] The fluid control module is used to calculate the fluid focusing ratio based on the topological permeability index and drive the microfluidic culture module to adjust the ratio of lateral sheath flow to central flow to optimize the permeation depth.

[0013] The drug analysis module is used to perform causal decoupling correction on cell activity data in drug screening based on the topological permeability efficiency index, and generate a standardized drug sensitivity score.

[0014] Preferably, the microfluidic culture module and boundary sensing module are specifically used for:

[0015] The microfluidic culture module is equipped with a microfluidic chip. The microfluidic chip has a mechanical trapping unit composed of a microcolumn array downstream of the convergence area of ​​the central flow channel and the lateral sheath flow channel. This unit is used to position the organoid within the virtual hydraulic boundary formed by the fluid convergence, while allowing the fluid to pass through the gaps between the microcolumns.

[0016] The boundary sensing module includes an imaging unit aligned with the visible area of ​​the microfluidic chip and a differential pressure sensing unit connected to the fluid pipeline.

[0017] The imaging unit tracks tracer particles in the fluid using an optical flow analysis algorithm and extracts the boundary velocity vectors near the channel wall and at the outer contour of the organoid. The differential pressure sensing unit simultaneously acquires the total pressure drop value generated by the fluid flowing through the porous medium of the organoid.

[0018] Preferably, the flow field reconstruction module includes a physical information neural network model, and the execution steps are as follows:

[0019] The physical information neural network model constructs a fully connected deep neural network as a function approximator of the fluid state, receiving spatiotemporal coordinates as input and outputting the corresponding velocity vector and pressure scalar.

[0020] The physical information neural network model embeds the fluid dynamics conservation law as a physical constraint term in the network's loss function. By introducing a permeation resistance term related to spatial location, it achieves a unified description of the Navier-Stokes equation and the Brinkman porous media equation.

[0021] The flow field reconstruction module uses automatic differentiation technology to calculate the partial derivative of the network output with respect to the input coordinates. By minimizing the matching error of the boundary observation data and the physical residual of the internal collocation points, it generates three-dimensional flow field distribution data that satisfies physical laws in the organoid internal region where there is no observation data.

[0022] Preferably, the topology parsing module includes a multi-scale geometric network model, and the execution steps are as follows:

[0023] The multi-scale geometric network model sets a flow threshold and filters out spatial nodes in the three-dimensional flow field distribution data whose flow velocity is greater than the flow threshold, generating an effective flow point cloud set representing an effective convection transmission channel.

[0024] The multi-scale geometric network model constructs a Vitoris-Lipps complex sequence based on an effective flow point cloud set, and connects discrete point clouds into a geometric network structure by introducing continuously varying filter scale parameters.

[0025] The topology analysis module captures the topological changes in fluid at the microscale as it bypasses the gaps between organoid villi by evolving the geometric network structure.

[0026] Preferably, the topology resolution module is specifically used for:

[0027] A continuous cohomology operation is performed on the multi-scale geometric network model to calculate the continuous plot of the one-dimensional Betty number as a function of the filtration scale parameter. The one-dimensional Betty number is used to represent the quantitative characteristics of the closed microcirculation pathways formed by the fluid.

[0028] The total volume of spatial nodes with flow velocities below the flow threshold in the statistical flow field is calculated, and the proportion of these nodes to the total spatial volume of the organoid is obtained to determine the dead zone volume ratio.

[0029] The topology analysis module extracts the integral features of the one-dimensional Betty number across the entire scale range and combines them with a penalty term for the proportion of dead zone volume to calculate a weighted topology permeability index, which is used to quantify the permeability quality of fluid into the organoid interior.

[0030] Preferably, the fluid control module includes an adaptive optimization algorithm, the execution steps of which are as follows:

[0031] The adaptive optimization algorithm defines the fluid focusing ratio as the ratio of the total volumetric flow rate of the lateral sheath flow channel to the volumetric flow rate of the central flow channel. By adjusting the ratio, the strength of the lateral extrusion stress in the fluid convergence zone is changed, thereby controlling the radial penetration depth of the central fluid in the organoid porous medium.

[0032] The adaptive optimization algorithm constructs an objective function that includes a permeation gain term and a shear damage penalty term. The permeation gain term is determined by the topological permeability performance index, and the shear damage penalty term is determined based on whether the maximum shear stress in the flow field exceeds a preset cell damage threshold.

[0033] The fluid control module uses the adaptive optimization algorithm to calculate the partial derivative of the objective function with respect to the fluid focusing ratio, and iteratively updates the set value of the fluid focusing ratio until the optimal operating point that maximizes the topological permeability index without exceeding the cell damage threshold is found.

[0034] Preferably, the fluid control module further includes a flow rate calculation unit, which performs the following steps:

[0035] The flow rate calculation unit, while maintaining a constant total perfusion flow rate or following a preset metabolic curve, calculates the target volumetric flow rate of the central flow channel and the lateral sheath flow channel in reverse based on the calculated optimal fluid focusing ratio.

[0036] The flow rate calculation unit sends control commands to the multi-channel injection pump group connected to the microfluidic culture module through the communication interface, and independently adjusts the propulsion speed of each channel.

[0037] Preferably, the drug analysis module includes a multi-source data spatiotemporal alignment unit, and the execution steps are as follows:

[0038] The multi-source data spatiotemporal alignment unit synchronously receives biological activity data reflecting cell survival rate or metabolic intensity collected by external detection equipment, as well as the topology penetration efficiency index output by the topology analysis module.

[0039] The multi-source data spatiotemporal alignment unit performs time integration processing on the instantaneous topological permeability index during the drug exposure window to calculate the cumulative permeability reflecting the fluid delivery capability throughout the entire experimental period.

[0040] The drug analysis module uses cumulative permeability as an intermediate variable connecting the physical flow field state and the biological phenotypic response, providing a pre-input for subsequent drug efficacy correction.

[0041] Preferably, the drug analysis module further includes a calibration calculation unit, which performs the following steps:

[0042] The correction calculation unit constructs a nonlinear permeation transfer function, which describes the mapping relationship between the effective utilization rate of the drug in the porous medium and the cumulative permeation efficiency.

