An AI-based laboratory data workflow management system and method

Abstract

This invention relates to the field of data security management technology, specifically disclosing an artificial intelligence-based laboratory data flow management method, comprising the following steps: collecting laboratory data; judging whether the experimental results meet expectations based on a pre-trained artificial intelligence model and experimental parameters; if not, marking them as abnormal data; if so, marking them as normal data; establishing a database to store normal data; starting from the moment a user accesses the database, collecting normal data browsed by the user during this access process at preset intervals, and using the normal data browsed by the user during this access process collected at the current moment as target data; generating coordinate points of the target data; clustering the coordinate points to obtain clusters; and determining whether a user is an abnormal user based on the clusters. This invention achieves behavior monitoring of logged-in users, improving data security.

Description

Technical Field

[0001] This invention relates to the field of data security management technology, specifically to an artificial intelligence-based laboratory data process management system and method. Background Technology

[0002] Laboratory experiments generate a large amount of data, such as absorbance values ​​of samples, chromatographic peak areas, pH value change curves, and real-time records of reaction temperature and pressure. This data covers both qualitative and quantitative dimensions and is the core basis for analyzing experimental phenomena and verifying hypotheses.

[0003] Experimental data involves crucial information such as experimental protocol details and raw test results, and its importance is self-evident. In existing technologies, access control for experimental data is mostly based on assigned accounts and passwords, achieving only "access control" at the basic level. This lack of real-time monitoring of logged-in user behavior and online user activity compromises the security of experimental data. Summary of the Invention

[0004] The purpose of this invention is to provide an artificial intelligence-based laboratory data flow management system and method to solve the above-mentioned technical problems.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] An artificial intelligence-based laboratory data workflow management method includes the following steps:

[0007] Collect laboratory data, including experimental parameters and results. Based on the pre-trained artificial intelligence model and experimental parameters, determine whether the experimental results meet expectations. If not, mark them as abnormal data; if so, mark them as normal data.

[0008] Establish a database to store normal data. Starting from the moment a user accesses the database, collect the normal data that the user browses during this access process at preset intervals, and use the normal data that the user browses during this access process collected at the current moment as the target data.

[0009] The coordinates of the target data are generated, and the coordinates are clustered based on the DBSCAN algorithm to obtain clusters. The number of clusters N is obtained. If the number N is greater than the preset threshold, the corresponding user is marked as an abnormal user and a prompt message is sent to the preset administrator.

[0010] As a further aspect of the present invention: the process of determining whether the experimental results meet expectations includes:

[0011] Laboratory data with the same experimental parameters as the latest collected laboratory data within the past n days are recorded as evaluation data. The mean and standard deviation of the experimental results of the evaluation data are calculated, where n is the preset number.

[0012] The latest collected experimental results, the experimental results of the evaluation data, and the mean and standard deviation of the experimental results of the evaluation data are input into the pre-trained artificial intelligence model. The artificial intelligence model outputs the judgment result, which includes whether it meets the expectations or does not meet the expectations.

[0013] If the judgment result is in line with expectations, then the experimental results in the latest collected laboratory data are considered to be in line with expectations.

[0014] As a further aspect of the present invention: the process of training an artificial intelligence model includes:

[0015] Historical laboratory data is collected. For each historical laboratory data point, the experimental results of the corresponding evaluation data, as well as the mean and standard deviation of the evaluation data, are collected. Each historical laboratory data point, the experimental results of the corresponding evaluation data, and the mean and standard deviation of the evaluation data are used as a sample. Based on manual labeling, labels are added to the sample, including those that meet expectations and those that do not.

[0016] An artificial intelligence model is built based on the random forest algorithm, and the model is trained and validated based on labeled samples.

[0017] As a further aspect of the present invention: the coordinate points for generating the target data include:

[0018] Generate the coordinates (X1, X2, ..., Xm) of the target data, where Xm represents the feature value of the m-th experimental parameter of the target data, and m represents the total number of types of experimental parameters;

[0019] If the m-th experimental parameter does not exist in the target data, then the eigenvalue Xm = 0; otherwise, the eigenvalue Xm = 1.

[0020] As a further aspect of the present invention, after obtaining normal data, the method of visualization processing of normal data includes line charts, scatter plots, heatmaps and bar charts.

[0021] As a further aspect of the present invention, the laboratory data also includes equipment serial numbers and sample serial numbers.

[0022] An artificial intelligence-based laboratory data workflow management system includes:

[0023] Data Judgment Module: Collects laboratory data, including experimental parameters and results. Based on a pre-trained artificial intelligence model and experimental parameters, it judges whether the experimental results meet expectations. If not, it marks them as abnormal data; if so, it marks them as normal data.

