In-vitro gastrointestinal tract simulation method and system

Abstract

The invention relates to the technical field of in-vitro simulation, and discloses an in-vitro gastrointestinal tract simulation method and system. The method comprises the following steps: acquiring real-time stomach parameters of a stomach simulation area, real-time small intestine parameters of a small intestine simulation area and real-time large intestine parameters of a large intestine simulation area in an in-vitro gastrointestinal tract simulation device; performing quantitative calculation on the acquired real-time parameters to generate a quantitative stomach data set of the stomach simulation area, a quantitative small intestine data set of the small intestine simulation area and a quantitative large intestine data set of the large intestine simulation area; executing simulation state evaluation operation according to the quantitative data set, and outputting a real-time digestion state index of the stomach simulation area, a real-time absorption state index of the small intestine simulation area and a real-time fermentation state index of the large intestine simulation area; and comparing the real-time state index with a predefined threshold interval, and starting a corresponding simulation condition correction operation according to a comparison result.

Description

Technical Field

[0001] This invention relates to the field of in vitro simulation technology, specifically to a method and system for simulating the gastrointestinal tract in vitro. Background Technology

[0002] Current in vitro gastrointestinal tract simulation methods primarily employ a single-region, independent simulation approach based on fixed parameter settings. Existing technologies often simulate different regions of the digestive tract independently, lacking the ability for coordinated parameter acquisition and analysis across multiple regions. Parameter monitoring typically focuses on a single or a few key indicators, failing to comprehensively reflect the complex physiological states of each region. Status assessment methods are simplistic, relying heavily on human experience or simple threshold alarms, making it difficult to quantitatively assess the real-time status of core functions such as digestion, absorption, and fermentation. Simulation condition adjustments are often delayed, requiring manual intervention and failing to achieve automatic correction based on real-time functional status. Existing methods need to address key technical challenges such as simultaneous monitoring of parameters across multiple regions, quantitative assessment of multifunctional status, and adaptive adjustment of simulation conditions.

[0003] Traditional in vitro simulation systems suffer from significant shortcomings in the comprehensiveness of data acquisition and the accuracy of state assessment. The multi-parameter synchronous acquisition mechanism is inadequate, leading to time asynchrony and lack of correlation among data from different regions. The simplistic quantitative calculation methods fail to effectively integrate multi-parameter information to form comprehensive functional indicators. The state assessment model is overly simplified, failing to accurately reflect the true state of complex physiological processes such as digestion, absorption, and fermentation. Static and fixed threshold settings are ill-suited to the specific needs of different simulation experiments. The rigid correction mechanism lacks intelligent decision-making capabilities based on multi-indicator collaborative analysis. Existing technologies require the establishment of a complete technological chain from multi-parameter acquisition to state assessment and intelligent correction. Summary of the Invention

[0004] The purpose of this invention is to provide an in vitro gastrointestinal tract simulation method and system to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a method for simulating the gastrointestinal tract in vitro, the method comprising: Real-time gastric parameters, small intestinal parameters, and large intestinal parameters were collected from the gastrointestinal tract simulation device. The real-time gastric parameters included real-time stomach pH, ​​real-time stomach temperature, and real-time stomach contents volume. The real-time small intestinal parameters included real-time small intestinal pH, real-time small intestinal enzyme activity, and real-time small intestinal peristalsis rate. The real-time large intestinal parameters included real-time large intestinal pH, real-time large intestinal flora count, and real-time large intestinal contents viscosity. The collected real-time stomach parameters, real-time small intestine parameters, and real-time large intestine parameters are quantified and calculated to generate quantified stomach data sets for the stomach simulation region, quantified small intestine data sets for the small intestine simulation region, and quantified large intestine data sets for the large intestine simulation region. Based on the quantitative stomach data set, the quantitative small intestine data set, and the quantitative large intestine data set, a simulation state evaluation calculation is performed, and the real-time digestion state index of the stomach simulation region, the real-time absorption state index of the small intestine simulation region, and the real-time fermentation state index of the large intestine simulation region are output. The real-time digestion status indicators of the stomach simulation region, the real-time absorption status indicators of the small intestine simulation region, and the real-time fermentation status indicators of the large intestine simulation region are compared with predefined threshold ranges, and the corresponding simulation condition correction operations are initiated based on the comparison results.

[0006] Preferably, the operation of collecting real-time stomach parameters in the stomach simulation region, real-time small intestine parameters in the small intestine simulation region, and real-time large intestine parameters in the large intestine simulation region of the in vitro gastrointestinal tract simulation device is achieved through the following steps: A multi-channel pH sensor and a thermistor array are deployed in the stomach simulation area to collect real-time stomach pH and temperature values ​​once per second. At the same time, the real-time stomach contents volume is measured by an ultrasonic flow meter. A titration analysis unit is introduced to dynamically calculate the real-time stomach acid secretion rate, and an optical marker tracking system is used to record the real-time stomach emptying cycle. A fiber optic pH probe and an enzyme activity biosensor are installed in the simulated small intestine area to continuously monitor the real-time pH and enzyme activity of the small intestine. A high-speed camera and a pressure sensor are used to capture the real-time peristalsis speed of the small intestine. A flow meter with an integrated bile injection module reads the real-time bile level of the small intestine, and a microdialysis system is used to calculate the real-time nutrient uptake rate of the small intestine. Microbial electrochemical sensors and rotational viscometers were installed in the colonic simulation area to obtain real-time pH, colonic flora count, and colonic contents viscosity values. An additional gas flow meter was used to monitor the real-time gas generation in the colon, and a timestamp recorder and contents tracker were used to determine the real-time transit time of the colon.

[0007] Preferably, the operation of quantifying the collected real-time stomach parameters, real-time small intestine parameters, and real-time large intestine parameters to generate quantized stomach data sets for the stomach simulation region, quantized small intestine data sets for the small intestine simulation region, and quantized large intestine data sets for the large intestine simulation region includes the following detailed steps: For each parameter value in the real-time gastric parameters, a sliding window average filter is performed to denoise the parameter values. Then, the minimum-maximum normalization method is used to scale the parameter values ​​to the range of zero to one, resulting in a standardized gastric parameter set. The standardized gastric parameters are input into a multivariate correlation analysis engine to calculate the covariance matrix between parameters. Key feature vectors are extracted through principal component analysis and combined to form a quantitative gastric dataset. The time series data in the real-time small intestine parameters are aligned using the dynamic time warping algorithm, then standardized and scaled, and finally representative features are extracted using the feature selection algorithm to construct a quantified small intestine dataset. The K-means clustering algorithm was used to group the real-time colon parameters into multiple typical state clusters. The centroid value of each cluster was calculated as a representative value and then summarized into a quantitative colon data set.

[0008] Preferably, the operation of performing simulated state evaluation calculations based on the quantified stomach data set, quantified small intestine data set, and quantified large intestine data set, and outputting real-time digestive state indicators for the simulated stomach region, real-time absorption state indicators for the simulated small intestine region, and real-time fermentation state indicators for the simulated large intestine region is completed through the following steps: For the simulated stomach region, a quantified stomach dataset is input into a pre-trained digestive state classification model. The digestive state classification model is based on a support vector machine algorithm and outputs a real-time digestive state index, which is used to represent the numerical level of the current digestive efficiency. For the small intestine simulation region, the quantified small intestine data set is imported into the absorption kinetics simulator. The absorption kinetics simulator uses an ordinary differential equation system to simulate the nutrient absorption process and solves to obtain real-time absorption state indicators, which are used to quantify the absorption rate. For the simulated coli region, the quantified coli dataset is fed into the fermentation reaction network simulator. The fermentation reaction network simulator calculates the accumulation rate of fermentation products based on the microbial metabolic kinetics model and generates real-time fermentation status indicators, which are used to assess the fermentation intensity.

