Mesenchymal stem cell culture monitoring system based on multiple parameters

Abstract

The invention provides a mesenchymal stem cell culture monitoring system based on multiple parameters, and relates to the technical field of data processing, and the mesenchymal stem cell culture monitoring system comprises a first-level control instruction generation module which is used for generating a second-level control instruction when the metabolism-morphological abnormality association intensity in a combined fuzzy state matrix is judged to reach a preset association threshold value through a fuzzy rule base; when the abnormal state flag bit indicates that the isolated metabolism is abnormal, a second-level control instruction is generated; the driving unit is used for responding to the first-stage control instruction and cooperatively driving the buffer solution injection unit and the nutrition supplementing pump; responding to the secondary control instruction, and dynamically adjusting the opening degrees of a stirring rate controller, a temperature feedback loop and a gas proportional valve of the bioreactor. The control efficiency and the cell culture stability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a mesenchymal stem cell culture monitoring system based on multiple parameters. Background Art

[0002] Large-scale culture of mesenchymal stem cells (MSCs) is a core component of regenerative medicine, and their quality directly impacts clinical efficacy. Current culture monitoring systems primarily rely on single-dimensional parameter threshold alarms (such as pH, dissolved oxygen, or microscopic image analysis), some of which have the following limitations: For example, traditional systems cannot dynamically correlate metabolic abnormalities (such as lactate accumulation) with cell morphological changes (such as pseudopodia contraction), resulting in delayed early warning; the cell metabolic rate rises nonlinearly in the middle and late stages of culture, and fixed thresholds are prone to false alarms (such as judging normal metabolic fluctuations in high-density culture as abnormal); regulatory instructions for local abnormalities (such as a single temperature drift) may aggravate the imbalance of the overall microenvironment (such as accelerated stirring leading to shear damage).

[0003] Therefore, in the existing technology, although some studies have attempted multi-parameter fusion analysis, it is still impossible to achieve graded and precise intervention, resulting in a certain batch of cell activity attenuation due to delayed remediation. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a mesenchymal stem cell culture monitoring system based on multiple parameters, thereby improving the control efficiency and cell culture stability.

[0005] In order to solve the above technical problems, the technical solutions of the present invention are as follows: First, a multi-parameter mesenchymal stem cell culture monitoring system, including: An acquisition module is used to obtain the real-time metabolic dynamic parameter set of the culture medium and the periodic morphological parameter set of the cell population; A generation module is used to train and generate a dynamic metabolic threshold library based on the historical data of the metabolic dynamic parameter set, and to construct a fuzzy morphological benchmark library based on the historical data of the morphological parameter set; The analysis module is used to compare the deviation of the real-time metabolic dynamic parameter set with the dynamic metabolic threshold library to generate metabolic abnormality membership; the real-time morphological parameter set is mapped to a preset feature space through a mirror projection algorithm, and the morphological abnormality confidence is calculated within this space based on the membership function of the fuzzy morphological reference library; the metabolic abnormality membership and morphological abnormality confidence are integrated to generate a joint fuzzy state matrix; A decision module is configured to generate a primary control instruction when the correlation strength between metabolic and morphological abnormalities in the joint fuzzy state matrix reaches a preset correlation threshold as determined by the fuzzy rule base; and to generate a secondary control instruction when the abnormal state flag indicates an isolated metabolic abnormality; The output module is used to respond to the first-level control instructions and coordinately drive the buffer injection unit and nutrient supplement pump; respond to the second-level control instructions and dynamically adjust the stirring rate controller, temperature feedback loop and gas proportional valve opening of the bioreactor.

[0006] Furthermore, the real-time metabolic dynamic parameter set of the culture medium and the periodic morphological parameter set of the cell population are obtained, including: The raw time-series data stream of the culture medium is collected through a biosensor array. A sliding window processor is used to perform segmented feature extraction on the raw time-series data stream to generate a primary metabolic feature set containing a mean vector, a standard deviation vector, and a parameter change rate sequence. Time dimension alignment and feature splicing are performed on the primary metabolic feature set to generate a structured metabolic dynamic parameter set. The microscopic imaging unit periodically captures a high-resolution image set of the cell population; and performs hierarchical processing on the high-resolution image set: Perform edge segmentation at the pixel level to generate a cell outline dataset. Extract geometric topological features at the single cell level based on the cell outline dataset to generate a single cell feature set. Generate population morphology indicators at the population level based on the spatial distribution relationship of the single cell feature set. The single-cell feature set and group morphology indicators are input into the multi-scale feature fusion, and cross-scale correlation features are generated through spatial distribution topology modeling; A structured morphological parameter set is constructed based on cross-scale correlation features and group morphological indicators.

[0007] Furthermore, a dynamic metabolic threshold library is generated based on the historical data of the metabolic dynamic parameter set, and a fuzzy morphological benchmark library is constructed based on the historical data of the morphological parameter set, including: The metabolic dynamic parameter sets of each culture cycle in the historical metabolic database are called; timestamp alignment is performed on the parameter sets to generate a time-series-associated metabolic feature tensor; a dynamic probability distribution model is constructed based on the metabolic feature tensor, and a dynamic confidence interval sequence is generated by extracting boundary values under a preset confidence level; The dynamic confidence interval series were integrated into a structured dynamic metabolic threshold library according to the culture stage; Extract morphological parameter sets across batches from a historical morphology database; perform time-series-aware incremental clustering on single-cell feature sets: Divide feature subsets according to the culture time window, perform density peak clustering on each subset to generate local morphological benchmarks, and fuse local benchmarks through time-attenuated weights to generate a global typical cell morphological benchmark vector. Modeling the dynamic evolution of group morphological indicators: The evolution trajectory of group indicators is constructed based on the timestamp index, and the Gaussian mixture model is fitted segment by segment according to the training stage to generate a dynamic benchmark interval set; The typical cell morphology benchmark vector and the dynamic benchmark interval set are input into the rule generator, and the adaptive membership function set for judging morphological parameter anomalies is defined. The membership function set and the corresponding benchmark data are associated and stored to generate a structured fuzzy morphology benchmark library.

[0008] Furthermore, the joint fuzzy state matrix includes: Obtain the real-time metabolic dynamic parameter set and call the threshold vector corresponding to the cultivation stage in the dynamic metabolic threshold library; Perform item-wise confidence interval deviation calculation on the mean vector, standard deviation vector and parameter change rate sequence of the parameter set to generate the original deviation set; Based on the original deviation set, a set of abnormal weight coefficients is generated based on the sensitivity of the parameters in the training stage; The original deviation set and the abnormal weight coefficient set are integrated to output the scalar metabolic abnormality membership; Obtain single cell feature sets and group morphology indicators in real-time morphological parameter sets; Call the typical cell morphology benchmark vector and dynamic benchmark interval set in the fuzzy morphology benchmark library; Project the single-cell feature set into the feature space spanned by the reference vector; calculate the Mahalanobis distance between the cell population distribution and the reference vector in the projected space to generate a single-cell deviation vector; perform dynamic interval out-of-bounds detection on the population morphology index to generate a population out-of-bounds flag set; According to the single cell deviation vector and the group out-of-bounds flag set, a scalar morphological abnormality confidence is obtained; The metabolic abnormality membership and morphological abnormality confidence are combined into a two-dimensional abnormal state vector. Based on the historical abnormal pattern library, an inter-dimensional correlation weight matrix is generated. Tensor multiplication operation is performed on the abnormal state vector and the correlation weight matrix, and the operation result is output as a joint fuzzy state matrix.