[0024] Behavior collection module: Establishes a database to store normal data. Starting from the moment the user accesses the database, it collects the normal data browsed by the user during this visit at preset intervals, and uses the normal data browsed by the user during this visit collected at the current moment as the target data.

[0025] Behavior judgment module: Generates the coordinate points of the target data, clusters the coordinate points based on the DBSCAN algorithm to obtain clusters, obtains the number of clusters N, and if the number N is greater than the preset threshold, marks the corresponding user as an abnormal user and sends a prompt message to the preset administrator.

[0026] The beneficial effects of this invention compared to the prior art are as follows:

[0027] This invention achieves effective screening and management of experimental data by collecting and classifying laboratory data and using the conformity of experimental results with expectations as the basic standard. During user access to the database, this invention can extract normal data viewed by users as target data and further generate corresponding coordinate points for cluster analysis. In this way, the purpose of user access to the database can be accurately reflected. If the access behavior is clearly targeted, the number of clusters is small, indicating a focused objective and reasonable access behavior. If the access behavior is abnormal, the number of clusters increases significantly, indicating a lack of purpose in the user's data access. Therefore, this invention enables real-time monitoring of the operational behavior of logged-in users, thereby further improving the security of experimental data usage based on data access control. Attached Figure Description

[0028] The invention will now be further described with reference to the accompanying drawings.

[0029] Figure 1 This is a flowchart illustrating a laboratory data flow management method based on artificial intelligence according to the present invention. Detailed Implementation

[0030] 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.

[0031] Please see Figure 1 As shown, this invention is an artificial intelligence-based laboratory data workflow management method, comprising the following steps:

[0032] Collect laboratory data, including experimental parameters and results. Based on the pre-trained artificial intelligence model and experimental parameters, determine whether the experimental results meet expectations. If not, mark them as abnormal data; if so, mark them as normal data.

[0033] Understandably, laboratory data also includes equipment serial numbers and sample serial numbers;

[0034] Taking a sample purity test in a chemical laboratory as an example, the experimental equipment is a high-performance liquid chromatograph. After the chromatograph completes the sample purity test, it generates the following laboratory data: equipment number "LC-001", sample number "Sample-20231001", experimental parameters "mobile phase ratio (methanol:water = 7:3), detection wavelength 254nm", and experimental result "purity 97.2%".

[0035] In a preferred embodiment of the present invention, the process of determining whether the experimental results meet expectations includes:

[0036] Laboratory data with the same experimental parameters as the latest collected laboratory data within the past n days are recorded as evaluation data. The mean and standard deviation of the experimental results of the evaluation data are calculated, where n is the preset number.

[0037] The latest collected experimental results, the experimental results of the evaluation data, and the mean and standard deviation of the experimental results of the evaluation data are input into the pre-trained artificial intelligence model. The artificial intelligence model outputs the judgment result, which includes whether it meets the expectations or does not meet the expectations.

[0038] If the judgment result is in line with expectations, then the experimental results in the latest collected laboratory data are judged to be in line with expectations.

[0039] In a preferred embodiment, the process of training the artificial intelligence model includes:

[0040] Historical laboratory data is collected. For each historical laboratory data point, the experimental results of the corresponding evaluation data, as well as the mean and standard deviation of the evaluation data, are collected. Each historical laboratory data point, the experimental results of the corresponding evaluation data, and the mean and standard deviation of the evaluation data are used as a sample. Based on manual labeling, labels are added to the sample, including those that meet expectations and those that do not.

[0041] An artificial intelligence model is built based on the random forest algorithm, and the model is trained and validated based on labeled samples.

[0042] It should be noted that in the process of training the artificial intelligence model, historical laboratory data is collected first. Historical data is data that has been generated and saved in past experiments. It truly records the correspondence between experimental parameters and experimental results, and can provide an empirical basis for the subsequent learning of the model.

[0043] After collecting historical laboratory data, corresponding evaluation data is found for each piece of laboratory data. Evaluation data refers to a set of data that is related to the historical data in time and has the same experimental parameters. This screening method can ensure that the data being compared are comparable.

[0044] The mean and standard deviation of the experimental results for these evaluation data are calculated. The mean can reflect the central tendency of the overall results of this type of experimental data, while the standard deviation can reflect the fluctuation of the data around the mean. These two indicators can comprehensively characterize the distribution characteristics of the evaluation data.

[0045] By integrating the experimental results of historical laboratory data and evaluation data, along with the mean and standard deviation, a sample is formed. This combination allows the model to not only consider a single experimental result during learning, but also to make judgments based on the statistical features of the context.

[0046] These samples are labeled manually, with labels divided into two categories: those that meet expectations and those that do not. This manual labeling uses domain knowledge to judge the experimental data, ensuring that the model has clear supervision signals during learning.