[0009] Preferably, the training and application of the digestive state classification model includes the following steps: Collect historical gastric parameter data and their corresponding digestive efficiency labels to construct a training dataset; Kernel functions are used to map the data to a high-dimensional space and optimize the classification hyperplane to complete model training; During the simulation, a set of quantified gastric data is input into the trained model in real time, and the real-time digestive status index is calculated through a decision function.

[0010] Preferably, the construction and operation of the absorption kinetics simulator includes the following steps: Establish a differential equation model for nutrient absorption, with parameters including enzyme activity, peristalsis rate, and bile level; Numerical integration methods, such as the Runge-Kutta method, are used to solve the differential equation system and the absorption state is dynamically updated. The quantified small intestine dataset is used as input parameters, and the simulator outputs real-time absorption status indicators.

[0011] Preferably, the implementation of the fermentation reaction network simulator includes the following steps: Define the reaction network of the microbial community's metabolic pathway, including the equations for substrate consumption and product formation rates; The steady-state approximation method is used to calculate the flow distribution of each node in the network; By substituting the quantified E. coli dataset into the network model, real-time fermentation status indicators are obtained through iterative calculation.

[0012] Preferably, the operation of comparing the real-time digestive status indicators of the stomach simulation region, the real-time absorption status indicators of the small intestine simulation region, and the real-time fermentation status indicators of the large intestine simulation region with predefined threshold intervals, and initiating simulation condition correction operations based on the comparison results, is performed through the following steps: The threshold ranges for gastric digestion, small intestinal absorption, and large intestinal fermentation are defined based on physiological standards. Continuously monitor real-time digestion status indicators. If the indicator value exceeds the gastric digestion threshold range, the acid-base regulator or temperature controller in the gastric simulation area will be triggered for correction. Continuously monitor real-time absorption status indicators. If the indicator value exceeds the small intestine absorption threshold range, activate the enzyme addition pump or peristalsis simulation mechanism in the small intestine simulation area for adjustment. Continuously monitor real-time fermentation status indicators. If the indicator value exceeds the coli fermentation threshold range, activate the coliform colony injection system or viscosity adjustment device in the coliform simulation area.

[0013] Preferably, the specific implementation steps of the correction operation include: When the real-time digestion status indicators of the simulated gastric region exceed the threshold, the pH control variable is adjusted by changing the simulated gastric acid injection rate, or the temperature control variable is adjusted by adjusting the power of the heating element. When the real-time absorption status index of the small intestine simulation area exceeds the threshold, the enzyme addition control variable is adjusted by controlling the opening of the enzyme solution delivery valve, or the peristalsis control variable is adjusted by changing the frequency of mechanical peristalsis. When the real-time fermentation status indicators of the coli simulation area exceed the threshold, the inoculation control variable is adjusted by modulating the bacterial solution injection volume, or the viscosity control variable is adjusted by operating the water addition pump.

[0014] Preferably, the present invention also includes an in vitro gastrointestinal tract simulation system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor, when executing the computer program, implements the steps of the above-described in vitro gastrointestinal tract simulation method.

[0015] Compared with the prior art, the beneficial effects of the present invention are: The quantitative calculation employs a multi-parameter weighted fusion algorithm, comprehensively considering the unit dimension, importance, and range of variation of each parameter. Stomach parameters include pH, temperature, and contents volume, which are quantitatively calculated to form a dataset reflecting the overall state of the stomach. Small intestine parameters include pH, enzyme activity, and peristalsis rate, with the quantitative calculation capturing the dynamic changes in the small intestinal environment. Large intestine parameters include pH, bacterial count, and contents viscosity, with the quantitative calculation characterizing the complex characteristics of the large intestine fermentation environment. This quantitative dataset provides a standardized and comparable data foundation for subsequent state assessment.

[0016] Based on a quantitative dataset, simulated state assessment calculations are performed, outputting real-time functional state indicators for each region. The state assessment calculations employ a multivariate statistical analysis model to establish a mapping relationship between parameter values ​​and functional states. The assessment calculation for the stomach simulated region outputs real-time digestive state indicators, which comprehensively reflect the intensity and efficiency of the digestive process within the stomach. The assessment calculation for the small intestine simulated region outputs real-time absorption state indicators, which quantitatively characterize the rate and extent of nutrient absorption. The assessment calculation for the large intestine simulated region outputs real-time fermentation state indicators, which assess the activity of microbial fermentation and the formation of products. The state indicators are output in standardized numerical form, facilitating comparisons between different regions and overall assessment.

[0017] Real-time functional status indicators are compared with predefined threshold ranges, and corresponding simulation condition correction operations are initiated based on the comparison results. The threshold ranges are set based on extensive experimental data and physiological knowledge, defining the normal range for each functional state. The comparison process calculates the degree and direction of deviation between the status indicators and the threshold ranges, identifying abnormal state types. Based on the comparison results, corresponding correction operations are automatically initiated, including specific measures such as adjusting pH, adjusting temperature, changing enzyme concentration, and controlling peristalsis speed. These correction operations precisely adjust specific regions and specific functional states, ensuring the simulation process remains within physiologically reasonable ranges. Through closed-loop control of real-time monitoring, quantitative evaluation, and automatic correction, precision and intelligence in in vitro gastrointestinal simulation are achieved. Attached Figure Description

[0018] Figure 1 This is a schematic diagram illustrating the working principle of the in vitro gastrointestinal tract simulation method described in this invention. Figure 2 A flowchart of the real-time parameter acquisition process; Figure 3 A flowchart of the parameter quantization calculation steps; Figure 4 A bar chart comparing the quantitative data dimensions of the three regions; Figure 5 A bar chart comparing the accuracy of gastrointestinal status assessment models. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. 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.

[0020] Please see Figure 1 This invention provides a method and system for simulating the gastrointestinal tract in vitro. The method includes: collecting real-time stomach parameters, small intestine parameters, and large intestine parameters in an in vitro gastrointestinal tract simulation device; collecting real-time stomach parameters, small intestine parameters, and large intestine parameters in a simulated stomach region; collecting real-time stomach parameters, small intestine temperature, and large intestine contents volume; collecting real-time small intestine parameters, small intestine pH, small intestine enzyme activity, and small intestine peristalsis rate; and collecting real-time large intestine parameters, large intestine pH, large intestine flora count, and large intestine contents viscosity. After collection, the real-time stomach, small intestine, and large intestine parameters are quantified to generate quantified stomach data sets, small intestine data sets, and large intestine data sets. Based on these quantified data sets, a simulation state evaluation operation is performed to output real-time digestive state indicators for the stomach region, real-time absorption state indicators for the small intestine region, and real-time fermentation state indicators for the large intestine region. Finally, the real-time digestive status indicators of the stomach simulation region, the real-time absorption status indicators of the small intestine simulation region, and the real-time fermentation status indicators of the large intestine simulation region are compared with predefined threshold ranges, and corresponding simulation condition correction operations are initiated based on the comparison results. The entire process ensures that the simulation environment is dynamically adjusted to approximate physiological conditions.

[0021] Example 1: See Figure 2In practice, parameter acquisition of the simulated stomach area is achieved by deploying a multi-channel pH sensor and a thermistor array. The multi-channel pH sensor employs a glass composite electrode structure, with the electrode probe directly contacting the contents of the simulated stomach area. Sensor signals are converted into digital values ​​via a high-precision analog-to-digital converter and acquired and recorded as a real-time stomach pH value data point once per second. The thermistor array consists of multiple negative temperature coefficient thermistor elements, evenly distributed within the outer wall layer of the simulated stomach area. Real-time stomach temperature is calculated by measuring changes in resistance. The readings from the thermistor array and the data from the multi-channel pH sensor are synchronously acquired using the same timer, ensuring temporal consistency between the real-time stomach pH and temperature values.