[0009] Furthermore, the metabolic abnormality membership and the morphological abnormality confidence are combined into a two-dimensional abnormal state vector; based on the historical abnormal pattern library, an inter-dimensional correlation weight matrix is generated, and a tensor multiplication operation is performed on the abnormal state vector and the correlation weight matrix to output the operation result as a joint fuzzy state matrix, including: The metabolic abnormality membership is used as the first dimension value, and the morphological abnormality confidence is used as the second dimension value; Generate a two-dimensional abnormal state vector by combining the dimensions in order; When the value of the first dimension exceeds the metabolic dominant threshold, the metabolic abnormality dominant factor is activated; when the value of the second dimension exceeds the morphological dominant threshold, the morphological abnormality dominant factor is activated; and a weight allocation instruction is generated according to the type of the activated dominant factor; In response to the weight allocation instruction, the historical abnormal pattern library is called; based on the correlation characteristics of similar dominant factors in the historical abnormal pattern library: According to metabolic abnormalities and morphological abnormalities, an adaptive weight matrix including self-action weights and cross-dimensional correlation weights is generated; A tensor convolution operation is performed on the two-dimensional abnormal state vector and the adaptive weight matrix, and the output result is used as the joint fuzzy state matrix.

[0010] Furthermore, when the correlation strength between metabolic and morphological abnormalities in the joint fuzzy state matrix reaches a preset correlation threshold as determined by the fuzzy rule base, a first-level control instruction is generated; when the abnormal state flag indicates an isolated metabolic abnormality, a second-level control instruction is generated, including: Extract the comprehensive intensity of metabolic abnormality and the comprehensive intensity of morphological abnormality from the joint fuzzy state matrix; Generate an abnormal intensity level identification set based on preset intensity grading rules; The abnormality intensity level identification set is input into the fuzzy rule base. When the metabolic and morphological abnormality intensities both reach the high-level threshold, the metabolic-morphological strong association flag is activated. When the metabolic abnormality intensity reaches the high level and the morphological abnormality intensity is lower than the decoupling threshold, the isolated metabolic abnormality flag is activated. In response to the strong metabolism-morphology correlation flag, a first-level control instruction generation pathway is triggered; In response to the isolated metabolic abnormality flag, a secondary control instruction generation pathway is triggered.

[0011] Furthermore, in response to the primary control instruction, the buffer injection unit and the nutrient supplement pump are cooperatively driven; in response to the secondary control instruction, the stirring rate controller, the temperature feedback loop and the gas proportional valve opening of the bioreactor are dynamically adjusted, including: In response to the first-level control instruction, the comprehensive intensity of metabolic abnormalities and the comprehensive intensity of morphological abnormalities in the joint fuzzy state matrix are extracted; The buffer control model processes the comprehensive intensity of metabolic abnormalities to generate a buffer pulse injection parameter set; it integrates the metabolic and morphological abnormality intensities to generate a nutrient supplement flow control curve; and it collaboratively drives the buffer injection unit and the nutrient supplement pump to perform linked operations. In response to the secondary control instruction, the parameter change rate sequence and metabolic abnormality membership in the real-time metabolic dynamic parameter set are called; Perform multi-parameter coupling analysis through a dynamic control strategy engine: Generate an adaptive correction value for the stirring rate based on the parameter change rate sequence; generate a temperature feedback compensation signal based on the metabolic abnormality membership; and generate a proportional valve opening increment sequence through the gas valve control model; Synchronously regulate the stirring rate controller, temperature feedback loop and gas proportioning valve.

[0012] In a second aspect, a computing device includes: one or more processors; The storage device is used to store one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the system.

[0013] According to a third aspect, a computer-readable storage medium stores a program, which implements the system when executed by a processor.

[0014] The above solution of the present invention includes at least the following beneficial effects: By acquiring the dynamic metabolic parameters of the culture medium and the periodic morphological parameters of the cell population in real time, the system overcomes the hysteresis limitations of traditional offline sampling monitoring and can instantly capture dynamic changes in cell metabolism and morphology. A dynamic metabolic threshold library constructed based on historical data dynamically adjusts the judgment criteria according to the cell growth stage, while a fuzzy morphological benchmark library quantifies morphological features at different cycles, resolving the problem of misjudgment caused by static thresholds and subjective judgment, thereby improving the accuracy of cell status monitoring.

[0015] Through automated parameter collection, analysis, and control instruction execution, manual sampling, observation, and operation steps are reduced, which not only reduces labor costs, but also reduces the interference of the external environment on the culture system, reduces the risk of contamination, and is conducive to ensuring the safety of mesenchymal stem cell culture. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 1 is a flow chart of a multi-parameter-based mesenchymal stem cell culture monitoring system provided by an embodiment of the present invention.

[0017] Figure 2 This is a flow chart of the present invention, which combines the metabolic abnormality membership and the morphological abnormality confidence into a two-dimensional abnormal state vector, generates an inter-dimensional correlation weight matrix based on a historical abnormal pattern library, performs a tensor multiplication operation on the abnormal state vector and the correlation weight matrix, and outputs the operation result as a joint fuzzy state matrix. DETAILED DESCRIPTION

[0018] The following will describe in more detail exemplary embodiments of the present disclosure with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, the embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0019] like Figure 1 As shown, an embodiment of the present invention provides a multi-parameter-based mesenchymal stem cell culture monitoring system, comprising: An acquisition module is used to obtain the real-time metabolic dynamic parameter set of the culture medium and the periodic morphological parameter set of the cell population; A generation module is used to train and generate a dynamic metabolic threshold library based on the historical data of the metabolic dynamic parameter set, and to construct a fuzzy morphological benchmark library based on the historical data of the morphological parameter set; The analysis module is used to compare the deviation of the real-time metabolic dynamic parameter set with the dynamic metabolic threshold library to generate metabolic abnormality membership; the real-time morphological parameter set is mapped to a preset feature space through a mirror projection algorithm, and the morphological abnormality confidence is calculated within this space based on the membership function of the fuzzy morphological reference library; the metabolic abnormality membership and morphological abnormality confidence are integrated to generate a joint fuzzy state matrix; A decision module is configured to generate a primary control instruction when the correlation strength between metabolic and morphological abnormalities in the joint fuzzy state matrix reaches a preset correlation threshold as determined by the fuzzy rule base; and to generate a secondary control instruction when the abnormal state flag indicates an isolated metabolic abnormality; The output module is used to respond to the first-level control instructions and coordinately drive the buffer injection unit and nutrient supplement pump; respond to the second-level control instructions and dynamically adjust the stirring rate controller, temperature feedback loop and gas proportional valve opening of the bioreactor.

[0020] In this embodiment of the present invention, by acquiring a set of metabolic dynamic parameters of the culture medium and a set of periodic morphological parameters of the cell population in real time, the hysteresis limitations of traditional offline sampling monitoring are overcome, enabling the immediate capture of dynamic changes in cell metabolism and morphology. A dynamic metabolic threshold library constructed based on historical data dynamically adjusts the judgment criteria according to the cell growth stage, while a fuzzy morphological benchmark library quantifies morphological features at different cycles, resolving the problem of misjudgment caused by static thresholds and subjective judgment, thereby improving the accuracy of cell status monitoring.

[0021] Through automated parameter collection, analysis, and control instruction execution, manual sampling, observation, and operation steps are reduced, which not only reduces labor costs, but also reduces the interference of the external environment on the culture system, reduces the risk of contamination, and is conducive to ensuring the safety of mesenchymal stem cell culture.

[0022] In a preferred embodiment of the present invention, obtaining a real-time metabolic dynamic parameter set of a culture medium and a periodic morphological parameter set of a cell population includes: The raw time-series data stream of the culture medium is collected through a biosensor array. A sliding window processor is used to perform segmented feature extraction on the raw time-series data stream to generate a primary metabolic feature set containing a mean vector, a standard deviation vector, and a parameter change rate sequence. Time dimension alignment and feature splicing are performed on the primary metabolic feature set to generate a structured metabolic dynamic parameter set. The microscopic imaging unit periodically captures a high-resolution image set of the cell population; and performs hierarchical processing on the high-resolution image set: Perform edge segmentation at the pixel level to generate a cell outline dataset. Extract geometric topological features at the single cell level based on the cell outline dataset to generate a single cell feature set. Generate population morphology indicators at the population level based on the spatial distribution relationship of the single cell feature set. The single-cell feature set and group morphology indicators are input into the multi-scale feature fusion, and cross-scale correlation features are generated through spatial distribution topology modeling; A structured morphological parameter set is constructed based on cross-scale correlation features and group morphological indicators.

