Stem cell preparation tracing management system

Abstract

The invention discloses a stem cell preparation traceability management system, and relates to the technical field of stem cell preparation. Comprises: a data acquisition module for acquiring environment, operation time sequence, cell growth and quality inspection result data for a stem cell preparation whole process, and establishing a unique association between the data and a preparation batch; the AI quality prediction and intelligent decision-making module is used for constructing a cell quality prediction model based on the full-chain data of historical successful / failed samples, and drawing a two-dimensional dual-mode cell growth curve through a multi-mode algorithm so as to realize abnormal growth risk judgment, passage optimal time window calculation and pre-release probability evaluation; the block chain data storage module is used for performing dual hash operation on the cell key auditing elements to generate a unique digital fingerprint; the query module is used for providing authorization query ports for related parties such as a preparation mechanism and a supervision department so as to display a whole-process tracing link; the system improves the stem cell preparation quality stability and supervision transparency.

Description

Technical Field

[0001] This invention belongs to the field of stem cell preparation technology, specifically relating to a stem cell preparation traceability management system. Background Technology

[0002] Clinical-grade stem cell preparation places extremely high demands on quality stability and regulatory traceability; however, the existing technological system has significant shortcomings, specifically: 1. At the data management level, the various types of data generated throughout the entire preparation process, such as environment, operation, cell growth, and quality inspection, form information silos and do not establish a unique and mandatory correlation with the stem cell preparation batch. When quality deviations occur, it is difficult to quickly trace the root cause of the problem, resulting in low localization efficiency. 2. At the quality assessment level, existing methods are mostly passive post-event quality inspections, lacking proactive prediction and monitoring of the cell growth process. Even if some solutions adopt artificial intelligence (AI) technology, their prediction models often rely only on the single dimension of cell number. The constructed growth curves cannot reflect the cell activity status and do not consider the dynamic impact of environmental fluctuations, resulting in blind spots in the early warning of latent growth abnormalities. 3. At the process decision-making and release judgment level, key operations such as the selection of cell passage time are mainly based on the personal experience of operators, lacking scientific calculations based on quantitative data such as cell interaction density, real-time nutrient consumption rate and stem expression uniformity. At the same time, the final release judgment standard is singular and fails to integrate multi-dimensional information such as environmental compliance, operational standardization, and growth curve conformity for comprehensive risk assessment, which may lead to the release of batches that are "inspection qualified" but "have hidden dangers".

[0003] In summary, existing technologies are insufficient to meet the demands of industrial-scale stem cell production for quality control and transparent oversight. Developing an integrated, intelligent, and reliable traceability management system has become an urgent need for the industry's development. Summary of the Invention

[0004] This invention aims to solve at least one of the technical problems existing in the prior art; to this end, this invention proposes a stem cell preparation traceability management system to solve the following technical problem: In current stem cell preparation processes, various data, including environmental data, are scattered across different devices or systems, lacking a mandatory link to each batch. This makes it difficult to quickly trace back to the specific stage when quality issues arise, leading to difficulties in localization. Furthermore, current technologies rely heavily on post-processing quality inspections for stem cell quality assessment, lacking process prediction capabilities. Even when AI prediction is introduced in some solutions, it only constructs cell growth curves based on single cell count data, failing to combine cell activity characteristics with dynamic environmental changes for comprehensive judgment, making it difficult to detect growth abnormalities in a timely manner. The determination of stem cell passage time relies heavily on operator experience, without quantitative calculations combining intercellular interaction density, nutrient consumption rate, and stemness maintenance status. In addition, pre-release judgment is based on a single quality inspection indicator, making it difficult to integrate multi-dimensional data, resulting in potential quality problems in qualified batches.

[0005] To address the above problems, this invention provides a stem cell preparation traceability management system, comprising the following modules: Data acquisition module: Collects environmental data, operation sequence data, cell growth data, and quality inspection results data for the entire process of stem cell preparation batches, and establishes a unique identifier for each batch; AI Quality Prediction and Intelligent Decision-Making Module: Collects full-chain data of successful and failed samples from historical stem cell preparation batches, extracts features to construct a cell quality prediction model, plots dual-mode cell growth curves, inputs the dual-mode cell growth curves into the cell quality prediction model, determines whether there is a risk of abnormal growth, and predicts the cell growth trend based on the cell quality prediction model, combined with historical stem cell growth data and various data collected in real time, calculates the optimal time window for passage, and simultaneously collects traceability data of the entire process of this batch of stem cell preparation and inputs it into the model, and calculates its pre-release probability based on preset evaluation indicators; Blockchain data storage module: Predefined key audit elements in the stem cell preparation process. The on-chain storage unit automatically extracts the corresponding key audit elements, performs hash operations on them to generate unique hash values, and synchronously uploads them to the blockchain network for consensus verification. Query module: Creates authorized query ports for various stem cell preparation stakeholders. Each authorized stakeholder is assigned a unique identity and access permissions. Query results display the entire process traceability chain.

[0006] Preferably, the data acquisition module includes: After cleaning and format conversion of the collected data, a unique association is established between the data and the stem cell preparation batch, and the data is simultaneously pushed to the AI ​​quality prediction and intelligent decision-making module and the blockchain data storage module.

[0007] Preferably, in the I-quality prediction and intelligent decision-making module, the cell quality prediction model includes: Collect the entire chain of data from successful and failed stem cell preparation samples throughout history. The data set comes from the data acquisition module, and each type of data is bound to a unique traceability identifier of the sample. The entire data chain is processed by feature engineering to extract key features. A machine learning algorithm is then used to construct a cell quality prediction model to output a normal or abnormal judgment. The key features include: spatiotemporal heterogeneous coupling characteristics of the microenvironment, operation-microenvironment response lag characteristics, cell self-organization pattern evolution characteristics, and dynamic adaptation characteristics of nutrient supply-growth demand. The model is trained using historical data, and then deployed to the system.