[0022] In some embodiments, an ultrasonic flow meter is installed at the pipe connection at the bottom of the simulated stomach area. The ultrasonic flow meter uses the time-of-flight measurement principle, with a transducer emitting and receiving ultrasonic signals. The time difference of the ultrasonic signal propagation in the fluid is proportional to the flow velocity, and the real-time stomach contents volume is calculated by combining this with the cross-sectional area of ​​the pipe. The measurement data from the ultrasonic flow meter is transmitted to a data recording unit via digital communication, and the data recording unit timestamps the real-time stomach contents volume value. The titration analysis unit includes a precision metering pump and a pH detection circuit. The precision metering pump quantitatively injects standard alkali solution into the simulated stomach area, and the pH detection circuit monitors the pH change curve during the reaction process. The dynamic calculation unit calculates the real-time stomach acid secretion rate based on the alkali consumption and the rate of pH change, and outputs the real-time stomach acid secretion rate in millimoles per minute.

[0023] In practice, the optical marker tracking system uses fluorescent microspheres as tracer particles, which emit light of a specific wavelength under illumination. A high-speed camera equipped with an optical filter captures the fluorescence signal, and the camera frame rate is set to 100 frames per second to record the trajectory of the fluorescent microspheres within the simulated stomach region. Image processing algorithms analyze the positional changes of the fluorescent microspheres in consecutive frames, and derive the real-time gastric emptying cycle value by calculating the centroid displacement. The real-time gastric emptying cycle value is defined as the time required for half of the markers to leave the simulated stomach region, and this value is obtained by fitting an emptying curve.

[0024] Understandably, the fiber optic pH probe in the small intestine simulation region is coated with a corrosion-resistant fluorinated polymer, and a hydrogen ion-selective membrane is built into the probe tip. The optical signal is transmitted through the fiber optic cable to a spectrometer, which detects the absorption peak shift and converts it into a real-time pH value for the small intestine. The enzyme activity biosensor operates based on electrochemical principles; a specific substrate membrane is immobilized on the sensor surface, and the substrate reacts with the enzyme to generate an electrical signal. This signal is amplified and read by a microcontroller, which outputs the real-time enzyme activity value of the small intestine according to the calibration curve. The real-time enzyme activity value is expressed in international units per milliliter.

[0025] In some embodiments, a high-speed camera is aimed at a transparent peristaltic segment in a simulated small intestine region. This segment is made of an elastic material and simulates intestinal wall muscle contractions. The camera records the deformation process of the segment wall, and an image processing algorithm calculates the velocity of the segment wall using optical flow to derive the real-time peristaltic velocity value of the small intestine. A pressure sensor array is embedded in the inner wall of the peristaltic segment. The pressure sensors employ piezoelectric sensing elements and measure local pressure fluctuations within the lumen. Pressure sensor data is fused and analyzed with image data to verify the accuracy of the real-time peristaltic velocity value of the small intestine. The flow meter in the bile injection module is an electromagnetic flow meter. This flow meter measures the volumetric flow rate of the conductive bile simulation fluid, and the flow meter reading is directly used as the real-time bile level value of the small intestine.

[0026] Optionally, the microdialysis system includes a dialysis probe, a perfusion pump, and a collector. The dialysis probe has a semi-permeable membrane structure that allows small molecules to pass through. The perfusion pump propels the perfusion fluid through the probe at a constant flow rate, where the perfusion fluid exchanges substances with the tissue fluid. The collector collects dialysis fluid samples at regular intervals. The dialysis fluid samples are analyzed for nutrient concentrations using high-performance liquid chromatography (HPLC), and the concentration change rate is used to calculate the real-time nutrient uptake rate in the small intestine. The real-time nutrient uptake rate in the small intestine is expressed in micrograms per minute per square centimeter, reflecting the absorption rate per unit area of ​​the intestinal tract.

[0027] In practice, the microbial electrochemical sensor in the coli simulation area employs a three-electrode system, with the working electrode modified with a nucleic acid probe sensitive to the intestinal flora. The metabolic activity of the flora generates an electrochemical signal, the signal intensity of which is proportional to the real-time coliform count. Sensor readings are processed by a differential amplifier circuit and recorded by a data acquisition card, which samples the real-time coliform count once per second. The rotor of a rotational viscometer is immersed in the contents of the coli simulation area, and the rotor is driven to rotate by a stepper motor. The torque of the stepper motor is related to the viscosity of the contents; a torque sensor measures the resistance torque and converts it into the real-time coliform content viscosity value, expressed in millipascal-seconds.

[0028] As can be understood, the gas flow meter uses a thermal mass flow measurement principle, where the sensing element is heated to a temperature higher than the ambient temperature. As gas flows through the sensing element, it carries away heat; this heat loss is related to the gas mass flow rate. The gas flow meter continuously monitors the volume of gas secreted at the top of the simulated large intestine area, outputting the real-time gas generation value. A timestamp recorder works in conjunction with a content tracker, which uses passive RFID tags to mark chyme particles. RFID readers are installed at the entrance and exit of the simulated large intestine area; when the reader detects a tag passing through, it sends a signal to the timestamp recorder. The timestamp recorder calculates the time the tag remains within the area, which is recorded as the real-time large intestine transit time value.

[0029] Optionally, all sensor data is synchronously acquired via a field-programmable gate array (FPGA), which generates a unified clock signal that is distributed to each sensor interface. Data time synchronization accuracy is controlled at the millisecond level, eliminating time jitter between data from different sensors. Acquired data packets are transmitted to the central processing unit (CPU) via an Ethernet interface, where a real-time operating system (RTOS) runs. The ROS allocates an independent data buffer for each parameter, and buffer data is processed according to a first-in, first-out (FIFO) principle to ensure data flow continuity. The CPU performs preliminary validity checks on the raw data, removing outliers that significantly exceed physiological ranges, providing a high-quality data source for subsequent quantitative calculations.

[0030] Example 2: See Figure 3 In practice, each parameter value in the real-time gastric parameters undergoes sliding window averaging filtering for noise reduction. The sliding window size is set to ten data points per processing unit. When a newly acquired parameter value enters the window, the oldest data point within the window is removed to maintain a constant window data capacity. The arithmetic mean of the ten data points within the window is used as the filtered output value at the current moment. This processing effectively suppresses random noise interference on real-time gastric pH, gastric temperature, and gastric contents. The filtered parameter values ​​then enter the standardization stage. The minimum-maximum standardization method uses the minimum and maximum parameters obtained from historical data statistics. Historical data comes from a database accumulated over long-term system operation, which is periodically updated to reflect parameter ranges under different simulation conditions. The standardization calculation formula linearly transforms each parameter value to a closed interval between zero and one, generating a standardized gastric parameter set. The standardized gastric parameter set contains standardized parameters in multiple dimensions, each dimension representing a dimensionless value of a gastric physiological parameter.