[0023] In an embodiment of the present invention, a biosensor array (monitoring key metabolic indicators such as pH value, dissolved oxygen concentration, glucose content, lactate concentration, etc. in the culture medium) continuously collects data at set time intervals (e.g., once per second) to form a raw data stream that changes continuously over time (each indicator corresponds to a timing curve).

[0024] The parameters of the sliding window processor are adjusted according to the degree of metabolic activity: in the logarithmic growth phase, the window length is set to 5 minutes (covering 600 data points in 300 seconds) and the sliding step is 1 minute (each movement is 60 seconds) to ensure intensive capture of changes; in the non-logarithmic growth phase, the window length is set to 10 minutes (covering 300 data points in 600 seconds) and the sliding step is 2 minutes to reduce redundant calculations.

[0025] For a certain type of indicator data within a single window (such as pH value within 5 minutes), the values of all data points are added up and divided by the number of data points to obtain the average level within the window. For example, there are 600 pH value data in a 5-minute window, and the sum is 4260, then the mean is 7.1. The window means of all indicators are integrated into a vector (such as [pH mean 7.1, dissolved oxygen mean 5.3 mg / L, glucose mean 4.8 mmol / L, lactate mean 1.1 mmol / L]). Calculate the deviation between a certain type of indicator data in a single window and the window mean, square all deviations and take the average, then square the root to get the standard deviation, which reflects the magnitude of data fluctuation. If the pH value fluctuates slightly within the window, the standard deviation may be 0.06; if the dissolved oxygen changes dramatically due to rapid cell consumption, the standard deviation may reach 0.5 mg / L. Take the mean of the same indicator in two adjacent windows, subtract the mean of the previous window from the mean of the latter window, and divide by the time interval between the two windows (i.e., the sliding step size) to obtain the rate of change of the indicator. For example, if the mean glucose value in the previous window is 5.0 mmol / L and the mean glucose value in the latter window is 4.7 mmol / L, and the step size is 1 minute, then the rate of change is (4.7-5.0) ÷ 1 = -0.3 mmol / (L・minute). Continuously calculate the rate of change of adjacent windows to form a rate sequence over time.

[0026] Since different sensors have different response delays (e.g., the pH sensor has a delay of approximately 0.3 seconds, and the glucose sensor has a delay of approximately 0.5 seconds), the characteristic sequence of each indicator is calibrated based on the system's unified clock: if the pH window feature has been recorded at a certain moment (e.g., the 1200th second), but the glucose feature has not yet been generated, the approximate value at the 1200th second is estimated based on the glucose feature values at the 1199th and 1201st seconds, ensuring that the four indicators have corresponding mean, standard deviation, and rate of change data at the same time point.

[0027] The characteristic data at each time point are spliced according to a fixed structure: first, the means of the four indicators (pH, dissolved oxygen, glucose, and lactic acid) are arranged; then the standard deviations of the four indicators are arranged; and finally, the rates of change of the four indicators are arranged.

[0028] A feature combination of 12 data items was formed (e.g., [7.1, 5.3, 4.8, 1.1, 0.06, 0.5, 0.2, 0.1, -0.02, -0.05, -0.3, 0.08]). The feature combinations of all time points were arranged in chronological order to form a structured metabolic dynamic parameter set. Each entry can fully reflect the metabolic state of the culture medium at a certain moment (average level, fluctuation amplitude, and change trend).

[0029] High-resolution images of the cell population are captured by a microscopic imaging unit at a fixed period (e.g., once per hour) to form a time series image set.

[0030] Perform edge segmentation on each image, identify and extract the contour boundary of each cell by distinguishing the grayscale difference and texture features between cells and background (such as using threshold segmentation and edge detection algorithm), and generate a cell contour dataset containing information such as contour pixel coordinates and boundary continuity.

[0031] The geometric topological features of individual cells are extracted based on the contour dataset, including: geometric features (area, perimeter, equivalent diameter, circularity, aspect ratio), topological features (the relative position of the cell nucleus and the cell contour, the degree of cell membrane wrinkling), etc. Each cell corresponds to a set of features, which are summarized as a single-cell feature set.

[0032] Analyze the spatial distribution relationship of single-cell feature sets and calculate population morphological indicators, including: cell density (number of cells per unit area), spatial distribution uniformity (such as average distance between cells, local aggregation coefficient), size / morphology population heterogeneity (such as standard deviation of cell size, proportion of cells with specific morphology), and spatial directionality (such as the dominant direction of cell arrangement).

[0033] The single-cell feature set and population morphology indicators are input into the multi-scale feature fusion device. By analyzing "how single-cell characteristics affect population distribution" (such as whether large-sized cells tend to gather in high-density areas, whether cells with specific morphology maintain a fixed distance from surrounding cells), a spatial distribution topological model (such as the association rules between local cell characteristics and population density) is constructed, and cross-scale correlation features (such as "population density gradient in large cell aggregation areas" and "spatial distribution entropy of highly circular cells") are generated.

[0034] The cross-scale correlation features and group morphological indicators are integrated in chronological order (corresponding to the time points of periodic image acquisition) to form a multi-dimensional morphological feature matrix corresponding to each time point (such as the average area of single cells, group density, and distribution entropy of large cell aggregation areas at a certain moment), that is, a set of structured morphological parameters.

[0035] Multi-dimensional features (mean, fluctuation, and rate of change) reflect the metabolic state of the culture medium in real time, capturing both overall trends and subtle fluctuations, providing a precise quantitative basis for regulating cellular metabolism. Hierarchical analysis, from pixels to single cells to populations, combined with cross-scale correlations, comprehensively reflects individual differences in cell morphology and the spatial organization of populations, avoiding the limitations of single-scale analysis. Structured parameter sets unify the temporal dimension, directly correlating metabolic environments with cell morphological changes at the same point in time.

[0036] In a preferred embodiment of the present invention, a dynamic metabolic threshold library is generated based on historical data of the metabolic dynamic parameter set, and a fuzzy morphological reference library is constructed based on historical data of the morphological parameter set, including: The metabolic dynamic parameter sets of each culture cycle in the historical metabolic database are called; timestamp alignment is performed on the parameter sets to generate a time-series-associated metabolic feature tensor; a dynamic probability distribution model is constructed based on the metabolic feature tensor, and a dynamic confidence interval sequence is generated by extracting boundary values under a preset confidence level; The dynamic confidence interval series were integrated into a structured dynamic metabolic threshold library according to the culture stage; Extract morphological parameter sets across batches from a historical morphology database; perform time-series-aware incremental clustering on single-cell feature sets: Divide feature subsets according to the culture time window, perform density peak clustering on each subset to generate local morphological benchmarks, and fuse local benchmarks through time-attenuated weights to generate a global typical cell morphological benchmark vector. Modeling the dynamic evolution of group morphological indicators: The evolution trajectory of group indicators is constructed based on the timestamp index, and the Gaussian mixture model is fitted segment by segment according to the training stage to generate a dynamic benchmark interval set; The typical cell morphology benchmark vector and the dynamic benchmark interval set are input into the rule generator, and the adaptive membership function set for judging morphological parameter anomalies is defined. The membership function set and the corresponding benchmark data are associated and stored to generate a structured fuzzy morphology benchmark library.

[0037] In an embodiment of the present invention, a structured metabolic dynamic parameter set of multiple culture cycles is extracted from a historical metabolic database, and the absolute time difference is eliminated by relative time conversion (with the unified starting time as "time 0") to generate a time-series-associated metabolic feature tensor (with the dimensions of "culture batch × relative time point × metabolic feature dimension").