[0043] The correction calculation unit uses the osmotic transfer function to weight and correct the input concentration of the drug, and estimates the effective bioavailability concentration of the drug in the intervillous space of organoids and deep tissues.

[0044] The drug analysis module calculates and generates a standardized drug sensitivity score based on the ratio of normalized cell activity data to the corrected effective bioavailability concentration.

[0045] Preferably, the drug analysis module is further used for:

[0046] A confidence interval for permeation performance based on physical limits is established to monitor the physical state of the fluid channel;

[0047] When the cumulative permeation efficiency is detected to be lower than the lower limit of the permeation efficiency confidence interval, it is determined that the current experimental group has a risk of flow channel blockage or structural failure.

[0048] The drug analysis module automatically marks the biological data generated by the current experimental group as invalid data and removes it to prevent false negative results caused by physical perfusion failure from being mixed into the final drug analysis report.

[0049] This invention provides an artificial intelligence-based assisted culture and analysis system for poultry intestinal organoids. It has the following beneficial effects:

[0050] 1. This invention utilizes the Navier-Stokes equations and Brinkman equations as physical constraints. It only requires the acquisition of flow velocity vectors and pressure drop data at the visible boundary of the chip to invert and calculate the three-dimensional flow field distribution inside the opaque porous medium region without damaging or interfering with the biological sample. This effectively overcomes the limitations of traditional optical detection, which is difficult to penetrate into the tissue, and implanted sensors, which are prone to interfering with the microenvironment. It provides accurate and visualized physical field data support for the dynamic culture process of organoids.

[0051] 2. This invention establishes a fluid adaptive control mechanism based on topological analysis, which improves the material exchange efficiency of deep organoid tissues. By constructing a multi-scale geometric network model and generating a topological permeability index, the system can quantitatively characterize the microscopic connectivity of fluids in the microvilli gaps and dynamically adjust the fluid focusing ratio of lateral sheath flow and central flow accordingly. This ensures that nutrients and drugs fully penetrate into the core region of the organoid while limiting the shear stress generated by the flow field below a safe threshold, achieving an optimal balance between material delivery depth and cellular biosafety, and extending the effective culture period of organoids.

[0052] 3. This invention employs a drug sensitivity analysis method based on causal decoupling, which improves the accuracy and reliability of drug screening data. The system calculates cumulative permeability efficiency and constructs a nonlinear permeability transfer function, separating fluid physical delivery efficiency as an independent variable from the biological response results. This achieves the correction of the effective bioavailability concentration of the drug and effectively eliminates deviations caused by channel blockage, uneven flow velocity distribution, or insufficient physical permeation. Attached Figure Description

[0053] Figure 1 This is a system architecture diagram of the present invention;

[0054] Figure 2 This is a flowchart of the method steps of the present invention;

[0055] Figure 3 This is a continuous diagram of the microcirculation channel characteristics of the multi-scale geometric network of the present invention;

[0056] Figure 4 This is a graph showing the relationship between the fluid focusing ratio, penetration efficiency, and shear stress of the present invention.

[0057] Among them, 100 is the microfluidic culture module; 200 is the boundary sensing module; 300 is the flow field reconstruction module; 400 is the topology analysis module; 500 is the fluid control module; and 600 is the drug analysis module. Detailed Implementation

[0058] The technical solutions in 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] See attached document Figure 1 , Figure 1 This is a system architecture diagram according to an embodiment of the present invention. The present invention provides an artificial intelligence-based assisted culture and analysis system for poultry intestinal organoids, including a microfluidic culture module 100, a boundary sensing module 200, a flow field reconstruction module 300, a topology analysis module 400, a fluid control module 500, and a drug analysis module 600.

[0060] The microfluidic culture module 100 is used to immobilize poultry intestinal organoids and provide a fluid culture environment. The microfluidic culture module 100 internally houses a microfluidic chip with a fluid-focusing channel structure, including a central flow channel to accommodate the organoid and lateral sheath flow channels symmetrically distributed on both sides of the central flow channel. The microfluidic culture module 100 is connected to an independently controlled multi-channel injection pump assembly to deliver culture medium to the central flow channel and the lateral sheath flow channels, respectively.

[0061] The boundary sensing module 200 is located outside the microfluidic culture module 100 and at the fluid pipeline connection points, and is used to collect physical environment data of the visible area and inlet / outlet of the microfluidic chip. The boundary sensing module 200 includes an imaging unit aligned with the transparent observation window of the microfluidic chip, used to acquire fluid motion images of the flow channel edge region and extract the boundary velocity vector; the boundary sensing module 200 also includes a differential pressure sensing unit connected to the inlet and outlet of the microfluidic chip, used to collect the total pressure drop value generated by the fluid flowing through the chip.

[0062] The flow field reconstruction module 300 is communicatively connected to the boundary sensing module 200 to receive the boundary velocity vector and total pressure drop values. The flow field reconstruction module 300 incorporates a physical information neural network model, which embeds the Navier-Stokes equations and the porous media Brinkman equation as physical constraints into the loss function. The flow field reconstruction module 300 is configured to use the boundary velocity vector and total pressure drop values ​​as boundary conditions, and through iterative solving of the physical information neural network model, calculate the three-dimensional flow field distribution data and pressure field distribution data of the porous media region inside the organoid.

[0063] The topology analysis module 400 is connected to the flow field reconstruction module 300 and is used to receive three-dimensional flow field distribution data. The topology analysis module 400 is configured to extract the effective flow region in the three-dimensional flow field based on the velocity distribution and construct a multi-scale geometric network model reflecting the connectivity of the fluid's microscopic spatial structure. The topology analysis module 400 analyzes the quantitative characteristics of the effective microcirculation channels in the geometric network model and, combined with the volume proportion of the static region in the flow field, calculates and generates a topology permeability efficiency index for evaluating the fluid's permeability to the organoid's interior.