[0008] Preferably, the key features include: The spatiotemporal heterogeneity coupling characteristics of the microenvironment are based on multi-point time-series data from a sensor array within the culture dish. The product coefficient of the spatial gradient variance and the temporal fluctuation variance of key environmental parameters is calculated, specifically as follows: ,in, The product coefficient, Let Variance be the spatial gradient. The variance is the time fluctuation. The operation-microenvironment response lag feature is based on a dynamic time warping algorithm to align operation events with cell growth rate time series, extracting the lag time of the growth rate response peak relative to the operation event, as well as the correlation coefficient between operation parameter deviation and the amplitude of environmental fluctuation parameters. ,in, The correlation coefficient is... This is due to deviation in operating parameters. For the amplitude of environmental fluctuation parameters; The cell self-organization pattern evolution features are based on gray-level co-occurrence matrix and morphological skeleton analysis of cell microscope images, extracting the image texture entropy value and the proportion of ring-shaped closed structures in all skeleton structures, specifically: ,in, The proportion of ring-shaped closed structures, This represents the total number of pixels in the ring structure. The total number of pixels in the overall skeleton structure; wherein, the cell microscope image is a cell population image taken by a phase contrast microscope, and the ring-shaped closed structure corresponds to the boundary of the cell island structure formed when the cell fusion degree is ≥70%; The dynamic adaptation feature of nutrient supply and growth demand is based on time-series data of culture medium component concentration and cell number, calculating the deviation of the ratio of nutrient consumption rate to cell population doubling rate, specifically: ,in, To address the dynamic adaptation deviation of nutrition, This represents the total number of time points. For time point indexing, For at a certain point in time The rate of nutrient consumption, For at a certain point in time The estimated cell population doubling rate, This represents the optimal nutrient-to-growth rate ratio.

[0009] Preferably, in the AI ​​quality prediction and intelligent decision-making module, plotting the dual-modal cell growth curve includes: Real-time reception of cell growth data, extraction of cell features and construction of dual-modal cell growth curves using multimodal image analysis and dynamic correlation algorithms, specifically including: Cell behavior feature extraction unit: It is used to receive continuous cell microscope image sequences and cell counting data, extract the cell population division synchronicity index and migration aggregation degree through optical flow field tracking algorithm, and extract the cell pseudopodia extension length and nucleocytoplasmic ratio through deep learning semantic segmentation model, which together constitute a dynamic cell activity state feature vector; Dual-dimensional growth curve construction unit: used to fuse the cell count data with the cell activity state feature vector to construct a dual-dimensional growth curve whose vertical axis simultaneously includes the cell number and the mean value of the activity feature vector; Environmental correction unit: used to access real-time environmental sensor data, establish a correlation model between environmental parameters and cell dynamic characteristics through an attention mechanism algorithm, and correct the prediction baseline of the two-dimensional growth curve in real time; Multi-mode deviation intelligent judgment and early warning module: It is used to compare the corrected real-time two-dimensional growth curve with the pre-stored successful model standard interval and execute the corresponding deviation mode judgment logic.

[0010] Preferably, the execution of the corresponding deviation mode determination logic includes: If only the cell number dimension deviates from the standard range, while the activity feature vector dimension is normal, it is judged as recoverable growth retardation. If the dimension of the activity feature vector deviates before the dimension of cell number, it is judged as a latent abnormality risk, triggering an orange alert, and the image backtracking algorithm is activated to generate an abnormality source heat map that identifies the source of the abnormality. If both dimensions deviate from the preset threshold range, it is determined to be an irreversible quality risk, triggering a red alert and automatically locking subsequent operation permissions for that batch of cells.

[0011] Preferably, the calculation of the optimal time window for generation includes: Based on real-time microscope image sequences, a graph neural network is used to construct an interaction map between cells, from which the number of effective proliferation contact pairs within the division cycle is extracted and compared with a passage critical interaction density threshold trained on historical data. The passage critical interaction density threshold is associated with the interaction density in historical data and the cell adhesion rate after passage. Based on the concentration changes of key nutrients in the culture medium and cell count data, the nutrient consumption rate is calculated. Using a long short-term memory network model, with the nutrient consumption rate and the current cell growth stage as input, the nutrient reserves are predicted to support the cells to complete the next full division cycle within a preset time period. Based on the prediction results, the adjustment instructions for the passage time window are output. Cell fluorescence images are processed, and the distribution variance of fluorescence signals in the cell population is calculated using image analysis algorithms as a coefficient of stemness expression evenness. A mapping table is established. When the mapping table indicates that the predicted stemness maintenance of the current index decreases by more than a preset stemness threshold, the current time window is excluded. At the same time, if the maximum predicted fluctuation range of any environmental parameter exceeds its corresponding static stress tolerance threshold, the current time window is excluded. The system receives the evaluation results and the feasible time window set, and takes the intersection of the feasible time window set to generate a candidate time window set. Based on the results of effective proliferation contact log comparison, nutritional support capacity prediction, stem expression evenness coefficient assessment, and environmental disturbance matching, the final optimal passage time window was determined by screening from candidate time windows.

[0012] Preferably, the calculation of the pre-release probability based on preset evaluation indicators includes: Collect data from the entire process of preparing this batch of stem cells and input it into the cell quality prediction model. The data from the entire process includes real-time dynamic data and statically recorded data. The cell quality prediction model calculates the pre-release probability of this batch of stem cells based on preset evaluation indicators, including environmental compliance rate, operational compliance rate, growth curve compliance rate, and stage quality inspection pass rate. The calculation of the pre-release probability is as follows:

[0013] in, This is the pre-release probability. For environmental compliance rate, To improve operational compliance rate, For the growth curve achievement rate, This refers to the pass rate of the phased quality inspection.

[0014] Preferably, the blockchain data storage module includes: The key audit elements include cell donor identification, key reagent batch number, cell operation timestamp, operator biovalidation data, and continuous readings of environmental sensors during key operation phases. The set of key audit elements is standardized and serialized to generate its corresponding unique digital fingerprint; The unique digital fingerprint is generated by adding the serialized key audit element data to an encrypted hash function to generate a first-level hash value, concatenating the first-level hash value with a random number generated by the hash of the preceding block header, and performing a second hash operation on the concatenated data to generate the final unique digital fingerprint.

[0015] Preferably, the query module includes: When a specific batch of stem cell traceability request is initiated, a query instruction is received through the authorized transparent query module, and the query instruction carries the unique identifier of the batch of stem cells. Based on the unique identifier, retrieve the full process data corresponding to this batch; Simultaneously, through the blockchain data storage module, the key audit elements corresponding to the batch and the unique digital fingerprint stored on the chain are matched according to the unique identifier to verify the consistency between the retrieved full-process data and the unique digital fingerprint, and to confirm that the data has not been tampered with. The retrieved data from the entire process are linked together according to the stem cell preparation timeline to form a complete traceability view.