[0038] A dynamic probability distribution model is constructed based on the metabolic feature tensor. For the metabolic features (such as pH mean and glucose change rate) at each relative time point, the numerical distribution of all historical batches is statistically analyzed. A preset confidence level of 90% to 95% is used (95% for strong regularity parameters such as glucose concentration and 90% for weak regularity parameters such as local pH fluctuations), and the boundary values under the corresponding confidence level are extracted. For example, at a confidence level of 95%, the upper and lower limits are calculated using the fitted normal distribution as "mean ± 1.96 × standard deviation", while at a confidence level of 90%, the upper and lower limits are "mean ± 1.64 × standard deviation", forming the normal range for a single time point. The data are arranged in relative time order to generate a dynamic confidence interval sequence.

[0039] Constructing a dynamic probability distribution model: Based on the metabolic feature tensor, "single relative time point + single metabolic feature" is used as the minimum analysis unit of the model, for example: "pH standard deviation at the 2nd hour of culture", "dissolved oxygen change rate at the 6th hour of culture", "glucose concentration at the 10th hour of culture", etc. Each unit corresponds to an independent analysis object to ensure the targetedness of the model; for each analysis unit (such as "lactate concentration at the 4th hour"), the corresponding data of all historical normal culture batches are screened from the metabolic feature tensor: assuming there are 80 historical normal batches, the lactate concentration values recorded for these 80 batches at the "4th hour of culture" are extracted to form an exclusive data set (80 values in total); only the data of "the same relative time point", "the same metabolic feature" and "normal batch" are retained, and the data of abnormal batches or other time points are excluded to ensure that the data only reflects normal fluctuations; observe the specific distribution status of the data set: for example, whether the values are concentrated around a certain central value (such as most of them are between 5.2-5 .6mM), whether it presents a symmetrical shape of "dense in the middle and sparse at both ends", or whether there is a skewness (for example, most values are low, a few are high but still normal); according to the distribution characteristics of the data, choose the corresponding method to quantitatively describe the fluctuation pattern: if the data presents a symmetrical and concentrated distribution (such as the conventional indicators of most metabolic characteristics), use the "central trend + dispersion" fitting logic: use the central value in the data set (such as the core value around which most values ​​are concentrated) to reflect the common range, and use the dispersion amplitude of the data (such as the average distance of most values ​​from the central value) to reflect the size of the fluctuation; if the data distribution is more complex (such as the presence of multiple fluctuation centers or asymmetric distribution), use the "density clustering" fitting logic: identify the numerical intervals with high frequency in the data (such as the two dense intervals of 3.1-3.3mM and 3.5-3.7mM), use the position and coverage of the interval to describe normal fluctuations, and record the probability of occurrence of values ​​outside the interval (for example, only 1% of normal data fall outside the interval).

[0040] Based on the above fitting logic, a unique distribution model is generated for each analysis unit: For the "lactate concentration at the 4th hour", if the logic of central tendency + dispersion is adopted, the model will clearly state: "The common central value of lactate concentration at this time point is 5.4mM, and most normal data (such as 90%) will fluctuate between 5.0-5.8mM. The farther it deviates from this range, the lower the probability of occurrence"; if the logic of density clustering is adopted, the model will clearly state: "The lactate concentration at this time point is mainly concentrated in 3.2-3.4mM (accounting for 60% of normal data) and 3.6-3.8mM (accounting for 30% of normal data). The values in these two ranges are all normal, and only 10% of normal data fall outside the range"; the exclusive distribution models of all analysis units are integrated according to the dimensions of "relative time point + metabolic characteristics" to form the final "dynamic probability distribution model".

[0041] The model as a whole is a structured set that includes the normal fluctuation patterns of all metabolic characteristics at all time points throughout the cultivation process. For example, from the 1st to the 48th hour of cultivation, the pH mean, dissolved oxygen content, and glucose concentration characteristics of each hour have corresponding distribution models that clearly record the normal numerical range, occurrence probability, and fluctuation pattern of the characteristic at that time point.

[0042] Dynamic confidence interval sequences are divided according to the culture stage (adaptation period, logarithmic growth period, etc.). In each stage, the confidence intervals of metabolic characteristics at all time points within that stage are integrated (the range of values fluctuates with the degree of discreteness of historical data; for example, the 95% confidence interval of glucose concentration in the logarithmic growth period may be 3.4-6.6 mM), forming a structured storage of "stage-time point-feature-threshold upper and lower limits".

[0043] Multiple batches of structured morphological parameter sets are extracted from the historical morphological database, including single-cell features, population indicators and cross-scale correlation features; they are divided into continuous time windows according to the culture time (such as 1 window every 4 hours), and the single-cell features in each window constitute a feature subset.

[0044] The feature subsets of each window are clustered by density peaks, and the highest point of feature space density is used as the cluster center to generate a local morphological benchmark (e.g., "typical cell area 20 μm in a certain window"). 2 , circularity 0.8”).

[0045] The local benchmarks were fused using time-decaying weights, with a single weight ranging from 0 to 1 and the sum of the weights of all time windows being 1. Recent time windows were given higher weights (e.g., if the current window was t, the weight of the t-1 window was 0.8, the t-2 window was 0.15, and the t-3 and earlier windows were 0.05). Alternatively, exponential decay was used (the decay coefficient τ was usually taken from 3 to 10 time windows to ensure that the weight dropped to less than 5% of the initial value after an interval of 3 windows). A global representative cell morphology benchmark vector covering the entire culture process was generated by weighted averaging.

[0046] The historical batch data of each group indicator (cell density, distribution uniformity, etc.) were integrated by timestamp to form a set of trajectories that changed over time. The data were segmented by culture stage, and the indicator distribution was fitted using a Gaussian mixture model. The 90%-95% confidence interval (95% for strong regularity indicators such as cell density and 90% for weak regularity indicators such as distribution uniformity) was extracted as the dynamic reference interval for that stage (for example, the 95% confidence interval for cell density in the stable period might be 1.2×10 6 ~2.5×10 6 cells / mL).

[0047] The global typical benchmark vector and the dynamic benchmark interval set are input into the rule generator to define the adaptive membership function set. The membership value range is 0-1, where 1 means the parameter "completely meets the normal benchmark", 0 means "completely deviates from the normal benchmark", and values between 0 and 1 represent "transitional state" (e.g., a membership of 0.3 means "slightly deviates from the normal"). The membership function boundaries are dynamically adjusted according to the benchmark (e.g., the typical cell area is 50 μm 2 When the “normal” membership is between 40 and 60 μm 2 The membership function set (value range 0-1) and the benchmark data are stored in association to form a structured fuzzy morphological benchmark library.

[0048] The dynamic metabolic threshold library dynamically adjusts the normal interval according to the culture stage and time, and the fuzzy morphological benchmark library captures the stage specificity of morphological characteristics through time-series-aware clustering and dynamic modeling. Both can adapt to the dynamic changes of the cell culture process and avoid misjudgment caused by fixed thresholds / benchmarks. The metabolic threshold is determined based on the probability distribution of historical data, and the morphological benchmark uses fuzzy membership functions to process the continuity and fuzziness of morphological characteristics (such as "close to abnormality but not exceeding the absolute threshold"), thereby improving the ability to identify early abnormalities and minor abnormalities; the benchmark library integrates historical rules across batches and can be used as a "gold standard" for the culture process. It is used to compare the deviation of real-time parameters from historical normal patterns, providing a quantitative basis for adjusting culture conditions (such as supplementing nutrition and optimizing environmental parameters) and reducing batch differences.