[0064] The fluid control module 500 is connected to the injection pump assembly of the topology analysis module 400 and the microfluidic culture module 100, respectively. The fluid control module 500 is configured to calculate the fluid focusing ratio using a gradient optimization algorithm based on the deviation between the topology permeability index and a preset target value. The fluid control module 500 sends control commands to the injection pump assembly to adjust the flow rate ratio between the lateral sheath flow channel and the central flow channel, thereby changing the cross-sectional shape and permeation depth of the central fluid using the lateral fluid pressure boundary.

[0065] The drug analysis module 600 is connected to the topology resolution module 400 and receives cell activity data acquired by external detection devices. The drug analysis module 600 is configured to simultaneously record drug concentration, cell activity detection results, and the topology permeability efficiency index output in real time by the topology resolution module 400 during drug screening experiments. The drug analysis module 600 analyzes the correlation between fluid permeability and cell activity, uses the topology permeability efficiency index to compensate for and correct deviations in cell activity data caused by uneven fluid permeability, and generates a drug sensitivity score that reflects the true efficacy of the drug.

[0066] See attached document Figure 2 , Figure 2 This is a flowchart of a method according to an embodiment of the present invention. The present invention provides an artificial intelligence-based method for assisted culture and analysis of poultry intestinal organoids, comprising the following steps:

[0067] S1, the microfluidic culture module 100 fixes poultry intestinal organoids in the central flow channel of the microfluidic chip, and the fluid control module 500 delivers culture medium with an initial flow rate to the central flow channel and the lateral sheath flow channel to establish a basic laminar flow environment;

[0068] S2, During the cultivation process, the boundary sensing module 200 collects the boundary velocity vector of the flow channel edge region through the imaging unit, and collects the total pressure drop value of the microfluidic chip inlet and outlet through the differential pressure sensing unit.

[0069] S3, the flow field reconstruction module 300 receives the boundary velocity vector and the total pressure drop value, inputs them into the physical information neural network model, uses the fluid dynamics equation as a constraint to perform inversion calculation, and outputs the three-dimensional flow field distribution data of the porous medium region inside the organoid.

[0070] S4, the topology analysis module 400 extracts the effective flow region in the three-dimensional flow field based on the velocity distribution, constructs a multi-scale geometric network model, analyzes the quantitative characteristics of the effective microcirculation channels and the proportion of the static region, and calculates the topology permeability index.

[0071] S5, the fluid control module 500 calculates the fluid focusing ratio based on the topological permeability efficiency index and adjusts the flow rate ratio between the lateral sheath flow channel and the central flow channel in the microfluidic culture module 100, and optimizes the penetration depth of the central fluid to the organoid using the lateral pressure boundary;

[0072] S6, In the drug screening experiment, the drug analysis module 600 simultaneously records cell activity data and topological permeability index, and uses the topological permeability index to compensate and correct fluid permeability deviation, generating a drug sensitivity score.

[0073] The microfluidic culture module 100 includes a microfluidic chip fabricated using biocompatible materials. Specifically, polydimethylsiloxane (PDMS) is selected as the upper channel structure layer, and standard optical glass is selected as the bottom support layer. The two are bonded together using an oxygen plasma bonding process to form a closed fluid chamber. PDMS material has good air permeability, which can meet the oxygen exchange requirements of poultry intestinal organoids during long-term culture. At the same time, its optical transparency provides an optical path basis for subsequent boundary sensing.

[0074] The microfluidic chip contains a micrometer-scale flow channel network etched internally. This network is configured as a three-dimensional fluid focusing structure, primarily consisting of a central flow channel, lateral sheath flow channels, and a fluid convergence zone. The central flow channel extends along the chip's longitudinal axis, with its inlet connected to a central nutrient solution reservoir for delivering culture medium containing a suspension of poultry intestinal organoids. The width of the central flow channel... and height The size is set according to the characteristic size of the organoid to be cultured, usually between 500 and 1000 micrometers, to ensure that the organoid can pass through smoothly without causing mechanical compression damage.

[0075] On both sides of the central flow channel, one or more pairs of lateral sheath flow channels are symmetrically distributed. The outlets of the lateral sheath flow channels intersect with the central flow channel in the fluid convergence zone, forming a specific physical angle between the lateral sheath flow channels and the central flow channel. The included angle The preferred value range is 30 degrees to 60 degrees. This geometric configuration aims to generate a fluid focusing effect using the principles of fluid dynamics. Specifically, it uses a high-speed fluid stream introduced through a lateral sheath flow channel to apply lateral compressive stress to a low-speed fluid stream in the central flow channel. This lateral stress forms a virtual hydraulic boundary in the fluid convergence zone, thereby changing the cross-sectional shape and streamline distribution of the central flow stream.

[0076] To achieve stable positioning and long-term observation of poultry intestinal organoids, a mechanical trapping unit is installed downstream of the fluid convergence zone in the central flow channel. This mechanical trapping unit is specifically a micropillar array structure, consisting of several micropillars arranged in a predetermined pattern perpendicular to the bottom surface of the flow channel. In a preferred embodiment, the micropillar array is arranged with U-shaped or V-shaped openings facing upstream of the fluid, forming a physical barrier structure similar to a "dam," with the gaps between the micropillars... Strictly designed to be smaller than the minimum statistical diameter of organoids At the same time, the size is larger than the characteristic size of metabolic waste in the culture medium and the diameter of the tracer particles. This size screening mechanism ensures that the organoids are flexibly confined in the culture chamber surrounded by the microcolumn array, while fluids and metabolites can freely pass through the gaps between the microcolumns, avoiding the formation of dead volume and the accumulation of local metabolic toxicity.

[0077] The fluid flow state within the microfluidic culture module 100 is strictly controlled within the laminar flow range to eliminate the shear damage of turbulence to the villous structure of the organoids and to satisfy the assumptions of the subsequent physical information neural network regarding the physical equations. The Reynolds number of the fluid flow... Defined by the following formula:

[0078] ;

[0079] in, Indicates the density of the fluid; Indicates the characteristic velocity of a fluid; Indicates the hydraulic diameter of the flow channel; This indicates the dynamic viscosity of the fluid.