[0031] In some embodiments, the standardized gastric parameter set is fed into a multivariate correlation analysis engine, which first calculates the covariance matrix between the parameters. The covariance matrix is ​​a symmetric square matrix; the diagonal elements are the variance of each parameter, and the off-diagonal elements are the covariance between different parameters. The covariance matrix characterizes the degree of linear correlation between parameters such as real-time gastric pH, real-time gastric temperature, and real-time gastric contents volume. The multivariate correlation analysis engine then performs eigenvalue decomposition on the covariance matrix to solve for the eigenvalues ​​and eigenvectors. The magnitude of the eigenvalues ​​reflects the degree of data variation along the principal component direction, and the eigenvectors indicate the spatial direction of the principal components. Principal component analysis extracts principal components with eigenvalues ​​greater than one as key eigenvectors, which are then sorted in descending order of their corresponding eigenvalues. These key eigenvectors are then linearly combined to project the original standardized gastric parameter set into a new feature space. The resulting combination of projected eigenvectors forms a quantified gastric dataset. The dimensionality of the quantified gastric dataset is lower than that of the original parameter set, but it retains the main variation information of the original dataset.

[0032] It is understandable that time-series data in real-time small intestine parameters exhibit slight asynchrony at different collection points. Dynamic time warping (VTW) algorithms align data by non-linearly scaling the time axis. The VTW constructs a cumulative distance matrix, where each element represents the Euclidean distance between two time-series points. The algorithm searches for an optimal path from the top left corner of the matrix to the bottom right corner, satisfying endpoint matching, continuity, and monotonicity constraints. The correspondence between points on the path determines the alignment between time series, resulting in aligned time series with the same number of data points. The aligned data series undergoes standardization and scaling, using the Z-score method to calculate the mean and standard deviation of each parameter. Subtracting the mean from each data point and dividing by the standard deviation transforms it into a standard normal distribution with a mean of zero and a variance of one. Standardization eliminates dimensional differences between parameters, making real-time small intestine pH, enzyme activity, and peristaltic velocity comparable.

[0033] Optionally, the feature selection algorithm uses recursive feature elimination to filter representative features from standardized data. Recursive feature elimination evaluates feature importance based on the weight coefficients of a support vector machine (SVM) model. First, an SVM model is trained on all features, and the model calculates the weight coefficient of each feature in the decision function. Features with smaller absolute values ​​of weight coefficients are considered to contribute less, and the recursive feature elimination technique removes the subset of features with the smallest weights. The SVM model is then retrained on the remaining features, and the feature removal process is repeated until a specified number of features remain. The selected features include the interaction term between real-time small intestinal enzyme activity and real-time small intestinal peristalsis speed, obtained by multiplying the two parameter values. The representative feature combination constitutes a quantized small intestinal dataset, which retains the most discriminative information from the original data.

[0034] In the specific implementation, real-time coliform parameters are grouped using the K-means clustering algorithm. The K-means clustering algorithm takes the real-time pH value, real-time coliform count, and real-time coliform content viscosity as input variables. The number of clusters is set to three, representing typical fermentation states in the coliform simulation region, including low, medium, and high fermentation states. During the algorithm initialization phase, three data points are randomly selected as initial cluster centers, and each data point is assigned to a corresponding cluster based on the Euclidean distance principle. After allocation, the centroid coordinates of each cluster are recalculated; the centroid is the average of all data points within the cluster. The data point allocation and centroid update process is iterative until the cluster allocation results no longer change or the maximum number of iterations is reached. Finally, the centroid value of each cluster serves as the parameter representative for that cluster, and the coordinate information of the three centroid values ​​is summarized into a quantized coliform dataset. The quantized coliform dataset represents the overall state of the coliform simulation region in vector form.

[0035] In some embodiments, the quantization calculation process is implemented on an embedded processor, which employs a multi-core architecture to process data from different simulation regions in parallel. Each processor core is allocated an independent data processing thread, and threads exchange data via shared memory. Computational tasks such as sliding window averaging filtering, min-maximum normalization, and principal component analysis are compiled into optimized machine instructions. The processor is equipped with a hardware floating-point unit to accelerate complex mathematical operations such as covariance matrix eigenvalue decomposition. The algorithm code employs a memory pre-allocation strategy to avoid runtime latency caused by dynamic memory allocation. The data processing pipeline uses a double-buffering mechanism, with one buffer handling data acquisition while the other handles algorithm calculations, ensuring the real-time performance of the quantization calculation.

[0036] It is understandable that aligning small intestine parameters in real time requires significant computational resources, and the computational complexity of the dynamic time warping algorithm is proportional to the square of the time series length. To improve computational efficiency, the system segments the long time series, dividing the original sequence into overlapping sub-sequence segments. Each sub-sequence segment is dynamically time-warped and aligned separately, and the aligned sub-sequences are then smoothly spliced ​​together into a complete sequence through the overlapping areas. This segmentation method significantly reduces memory usage and computation time, enabling the algorithm to meet the requirements of real-time data processing. Alignment quality is evaluated by measuring the cumulative distance between sequences; if the cumulative distance is less than a set threshold, the alignment result is considered valid.

[0037] Optionally, the iterative convergence speed of the K-means clustering algorithm is affected by the selection of initial cluster centers. The system employs a K-means initialization strategy to optimize the selection of initial center points. The K-means strategy first randomly selects a data point as the first cluster center, and then selects data points that are far away from the selected center point for subsequent cluster centers. This initialization method ensures that the initial cluster centers are evenly distributed throughout the data space, reducing the number of algorithm iterations and avoiding getting trapped in local optima. The effectiveness of the clustering results is evaluated using the silhouette coefficient, which measures the density of each data point among data points of the same class and the separation degree between data points of other classes. The silhouette coefficient value ranges from -1 to +1, with a larger value indicating a better clustering effect. The system sets a silhouette coefficient threshold; when the silhouette coefficient of the clustering result falls below the threshold, the number of clusters is automatically adjusted and the calculation is repeated.

[0038] See Figure 4This figure focuses on the quantitative calculation stage of in vitro gastrointestinal tract simulation, comparing standardized quantitative data of the simulated stomach, small intestine, and large intestine from three dimensions: pH correlation, enzyme / microbiota correlation, and physical state correlation. Through multi-dimensional standardized data comparison, the quantitative differences in different physiological characteristics of each region are intuitively presented: the stomach emphasizes the quantitative characterization of the acid-base environment, the small intestine considers both enzyme activity and basic microbiota quantification, and the large intestine, dominated by fermentation function, shows a significant prominence in enzyme / microbiota correlation quantification values. This figure is a visual representation of the multi-parameter weighted fusion quantification algorithm. After quantitative processing of the original parameters of the stomach, small intestine, and large intestine through sliding filtering, standardization, principal component analysis / feature selection / K-means clustering, a comparable multi-dimensional quantitative data set is formed. This provides a standardized and structured data foundation for subsequent simulation state assessment, effectively supporting the transformation logic from original parameters to functional state indicators, and demonstrating the technical value of multi-region parameter collaborative quantification.

[0039] Example 3: In specific implementation, a pre-trained digestive state classification model is input into a quantized stomach dataset representing the simulated stomach region. This model is constructed based on the Support Vector Machine (SVM) algorithm. The SVM algorithm uses radial basis functions (RBFs) as kernel functions, which map the quantized stomach dataset from the original feature space to a high-dimensional feature space. An optimal classification hyperplane is found in the high-dimensional feature space. The parameters of the optimal hyperplane are determined by solving a convex optimization problem, where the objective function simultaneously considers maximizing the classification margin and minimizing the classification error. The quantized stomach dataset serves as the input feature vector for the SVM model, with each dimension corresponding to a principal component score of the quantized stomach dataset. The output of the SVM model is a decision function value, which is converted into a class label using a sign function. However, in this application, the numerical value of the decision function is directly output as a real-time digestive state indicator. The real-time digestive state indicator is a continuous numerical value, theoretically ranging from negative infinity to positive infinity. However, in practical applications, it is compressed to the range of zero to one using a sigmoid function. The closer the value is to one, the higher the digestive efficiency.