[0049] In a preferred embodiment of the present invention, the joint fuzzy state matrix includes: Obtain the real-time metabolic dynamic parameter set and call the threshold vector corresponding to the cultivation stage in the dynamic metabolic threshold library; Perform item-wise confidence interval deviation calculation on the mean vector, standard deviation vector and parameter change rate sequence of the parameter set to generate the original deviation set; Based on the original deviation set, a set of abnormal weight coefficients is generated based on the sensitivity of the parameters in the training stage; The original deviation set and the abnormal weight coefficient set are integrated to output the scalar metabolic abnormality membership; Obtain single cell feature sets and group morphology indicators in real-time morphological parameter sets; Call the typical cell morphology benchmark vector and dynamic benchmark interval set in the fuzzy morphology benchmark library; Project the single-cell feature set into the feature space spanned by the reference vector; calculate the Mahalanobis distance between the cell population distribution and the reference vector in the projected space to generate a single-cell deviation vector; perform dynamic interval out-of-bounds detection on the population morphology index to generate a population out-of-bounds flag set; According to the single cell deviation vector and the group out-of-bounds flag set, a scalar morphological abnormality confidence is obtained; The metabolic abnormality membership and morphological abnormality confidence are combined into a two-dimensional abnormal state vector. Based on the historical abnormal pattern library, an inter-dimensional correlation weight matrix is generated. Tensor multiplication operation is performed on the abnormal state vector and the correlation weight matrix, and the operation result is output as a joint fuzzy state matrix.

[0050] In an embodiment of the present invention, a real-time metabolic dynamic parameter set (including a mean vector, a standard deviation vector, and a parameter change rate sequence) at the current culture moment is obtained, and according to the current culture stage (such as the logarithmic growth phase), the threshold vector of the corresponding stage (i.e., the normal confidence interval of each metabolic parameter in this stage, such as glucose concentration 3.4~6.6mM, pH 6.8~7.2) is called from the dynamic metabolic threshold library.

[0051] A "confidence interval deviation" calculation is performed on each parameter in the metabolic parameter set. If the parameter value is within the threshold interval, the deviation is 0; if it exceeds the interval (such as glucose concentration of 7.0mM, which is higher than the upper limit of 6.6mM), the degree of deviation is calculated (such as the ratio of the excess part to the interval width, 7.0-6.6=0.4mM, interval width 6.6-3.4=3.2mM, and the deviation is 0.4 / 3.2=0.125); the deviations of all parameters are summarized to generate the original deviation set (such as containing glucose deviation 0.125, pH deviation 0, dissolved oxygen deviation 0.05, etc.).

[0052] According to the physiological sensitivity of each parameter in the current culture stage (for example, the effect of glucose concentration on cell proliferation in the logarithmic growth phase is much greater than that of local pH fluctuations), an abnormal weight coefficient (range 0-1, the higher the sensitivity, the greater the weight, such as glucose weight 0.8, pH weight 0.2) is assigned to each parameter to form an abnormal weight coefficient set.

[0053] The original deviation set and the abnormal weight coefficient set are weightedly fused (for example, the deviation of each parameter is multiplied by the corresponding weight and then summed up) to obtain the scalar metabolic abnormality membership (value range 0-1, 0 indicates no abnormality, 1 indicates severe abnormality, such as the calculated result of 0.15 indicates mild metabolic abnormality).

[0054] Obtain the real-time morphological parameter set at the current culture moment (including single cell feature sets such as area and circularity, and group morphological indicators such as cell density and distribution uniformity), and call the typical cell morphology benchmark vector of the corresponding stage from the fuzzy morphology benchmark library (such as the typical cell area of 50μm in this stage). 2 , circularity 0.8) and a population dynamics benchmark interval set (e.g., cell density 1.2×10 6 ~2.5×10 6 cells / mL).

[0055] The real-time single-cell feature set (area, circularity, etc. of each cell) is projected into a feature space spanned by a reference vector of typical cell morphology (i.e., aligned with the dimensions of the reference vector to ensure feature comparability). Within the projected space, the Mahalanobis distance (a distance that takes into account inter-feature correlations, such as the association between area and circularity) between the feature distribution of the cell population and the reference vector is calculated to obtain the degree of deviation of each cell and summarize it into a single-cell deviation vector (element value range 0-1, with larger values indicating greater deviation from the reference).

[0056] Perform "dynamic interval crossing detection" on the population morphology index. If the index is within the dynamic reference interval (e.g. cell density 2.0×10 6 cells / mL, 1.2~2.5×10 6 If the range is exceeded (e.g. 3.0×10 6 cells / mL, a mark value of 0 to 1 is assigned according to the degree of excess (e.g., if the upper limit is exceeded by 0.5×10 6 , interval width 1.3×10 6 , flag value 0.5 / 1.3≈0.38), generating a group out-of-bounds flag set.

[0057] The average value of the single-cell deviation vector (reflecting the overall deviation degree of the group) is combined with the weighted sum of the group out-of-bounds flag set (the comprehensive impact of the group indicator out-of-bounds) to obtain a scalar morphological abnormality confidence (value range 0-1, such as a calculated result of 0.2 indicates mild morphological abnormality).

[0058] The metabolic abnormality membership (such as 0.15) and the morphological abnormality confidence (such as 0.2) are combined into a two-dimensional abnormality state vector (such as [0.15, 0.2]), and the vector elements correspond to the degree of metabolic and morphological abnormalities, respectively.

[0059] Extract the inter-dimensional correlation weight matrix from the historical abnormality pattern library - this matrix is generated based on the correlation pattern between metabolic abnormalities and morphological abnormalities in historical data (for example, if the probability of "metabolic abnormalities accompanied by morphological abnormalities" in history is high, the corresponding element weight in the matrix will be high), and the matrix element value range is 0~1 (for example, the weight matrix is [[0.9, 0.3], [0.3, 0.8]], indicating that the metabolic abnormality has a weight of 0.9 on its own state and a correlation weight of 0.3 on the morphological state).

[0060] A tensor multiplication operation is performed on the two-dimensional abnormal state vector and the associated weight matrix (i.e., element-wise weighted combination of the vector and the matrix). The output result is a joint fuzzy state matrix (e.g., a 2×2 matrix, whose element values reflect the comprehensive abnormal state of "metabolism-metabolism", "metabolism-morphology", "morphology-metabolism", and "morphology-morphology", with a value range of 0~1).

[0061] This dual-dimensional abnormality assessment of metabolism and morphology avoids the limitations of misjudgment based on a single indicator (e.g., metabolism or morphology alone) and more comprehensively reflects the true state of cells. The correlation weight matrix incorporates historical abnormality patterns to quantify the inherent correlation between metabolic and morphological abnormalities (e.g., insufficient metabolic substrates are often accompanied by shrunken cell morphology), enhancing the logic of abnormality determination. The degree of abnormality is described using a continuous value from 0 to 1, rather than a simple "normal / abnormal" binary, providing a more refined quantitative basis for early warning and precise regulation of cell culture processes.

[0062] In a preferred embodiment of the present invention, the metabolic abnormality membership and the morphological abnormality confidence are combined into a two-dimensional abnormal state vector; a dimension-to-dimensional correlation weight matrix is generated based on a historical abnormal pattern library; a tensor multiplication operation is performed on the abnormal state vector and the correlation weight matrix, and the operation result is output as a joint fuzzy state matrix, including: The metabolic abnormality membership is used as the first dimension value, and the morphological abnormality confidence is used as the second dimension value; Generate a two-dimensional abnormal state vector by combining the dimensions in order; When the value of the first dimension exceeds the metabolic dominant threshold, the metabolic abnormality dominant factor is activated; when the value of the second dimension exceeds the morphological dominant threshold, the morphological abnormality dominant factor is activated; and a weight allocation instruction is generated according to the type of the activated dominant factor; In response to the weight allocation instruction, the historical abnormal pattern library is called; based on the correlation characteristics of similar dominant factors in the historical abnormal pattern library: According to metabolic abnormalities and morphological abnormalities, an adaptive weight matrix including self-action weights and cross-dimensional correlation weights is generated; A tensor convolution operation is performed on the two-dimensional abnormal state vector and the adaptive weight matrix, and the output result is used as the joint fuzzy state matrix.