[0016] The beneficial effects of this invention are: This invention collects basic data during the stem cell preparation process and extracts features such as spatiotemporal heterogeneity coupling of the microenvironment, operation-microenvironment response lag, cell self-organization pattern evolution, and dynamic adaptation of nutrient supply and growth requirements through feature engineering. This allows the AI ​​model to capture early and subtle signals of cell state decay, enabling it to identify abnormalities by observing deviations in the activity feature vector before significant changes in cell number occur. This represents a shift from post-event inspection to in-event prediction and early warning. This invention constructs a two-dimensional growth curve with the vertical axis containing both cell number and cell activity state feature vectors, and defines judgment logic and response strategies for different deviation modes (recoverable growth retardation, latent abnormality risk, and irreversible quality risk). It effectively avoids misjudgment caused by judging solely by cell number, can distinguish different levels of risk, and trigger different levels of response, thus achieving more refined and automated management decisions. This invention analyzes the results of four dimensions: cell interaction map, nutritional support capacity prediction, stemness expression uniformity and environmental disturbance matching. By taking the intersection and intelligent screening, the optimal time window is determined. The vague experience is transformed into a scientific decision-making process that optimizes multiple objectives. This greatly improves the cell adhesion rate, activity and stemness maintenance capacity after passage, thereby improving the consistency of final product quality and the success rate of preparation. This invention predefines key audit elements and performs two-level hash operations by concatenating their serialized data with random numbers generated from the hash of the preceding block header to generate a strongly unique digital fingerprint, which is then uploaded to the blockchain. This ensures the authenticity and integrity of all key operational, personnel, and material data. Any tampering with the original data will result in a mismatch between its hash value and the on-chain evidence, making the final generated full-chain traceability view meet the highest level of auditing requirements in the industry. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the module flow of the present invention; Figure 2 This is a schematic diagram of the process for constructing a dual-mode cell growth curve according to the present invention. Detailed Implementation

[0018] 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.

[0019] Please see Figure 1 As shown, this invention is a stem cell preparation traceability management system, comprising the following modules: Data acquisition module: Collects environmental data, operation sequence data, cell growth data, and quality inspection results data for the entire process of stem cell preparation batches, and establishes a unique identifier for each batch; AI Quality Prediction and Intelligent Decision-Making Module: Collects full-chain data of successful and failed samples from historical stem cell preparation batches, extracts features to construct a cell quality prediction model, plots dual-mode cell growth curves, inputs the dual-mode cell growth curves into the cell quality prediction model, determines whether there is a risk of abnormal growth, and predicts the cell growth trend based on the cell quality prediction model, combined with historical stem cell growth data and various data collected in real time, calculates the optimal time window for passage, and simultaneously collects traceability data of the entire process of this batch of stem cell preparation and inputs it into the model, and calculates its pre-release probability based on preset evaluation indicators; Blockchain data storage module: Predefined key audit elements in the stem cell preparation process. The on-chain storage unit automatically extracts the corresponding key audit elements, performs hash operations on them to generate unique hash values, and synchronously uploads them to the blockchain network for consensus verification. Query module: Creates authorized query ports for various stem cell preparation stakeholders. Each authorized stakeholder is assigned a unique identity and access permissions. Query results display the entire process traceability chain.

[0020] Specifically, throughout the entire stem cell preparation process, four types of data are collected in real time using sensors, equipment logs, microscopes, and quality control equipment. The collected raw data is cleaned, formatted, and bound to a unique identifier for the current stem cell preparation batch, then synchronously pushed to the subsequent AI decision-making and blockchain evidence storage modules. The system collects full-chain data from historical successful and failed batches, extracts key features such as spatiotemporal heterogeneity of the microenvironment and operation-microenvironment response lag through feature engineering, and trains a cell quality prediction model using machine learning algorithms. The system receives real-time cell growth data, and through image analysis and behavioral analysis, plots a dual-modal growth curve that simultaneously reflects cell number and cell viability. This curve is compared with a standard success model to determine the growth status in real time and trigger corresponding warnings. By comprehensively predicting cell interactions, nutrient consumption, stemness maintenance, and environmental disturbances, the optimal passage operation time window is calculated. At key preparation nodes or at the end of the process, the system... The system automatically calculates the pre-release probability of a batch of cells based on indicators such as environmental compliance rate, operational compliance rate, and growth curve compliance rate. It predefines key audit elements in stem cell preparation; when key operations are completed, the on-chain evidence storage unit is automatically triggered to extract these elements, standardize and serialize them, and then generate a unique digital fingerprint for the batch data using a cryptographic hash function. This fingerprint is then simultaneously uploaded to the blockchain network for consensus verification and permanent evidence storage, ensuring the data is tamper-proof. Different authorized query ports are created for different stakeholders such as regulatory agencies and hospitals. When a user initiates a query by entering the unique identifier of a batch of stem cells, the system first retrieves complete end-to-end data from the off-chain database and simultaneously obtains the corresponding unique digital fingerprint from the blockchain for consistency verification. After successful verification, the system connects all data related to environment, operation, growth, and quality inspection in a timeline to generate a complete and reliable end-to-end traceability view for the querying party.

[0021] In one embodiment of the present invention, the data acquisition module includes: After cleaning and format conversion of the collected data, a unique association is established between the data and the stem cell preparation batch, and the data is simultaneously pushed to the AI ​​quality prediction and intelligent decision-making module and the blockchain data storage module.

[0022] Specifically, environmental data includes data on environmental parameters such as culture temperature, humidity, carbon dioxide concentration, and cleanliness; operational sequence data includes time, parameters, and operator information for each step; cell growth data includes periodically captured cell images and data on cell number and survival rate to form raw cell growth data; and quality control result data includes data from the stem cell preparation stage and the final quality control.

[0023] In one embodiment of the present invention, the cell quality prediction model in the AI ​​quality prediction and intelligent decision-making module includes: The entire data chain is processed by feature engineering to extract key features. A machine learning algorithm is then used to construct a cell quality prediction model to output a normal or abnormal judgment. The key features include: spatiotemporal heterogeneity coupling characteristics of the microenvironment, operation-microenvironment response lag characteristics, cell self-organization pattern evolution characteristics, and dynamic adaptation characteristics of nutrient supply-growth demand. The model is trained using historical data, and the trained model is then deployed to the system.