[0080] The microfluidic culture module 100 relies on a connected multi-channel injection pump assembly. This assembly comprises at least two independently controlled pumping units, connected to the central flow channel inlet and the lateral sheath flow channel inlet of the microfluidic chip, respectively. The central pumping unit provides a basal perfusion flow carrying nutrients, while the lateral pumping unit provides a cell-free sheath fluid flow. The two pumping units are hermetically connected to the chip interface via rigid polytetrafluoroethylene (PTFE) conduits, enabling a constant volumetric flow rate from nanoliters to microliters without pulsation. This independent dual-path drive design allows for subsequent adjustment of the fluid focusing ratio. It provides the hardware execution foundation, enabling the system to arbitrarily adjust the relative ratio of the central flow and the lateral sheath flow while keeping the total flow rate constant or changing, thereby controlling the spatial distribution pattern of the fluid in the culture chamber.

[0081] The fabrication process of microfluidic chips, including photomask fabrication, soft photolithography molding, drilling and bonding, as well as the electrical control interface standard of the injection pump assembly, are conventional technical means in the field of microfluidic chip manufacturing and electromechanical control. Those skilled in the art can select and implement them according to specific experimental needs, and will not be elaborated here.

[0082] The boundary sensing module 200 is physically coupled to the microfluidic culture module 100 through the optical imaging path and the fluid pipeline interface. It is used to overcome the optical occlusion limitation of poultry intestinal organoids in the microfluidic chip and obtain the boundary condition data and global constraint data required for flow field reconstruction. Specifically, it includes the two-dimensional velocity field data of the visible area and the total pressure drop data generated by the fluid flowing through the culture chamber.

[0083] The boundary sensing module 200 includes an imaging unit whose optical field of view is calibrated to cover the convergence areas of the central flow channel and the lateral sheath flow channels of the microfluidic chip. The imaging unit, along the optical path, sequentially includes a light source assembly, an objective lens group, and a high frame rate complementary metal-oxide-semiconductor (CMOS) image sensor. The light source assembly is configured to provide excitation light or bright-field illumination light of a specific wavelength to penetrate the transparent top cover layer of the microfluidic chip. To characterize the fluid motion, biocompatible tracer particles are pre-suspended in the culture medium delivered to the microfluidic chip. The density of the tracer particles is modulated to match the density of the culture medium to follow the fluid movement. The particle size is selected from 1 micrometer to 5 micrometers to ensure sufficient resolvable pixel area under imaging without interfering with the physiological activity of the organoids.

[0084] The imaging unit acquires continuous frame images of fluid motion within the microfluidic channel, generating a time-series image set. The imaging unit integrates an optical flow computing processor configured to perform sparse optical flow analysis or particle image velocimetry (PIV) operations on the time-series image set. The optical flow computing processor first identifies high-contrast feature points in the image, i.e., the positions of tracer particles in the image coordinate system. Then, it tracks the displacement vector of the same feature point between two adjacent frames. Based on the time interval of image acquisition and the magnification of the imaging system, the processor calculates the instantaneous velocity vector field of the visible area within the channel, i.e., the free-flow region outside the organoid edge.

[0085] For the extraction of boundary velocity vectors, the optical flow computation processor follows the optical flow conservation equation, assuming that the brightness of the tracer particles remains constant over extremely short time intervals. Any pixel within the visible area... At any moment grayscale value The following constraints must be satisfied:

[0086] ;

[0087] in, and These represent the fluid in the image plane. shaft and Velocity components in the axial direction , and These represent the partial derivatives of the image grayscale in space and time, respectively. By simultaneously solving this equation within a local neighborhood, the boundary perception module 200 outputs a set of velocity vectors for the visible boundary region. This velocity vector set excludes the internal regions obscured by the organoids and only includes flow field data near the channel wall and at the outer contour of the organoids, serving as the spatial boundary conditions for subsequent flow field reconstruction.

[0088] The boundary sensing module 200 also includes a differential pressure sensing unit for acquiring the global resistance characteristics of the fluid system. This unit consists of two high-sensitivity pressure transmitters, connected to the total fluid inlet and outlet of the microfluidic chip via low dead volume three-way fluid interfaces. The inlet pressure sensor is located between the syringe pump assembly and the chip inlet, while the outlet pressure sensor is located between the chip outlet and the waste liquid collection tank.

[0089] The differential pressure sensing unit synchronously acquires the inlet absolute pressure at a sampling rate higher than the flow field pulsation frequency. absolute pressure on exports The internal processing circuit of this differential pressure sensing unit performs differential operations and outputs the total voltage drop value across the microfluidic chip. The total pressure drop is determined by the following formula:

[0090] ;

[0091] Total pressure drop value This not only reflects the frictional resistance of the microchannel itself, but more importantly, it includes the osmotic resistance generated when fluid flows through the porous medium of poultry intestinal organoids. During the flow field reconstruction process, since the boundary velocity field alone cannot uniquely determine the permeability distribution of the internal porous medium, this total pressure drop value provides the necessary global energy constraint, ensuring that the subsequently constructed physical model is closed-loop in the pressure dimension.

[0092] The boundary sensing module 200 is also equipped with a data synchronization controller to coordinate the working timing of the imaging unit and the differential pressure sensing unit. The data synchronization controller generates a unified timestamp trigger signal to ensure that each frame of the flow field velocity vector diagram corresponds to an accurate differential pressure reading, thereby constructing a synchronized dataset composed of spatial boundary data and global physical quantity data, which is transmitted to the subsequent flow field reconstruction module 300 via a data bus. The specific construction method of the imaging optical path and the circuit connection details of the pressure sensor are conventional technologies in the fields of optical engineering and sensor applications. Those skilled in the art can select and implement them according to the actual equipment, and will not be elaborated here.