[0063] In an embodiment of the present invention, the previously generated scalar metabolic abnormality membership (value range 0~1, such as 0.6) is used as the first dimension value of the two-dimensional vector (representing the degree of metabolic abnormality); the scalar morphological abnormality confidence (value range 0~1, such as 0.3) is used as the second dimension value (representing the degree of morphological abnormality).

[0064] Vector combination: In the order of "metabolic dimension first, morphological dimension last", the two dimension values are combined into a two-dimensional abnormal state vector (such as [0.6, 0.3]). The vector directly reflects the current abnormal level of metabolism and morphology.

[0065] A metabolic dominance threshold (e.g., 0.5, adjustable based on historical data) and a morphological dominance threshold (e.g., 0.5) are preset. If the first dimension of the two-dimensional vector (the degree of membership of metabolic anomaly) exceeds the metabolic dominance threshold (e.g., 0.6 > 0.5), the "metabolic anomaly dominance factor" is activated; if the second dimension (the confidence level of morphological anomaly) exceeds the morphological dominance threshold (e.g., 0.6 > 0.5), the "morphological anomaly dominance factor" is activated. If both exceed the threshold, both dominance factors are activated simultaneously.

[0066] Instructions are generated based on the type of activated dominant factor. For example, when only the metabolic dominant factor is activated, the instruction is "enhance the metabolic dimension's own weight and adjust the cross-dimensional correlation weight of metabolism to morphology"; when only the morphological dominant factor is activated, the instruction is "enhance the morphological dimension's own weight and adjust the cross-dimensional correlation weight of morphology to metabolism"; when both factors are activated, the instruction is "balance the weights of the two and strengthen the two-way cross-dimensional correlation weight."

[0067] In response to the weight allocation instruction, the correlation characteristics of the same type of dominant factors are extracted from the historical abnormality patterns. For example, the historical data of "metabolic dominant abnormality" is called to obtain the correlation law between metabolic abnormality and morphological abnormality in this pattern (such as the correlation strength between the degree of metabolic abnormality and cell morphological shrinkage when the metabolic substrate is insufficient); the historical data of "morphological dominant abnormality" is called to obtain the influence of morphological abnormality (such as cell aggregation) on the metabolic environment (such as aggregation leading to local metabolic waste accumulation).

[0068] An adaptive weight matrix (2×2 matrix) is generated based on historical association features, including self-action weights: the influence weight of the metabolic dimension on itself (e.g., set to 0.8 when metabolism is dominant) and the influence weight of the morphological dimension on itself (e.g., set to 0.4 when metabolism is dominant); cross-dimensional association weights: the influence weight of the metabolic dimension on the morphological dimension (e.g., set to 0.3 when metabolism is dominant, reflecting the degree of involvement of metabolic abnormalities on morphology), and the influence weight of the morphological dimension on the metabolic dimension (e.g., set to 0.2 when metabolism is dominant, reflecting the reverse effect of morphology on metabolism); weight values are dynamically adjusted according to historical patterns (value range 0–1, with the dominant dimension’s self-weight and related cross-dimensional weights being higher).

[0069] A tensor convolution operation is performed on a two-dimensional abnormal state vector (e.g., [0.6, 0.3]) with an adaptive weight matrix. This operation weights and fuses the two dimensions of the vector using the matrix elements. This not only preserves the abnormal information of each dimension but also incorporates the influence of inter-dimensional correlations (e.g., metabolic abnormalities influence the assessment of morphological abnormalities through cross-dimensional weights). The output of this operation is a joint fuzzy state matrix (2×2 matrix, with element values ranging from 0 to 1), in which each element represents the combined abnormal state of "metabolism-metabolism," "metabolism-morphology," "morphology-metabolism," and "morphology-morphology."

[0070] By activating dominant factors, weight distribution is tilted toward the currently more significant abnormal dimension (e.g., when metabolic abnormalities are more severe, metabolic-related weights are prioritized), preventing secondary abnormalities from interfering with core judgments. An adaptive weight matrix is generated based on a historical abnormality pattern library, quantifying the potential associations between metabolic and morphological abnormalities (e.g., metabolic disorders often cause morphological changes) as weights, making abnormality assessment more consistent with the actual laws of biological processes. The joint fuzzy state matrix not only reflects abnormalities in a single dimension but also, through the quantification of "inter-dimensional interactions," comprehensively depicts the complex abnormal patterns of cellular states, providing more specific targets for precise intervention in the culture process (e.g., adjusting nutrient supply and culture environment for abnormalities in the "metabolism-morphology" relationship).

[0071] In a preferred embodiment of the present invention, when the correlation strength between metabolic and morphological abnormalities in the joint fuzzy state matrix reaches a preset correlation threshold as determined by the fuzzy rule base, a primary control instruction is generated; when the abnormal state flag indicates an isolated metabolic abnormality, a secondary control instruction is generated, including: Extract the comprehensive intensity of metabolic abnormality and the comprehensive intensity of morphological abnormality from the joint fuzzy state matrix; Generate an abnormal intensity level identification set based on preset intensity grading rules; The abnormality intensity level identification set is input into the fuzzy rule base. When the metabolic and morphological abnormality intensities both reach the high-level threshold, the metabolic-morphological strong association flag is activated. When the metabolic abnormality intensity reaches the high level and the morphological abnormality intensity is lower than the decoupling threshold, the isolated metabolic abnormality flag is activated. In response to the strong metabolism-morphology correlation flag, a first-level control instruction generation pathway is triggered; In response to the isolated metabolic abnormality flag, a secondary control instruction generation pathway is triggered.

[0072] In this embodiment of the present invention, the combined strength of metabolic abnormalities and the combined strength of morphological abnormalities are extracted from the joint fuzzy state matrix. The combined strength of metabolic abnormalities integrates the values of the "metabolism-metabolism" and "morphology-metabolism" elements in the matrix (reflecting the sum of abnormalities in metabolism itself and the associated abnormalities in morphology). The combined strength of morphological abnormalities integrates the values of the "morphology-morphology" and "metabolism-morphology" elements (reflecting the sum of abnormalities in morphology itself and the associated abnormalities in metabolism and morphology). Both values range from 0 to 1.

[0073] Based on preset intensity grading rules (e.g., 0-0.3 is "low," 0.3-0.7 is "medium," and 0.7-1 is "high"), the combined intensity of metabolism and morphology is graded. For example, a combined metabolic intensity of 0.8 is graded "high," while a combined morphological intensity of 0.2 is graded "low." This generates an abnormal intensity grade set containing both dimensional grades (e.g., [high, low]).

[0074] The abnormality intensity level identification set is input into the fuzzy rule base, and the following judgment is performed: when the comprehensive intensity of metabolic abnormalities reaches a "high level" (such as ≥0.7) and the comprehensive intensity of morphological abnormalities also reaches a "high level" (such as ≥0.7), it is determined that there is a strong correlation between the two abnormalities, and the "metabolic-morphological strong correlation flag" is activated.

[0075] When the comprehensive intensity of metabolic abnormalities reaches a "high level" (such as ≥0.7), but the comprehensive intensity of morphological abnormalities is lower than the preset decoupling threshold (such as ≤0.3, i.e. "low level"), the metabolic abnormalities are judged to be independent of the morphological abnormalities, and the "isolated metabolic abnormality flag" is activated.

[0076] In response to the "strong metabolic-morphological correlation flag": trigger the first-level control instruction generation pathway to generate comprehensive regulatory instructions for metabolic and morphological coordinated abnormalities (such as adjusting nutrient supply and culture environment parameters at the same time to alleviate the linkage abnormalities between the two); in response to the "isolated metabolic abnormality flag": trigger the second-level control instruction generation pathway to generate precise regulatory instructions only for metabolic abnormalities (such as supplementing metabolic substrates separately to avoid unnecessary interference with cells with normal morphology).