[0024] Specifically, the full-chain data consists of historical data received from the data acquisition module, each uniquely identified and bound to a specific preparation batch. This includes environmental data, operational timeline data, cell growth data, and quality inspection result data. Based on the preprocessed data, four key feature vectors are extracted using different algorithms. All the features extracted from each historical batch are concatenated into a joint feature vector. This joint feature vector is then associated with the corresponding batch's quality label (success / failure). A gradient boosting decision tree algorithm is used, with the joint feature vector as input and the quality label as the prediction target. The model is trained on 70% of the historical data and validated on the remaining 30% of the test data, requiring a recall rate of at least 85% for failed batches. The trained model is then encapsulated as a RESTful API microservice and deployed on the system's server. When a new batch enters real-time monitoring, the system extracts its features in real time according to the above process and sends the resulting joint feature vector to the API.

[0025] In one embodiment of the present invention, the key features include: The spatiotemporal heterogeneity coupling characteristics of the microenvironment are based on multi-point time-series data from a sensor array within the culture dish. The product coefficient of the spatial gradient variance and the temporal fluctuation variance of key environmental parameters is calculated, specifically as follows: ,in, The product coefficient, Let Variance be the spatial gradient. The time-varying variance; The operation-microenvironment response lag feature is based on a dynamic time warping algorithm to align operation events with cell growth rate time series, extracting the lag time of the growth rate response peak relative to the operation event, as well as the correlation coefficient between operation parameter deviation and the amplitude of environmental fluctuation parameters. ,in, The correlation coefficient, This is due to deviation in operating parameters. For the amplitude of environmental fluctuation parameters; The cell self-organization pattern evolution features are based on gray-level co-occurrence matrix and morphological skeleton analysis of cell microscope images, extracting the image texture entropy value and the proportion of ring-shaped closed structures in all skeleton structures, specifically: ,in, The proportion of ring-shaped closed structures, This represents the total number of pixels in the ring structure. The total number of pixels in the overall skeleton structure; wherein, the cell microscope image is a cell population image taken by a phase contrast microscope, and the ring-shaped closed structure corresponds to the boundary of the cell island structure formed when the cell fusion degree is ≥70%; The dynamic adaptation feature of nutrient supply to growth demand is based on time-series data of culture medium component concentration and cell number, calculating the deviation of the ratio of nutrient consumption rate to cell population doubling rate, specifically: ,in, To address the dynamic adaptation deviation of nutrition, This represents the total number of time points. For time point indexing, For at a certain point in time The rate of nutrient consumption, For at a certain point in time The estimated cell population doubling rate, This represents the optimal nutrient-to-growth rate ratio.

[0026] Specifically, the spatiotemporal heterogeneity coupling characteristics of the microenvironment are as follows: by acquiring time-series data of key environmental parameters recorded by a 3*3 grid of micro-sensor arrays in the culture dish, for each sampling time point, the variance of all sensor readings at that time is calculated as the spatial gradient variance. The spatial gradient variances of all time points throughout the entire culture cycle are combined into a new time series, and the variance of this series is calculated as the temporal fluctuation variance. The product of the above two variances is calculated, and the coefficient is used as a scalar feature. The larger the value of the coefficient, the higher the synergistic risk to cell growth, indicating that the microenvironment is both spatially heterogeneous and temporally unstable. The operation-microenvironment response lag characteristic is based on a specific operation event (e.g., medium change). Using a dynamic time warping algorithm, the time point of the operation event is non-linearly aligned with the subsequent cell population doubling rate time series. In the aligned sequence, the first response peak of cell growth rate after the operation is located, and the difference between the peak time point and the operation time point is calculated as the operation-growth lag time. The absolute deviation of the actual operation parameters (e.g., digestion time) and standard deviation is calculated. At the same time, the standard deviation of the fluctuation of key environmental parameters (e.g., carbon dioxide concentration, etc.) within a specific window (e.g., 12 hours) after the operation is calculated as its fluctuation amplitude. The Pearson correlation coefficient between the operation parameter deviation and the environmental fluctuation amplitude is calculated. The lag time and the perturbation Pearson correlation coefficient are combined into a two-dimensional feature vector to jointly quantify the quality of the operation and its impact on the culture system. Cellular self-organization pattern evolution characteristics: When the cell confluence reaches ≥70%, a phase contrast microscope is used to capture images of the cell population. At this time, the cells begin to form continuous cell island structures. Gray-level co-occurrence matrix analysis is performed on the images to calculate their texture entropy. At the same time, the images are binarized and morphological skeletonized to obtain the cellular skeleton structure map. All ring-shaped closed structures in the skeleton are identified. The ring-shaped closed structures correspond to the boundaries of cell islands and are markers of healthy intercellular connections and functionalization. The ratio of the total number of pixels of the ring-shaped closed structures to the total number of pixels of the entire skeleton structure is calculated to obtain the ring structure proportion. The texture entropy and the ring structure proportion are combined into a two-dimensional feature vector. A healthy culture sample usually exhibits low texture entropy and high ring structure proportion. Dynamic adaptation characteristics of nutrient supply to growth requirements: Based on time-series data of glucose concentration and cell number in the culture medium, the glucose consumption rate and cell population doubling rate between adjacent time points are calculated. From the data of historical successful batches, an optimal nutrient-to-growth rate ratio is determined through regression analysis. For the current batch, the deviation between the actual rate ratio (the ratio of glucose consumption rate to cell population doubling rate) and the optimal nutrient-to-growth rate ratio at each time point throughout the entire culture cycle is calculated. The average absolute error of the above two deviations is calculated as the dynamic adaptation deviation characteristic. The smaller this value, the better the nutrient supply strategy matches the cell growth requirements and the more stable the process.