[0093] The flow field reconstruction module 300, as the core computing unit of the system, is equipped with a high-performance computing processor. It runs a deep learning algorithm based on a physical information neural network. Under conditions with only sparse boundary observation data, this module 300 solves the three-dimensional flow and pressure fields inside the microfluidic chip, particularly in the region obscured by poultry intestinal organoids, by incorporating fluid dynamics conservation laws. The flow field reconstruction module 300 first constructs a fully connected deep neural network model in the computational domain. This network model acts as a function approximator of the fluid state. The input layer of the network model receives spatial coordinate vectors. and time variables After nonlinear transformations through multiple hidden layers, the output layer generates predicted fluid state values ​​at corresponding spatiotemporal points, including three-dimensional velocity vectors. and scalar pressure To ensure the physical smoothness and differentiability of the network output, the activation function of the hidden layer is preferably a hyperbolic tangent function or a Swish function, in order to support subsequent higher-order automatic differentiation operations.

[0094] The flow field reconstruction module 300 introduces physical equation constraints during network training, treating poultry intestinal organoids as porous media regions and external flow channels as free-flow regions. This module 300 employs a unified momentum conservation equation to describe global fluid behavior and achieves a smooth transition between the Navier-Stokes equations and the Brinkman equations by introducing spatially relevant permeability resistance terms. Defined physical residual functions... and mass conservation residual function Given by the following formula:

[0095] ;

[0096] ;

[0097] in, Indicates fluid density, Represents the velocity vector. Represents the Hamiltonian operator. Indicates fluid dynamic viscosity, This represents the location-dependent reverse osmosis coefficient. The flow field reconstruction module 300 sets the parameters based on the pre-input organoid morphology mask. Value: When coordinates When located in the external free-flow region, When the coordinates are set to 0, the equation degenerates into the standard Navier-Stokes equations; When located in the internal region of organoids, Set to a positive value ( (where the permeability is 0), at this point the equation becomes the Brinkman porous medium flow equation, which reflects the combined effect of Darcy resistance and viscous shear force on the fluid in the microfibrillary gaps.

[0098] To solve the above physical model, the flow field reconstruction module 300 uses automatic differentiation technology to differentiate the output variable of the neural network with respect to the input coordinates and calculate the partial derivative terms in the above residual function without the need for grid discretization. The flow field reconstruction module 300 generates two sets of points in the computational domain: one set is the boundary data point set corresponding to the measurement position of the boundary sensing module 200, and the other set is the collocation set randomly distributed within the computational domain (including inside the organoid). The collocation set is used to force the network to follow the physical conservation law in areas without observation data.

[0099] The flow field reconstruction module 300 constructs a composite loss function. To guide the parameter optimization of the neural network, the loss function is composed of a weighted sum of a data matching error term and a physical residual term. The data matching error term constrains the network's predicted velocity at the visible boundary and the boundary flow velocity vector measured by the imaging unit. To maintain consistency, while constraining the integral pressure difference between the inlet and outlet of the computational domain and the total pressure drop measured by the differential pressure sensing unit. To maintain consistency, the physical residual term constrains the network at all collocation points. and Approaching zero. Composite loss function. The mathematical expression is as follows:

[0100] ;

[0101] in, The number of boundary data points. For the number of internal distribution points, For network speed prediction, and These are the average pressures at the inlet and outlet sections predicted by the network. These are the weighting coefficients for each loss term; Indicates the first Spatial coordinates of the internal collocation points; This indicates the total pressure drop value; Indicates the first Spatial coordinates of boundary data points.

[0102] The flow field reconstruction module 300 employs a gradient-based optimization algorithm (such as the Adam optimizer or the L-BFGS optimizer) to minimize the aforementioned loss function. As the iteration process progresses, the weight parameters of the neural network are continuously updated, ensuring that the output flow field not only conforms to the external observation data but also satisfies the laws of fluid mechanics and porous media flow in the internal, invisible region. When the loss function converges, the flow field reconstruction module 300 outputs high-resolution three-dimensional velocity field data covering the entire internal region of the organoid. and pressure field data This provides an accurate digital physical model for subsequent topology analysis and fluid control. The specific programming implementation of the deep learning framework is a standard practice in the field of computer application technology, and those skilled in the art can configure it according to available computing resources; therefore, it will not be elaborated upon here.

[0103] See attached document Figure 3 The topology analysis module 400 applies the principles of algebraic topology to transform the continuous three-dimensional flow field data output by the flow field reconstruction module 300 into discrete topological feature descriptors, thereby quantifying the permeability of fluid in the microstructure of poultry intestinal organoids. The topology analysis module 400 first discretizes the effective flow region of the input reconstructed flow field, mapping the three-dimensional velocity field data into a point cloud set in a high-dimensional space. This module sets a non-zero flow threshold. This threshold is determined based on the rate of substance exchange required for organoid cells to maintain minimum metabolism. The topology analysis module 400 traverses each spatial node in the flow field reconstruction module 300 mesh. Filter out those that meet the requirements The nodes constitute an effective flow point cloud set. This point cloud set represents the physical channels through which the nutrient solution can actually undergo effective convective transport in geometric space, eliminating invalid areas with extremely low flow rates or stasis.

[0104] Topology parsing module 400 is based on effective flow point cloud set A multi-scale geometric network model is constructed, specifically implemented as a Vitoris-Lipps complex sequence. This topology analytical module 400 introduces a continuously varying scale parameter. (Also known as the filtering parameter), for any two points in the point cloud, if their Euclidean distance is less than 1 / 2, then the filtering parameter is selected. Then, construct an edge (1-simulacra) between the two points; if the distance between any two of the three points is less than 1 / 3, then construct an edge (1-simulacra) between them. This forms a triangular facet (2-simula), and so on, to construct higher-dimensional simplexes. As the scale parameter... As the parameters gradually increase from 0, the discrete points gradually connect to form a network, eventually filling in the gaps to form a volume. During this process, the topology of the point cloud set changes with the parameters. This evolution has led to the formation of a series of nested geometric network structures, namely, the multi-scale geometric network model.

[0105] The topology analysis module 400 performs continuous homology operations on the aforementioned multi-scale geometric network model to extract the connectivity features of the fluid channels. In algebraic topology, the... The rank of a homology group is called the rank of the 1st homology group. The number of Bettis (BettiNumber, denoted as BettiNumber) For one-dimensional Betty numbers In a physical sense, it corresponds to the number of "rings" or "holes" in a geometric structure.