[0077] By distinguishing between strongly correlated and isolated anomalies, ineffective interventions can be reduced, minimizing disturbances in the culture system. Rapid generation of primary instructions for strongly correlated metabolic-morphological anomalies prevents the spread of related anomalies (e.g., metabolic disturbances leading to morphological deterioration, which further exacerbates metabolic anomalies) and enables early containment. Secondary instructions target isolated metabolic anomalies, regulating only the necessary steps, conserving regulatory resources (e.g., eliminating the need to adjust morphologically related equipment parameters), and improving the efficiency and stability of the cell culture process.

[0078] In a preferred embodiment of the present invention, in response to the primary control instruction, the buffer injection unit and the nutrient supplement pump are cooperatively driven; in response to the secondary control instruction, the stirring rate controller, the temperature feedback loop and the gas proportional valve opening of the bioreactor are dynamically adjusted, including: In response to the first-level control instruction, the comprehensive intensity of metabolic abnormalities and the comprehensive intensity of morphological abnormalities in the joint fuzzy state matrix are extracted; The buffer control model processes the comprehensive intensity of metabolic abnormalities to generate a buffer pulse injection parameter set; it integrates the metabolic and morphological abnormality intensities to generate a nutrient supplement flow control curve; and it collaboratively drives the buffer injection unit and the nutrient supplement pump to perform linked operations. In response to the secondary control instruction, the parameter change rate sequence and metabolic abnormality membership in the real-time metabolic dynamic parameter set are called; Perform multi-parameter coupling analysis through a dynamic control strategy engine: Generate an adaptive correction value for the stirring rate based on the parameter change rate sequence; generate a temperature feedback compensation signal based on the metabolic abnormality membership; and generate a proportional valve opening increment sequence through the gas valve control model; Synchronously regulate the stirring rate controller, temperature feedback loop and gas proportioning valve.

[0079] In an embodiment of the present invention, in response to the first-level control instruction, the comprehensive intensity of metabolic abnormalities (such as 0.8) and the comprehensive intensity of morphological abnormalities (such as 0.7) are extracted from the joint fuzzy state matrix, both of which reflect the severity of the abnormalities associated with metabolism and morphology.

[0080] The comprehensive intensity of metabolic abnormalities is input into the buffer control model (which is constructed based on the correlation between metabolic abnormalities and buffer requirements in historical data) to generate a set of buffer pulse injection parameters - including single injection volume (such as an abnormality intensity of 0.8 corresponds to an injection of 50 mL), injection interval (such as once every 10 minutes), and duration (such as 30 minutes) to quickly alleviate metabolic environment imbalances (such as pH abnormalities).

[0081] The comprehensive intensity of metabolic abnormalities and the comprehensive intensity of morphological abnormalities are integrated (e.g., weighted at 7:3), input into the nutrient supplementation model, and a nutrient supplementation flow control curve is generated. The curve dynamically adjusts the flow over time (e.g., the initial flow is 2 mL / min, which gradually decreases to 0.5 mL / min as the abnormality intensity decreases), while taking into account the metabolic substrate demand and the cell morphology's ability to absorb nutrients (e.g., avoiding toxic accumulation caused by excessive nutrition when the morphology is abnormal).

[0082] The buffer pulse injection parameter set is sent to the buffer injection unit, and the nutrient supplement flow control curve is sent to the nutrient supplement pump to control the two units to perform operations synchronously according to time (for example, the nutrient flow is adjusted synchronously during buffer injection to avoid the superimposed disturbance of the two on the culture medium environment).

[0083] In response to the secondary control instruction, the parameter change rate sequence (such as glucose concentration decreasing by 0.6 mM per hour, dissolved oxygen content decreasing by 5% per hour) and metabolic abnormality membership (such as 0.6) in the real-time metabolic dynamic parameter set are called.

[0084] Adaptive correction of stirring rate: Based on the parameter change rate sequence, the uniformity of the culture medium is judged (for example, large fluctuations in the change rate of glucose concentration indicate local uneven concentration) and the stirring rate correction is generated (for example, the current rate is 200 rpm, which is corrected to 220 rpm to enhance mixing); Temperature feedback compensation signal: Based on the membership of metabolic abnormality, the temperature adjustment amplitude is determined (for example, a membership of 0.6 corresponds to a compensation of +0.5°C to alleviate abnormal metabolic enzyme activity) and a compensation signal for the temperature feedback loop is generated; Gas proportional valve opening increment sequence: Through the gas valve control model (which links the dissolved oxygen content change rate with the gas ratio requirement), the opening increment of the oxygen / carbon dioxide proportional valve is generated (for example, if the dissolved oxygen content decreases rapidly, the oxygen valve opening is increased by 5%, while the carbon dioxide valve opening remains unchanged).

[0085] The stirring rate correction is sent to the stirring rate controller, the temperature compensation signal is input into the temperature feedback loop, and the opening increment sequence is sent to the gas proportional valve. The operating parameters of the three devices are adjusted synchronously to improve metabolic abnormalities in a targeted manner (avoiding interference with normal cell morphology due to adjustment of morphological related parameters).

[0086] Primary control instructions, through the coordinated regulation of buffers and nutrients, simultaneously improve anomalies associated with the metabolic environment and cell morphology (e.g., cell shrinkage due to insufficient metabolic substrates), avoiding secondary imbalances caused by single-factor regulation. Secondary control instructions adjust only reactor parameters directly related to metabolism (agitation, temperature, and gases), without intervening in morphological factors. This minimizes disturbances to cells with normal morphology and improves control efficiency. Control values are dynamically generated based on the real-time rate of change of metabolic parameters and the intensity of anomalies, rather than fixed values. This allows adjustments to be more tailored to the current culture state (e.g., automatically increasing the control range when metabolic anomalies intensify), improving the stability and controllability of the cell culture process.

[0087] An embodiment of the present invention further provides a computing device comprising: a processor and a memory storing a computer program, wherein when the computer program is executed by the processor, the system described above is executed. All implementations in the above system embodiments are applicable to this embodiment and can achieve the same technical effects.

[0088] The embodiment of the present invention further provides a computer-readable storage medium storing instructions, which, when executed on a computer, cause the computer to execute the system described above. All implementations in the above system embodiments are applicable to this embodiment and can achieve the same technical effects.

[0089] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A multi-parameter mesenchymal stem cell culture monitoring system, characterized in that: include: An acquisition module is used to obtain the real-time metabolic dynamic parameter set of the culture medium and the periodic morphological parameter set of the cell population; A generation module is used to train and generate a dynamic metabolic threshold library based on the historical data of the metabolic dynamic parameter set, and to construct a fuzzy morphological benchmark library based on the historical data of the morphological parameter set; An analysis module is used to compare the deviation of the real-time metabolic dynamic parameter set with the dynamic metabolic threshold library to generate metabolic abnormality membership; the real-time morphological parameter set is mapped to a preset feature space through a mirror projection algorithm, and the morphological abnormality confidence is calculated in this space based on the membership function of the fuzzy morphological reference library; The membership of metabolic abnormality and the confidence of morphological abnormality are integrated to generate a joint fuzzy state matrix; A decision module is configured to generate a primary control instruction when the correlation strength between metabolic and morphological abnormalities in the joint fuzzy state matrix reaches a preset correlation threshold as determined by the fuzzy rule base; and to generate a secondary control instruction when the abnormal state flag indicates an isolated metabolic abnormality; The output module is used to respond to the first-level control instructions and coordinately drive the buffer injection unit and nutrient supplement pump; respond to the second-level control instructions and dynamically adjust the stirring rate controller, temperature feedback loop and gas proportional valve opening of the bioreactor.