[0027] In one embodiment of the present invention, the AI ​​quality prediction and intelligent decision-making module plots a dual-modal cell growth curve, including: Real-time reception of cell growth data, extraction of cell features and construction of dual-modal cell growth curves using multimodal image analysis and dynamic correlation algorithms, specifically including: Cell behavior feature extraction unit: It is used to receive continuous cell microscope image sequences and cell counting data, extract the cell population division synchronicity index and migration aggregation degree through optical flow field tracking algorithm, and extract the cell pseudopodia extension length and nucleocytoplasmic ratio through deep learning semantic segmentation model, which together constitute a dynamic cell activity state feature vector; Dual-dimensional growth curve construction unit: used to fuse the cell count data with the cell activity state feature vector to construct a dual-dimensional growth curve whose vertical axis simultaneously includes the cell number and the mean value of the activity feature vector; Environmental correction unit: used to access real-time environmental sensor data, establish a correlation model between environmental parameters and cell dynamic characteristics through an attention mechanism algorithm, and correct the prediction baseline of the two-dimensional growth curve in real time; Multi-mode deviation intelligent judgment and early warning module: It is used to compare the corrected real-time two-dimensional growth curve with the pre-stored successful model standard interval and execute the corresponding deviation mode judgment logic.

[0028] Specifically, in the cell behavior feature extraction unit, the system receives a continuous sequence of cell images from an automated microscope (e.g., one frame per second) and cell count and viability data from image analysis software in real time. It applies an optical flow tracing algorithm (e.g., Farneback algorithm) to the continuous image sequence to calculate the motion vector of each pixel in the image, thereby tracking the movement trajectory of the cell population. The division synchronicity index is calculated by analyzing the spatiotemporal consistency of morphological changes in cells during division (intense cytoplasmic contraction) within the population. A higher index indicates more synchronized and healthy population growth. Migration aggregation is calculated by measuring the average aggregation degree of all cell movement trajectories (e.g., orbital density). The aggregation degree is quantified by the reciprocal of the average nearest neighbor distance between traces. An abnormally high aggregation degree may indicate a stress response, while an abnormally low aggregation degree may indicate insufficient cell viability. A pre-trained deep learning semantic segmentation model is used to process single cell images to accurately segment the cell outline and nucleus. The average length of pseudopodia is calculated from the segmented cell outline. The longer the pseudopodia, the stronger the cell migration and adhesion ability is usually indicated by the ratio of the cell nucleus pixel area to the total cytoplasm pixel area. The above four indicators, including the division synchronicity index, migration aggregation degree, pseudopodia extension length and nucleoplasm ratio, are combined into a four-dimensional dynamic cell activity state feature vector at each time point. In the dual-dimensional growth curve construction unit, cell count data is fused with dynamic cell activity state feature vectors to construct a growth curve with time as the horizontal axis and two vertical axes as the core. Specifically, the first vertical axis displays the normalized cell count, and the second vertical axis displays the mean of the dynamic cell activity state feature vector, which comprehensively reflects the overall level of four activity indicators. The system interface simultaneously displays these two curves, allowing operators to intuitively compare cell count growth with changes in intrinsic activity. In the environmental correction unit, real-time sensor data from the culture environment, including temperature, pH, and dissolved oxygen concentration, is input. Using an attention mechanism algorithm, the environmental parameters and cell activity feature vectors from the previous moment are used as input to predict the cell count and activity at the current moment. Based on the established correlation model, the predicted baseline of the standard growth curve is fine-tuned in real time. For example, when the system detects a brief but significant positive fluctuation in temperature, it uses the model to predict its potential stimulating effect on growth and adjusts the predicted baseline accordingly.

[0029] In one embodiment of the present invention, the execution of the corresponding deviation mode determination logic includes: If only the cell number dimension deviates from the standard range, while the activity feature vector dimension is normal, it is judged as recoverable growth retardation. If the dimension of the activity feature vector deviates before the dimension of cell number, it is judged as a latent abnormality risk, triggering an orange alert, and the image backtracking algorithm is activated to generate an abnormality source heat map that identifies the source of the abnormality. If both dimensions deviate from the preset threshold range, it is determined to be an irreversible quality risk, triggering a red alert and automatically locking subsequent operation permissions for that batch of cells.

[0030] Specifically, complete data from at least 200 historical successful batches (those that passed final quality inspection) were collected, including their complete two-dimensional growth curve data (cell number sequence and activity feature vector sequence). For the normalized cell number at each time point, the mean A and standard deviation B of all successful batches at that point were calculated. The standard interval was set as A±2B, which covers approximately 95% of the successful batch data and serves as the normal range. The red alert threshold (lower limit) was set as A-3B. When the cell number is consistently below this lower limit (e.g., in three consecutive monitoring points, the activity feature vector exceeds the same side of its normal range at two points), it indicates a serious deviation from the normal population, triggering a red alert. The activity dimension threshold... The activity feature vector is set as a multivariate, and a threshold is set separately for each component (persistent deviation: such as splitting synchronicity index, nucleo-cytoplasmic ratio, etc.). Similarly, the mean 'a' and standard deviation 'b' of each activity index at each time point are calculated, and the normal range is set to a ± 2b. For key indicators such as nucleo-cytoplasmic ratio, a one-way threshold is set. For example, the upper limit of the red warning for nucleo-cytoplasmic ratio can be set to the mean nucleo-cytoplasmic ratio ± 2.5 times the standard deviation of nucleo-cytoplasmic ratio, because an abnormal increase in nucleo-cytoplasmic ratio is a risk signal. The preceding time-series logic is defined as follows: the system maintains two status flags for each batch, including a data anomaly flag and an activity anomaly flag. When the above-mentioned persistent deviation rule is triggered for the first time in the activity dimension, the system records the time and sets the activity anomaly flag.

[0031] In one embodiment of the present invention, the calculation of the optimal time window for succession includes: Based on real-time microscope image sequences, a graph neural network is used to construct an interaction map between cells, from which the number of effective proliferation contact pairs within the division cycle is extracted and compared with a passage critical interaction density threshold trained on historical data. The passage critical interaction density threshold is associated with the interaction density in historical data and the cell adhesion rate after passage. Based on the concentration changes of key nutrients in the culture medium and cell count data, the nutrient consumption rate is calculated. Using a long short-term memory network model, with the nutrient consumption rate and the current cell growth stage as input, the nutrient reserves are predicted to support the cells to complete the next full division cycle within a preset time period. Based on the prediction results, the adjustment instructions for the passage time window are output. Cell fluorescence images are processed, and the distribution variance of fluorescence signals in the cell population is calculated using image analysis algorithms as a coefficient of stemness expression evenness. A mapping table is established. When the mapping table indicates that the predicted stemness maintenance of the current index decreases by more than a preset stemness threshold, the current time window is excluded. At the same time, if the maximum predicted fluctuation range of any environmental parameter exceeds its corresponding static stress tolerance threshold, the current time window is excluded. The system receives the evaluation results and the feasible time window set, and takes the intersection of the feasible time window set to generate a candidate time window set. Based on the results of effective proliferation contact log comparison, nutritional support capacity prediction, stem expression evenness coefficient assessment, and environmental disturbance matching, the final optimal passage time window was determined by screening from candidate time windows.