[0106] In the microfluidic perfusion scenario of poultry intestinal organoids, high-dimensional Eigenvalues ​​characterize the number of closed microcirculatory pathways formed by fluid bypassing the intestinal microvilli structure. When the fluid can sufficiently surround and penetrate into the intervilli interstices, it will form a large number of ring-shaped features in the topological space; conversely, if the fluid merely slides across the organoid surface, then... The value is low. The topology parsing module 400 calculates the value across the entire range of filtering parameters. Inside Follow The changes were plotted continuously, and the integral characteristics of the entire life cycle were statistically analyzed to quantify the richness of microcirculatory pathways.

[0107] To comprehensively evaluate the infusion effect, the topology analysis module 400 simultaneously calculates the proportion of dead zone volume in the flow field. This topology analysis module 400 also counts flow velocities below a flow threshold. The total volume of a spatial node is defined as the dead zone volume. And obtain the total space volume occupied by organoids. The percentage of dead zone volume directly reflects the proportion of nutrient-deficient areas caused by insufficient perfusion.

[0108] The topology parsing module 400 combines the aforementioned topological connectivity features and dead zone physical features to generate a topological penetration efficiency index through weighted calculation. The topological permeability efficiency index is used to characterize the current fluid perfusion state in a single-valued manner, and its calculation formula is as follows:

[0109] ;

[0110] in, This represents the maximum filtering scale during the construction of the Vitoris-Lips complex. This indicates that the scale parameter is One-dimensional Betty number at time, Indicates the dead zone penalty coefficient ( The topological permeability index is used to adjust the sensitivity to quiescent regions. This formula indicates that the more microcirculation pathways (larger integral term) and the fewer dead zones (larger exponential term), the higher the topological permeability efficiency index. The higher the value, the more the topological permeability efficiency index is used as a dimensionless control feedback variable, which is directly output to the fluid control module 500 for closed-loop regulation and to the drug analysis module 600 for data correction.

[0111] For the specific mathematical library calls for the continuous homology algorithm, the C++-based GUDHI library or Dionysus library can be used. Those skilled in the art can choose according to the computational efficiency requirements, and will not be elaborated here.

[0112] The fluid control module 500 constructs a closed-loop control circuit connecting the topology analysis results with the physical fluid drive hardware. This fluid control module 500 achieves non-invasive deep perfusion of poultry intestinal organoids by dynamically adjusting the fluid dynamics boundary conditions within the microfluidic chip. The fluid control module 500 receives the topology permeability efficiency index output by the topology analysis module 400. And combined with the maximum shear stress value calculated by the flow field reconstruction module 300 To implement a multi-objective optimization control strategy.

[0113] Fluid control module 500 defines a dimensionless control variable: fluid focusing ratio. Fluid focusing ratio The relative strength of the lateral sheath flow and the central mainstream flow in the microfluidic chip was characterized by the following formula:

[0114] ;

[0115] in, This represents the total volumetric flow rate of all lateral sheath channels. This indicates the volumetric flow rate of the central flow channel. The fluid control module 500 adjusts the fluid focusing ratio. This involves constructing and modulating a "fluid cage" effect in the fluid convergence region of a microfluidic chip. Specifically, when the fluid focusing ratio... As the flow rate increases, the momentum flux of the lateral sheath flow increases, enhancing the lateral compressive stress exerted on the central flow bundle. This causes the streamlines of the central flow bundle to deflect radially inward, forming a contracting virtual hydraulic boundary. This flow field shaping effect forces the nutrient-carrying central fluid to overcome the permeability resistance of the organoid porous medium and penetrate into the deep tissues of the organoid, thereby altering the topological connectivity of the internal flow field. Conversely, when the fluid focusing ratio decreases... When the flow is reduced, the lateral confinement weakens, the central stream diffuses laterally, and the penetration depth decreases.

[0116] The fluid control module 500 incorporates a gradient-ascent-based adaptive optimization algorithm to search for the optimal focusing ratio that maximizes penetration efficiency without damaging cell viability. This adaptive optimization algorithm constructs the objective function. This objective function comprehensively considers both penetration gain and shear damage risk. Defined as:

[0117] ;

[0118] in, It is the topological penetration efficiency index under the current focus ratio. It is the maximum shear stress in the current flow field. It is a preset cell shear damage threshold (set according to the physiological characteristics of poultry intestinal epithelial cells, for example, 0.5 dyne / cm² to 2 dyne / cm²). It is the penalty weighting coefficient. This is the Herveyd step function. The formula shows that when the flow field shear stress is below the safety threshold, the penalty term is zero, and the control objective depends entirely on the permeability; once the shear stress exceeds the safety threshold, the penalty term increases quadratically, forcing the objective function value to drop sharply, thereby guiding the system to avoid high shear conditions.

[0119] In each control cycle, the fluid control module 500 performs the operation according to the objective function. Focus ratio partial derivatives To update the focus ratio setting for the next moment. The update rule follows the gradient ascent iteration formula:

[0120] ;

[0121] in, The focus ratio at the current moment, This is the learning rate or step size parameter. To prevent severe flow field oscillations, the fluid control module 500 pairs... The amplitude of the change is limited to ensure that the flow field change is in a quasi-static process.

[0122] The fluid control module 500 includes a flow rate calculation unit for calculating the optimal focusing ratio. This is then translated into specific pump control commands. To eliminate the interference of total flow fluctuations on the experiment, the flow rate calculation unit typically sets the total flow rate. For constant values ​​or following a specific metabolic demand curve, based on and The flow rate calculation unit calculates the target flow rates for the center flow pump and the sheath flow pump:

[0123] ;

[0124] The fluid control module 500 sends data via standard communication protocols (such as RS-232 or USB) to the multichannel syringe pump assembly connected to the microfluidic culture module 100. and Numerical control commands are used. The drive motor of the injection pump unit responds to the commands to adjust the propulsion speed, thereby physically altering the fluid boundary conditions within the microfluidic chip and achieving closed-loop adaptive control of the microenvironment for culturing poultry intestinal organoids. The specific code implementation of the gradient optimization algorithm and the writing of the pump unit communication driver are conventional techniques in the fields of automatic control and embedded systems. Those skilled in the art can implement them according to specific hardware platforms, and will not be elaborated here.