2. The multi-parameter mesenchymal stem cell culture monitoring system according to claim 1, characterized in that: Obtain real-time metabolic dynamic parameter sets of the culture medium and periodic morphological parameter sets of the cell population, including: The raw time-series data stream of the culture medium is collected through a biosensor array. A sliding window processor is used to perform segmented feature extraction on the raw time-series data stream to generate a primary metabolic feature set containing a mean vector, a standard deviation vector, and a parameter change rate sequence. Time dimension alignment and feature splicing are performed on the primary metabolic feature set to generate a structured metabolic dynamic parameter set. The microscopic imaging unit periodically captures a high-resolution image set of the cell population; and performs hierarchical processing on the high-resolution image set: Perform edge segmentation at the pixel level to generate a cell outline dataset. Extract geometric topological features at the single cell level based on the cell outline dataset to generate a single cell feature set. Generate population morphology indicators at the population level based on the spatial distribution relationship of the single cell feature set. The single-cell feature set and group morphology indicators are input into the multi-scale feature fusion, and cross-scale correlation features are generated through spatial distribution topology modeling; A structured morphological parameter set is constructed based on cross-scale correlation features and group morphological indicators.

3. The multi-parameter mesenchymal stem cell culture monitoring system according to claim 2, characterized in that: A dynamic metabolic threshold library is generated based on the historical data of the metabolic dynamic parameter set. At the same time, a fuzzy morphological benchmark library is constructed based on the historical data of the morphological parameter set, including: The metabolic dynamic parameter sets of each culture cycle in the historical metabolic database are called; timestamp alignment is performed on the parameter sets to generate a time-series-associated metabolic feature tensor; a dynamic probability distribution model is constructed based on the metabolic feature tensor, and a dynamic confidence interval sequence is generated by extracting boundary values under a preset confidence level; The dynamic confidence interval series were integrated into a structured dynamic metabolic threshold library according to the culture stage; Extract morphological parameter sets across batches from a historical morphology database; perform time-series-aware incremental clustering on single-cell feature sets: Divide feature subsets according to the culture time window, perform density peak clustering on each subset to generate local morphological benchmarks, and fuse local benchmarks through time-attenuated weights to generate a global typical cell morphological benchmark vector. Modeling the dynamic evolution of group morphological indicators: The evolution trajectory of group indicators is constructed based on the timestamp index, and the Gaussian mixture model is fitted segment by segment according to the training stage to generate a dynamic benchmark interval set; The typical cell morphology benchmark vector and the dynamic benchmark interval set are input into the rule generator to define the adaptive membership function set for judging morphological parameter anomalies. The membership function set and the corresponding benchmark data are associated and stored to generate a structured fuzzy morphology benchmark library.

4. The multi-parameter mesenchymal stem cell culture monitoring system according to claim 3, characterized in that: The joint fuzzy state matrix includes: Obtain the real-time metabolic dynamic parameter set and call the threshold vector corresponding to the cultivation stage in the dynamic metabolic threshold library; Perform item-wise confidence interval deviation calculation on the mean vector, standard deviation vector and parameter change rate sequence of the parameter set to generate the original deviation set; Based on the original deviation set, a set of abnormal weight coefficients is generated based on the sensitivity of the parameters in the training stage; The original deviation set and the abnormal weight coefficient set are integrated to output the scalar metabolic abnormality membership; Obtain single cell feature sets and group morphology indicators in real-time morphological parameter sets; Call the typical cell morphology benchmark vector and dynamic benchmark interval set in the fuzzy morphology benchmark library; Project the single-cell feature set into the feature space spanned by the reference vector; calculate the Mahalanobis distance between the cell population distribution and the reference vector in the projected space to generate a single-cell deviation vector; perform dynamic interval out-of-bounds detection on the population morphology index to generate a population out-of-bounds flag set; According to the single cell deviation vector and the group out-of-bounds flag set, a scalar morphological abnormality confidence is obtained; The metabolic abnormality membership and morphological abnormality confidence are combined into a two-dimensional abnormal state vector. Based on the historical abnormal pattern library, an inter-dimensional correlation weight matrix is generated. Tensor multiplication operation is performed on the abnormal state vector and the correlation weight matrix, and the operation result is output as a joint fuzzy state matrix.

5. The multi-parameter mesenchymal stem cell culture monitoring system according to claim 4, characterized in that: The metabolic abnormality membership and morphological abnormality confidence are combined into a two-dimensional abnormal state vector; based on the historical abnormal pattern library, an inter-dimensional correlation weight matrix is generated, and a tensor multiplication operation is performed on the abnormal state vector and the correlation weight matrix to output the operation result as a joint fuzzy state matrix, including: The metabolic abnormality membership is used as the first dimension value, and the morphological abnormality confidence is used as the second dimension value; Generate a two-dimensional abnormal state vector by combining the dimensions in order; When the value of the first dimension exceeds the metabolic dominant threshold, the metabolic abnormality dominant factor is activated; when the value of the second dimension exceeds the morphological dominant threshold, the morphological abnormality dominant factor is activated; and a weight allocation instruction is generated according to the type of the activated dominant factor; In response to the weight allocation instruction, the historical abnormal pattern library is called; based on the correlation characteristics of similar dominant factors in the historical abnormal pattern library: According to metabolic abnormalities and morphological abnormalities, an adaptive weight matrix including self-action weights and cross-dimensional correlation weights is generated; A tensor convolution operation is performed on the two-dimensional abnormal state vector and the adaptive weight matrix, and the output result is used as the joint fuzzy state matrix.

6. The multi-parameter mesenchymal stem cell culture monitoring system according to claim 5, characterized in that: When the correlation strength between metabolic and morphological abnormalities in the joint fuzzy state matrix reaches a preset correlation threshold as determined by the fuzzy rule base, a first-level control instruction is generated; When the abnormal status flag indicates an isolated metabolic abnormality, a secondary control instruction is generated, including: Extract the comprehensive intensity of metabolic abnormality and the comprehensive intensity of morphological abnormality from the joint fuzzy state matrix; Generate an abnormal intensity level identification set based on preset intensity grading rules; The abnormality intensity level identification set is input into the fuzzy rule base. When the metabolic and morphological abnormality intensities both reach the high-level threshold, the metabolic-morphological strong association flag is activated. When the metabolic abnormality intensity reaches the high level and the morphological abnormality intensity is lower than the decoupling threshold, the isolated metabolic abnormality flag is activated. In response to the strong metabolism-morphology correlation flag, a first-level control instruction generation pathway is triggered; In response to the isolated metabolic abnormality flag, a secondary control instruction generation pathway is triggered.

7. The multi-parameter mesenchymal stem cell culture monitoring system according to claim 6, characterized in that: Responding to the primary control instructions, it collaboratively drives the buffer injection unit and the nutrient supplement pump; responding to the secondary control instructions, it dynamically adjusts the bioreactor's stirring rate controller, temperature feedback loop, and gas proportional valve opening, including: In response to the first-level control instruction, the comprehensive intensity of metabolic abnormalities and the comprehensive intensity of morphological abnormalities in the joint fuzzy state matrix are extracted; The buffer control model processes the comprehensive intensity of metabolic abnormalities to generate a buffer pulse injection parameter set; it integrates the metabolic and morphological abnormality intensities to generate a nutrient supplement flow control curve; and it collaboratively drives the buffer injection unit and the nutrient supplement pump to perform linked operations. In response to the secondary control instruction, the parameter change rate sequence and metabolic abnormality membership in the real-time metabolic dynamic parameter set are called; Perform multi-parameter coupling analysis through a dynamic control strategy engine: Generate an adaptive correction value for the stirring rate based on the parameter change rate sequence; generate a temperature feedback compensation signal based on the metabolic abnormality membership; and generate a proportional valve opening increment sequence through the gas valve control model; Synchronously regulate the stirring rate controller, temperature feedback loop and gas proportioning valve.

8. A computing device, characterized in that include: one or more processors; A storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the system according to any one of claims 1 to 7.

Images