[0040] In some embodiments, the decision function of the support vector machine model is determined by a linear combination of support vectors, which are key sample points in the training data located near the classification boundary. The specific form of the decision function is: in: This represents the input vector of quantized gastric data. Indicates the first Support vectors, This represents the category label corresponding to the support vector. The solution to the dual problem is the Lagrange multiplier. It is a radial basis function kernel function. It is a bias term. In this implementation... The function has been removed; use it directly. The value is used as the base value for real-time digestion status indicators.

[0041] Understandably, a quantified small intestinal dataset targeting the simulated small intestine region is imported into an absorption kinetics simulator. This simulator uses an ordinary differential equation (ODE) system to simulate the nutrient absorption process. The ODE system comprises multiple interrelated differential equations, each describing the rate of change of a nutrient concentration. Equation variables include the spatiotemporal distribution of major nutrients such as glucose, amino acid, and fatty acid concentrations. Equation parameters are derived from the quantified small intestinal dataset. Real-time small intestinal enzyme activity affects the enzymatic reaction rate constant, real-time small intestinal peristalsis affects the convection-diffusion term coefficient, and real-time small intestinal bile levels affect the lipid emulsification rate. The ODE system is solved numerically, and the absorption kinetics simulator uses the fourth-order Runge-Kutta method for numerical integration. The Runge-Kutta method calculates at fixed time steps, performing four slope estimates within each time step, and updating the nutrient concentration values ​​after weighted averaging. The simulator outputs a real-time absorption status index, which is the total amount of nutrients passing through the intestinal wall per unit time, quantified as the absorption rate in millimoles per minute.

[0042] Optionally, the ordinary differential equation system of the absorption kinetics simulator is established based on the law of conservation of mass, considering three main transport mechanisms: convective transport, diffusion transport, and active transport. The convective transport term is proportional to the real-time peristaltic velocity of the small intestine, describing the transport of nutrients resulting from the overall movement of the contents. The diffusion transport term is proportional to the concentration gradient, reflecting the random molecular motion caused by the chemical potential difference. The active transport term uses the Michaelis-Menten equation to describe the carrier protein-mediated active transport process, including two parameters: the maximum transport rate and the half-saturation constant. The equation system also includes enzymatic reaction terms, which break down large nutrient molecules into absorbable small molecule products; the reaction rate is regulated by the real-time enzyme activity of the small intestine. The role of bile is reflected in the lipid emulsification term, where bile acids form micelles that promote the dissolution and absorption of lipid hydrolysis products.

[0043] In practice, a quantified coliform dataset for the simulated coli region is fed into a fermentation reaction network simulator, which operates based on a microbial metabolic kinetic model. This model defines a complex fermentation reaction network encompassing multiple metabolic pathways, including carbohydrate fermentation, protein breakdown, and short-chain fatty acid production. Each pathway consists of a series of biochemical reactions, and the reaction rate is described using the Monod equation to account for the relationship between microbial growth and substrate consumption. The real-time coliform count in the quantified coliform dataset serves as the initial value for microbial biomass, while the real-time pH value affects enzyme activity, thus regulating the reaction rate constant. The fermentation reaction network simulator calculates the production rates of major fermentation products, including acetic acid, propionic acid, and butyric acid. The product production rates are integrated to obtain real-time fermentation state indices. These indices are expressed as the total short-chain fatty acid yield in millimoles per hour (mmol / H), with higher values ​​reflecting the fermentation intensity of the simulated coli region.

[0044] In some embodiments, the fermentation network simulator simplifies calculations using a steady-state approximation method, which assumes that the concentration change rate of metabolic intermediates is zero. The steady-state approximation transforms the differential equation system into an algebraic equation system, which is solved using the Newton-Raphson iterative method. The Newton-Raphson iterative method requires calculating the Jacobian matrix, where each element is the partial derivative of each equation with respect to each variable. The iterative process starts with an initial guess and gradually approximates the true solution of the equation system. Convergence is considered achieved when the difference between two iterations is less than a tolerance threshold. The fermentation network simulator also includes a mass balance verification module, which verifies the conservation of carbon and nitrogen elements in the reaction network, ensuring the reasonableness of the simulation results.

[0045] Understandably, the simulated state assessment calculations are performed on an industrial computer equipped with a multi-core CPU and large-capacity memory. The assessment task employs a parallel computing architecture, with assessment calculations for the stomach, small intestine, and large intestine simulation regions distributed across different processor cores for simultaneous execution. Data is exchanged between processor cores via a high-speed bus, and a shared data area is allocated in memory to store the quantified stomach, small intestine, and large intestine data sets. The operating system uses a real-time Linux kernel, which provides precise timer interrupts and task scheduling capabilities to ensure real-time performance of the assessment calculations. The assessment results are transmitted to a display terminal and control system via a network interface. The display terminal graphically shows the changing trends of real-time digestion, absorption, and fermentation state indicators.

[0046] Optionally, the training data for the Support Vector Machine (SVM) model comes from historical experimental records, which include quantified gastric data sets and corresponding digestion efficiency labels under different simulation conditions. Digestion efficiency labels are obtained using standard in vitro digestibility assays, where digestibility is defined as the percentage of substrate degradation relative to the initial amount. The training process employs five-fold cross-validation to determine the model's hyperparameters, including the width of the radial basis function kernel and the regularization coefficient. The trained SVM model is serialized into a file and stored on a solid-state drive (SSD). Upon system startup, the model file is loaded into memory, and the model instance in memory responds to real-time evaluation requests. The model is periodically updated incrementally using new data. These incremental updates are implemented through an online learning algorithm, which adjusts the support vector set and classification hyperplane parameters to adapt the model to the slow changes in simulation conditions.

[0047] See Figure 5 This figure visualizes the key results of the state assessment step, intuitively presenting the accuracy performance of three types of gastrointestinal state assessment models. The vertical axis represents accuracy; the horizontal axis represents the state assessment models, including three types: digestive state classification model, absorption kinetics simulator, and fermentation reaction network simulator. This figure clearly compares the performance differences of the assessment models for the three key stages of digestion, absorption, and fermentation, providing data support for optimizing the accuracy of the gastrointestinal simulation system and representing the core achievement of the multi-dimensional state assessment step.

[0048] Example 4: In specific implementation, the training and application of the digestive state classification model includes collecting historical gastric parameter data and their corresponding digestive efficiency labels. The historical gastric parameter data comes from complete records of multiple independent simulation experiments. Each simulation experiment record includes time-series data of multiple parameters such as real-time gastric pH, real-time gastric temperature, and real-time gastric contents volume. The data is recorded at a fixed sampling frequency and stored in a relational database. The digestive efficiency label is obtained through a standard in vitro digestibility assay, which uses the degradation rate of a specific substrate in simulated gastrointestinal fluid as a quantitative indicator. The degradation rate is calculated by detecting substrate residues, and detection methods include reducing sugar determination, free amino nitrogen determination, or high-performance liquid chromatography analysis. Each historical data sample consists of a set of gastric parameter data and a digestive efficiency label value, forming a training data pair. The number of training data pairs must meet statistical significance requirements, typically not less than several hundred pairs, to cover different physiological states and simulation conditions. Data samples must undergo rigorous preprocessing, including outlier removal, missing value imputation, and data standardization, to ensure data quality meets the requirements for model training.