[0032] Specifically, based on real-time acquired microscope image sequences, a pre-trained graph neural network is used to construct a cell-cell interaction map. In this map, each cell is considered a node, and physical contact or close interaction between cells is considered an edge. From the constructed interaction map, the number of cell pairs in prophase and metaphase of cell division within the current cell division cycle that have stable contact with other cells is counted. This is the effective proliferation contact pair count, which can intuitively reflect the activity level of cell populations in communicating and coordinating proliferation through contact inhibition. The real-time calculated effective proliferation contact pair count is compared with the passage critical value trained based on historical data. Interaction density thresholds were compared, with the critical interaction density threshold for passage determined by analyzing the correlation between interaction density and cell adhesion rate after subsequent passages in historical data. When the real-time interaction density is below this threshold, it indicates that the cells still have sufficient independent growth space, and passage is not urgent; when it approaches or reaches this threshold, it indicates that the cells begin to slow down division due to contact inhibition, which is an ideal signal for passage. Based on the time-series monitoring data of the concentration of key nutrients (e.g., glucose, glutamine) in the culture medium and cell count data, the nutrient consumption rate per unit time was calculated, and the nutrient consumption rate was compared with the current cell growth stage (e.g., ...). The cells (in the late logarithmic growth phase) are fed into a pre-trained Long Short-Term Memory (LSTM) network model. This model is trained to predict whether the current nutrient reserves are sufficient to support the cell population to complete the next full division cycle within a preset time period (e.g., the next 24 hours). The LSTM model outputs a nutrient support capacity score. Based on the score, the system outputs instructions to adjust the passage time window: if the score is high, the time window can be appropriately extended; if the score drops rapidly, passage is strongly recommended before nutrient depletion. Stem uniformity is assessed by fluorescent staining of cells with pluripotency markers and imaging. The fluorescence image is analyzed using an image analysis algorithm to calculate the distribution variance of fluorescence signal intensity in all cells as the stemness expression evenness coefficient. The smaller the coefficient, the more uniform the stemness of the cell population and the better its condition. A pre-established mapping table is consulted (the mapping table is learned from historical data and describes the correspondence between different evenness coefficients and the stemness maintenance after passage). When the mapping table indicates that the predicted decrease in stemness maintenance based on the current coefficient exceeds the preset stemness threshold (95% of the stemness expression evenness coefficient), the current and adjacent time windows are excluded to avoid the exacerbation of stemness loss due to passage operations.The environmental disturbance assessment involves accessing an environmental prediction system to obtain the maximum predicted fluctuation range of various environmental parameters over the next 24 hours. This fluctuation range is then compared with the static stress tolerance threshold of the cell line, which is determined experimentally beforehand. (This threshold represents the maximum static fluctuation range of a single environmental parameter that a specific stem cell line can withstand during its most vulnerable recovery period after passage (typically 6-24 hours post-passage) without a statistically significant decrease in adhesion rate, survival rate, or proliferation capacity.) If the maximum predicted fluctuation range of any environmental parameter exceeds the corresponding static stress tolerance threshold, the environmental risk for that period is considered too high, and this time window is excluded to ensure that cells can survive after passage. Recovery in a stable environment; receiving the evaluation results from the above steps and the output set of feasible time windows (i.e., time periods where nutritional support capacity allows and are not excluded by stemness or environmental factors), the system performs an intersection operation on all the above feasible time window sets to generate a preliminary candidate time window set; finally, the system selects the final optimal passage time window from the candidate time window set based on the following priorities: within the candidate window, the earliest time point with the closest effective proliferation contact pairs but not exceeding the passage critical interaction density threshold is selected; if there are multiple time points that meet the primary priority, the time point with the highest nutritional support capacity score and the best stemness expression evenness coefficient (smallest variance) is selected.

[0033] In one embodiment of the present invention, the calculation of the pre-release probability based on a preset evaluation index includes: Collect data from the entire process of preparing this batch of stem cells and input it into the cell quality prediction model. The data from the entire process includes real-time dynamic data and statically recorded data. The cell quality prediction model calculates the pre-release probability of this batch of stem cells based on preset evaluation indicators, including environmental compliance rate, operational compliance rate, growth curve compliance rate, and stage quality inspection pass rate. The calculation of the pre-release probability is as follows:

[0034] in, This is the pre-release probability. For environmental compliance rate, To improve operational compliance rate, For the growth curve achievement rate, This refers to the pass rate of the phased quality inspection.

[0035] Specifically, the real-time dynamic data includes time-series data from sensors and devices, such as all environmental parameter records, cell growth curve data, and online microscope image sequences; static recorded data includes static data entered by operators or provided by the equipment information system, such as cell donor information, key reagent batch numbers, operator records, and offline quality inspection reports at each stage; the system calls predefined algorithms to calculate scores for four core evaluation indicators, specifically: Environmental compliance rate calculation: calculating the percentage of data points where all environmental parameters (temperature, etc.) fall within their preset standard range throughout the entire preparation cycle; Operational compliance rate calculation: comparing actual operation records with standard operating procedures to calculate key operational steps. (e.g., digestion, passage) The percentage of steps that fully comply with the SOP in terms of operation time, reagent dosage, and reaction time; Growth curve compliance rate calculation: The bimodal cell growth curve of this batch is compared with the pre-stored standard interval of successful batches. This compliance rate can be defined as the percentage of time points when both the cell number and activity feature vector curves are within the standard interval out of the total monitoring time points; Stage quality inspection pass rate calculation: Based on the completed stage quality inspection reports, the percentage of qualified items out of the total number of tested items is calculated; The scores of the above four indicators are input into the pre-trained cell quality prediction model. The model assigns different weights to each indicator according to its influence on the final quality, and finally obtains the pre-release probability value.