[0125] See attached document Figure 4 The horizontal axis represents the fluid focusing ratio. The left vertical axis represents the topological penetration efficiency index. The right vertical axis represents the maximum shear stress. The solid curve in the figure shows that as the fluid focusing ratio... With the increase in lateral compression, fluid is forced into the deeper layers of the organoid. It shows an upward trend; however, the dashed curve indicates shear stress. This also increases exponentially. The adaptive optimization algorithm of this invention is shown in the figure. Equilibrium is achieved at a point where the shear stress reaches the safety threshold. Previously, the permeability efficiency reached a local maximum, thus proving that the system can achieve optimal perfusion while ensuring cell safety.

[0126] The drug analysis module 600 communicates with the topology resolution module 400 and external biochemical detection equipment (such as an enzyme-linked immunosorbent assay (ELISA) reader, high-content imaging system or ATP luminescence detector) through a data interface to eliminate the interference of drug delivery differences caused by organoid three-dimensional structure on efficacy evaluation and output a standardized drug sensitivity score.

[0127] The drug analysis module 600 includes a multi-source data spatiotemporal alignment unit for synchronously processing physical domain data and biological domain data. During the drug screening experiment, external biochemical detection equipment acquires cell activity data of poultry intestinal organoids at a preset sampling frequency. This cell activity data is typically expressed as cell viability, fluorescence intensity, or metabolite concentration. The topology resolution module 400 continuously outputs the instantaneous topology permeability index. Given the cumulative effect of drug-induced cell killing or inhibition over time, the instantaneous flow field state alone is insufficient to explain the current biological phenotype. The multi-source data spatiotemporal alignment unit is configured to perform time integration on the instantaneous topological permeability index to calculate the cumulative permeability within the dosing window. Cumulative penetration efficiency Determined by the following formula:

[0128] ;

[0129] in, This represents the duration of drug exposure. This step transforms the dynamically changing fluid perfusion mass into a comprehensive physical scalar reflecting the drug delivery capability throughout the entire experimental period, providing a stable antecedent variable for subsequent causal inference.

[0130] The drug analysis module 600 incorporates a calibration calculation unit based on a structural causal model. This calibration calculation unit processes experimentally observed cell activity data. Treating drug concentration as an outcome variable Treating it as a treatment variable will accumulate penetration effectiveness. This is considered a confounding variable. Underestimating osmotic efficiency can lead to a lower actual drug concentration within the organoid than the input concentration, resulting in "pseudo-resistance." The calibration unit introduces the osmotic transfer function... Estimate the effective bioavailability concentration of the drug in the intervillous space and deep tissues of poultry intestinal organoids. .

[0131] ;

[0132] in, The permeability sensitivity coefficient, The permeation threshold parameter is obtained by fitting standard control data from preliminary experiments. This formula indicates that when the cumulative permeation efficiency... At lower concentrations, drugs have difficulty penetrating deep into organoids, resulting in lower effective concentrations. Corrected to below input concentration Conversely, when the penetration efficiency is good, Approaching .

[0133] Based on the corrected effective bioavailability concentration, the drug analysis module 600 calculates a standardized drug sensitivity score. This drug sensitivity score aims to reflect the true efficacy of drug molecules binding to cellular targets after overcoming physical delivery barriers. Standardized drug sensitivity scores... Defined by the following formula:

[0134] ;

[0135] in, Normalized cell viability (values ​​range from 0 to 1, where 1 indicates complete survival and 0 indicates complete death). The higher the value, the stronger the killing power of the drug at a unit effective concentration, that is, the higher the sensitivity.

[0136] The drug analysis module 600 also includes an anomaly data removal and report generation unit, which sets a confidence interval for penetration efficacy. When anomaly data removal and report generation is detected... When the data falls below the system's physical limit threshold, it indicates severe blockage of the fluid channels or failure of the microfluidic chip structure. The resulting biological data is not statistically significant. The drug analysis module 600 will automatically mark this set of experimental data as invalid and remove it from the final drug analysis report, preventing false negative results caused by physical experimental environment failures from misleading subsequent drug screening decisions. The parsing of data interface protocols and database storage operations for external biochemical detection equipment are standard techniques in the fields of laboratory automation and bioinformatics. Those skilled in the art can configure them according to the equipment manual, and will not be elaborated upon here.