[0049] In some embodiments, the training dataset is constructed as a combination of a feature matrix and label vectors. Each row of the feature matrix corresponds to a historical data sample, and each column corresponds to a feature of a gastric parameter. The feature matrix may contain the original parameter values, parameter statistics (such as mean and variance), and interaction terms between parameters. The label vectors are column vectors of digestive efficiency label values, which can be continuous numerical values ​​or discrete levels. The Support Vector Machine (SVM) algorithm uses a kernel function to map the data to a high-dimensional feature space. The kernel function chosen is a radial basis function (RBF), which can handle non-linearly separable data patterns. Hyperplane optimization is achieved by minimizing structural risk, which includes an empirical risk term and a regularization term. The empirical risk term measures the model's classification error on the training data, and the regularization term controls the model complexity to prevent overfitting. The optimization problem is transformed into solving a dual problem involving the optimization of Lagrange multipliers, which is efficiently solved using a sequential minimum optimization algorithm. The trained model parameters include a set of support vectors, Lagrange multiplier values, and bias terms, which are saved as structured text files or binary files.

[0050] Understandably, during the simulation, a quantized stomach dataset is input into the trained digestive state classification model in real time. This dataset consists of feature vectors reduced in dimensionality through principal component analysis. The model calculates the kernel function value between the quantized stomach dataset and each support vector; the kernel function value reflects the similarity between the input features and the support vectors. The kernel function values ​​of all support vectors are weighted and summed, and a bias term is added to obtain the output value of the decision function. The output value of the decision function is converted into a probability value using a sigmoid function, and this probability value serves as a real-time digestive state indicator. The real-time digestive state indicator is a value between zero and one, representing the confidence level that the current digestive state belongs to the efficient digestion category. The model inference process is performed in memory, where model parameters and inference code are loaded to ensure low latency in real-time response. The system records the input and output data for each inference iteration; this recorded data is used for model performance monitoring and subsequent iterative updates.

[0051] The construction and operation of the absorption kinetics simulator involves establishing a differential equation model of nutrient absorption, based on the principles of matter conservation and reaction kinetics. The model's state variables include intraluminal nutrient concentration, intestinal boundary layer nutrient concentration, and blood-side nutrient concentration. The differential equations describe the rate of change of nutrient concentration over time, determined by processes such as convection, diffusion, enzymatic hydrolysis, and transmembrane transport. Parameter estimation utilizes the least squares method to fit historical experimental data, which includes time curves of nutrient concentration under different conditions. Differential equation systems are typically rigid, requiring implicit numerical integration methods to ensure stability. The absorption kinetics simulator employs the Runge-Kutta method, an explicit method with high accuracy for non-rigid equations. A quantified small intestinal dataset is used as input parameters, including key parameters such as real-time small intestinal enzyme activity and peristaltic velocity. The simulator outputs real-time absorption state indices, which can be the absorption flux of a specific nutrient or an overall absorption efficiency index.

[0052] Optionally, the differential equation system of the absorption kinetics simulator contains multiple coupled ordinary differential equations, each describing the kinetic behavior of a nutrient or metabolite. Typical equation forms include convection, diffusion, reaction, and absorption terms. The convection term is related to the real-time peristaltic velocity of the small intestine and uses a one-dimensional flow tube model to describe the axial movement of the contents. The diffusion term is proportional to the concentration gradient and considers axial and radial diffusion effects. The reaction term simulates the enzymatic hydrolysis process, described using the Michaelis-Menten equation, with parameters regulated by the real-time enzyme activity of the small intestine. The absorption term simulates the active and passive transport processes of intestinal epithelial cells, using a saturation kinetic model for the transport rate. Solving the equation system requires initial conditions, determined by the nutrient composition of the contents at the start of the simulation. Boundary conditions are set such that the nutrient concentration at the inlet is known, and a zero-gradient condition is used at the outlet. Numerical solutions use a variable step-size algorithm, which dynamically adjusts the step size based on local truncation errors, balancing computational efficiency and accuracy.

[0053] Table 1: Parameters of the Absorption Kinetics Model In its implementation, the fermentation reaction network simulator defines a reaction network based on known gut microbiota metabolic pathways. These pathways include major pathways such as carbohydrate fermentation, amino acid fermentation, fatty acid metabolism, and gas generation. Each pathway consists of a series of biochemical reaction steps, involving substrate consumption, intermediate product generation, and final product secretion. Rate equations are based on stoichiometry and kinetics, such as the Monod equation for microbial growth and the Michaelis-Menten equation for enzymatic reactions. Equation parameters are derived from microbiology literature and in vitro experimental measurements, and parameter values ​​are adjusted within a certain range according to the microbial composition. The fermentation reaction network simulator uses a steady-state approximation to simplify calculations, assuming a zero net rate of metabolic intermediate generation. The steady-state approximation transforms a system of differential-algebraic equations into a system of pure algebraic equations, which are then solved numerically. A quantified coli dataset is substituted into the network model, containing state parameters such as real-time coliform count and pH. The model iteratively calculates the metabolic flux distribution, reflecting the activity level of each pathway. Generate real-time fermentation status indicators, which can be the total yield of short-chain fatty acids, gas yield, or carbon conversion efficiency.

[0054] In some embodiments, the reaction network structure of the fermentation reaction network simulator is represented in matrix form, with rows corresponding to metabolites and columns corresponding to biochemical reactions. The stoichiometric matrix elements represent the stoichiometric coefficients of metabolites in the reaction; negative values ​​represent reactants, and positive values ​​represent products. The metabolic flux vector contains the net flux value for each reaction, which must satisfy mass balance constraints. The kinetic equation describes the relationship between flux values ​​and metabolite concentrations, with parameters including the maximum reaction rate and the half-saturation constant. The steady-state approximation method requires that the product of the stoichiometric matrix and the flux vector be zero, constituting a linear equality constraint. The optimization problem seeks a flux distribution that satisfies the constraints and conforms to the kinetic equation; the objective function can be to maximize biomass yield or minimize substrate consumption. The solution algorithm employs linear programming or quadratic programming, combined with a local search strategy to avoid getting trapped in local optima. The model output includes predicted concentrations of all metabolites and reaction flux values, with key output variables serving as real-time fermentation state indicators. The simulator integrates sensitivity analysis functionality, which assesses the impact of parameter uncertainties on the output results.

[0055] It is understandable that the implementation of the digestion state classification model, absorption kinetics simulator, and fermentation reaction network simulator all rely on the computational software platform. The computational software platform adopts a modular design, with each simulator developed as an independent dynamic link library. The dynamic link library provides a standard application programming interface (API), which includes methods for initialization, parameter setting, execution of calculations, and result acquisition. The main control program calls the APIs of each simulator at fixed intervals, synchronized with the data acquisition cycle. The platform employs a multi-threaded programming model, with each simulator running in an independent execution thread, and threads exchanging information through thread-safe data structures. Computational tasks are executed on industrial computers equipped with multi-core CPUs, high-speed memory, and solid-state drives. The operating system is a real-time operating system, providing deterministic scheduling and hard real-time response capabilities. The software development environment includes numerical computation libraries, optimization algorithm libraries, and linear algebra libraries; these fundamental libraries accelerate the model solving process. The system has logging and error handling mechanisms; the log records the running status and abnormal events, and the error handling mechanism provides backup strategies in case of computational failure.

[0056] Example 5: In specific implementation, three threshold ranges are set: gastric digestion threshold range, small intestinal absorption threshold range, and large intestinal fermentation threshold range. The gastric digestion threshold range is defined based on the physiological range of human gastric digestion. This physiological range is derived from clinical medical research data, which statistically analyzes the fluctuation range of digestive efficiency in healthy individuals under standard dietary conditions. The gastric digestion threshold range is represented as a numerical interval, with the lower limit taken as the 5th percentile of the digestive efficiency distribution in healthy individuals and the upper limit as the 95th percentile. The small intestinal absorption threshold range considers normal fluctuations in nutrient absorption rates, based on studies of the absorption kinetics of isotope-labeled nutrients. The large intestinal fermentation threshold range references intestinal fermentation product concentration standards, derived from population research data on fecal short-chain fatty acid content. The threshold ranges are stored in non-volatile memory as configuration files, using JSON format to record the upper and lower limits and units of measurement for each threshold. Upon system startup, the configuration file content is read and loaded into a threshold register group in memory. The threshold register group allocates an independent storage unit for each status indicator, and the content of each storage unit remains unchanged during operation.