[0036] In one embodiment of the present invention, the blockchain data storage module includes: The key audit elements include cell donor identification, key reagent batch number, cell operation timestamp, operator biovalidation data, and continuous readings of environmental sensors during key operation phases. The set of key audit elements is standardized and serialized to generate its corresponding unique digital fingerprint; The unique digital fingerprint is generated by adding the serialized key audit element data to an encrypted hash function to generate a first-level hash value. This first-level hash value is then concatenated with a random number generated from a preceding block header hash. A second hash operation is performed on the concatenated data to generate the final unique digital fingerprint. Specifically, when the stem cell preparation process reaches a predefined key node (e.g., completion of cell resuscitation, end of passage, and final harvest), the system automatically triggers the evidence storage process. The on-chain evidence storage unit automatically extracts the following key audit elements associated with the unique identifier of that batch from the data center: cell donor identifier, key reagent batch number, cell operation timestamp, operator bio-verification data (hash value of fingerprint or facial recognition feature data verified when the operator logs into the system), and continuous readings from environmental sensors during the key operation phase (e.g., 10 minutes before the start of the operation to 10 minutes after the end, all raw readings from temperature sensors or other sensors in the incubator). The extracted data is converted into a unified format, all strings are converted to UTF-8 encoding, and numerical data is standardized in precision. The standardized key audit elements are then assembled into a single fingerprint according to a predefined, fixed key-value pair sequence. Structured data objects are converted into a unique binary data stream using serialization tools such as JOSN or Protocol Buffers. This ensures that even if the data content is the same, different orders will generate different hash values. The generated serialized binary data stream is used as input and processed by a cryptographic hash function to generate a fixed-length first-level hash value, representing the data content of key audit elements. The block header hash of the latest block is obtained from the blockchain network, combined with a current timestamp, and then hashed again to generate a random number, ensuring that the fingerprint generated for each batch of data is unique. The first-level hash value is concatenated with the random tree generated in the previous step. The new string obtained after concatenation is hashed again using the same cryptographic hash function. The final result of this operation is the unique digital fingerprint of the key stem cell data for that batch.

[0037] In one embodiment of the present invention, the query module includes: When a specific batch of stem cell traceability request is initiated, a query instruction is received through the authorized transparent query module, and the query instruction carries the unique identifier of the batch of stem cells. Based on the unique identifier, retrieve the full process data corresponding to this batch; Simultaneously, through the blockchain data storage module, the key audit elements corresponding to the batch and the unique digital fingerprint stored on the chain are matched according to the unique identifier to verify the consistency between the retrieved full-process data and the unique digital fingerprint, and to confirm that the data has not been tampered with. The retrieved data from the entire process are linked together according to the stem cell preparation timeline to form a complete traceability view.

[0038] Specifically, when a stem cell batch is created in the system, it is first assigned a unique identifier. Subsequently, all collected data throughout the entire process is associated with this unique identifier and stored in an (off-chain) database. At key nodes in the preparation process or when the batch is completed, the system automatically triggers the evidence storage process. Specifically, according to predefined rules, key audit elements are extracted from all data associated with the batch's unique identifier. After standardization and serialization, a unique digital fingerprint for the batch is generated through hash calculation. The generated unique digital fingerprint is uploaded to the blockchain. During the upload, the unique digital fingerprint is bound together with the batch's unique identifier and recorded on the blockchain, so that the corresponding unique digital fingerprint can be found on the blockchain through the unique identifier.

[0039] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A stem cell preparation traceability management system, characterized by, The system comprises the following modules: A data collection module: for the whole process of stem cell preparation batch, collecting environmental data, operation timing data, cell growth data and quality inspection result data, and establishing a unique identification for each batch; An AI quality prediction and intelligent decision-making module: collects the whole chain data of successful and failed samples of historical stem cell preparation batches, extracts features to build a cell quality prediction model, draws a bimodal cell growth curve, inputs the bimodal cell growth curve into the cell quality prediction model to determine whether there is a risk of abnormal growth, predicts the growth trend of the cells according to the cell quality prediction model combined with historical stem cell growth data and real-time collected various data, calculates the optimal time window for subculture, and collects the traceability data of the whole process of the batch of stem cells and inputs it into the model to calculate the pre-release probability based on the preset evaluation index; A blockchain data storage module: predefines key audit elements in the stem cell preparation process, and the on-chain storage unit automatically extracts the corresponding key audit elements, performs hash operation to generate a unique hash value, and synchronously uploads it to the blockchain network for consensus verification; A query module: creates an authorized query port for various stem cell preparation related parties, assigns a unique identity and access permission to each authorized party, and displays the whole process traceability link in the query result.

2. The stem cell preparation traceability management system according to claim 1, wherein, The data collection module comprises: After cleaning and format conversion of the collected various data, the unique association between the data and the stem cell preparation batch is established, and the data is synchronously pushed to the AI quality prediction and intelligent decision-making module and the blockchain data storage module.

3. The stem cell preparation traceability management system according to claim 1, wherein, In the AI quality prediction and intelligent decision-making module, the cell quality prediction model comprises: Collecting the whole chain data of successful and failed samples of historical stem cell preparation, the data set from the data collection module, and each type of data is bound with the unique traceability identifier of the sample; Performing feature engineering processing on the whole chain data, extracting key features, and using machine learning algorithm to build a cell quality prediction model to output normal or abnormal judgment; The key features include: microenvironment spatiotemporal heterogeneity coupling feature, operation-microenvironment response lag feature, cell self-organization pattern evolution feature and nutrient supply-growth demand dynamic adaptation feature; And train the model through historical data, and deploy the trained model to the system.