[0137] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An artificial intelligence-based assisted culture and analysis system for poultry intestinal organoids, characterized in that, include: A microfluidic culture module for immobilizing poultry intestinal organoids and providing a fluid culture environment with central flow channels and lateral sheath flow channels; The boundary sensing module is used to collect the flow velocity vector and the total pressure drop at the inlet and outlet of the visible boundary region of the microfluidic culture module. The flow field reconstruction module is used to invert and calculate the three-dimensional flow field distribution data of the porous media region inside the organoid based on the flow velocity vector and the total pressure drop value using a deep learning model constrained by embedded physical equations. The topology analysis module is used to extract the effective flow region from the three-dimensional flow field distribution data, construct a geometric network model that reflects microscopic connectivity, and generate a topology penetration efficiency index accordingly. The fluid control module is used to calculate the fluid focusing ratio based on the topological permeability index and drive the microfluidic culture module to adjust the ratio of lateral sheath flow to central flow to optimize the permeation depth. The drug analysis module is used to perform causal decoupling correction on cell activity data in drug screening based on the topological permeability efficiency index, and generate a standardized drug sensitivity score. The microfluidic culture module and boundary sensing module are specifically used for: The microfluidic culture module is equipped with a microfluidic chip. The microfluidic chip has a mechanical trapping unit composed of a microcolumn array downstream of the convergence area of ​​the central flow channel and the lateral sheath flow channel. This unit is used to position the organoid within the virtual hydraulic boundary formed by the fluid convergence, while allowing the fluid to pass through the gaps between the microcolumns. The boundary sensing module includes an imaging unit aligned with the visible area of ​​the microfluidic chip and a differential pressure sensing unit connected to the fluid pipeline. The imaging unit tracks tracer particles in the fluid using an optical flow analysis algorithm and extracts the boundary velocity vectors near the channel wall and the outer contour of the organoid. The differential pressure sensing unit simultaneously collects the total pressure drop value generated by the fluid flowing through the porous medium of the organoid. The flow field reconstruction module includes a physical information neural network model, and the execution steps are as follows: The physical information neural network model constructs a fully connected deep neural network as a function approximator of the fluid state, receiving spatiotemporal coordinates as input and outputting the corresponding velocity vector and pressure scalar. The physical information neural network model embeds the fluid dynamics conservation law as a physical constraint term in the network's loss function. By introducing a permeation resistance term related to spatial location, it achieves a unified description of the Navier-Stokes equation and the Brinkman porous media equation. The flow field reconstruction module uses automatic differentiation technology to calculate the partial derivative of the network output with respect to the input coordinates. By minimizing the matching error of the boundary observation data and the physical residual of the internal collocation points, it generates three-dimensional flow field distribution data that satisfies physical laws in the organoid internal region where there is no observation data. The topology parsing module is specifically used for: A continuous cohomology operation is performed on the multi-scale geometric network model to calculate the continuous plot of the one-dimensional Betti number as a function of the filtration scale parameter. The one-dimensional Betti number is used to represent the quantitative characteristics of the closed microcirculation pathways formed by the fluid. The total volume of spatial nodes with flow velocities below the flow threshold in the statistical flow field is calculated, and the proportion of these nodes to the total spatial volume of the organoid is obtained to determine the dead zone volume ratio. The topology analysis module extracts the integral features of the one-dimensional Betty number across the entire scale range and combines them with a penalty term for the proportion of dead zone volume to calculate a single-valued topological permeability index, which is used to quantify the permeability quality of fluid into the organoid. The fluid control module is equipped with an adaptive optimization algorithm, and the execution steps are as follows: The adaptive optimization algorithm defines the fluid focusing ratio as the ratio of the total volumetric flow rate of the lateral sheath flow channel to the volumetric flow rate of the central flow channel. By adjusting the ratio, the strength of the lateral extrusion stress in the fluid convergence zone is changed, thereby controlling the radial penetration depth of the central fluid in the organoid porous medium. The adaptive optimization algorithm constructs an objective function that includes a permeation gain term and a shear damage penalty term. The permeation gain term is determined by the topological permeability performance index, and the shear damage penalty term is determined based on whether the maximum shear stress in the flow field exceeds a preset cell damage threshold. The fluid control module uses the adaptive optimization algorithm to calculate the partial derivative of the objective function with respect to the fluid focusing ratio, and iteratively updates the set value of the fluid focusing ratio until the optimal operating point that maximizes the topological permeability efficiency index is found without exceeding the cell damage threshold. The drug analysis module also includes a calibration calculation unit, which performs the following steps: The correction calculation unit constructs a nonlinear permeation transfer function, which describes the mapping relationship between the effective utilization rate of the drug in the porous medium and the cumulative permeation efficiency. The correction calculation unit uses the osmotic transfer function to weight and correct the input concentration of the drug, and estimates the effective bioavailability concentration of the drug in the intervillous space of organoids and deep tissues. The drug analysis module calculates and generates a standardized drug sensitivity score based on the ratio of normalized cell activity data to the corrected effective bioavailability concentration.

2. The artificial intelligence-based poultry intestinal organoid-assisted culture and analysis system according to claim 1, characterized in that, The topology parsing module includes a multi-scale geometric network model, and the execution steps are as follows: The multi-scale geometric network model sets a flow threshold, filters out spatial nodes in the three-dimensional flow field distribution data whose flow velocity is greater than the flow threshold, and generates an effective flow point cloud set representing an effective convection transmission channel. The multi-scale geometric network model constructs a Vitoris-Lipps complex sequence based on an effective flow point cloud set, and connects discrete point clouds into a geometric network structure by introducing continuously varying filter scale parameters. The topology analysis module captures the topological changes in fluid at the microscale as it bypasses the gaps between organoid villi by evolving the geometric network structure.

3. The artificial intelligence-based poultry intestinal organoid-assisted culture and analysis system according to claim 1, characterized in that, The fluid control module also includes a flow rate calculation unit, which performs the following steps: The flow rate calculation unit, while maintaining a constant total perfusion flow rate or following a preset metabolic curve, calculates the target volumetric flow rate of the central flow channel and the lateral sheath flow channel in reverse based on the calculated optimal fluid focusing ratio. The flow rate calculation unit sends control commands to the multi-channel injection pump group connected to the microfluidic culture module through the communication interface, and independently adjusts the propulsion speed of each channel.

4. The artificial intelligence-based poultry intestinal organoid-assisted culture and analysis system according to claim 1, characterized in that, The drug analysis module includes a multi-source data spatiotemporal alignment unit, and the execution steps are as follows: The multi-source data spatiotemporal alignment unit synchronously receives biological activity data reflecting cell survival rate or metabolic intensity collected by external detection equipment, as well as the topology penetration efficiency index output by the topology analysis module. The multi-source data spatiotemporal alignment unit performs time integration processing on the instantaneous topological permeability index during the drug exposure window to calculate the cumulative permeability reflecting the fluid delivery capability throughout the entire experimental period. The drug analysis module uses cumulative permeability as an intermediate variable connecting the physical flow field state and the biological phenotypic response, providing a pre-input for subsequent drug efficacy correction.

5. The artificial intelligence-based poultry intestinal organoid-assisted culture and analysis system according to claim 1, characterized in that, The drug analysis module is also used for: A confidence interval for permeation performance based on physical limits is established to monitor the physical state of the fluid channel; When the cumulative permeation efficiency is detected to be lower than the lower limit of the permeation efficiency confidence interval, it is determined that the current experimental group has a risk of flow channel blockage or structural failure. The drug analysis module automatically marks the biological data generated by the current experimental group as invalid data and removes it to prevent false negative results caused by physical perfusion failure from being mixed into the final drug analysis report.

Images