[0057] In some embodiments, real-time digestive status indicators are continuously monitored. These indicators are periodically output by a digestive status classification model, with the output frequency matching the data acquisition frequency. The real-time digestive status indicator values ​​are compared with gastric digestion threshold ranges. This comparison is implemented using a hardware comparator circuit. The hardware comparator circuit has two analog voltage input ports: one connected to the real-time digestive status indicator voltage signal output from the digital-to-analog converter, and the other connected to a reference voltage for the threshold range. When the real-time digestive status indicator voltage signal is lower than the reference voltage corresponding to the lower limit of the gastric digestion threshold range, the hardware comparator circuit outputs a low-level alarm signal. When the real-time digestive status indicator voltage signal is higher than the reference voltage corresponding to the upper limit of the gastric digestion threshold range, the hardware comparator circuit outputs a high-level alarm signal. The alarm signal is transmitted to a programmable logic controller (PLC), which triggers a corresponding correction operation based on the alarm signal type. This triggers a correction operation by the pH regulator or temperature controller in the simulated gastric region. The pH regulator is a precision metering pump, and the temperature controller is a PID temperature control module.

[0058] It is understandable that continuous monitoring of real-time absorption status indicators is performed. These indicators are calculated by the absorption kinetics simulator, and the update interval of the calculation results is synchronized with the simulation time step. The real-time absorption status indicator value is compared with the small intestinal absorption threshold range; the comparison logic is implemented in the microcontroller's firmware. The microcontroller periodically reads the real-time absorption status indicator value register, which is updated by the absorption kinetics simulator via memory mapping. The microcontroller performs an arithmetic comparison between the read value and the upper and lower limits of the small intestinal absorption threshold range, and stores the comparison result in the status flag register. If the real-time absorption status indicator value exceeds the small intestinal absorption threshold range, the overflow bit in the status flag register is set. The overflow bit triggers an interrupt service routine, which activates the enzyme addition pump or peristaltic simulation mechanism in the small intestinal simulation area via the digital output interface. The enzyme addition pump is a precision injection pump driven by a stepper motor, and the peristaltic simulation mechanism is a mechanical compression device controlled by a servo motor.

[0059] Optionally, real-time fermentation status indicators are continuously monitored. These indicators are derived from the output variables of the fermentation reaction network simulator, and are passed to the monitoring process via shared memory. The monitoring process runs in the user space of the real-time operating system, and is scheduled with a fixed priority to ensure timely response. The real-time fermentation status indicator values ​​are compared with the coliform fermentation threshold range. The comparison function uses inline assembly code to optimize execution efficiency. The comparison result is sent to the control process via a message queue. The control process parses the message content and generates execution instructions. If the real-time fermentation status indicator value exceeds the coliform fermentation threshold range, the control instruction activates the coliform microbiota injection system or viscosity adjustment device in the coliform simulation area. The microbiota injection system is a pneumatic diaphragm pump system, and the viscosity adjustment device is a variable frequency centrifugal pump unit. The entire monitoring and comparison process forms a closed-loop control loop, with the control loop cycle completely closed from data acquisition to actuator action.

[0060] In practice, the correction operation is achieved through a feedback control loop, which is based on the principle of deviation adjustment. The deviation value is the difference between the measured value of the state index and the median of the threshold interval. The deviation value is input to the PID controller to calculate the control quantity. The PID controller includes a proportional unit, an integral unit, and a derivative unit. The proportional unit responds to the current deviation, the integral unit eliminates accumulated errors, and the derivative unit predicts the trend of change. The control signal is generated by the PID controller and can be a standard current signal or a pulse width modulation (PWM) signal. The standard current signal ranges from 4 to 20 mA, and the current value is linearly related to the control quantity. The duty cycle of the PWM signal is proportional to the control quantity, and the frequency is fixed at 1 kHz. The control signal drives the actuators, including regulating valves, frequency converters, and servo drives. The parameter tuning of the feedback control loop is completed using the Ziegler-Nichols method, which ensures fast system response and small overshoot.

[0061] In some embodiments, when the real-time digestive status indicators of the simulated gastric region exceed a threshold, the pH control variable is adjusted by changing the simulated gastric acid injection rate. The simulated gastric acid injection rate is controlled by the rotational speed of a metering pump, which is a peristaltic pump. The control signal output from the PID controller is connected to the frequency converter of the peristaltic pump, which converts the signal into the rotational speed of a three-phase motor. The rotational speed is linearly related to the injection rate, which varies from 0.1 ml / min to 10 ml / min. The temperature control variable is adjusted by regulating the power of the heating element, which is an armored electric heating tube. The PID controller output signal is connected to a solid-state relay, which uses a zero-crossing triggering method to adjust the energizing time of the heating tube. The power adjustment range is from zero to a rated power of 3 kW, and the temperature control accuracy reaches ±0.5 degrees Celsius.

[0062] It is understandable that when the real-time absorption status indicators of the simulated small intestine region exceed the threshold, the control variable for enzyme addition is adjusted by controlling the opening of the enzyme solution delivery valve. The enzyme solution delivery valve is an electromagnetic regulating valve, and the valve core opening is proportional to the input current signal. The PID controller outputs a 4-20 mA current signal to drive the valve positioner, which precisely controls the valve stem displacement. The valve opening linearly adjusts the enzyme solution flow rate from fully closed to fully open, covering a flow range of 0.5 liters to 5 liters per hour. The peristalsis control variable is adjusted by changing the mechanical peristalsis frequency, which is determined by the servo motor speed. The servo motor receives pulse frequency commands, and the pulse frequency is proportional to the speed. The PID controller outputs the pulse frequency, which is sent to the servo driver via a motion control card. The peristalsis frequency adjustment range is 2 to 12 contractions per minute.

[0063] Optionally, when the real-time fermentation status indicators of the simulated coliform region exceed the threshold, the inoculation control variable is adjusted by modulating the bacterial solution injection volume. The bacterial solution injection volume is determined by the stroke volume and frequency of the injection pump, which is a dual-plunger parallel structure. The PID controller output signal controls the stepper motor's step number and pulse frequency; the step number determines the single injection volume, and the pulse frequency determines the number of injections. The injection volume resolution reaches 0.1 μL, and the injection frequency can reach up to 10 times per minute. The viscosity control variable is adjusted by operating the water addition pump, which is a diaphragm metering pump. The PID controller output current signal adjusts the diaphragm stroke length, which is proportional to the single water delivery volume. The water addition rate is continuously adjustable from 1 mL to 100 mL per minute, and a static mixer is installed at the outlet of the water addition pump to promote uniform water distribution.

[0064] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0065] 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. A method for simulating the gastrointestinal tract in vitro, characterized in that, The method is implemented through the following sequential operations: Real-time gastric parameters, small intestinal parameters, and large intestinal parameters were collected from the gastrointestinal tract simulation device. The real-time gastric parameters included real-time stomach pH, ​​real-time stomach temperature, and real-time stomach contents volume. The real-time small intestinal parameters included real-time small intestinal pH, real-time small intestinal enzyme activity, and real-time small intestinal peristalsis rate. The real-time large intestinal parameters included real-time large intestinal pH, real-time large intestinal flora count, and real-time large intestinal contents viscosity. The collected real-time stomach parameters, real-time small intestine parameters, and real-time large intestine parameters are quantified and calculated to generate quantified stomach data sets for the stomach simulation region, quantified small intestine data sets for the small intestine simulation region, and quantified large intestine data sets for the large intestine simulation region. Based on the quantitative stomach data set, the quantitative small intestine data set, and the quantitative large intestine data set, a simulation state evaluation calculation is performed, and the real-time digestion state index of the stomach simulation region, the real-time absorption state index of the small intestine simulation region, and the real-time fermentation state index of the large intestine simulation region are output. The real-time digestion status indicators of the stomach simulation region, the real-time absorption status indicators of the small intestine simulation region, and the real-time fermentation status indicators of the large intestine simulation region are compared with predefined threshold ranges, and the corresponding simulation condition correction operations are initiated based on the comparison results.