4. The stem cell preparation traceability management system according to claim 3, wherein, The key features include: The microenvironment spatiotemporal heterogeneity coupling feature is based on multi-point time series data of a sensor array in a culture dish, and a product coefficient of spatial gradient variance of key environmental parameters and time fluctuation variance thereof is calculated, specifically: wherein, is the product coefficient, is the spatial gradient variance, is the time fluctuation variance; The operation-microenvironment response lag feature aligns operation events with cell growth rate time series based on a dynamic time warping algorithm, extracts a lag time of a growth rate response peak relative to an operation event, and a correlation coefficient of an operation parameter deviation and an environmental fluctuation parameter amplitude wherein, is a correlation coefficient, is an operation parameter deviation, is an environmental fluctuation parameter amplitude; The cell self-organization pattern evolution characteristics are based on a gray level co-occurrence matrix and a morphological skeleton analysis of a cell microscope image, and extract an image texture entropy value and a proportion of a ring closed structure in all skeleton structures, specifically: wherein, is the proportion of the ring closed structure, is a total number of pixels of the ring structure, is a total number of pixels of the total skeleton structure; wherein the cell microscope image is a cell population image photographed by a phase contrast microscope, and the ring closed structure corresponds to a cell island structure boundary formed when a cell fusion degree is greater than or equal to 70%. The nutritional supply-growth demand dynamic adaptation feature is based on medium composition concentration time series data and cell number time series data, calculates a ratio deviation of a nutrient consumption rate and a cell population doubling rate, specifically: wherein, is a nutrient dynamic adaptation deviation, is a total number of time points, is a time point index, is a nutrient consumption rate at time point is an estimate of a cell population doubling rate at time point is an estimate of a cell population doubling rate at time point is an estimate of a cell population doubling rate at time point is an optimal nutrient-growth rate ratio.

5. The stem cell preparation traceability management system according to claim 1, wherein, In the AI quality prediction and intelligent decision-making module, the bimodal cell growth curve comprises: Real-time receiving of cell growth data, using multi-modal image analysis and dynamic correlation algorithm to extract cell features and construct bimodal cell growth curve, specifically including: A cell behavior feature extraction unit: used for receiving continuous cell microscope image sequence and cell counting data, extracting cell population division synchrony index and migration aggregation degree through optical flow field tracking algorithm, and extracting cell pseudopod extension length and nuclear-cytoplasmic ratio through deep learning semantic segmentation model to form a dynamic cell activity state feature vector; A two-dimensional growth curve construction unit is configured to fuse the cell count data and the cell activity state feature vector to construct a two-dimensional growth curve whose longitudinal axis contains both the cell quantity and the average of the cell activity feature vector; An environment correction unit is configured to access real-time environment sensor data, establish an association model between the environment parameters and the cell dynamic characteristics through an attention mechanism algorithm, and correct the prediction baseline of the two-dimensional growth curve in real time; A multi-mode deviation intelligent judgment and early warning module is configured to compare the corrected real-time two-dimensional growth curve with the pre-stored successful model standard interval, and execute corresponding deviation mode judgment logic.

6. The stem cell preparation traceability management system according to claim 5, wherein, The execution of the corresponding deviation mode judgment logic includes: If only the cell quantity dimension deviates from the standard interval, and the activity feature vector dimension is normal, it is determined as recoverable growth retardation; If the activity feature vector dimension deviates before the cell quantity dimension, it is determined as an implicit abnormal risk, an orange early warning is triggered, and an image backtracking algorithm is started to generate an abnormal source tracing heat map; If both dimensions deviate beyond the pre-set threshold range, it is determined as an irreversible quality risk, a red early warning is triggered, and the subsequent operation permission of the batch of cells is automatically locked.

7. The stem cell preparation traceability management system according to claim 1, wherein The calculation of the optimal time window for subculture includes: Based on the real-time microscope image sequence, an interaction map between cells is constructed through a graph neural network to extract the number of effective proliferation contact pairs in the division cycle, and a subculture critical interaction density threshold trained based on historical data is compared, wherein the subculture critical interaction density threshold is related to the interaction density in the historical data and the cell adhesion rate after subculture; Based on the concentration change data of key nutrient factors in the culture medium and the cell count data, the nutrient consumption rate is calculated, and through a long short-term memory network model, the support ability of nutrient reserves for cells to complete the next complete division cycle in a future preset period is predicted with the nutrient consumption rate and the current cell growth stage as inputs. According to the prediction result, an adjustment instruction for the subculture time window is outputted; The cell fluorescence image is processed, the distribution variance of the fluorescence signal in the cell population is calculated as the sternness expression uniformity coefficient through an image analysis algorithm, and a mapping relationship table is established. When the mapping relationship table indicates that the predicted sternness maintenance degree of the current exponential phase decreases by more than a pre-set sternness threshold, the current time window is excluded. At the same time, if the maximum predicted fluctuation range of any environmental parameter exceeds its corresponding static stress tolerance threshold, the current time window is excluded; The received evaluation results and feasible time window set are intersected to generate a candidate time window set; Based on the comparison results of the number of effective proliferation contact pairs, the prediction results of the nutrient support ability, the evaluation results of the sternness expression uniformity coefficient, and the environmental disturbance matching results, the final optimal subculture time window is determined from the candidate time window set.

8. The stem cell preparation traceability management system of claim 1, wherein, The calculation of the pre-release probability based on the pre-set evaluation index includes: Collecting the whole process data of the batch of stem cells and inputting them into the cell quality prediction model, the whole process data includes real-time collected dynamic data and statically recorded data; The cell quality prediction model calculates a pre-release probability of the batch of stem cells based on preset evaluation indexes, and the preset evaluation indexes include an environmental compliance rate, an operation compliance rate, a growth curve compliance rate, and a stage quality inspection pass rate. The calculation of the pre-release probability is specifically as follows: wherein, is a pre-release probability, is an environmental compliance rate, is an operational compliance rate, is a growth curve compliance rate, is a stage quality inspection pass rate.

9. The stem cell preparation traceability management system according to claim 1, wherein, The blockchain data storage module includes: The key audit elements include a cell donor identifier, a key reagent batch number, a cell operation timestamp, operation personnel biological authentication data, and continuous readings of an environmental sensor during a key operation stage; The key audit element set is standardized and serialized to generate a corresponding unique digital fingerprint; The unique digital fingerprint is generated by adding the serialized key audit element data to an encryption hash function to generate a first-level hash value, concatenating the first-level hash value with a random number generated by a pre-sequence block header hash, and performing a second hash operation on the concatenated data to generate a final unique digital fingerprint.

10. The stem cell preparation traceability management system according to claim 1, wherein, The query module includes: When a specific batch of stem cell traceability request is initiated, a query instruction is received through the authorized transparent query module, and the query instruction carries a unique identifier of the batch of stem cells; Based on the unique identifier, the corresponding whole-process data of the batch is called; The whole-process data is verified for consistency with the unique digital fingerprint by matching the key audit elements and the unique digital fingerprint stored on the chain based on the unique identifier through the blockchain data storage module, and it is confirmed that the data has not been tampered with; The called whole-process data is concatenated according to the stem cell preparation timeline to form a whole-chain traceability view.

Images