2. The in vitro gastrointestinal tract simulation method according to claim 1, characterized in that, The operation of collecting real-time stomach parameters, real-time small intestine parameters, and real-time large intestine parameters in the in vitro gastrointestinal tract simulation device are achieved through the following steps: A multi-channel pH sensor and a thermistor array are deployed in the stomach simulation area to collect real-time stomach pH and temperature values ​​once per second. At the same time, the real-time stomach contents volume is measured by an ultrasonic flow meter. A titration analysis unit is introduced to dynamically calculate the real-time stomach acid secretion rate, and an optical marker tracking system is used to record the real-time stomach emptying cycle. A fiber optic pH probe and an enzyme activity biosensor are installed in the simulated small intestine area to continuously monitor the real-time pH and enzyme activity of the small intestine. A high-speed camera and a pressure sensor are used to capture the real-time peristalsis speed of the small intestine. A flow meter with an integrated bile injection module reads the real-time bile level of the small intestine, and a microdialysis system is used to calculate the real-time nutrient uptake rate of the small intestine. Microbial electrochemical sensors and rotational viscometers were installed in the colonic simulation area to obtain real-time pH, colonic flora count, and colonic contents viscosity values. An additional gas flow meter was used to monitor the real-time gas generation in the colon, and a timestamp recorder and contents tracker were used to determine the real-time transit time of the colon.

3. The in vitro gastrointestinal tract simulation method according to claim 2, characterized in that, The operation of quantifying the collected real-time stomach parameters, real-time small intestine parameters, and real-time large intestine parameters to generate quantified stomach data sets for the stomach simulation region, quantified small intestine data sets for the small intestine simulation region, and quantified large intestine data sets for the large intestine simulation region includes the following detailed steps: For each parameter value in the real-time gastric parameters, a sliding window average filter is performed to denoise the parameter values. Then, the minimum-maximum normalization method is used to scale the parameter values ​​to the range of zero to one, resulting in a standardized gastric parameter set. The standardized gastric parameters are input into a multivariate correlation analysis engine to calculate the covariance matrix between parameters. Key feature vectors are extracted through principal component analysis and combined to form a quantitative gastric dataset. The time series data in the real-time small intestine parameters are aligned using the dynamic time warping algorithm, then standardized and scaled, and finally representative features are extracted using the feature selection algorithm to construct a quantified small intestine dataset. The K-means clustering algorithm was used to group the real-time colon parameters into multiple typical state clusters. The centroid value of each cluster was calculated as a representative value and then summarized into a quantitative colon data set.

4. The in vitro gastrointestinal tract simulation method according to claim 3, characterized in that, The operation of performing simulated state evaluation calculations based on quantified stomach data sets, quantified small intestine data sets, and quantified large intestine data sets, and outputting real-time digestive state indicators for the simulated stomach region, real-time absorption state indicators for the simulated small intestine region, and real-time fermentation state indicators for the simulated large intestine region, is completed through the following steps: For the simulated stomach region, a quantified stomach dataset is input into a pre-trained digestive state classification model. The digestive state classification model is based on a support vector machine algorithm and outputs a real-time digestive state index, which is used to represent the numerical level of the current digestive efficiency. For the small intestine simulation region, the quantified small intestine data set is imported into the absorption kinetics simulator. The absorption kinetics simulator uses an ordinary differential equation system to simulate the nutrient absorption process and solves to obtain real-time absorption state indicators, which are used to quantify the absorption rate. For the simulated coli region, the quantified coli dataset is fed into the fermentation reaction network simulator. The fermentation reaction network simulator calculates the accumulation rate of fermentation products based on the microbial metabolic kinetics model and generates real-time fermentation status indicators, which are used to assess the fermentation intensity.

5. The in vitro gastrointestinal tract simulation method according to claim 4, characterized in that, The training and application of the digestive state classification model includes the following steps: Collect historical gastric parameter data and their corresponding digestive efficiency labels to construct a training dataset; Kernel functions are used to map the data to a high-dimensional space and optimize the classification hyperplane to complete model training; During the simulation, a set of quantified gastric data is input into the trained model in real time, and the real-time digestive status index is calculated through a decision function.

6. The in vitro gastrointestinal tract simulation method according to claim 4, characterized in that, The construction and operation of the absorption kinetics simulator includes the following steps: Establish a differential equation model for nutrient absorption, with parameters including enzyme activity, peristalsis rate, and bile level; Numerical integration methods, such as the Runge-Kutta method, are used to solve the differential equation system and the absorption state is dynamically updated. The quantified small intestine dataset is used as input parameters, and the simulator outputs real-time absorption status indicators.

7. The in vitro gastrointestinal tract simulation method according to claim 4, characterized in that, The implementation of the fermentation reaction network simulator includes the following steps: Define the reaction network of the microbial community's metabolic pathway, including the equations for substrate consumption and product formation rates; The steady-state approximation method is used to calculate the flow distribution of each node in the network; By substituting the quantified E. coli dataset into the network model, real-time fermentation status indicators are obtained through iterative calculation.

8. The method for simulating the in vitro gastrointestinal tract according to claim 1, characterized in that, The operation of comparing the real-time digestive status indicators of the stomach simulation region, the real-time absorption status indicators of the small intestine simulation region, and the real-time fermentation status indicators of the large intestine simulation region with predefined threshold intervals, and initiating simulation condition correction operations based on the comparison results, is performed through the following steps: The threshold ranges for gastric digestion, small intestinal absorption, and large intestinal fermentation are defined based on physiological standards. Continuously monitor real-time digestion status indicators. If the indicator value exceeds the gastric digestion threshold range, the acid-base regulator or temperature controller in the gastric simulation area will be triggered for correction. Continuously monitor real-time absorption status indicators. If the indicator value exceeds the small intestine absorption threshold range, activate the enzyme addition pump or peristalsis simulation mechanism in the small intestine simulation area for adjustment. Continuously monitor real-time fermentation status indicators. If the indicator value exceeds the coli fermentation threshold range, activate the coliform colony injection system or viscosity adjustment device in the coliform simulation area.

9. The method for simulating the in vitro gastrointestinal tract according to claim 8, characterized in that, The specific steps for implementing the correction operation include: When the real-time digestion status indicators of the simulated gastric region exceed the threshold, the pH control variable is adjusted by changing the simulated gastric acid injection rate, or the temperature control variable is adjusted by adjusting the power of the heating element. When the real-time absorption status index of the small intestine simulation area exceeds the threshold, the enzyme addition control variable is adjusted by controlling the opening of the enzyme solution delivery valve, or the peristalsis control variable is adjusted by changing the frequency of mechanical peristalsis. When the real-time fermentation status indicators of the coli simulation area exceed the threshold, the inoculation control variable is adjusted by modulating the bacterial solution injection volume, or the viscosity control variable is adjusted by operating the water addition pump.

10. An in vitro gastrointestinal tract simulation system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the in vitro gastrointestinal tract simulation method as described in any one of claims 1 to 9.

Images