[0047] We chose to build an artificial intelligence model based on the random forest algorithm and used labeled samples to train and validate the model. The random forest algorithm is an ensemble learning method based on multiple decision trees. It reduces the bias of a single model by using the voting results of multiple decision trees and can learn the patterns in the samples more stably.

[0048] Historical data contains a wealth of experimental experience. By extracting data that aligns with experimental parameters and calculating the mean and standard deviation, individual data points can be linked to group characteristics, thus more accurately expressing whether the data is reasonable. Manual annotation provides the model with clear criteria for judgment, enabling the model to continuously compare the differences between "meeting expectations" and "not meeting expectations" during the learning process, thereby learning patterns that are closer to human judgment logic. The random forest algorithm is used to build the model because it maintains strong stability when facing high-dimensional features and can effectively avoid overfitting.

[0049] Establish a database to store normal data. Starting from the moment a user accesses the database, collect the normal data that the user browses during this access process at preset intervals, and use the normal data that the user browses during this access process collected at the current moment as the target data.

[0050] The coordinates of the target data are generated, and the coordinates are clustered based on the DBSCAN algorithm to obtain clusters. The number of clusters N is obtained. If the number N is greater than the preset threshold, the corresponding user is marked as an abnormal user and a prompt message is sent to the preset administrator.

[0051] In a preferred embodiment of the present invention, the coordinate points for generating the target data include:

[0052] Generate the coordinates (X1, X2, ..., Xm) of the target data, where Xm represents the feature value of the m-th experimental parameter of the target data, and m represents the total number of types of experimental parameters;

[0053] If the m-th experimental parameter does not exist in the target data, then the eigenvalue Xm = 0; otherwise, the eigenvalue Xm = 1.

[0054] Specifically, first, determine the total number of experimental parameter types, *m*. This total number refers to the set of all possible experimental parameters that may appear in the entire experimental system. For example, different experiments in the laboratory may involve temperature, humidity, pressure, solution concentration, stirring speed, reaction time, etc. If there are a total of one hundred experimental parameters, then *m* equals one hundred. After determining the total number, for the target data collected during a user visit, it is necessary to check whether each of the *m* experimental parameters appears in the current target data. If it appears, for example, if the target data records twenty parameters such as temperature, solution concentration, and reaction time, the feature value is set to 1 at the corresponding position. For other experimental parameters that do not appear, such as stirring speed or pressure not recorded in this experiment, the feature value is set to 0 at the corresponding position. In this way, a coordinate point composed of *m* values ​​can be obtained. The position of each value is fixed, corresponding to one type of experimental parameter. A value of 1 indicates that the parameter appears in the target data, and a value of 0 indicates that the parameter does not appear. In this way, the target data is abstracted into a vectorized point, which can be used in subsequent cluster analysis.

[0055] It is important to note that DBSCAN is a density-based clustering method. Its basic principle is to determine whether points belong to the same cluster by calculating the distance between coordinate points. When the number of neighboring points within a preset radius of a point exceeds a set threshold, the point and its neighbors are grouped into the same cluster, and this cluster is further expanded to form a larger cluster. If a point has insufficient neighbors within the preset radius, it is considered a noise point or an isolated point and will not be merged into an existing cluster.

[0056] Since the coordinate points generated in this invention are constructed based on the presence or absence of experimental parameters, the formation of the same type of cluster reflects the similarity of the data browsed by users when accessing the database in terms of parameter combinations. That is, the user's browsing behavior has a certain focus and purpose, while different clusters represent differences in the user's access behavior, involving data in different directions or with different purposes.

[0057] When normal users access experimental data, they typically focus on a specific experimental question or research objective. Therefore, the data they browse should exhibit strong concentration in terms of parameters. After DBSCAN clustering, the number of clusters formed should be relatively small, reflecting the clear purpose of their access behavior. Conversely, if a user forms a significantly larger number of clusters during their access, it indicates that the data they browse is scattered in terms of parameter distribution, lacking connection between different data points, and exhibiting aimless or abnormal characteristics. Therefore, when the number of clusters exceeds a preset threshold, the user can be marked as an abnormal user, and a prompt message can be sent to the administrator in a timely manner. This allows for real-time monitoring of the user's behavior even when they are logged in and have access permissions, ensuring the security of the experimental data.

[0058] In another preferred embodiment of the present invention, after obtaining normal data, the method further includes visualization processing of the normal data, including line charts, scatter plots, heatmaps and bar charts;

[0059] Understandably, line graphs are mainly used to show the trend of experimental parameters over time. Since experiments often require long-term tracking and observation, line graphs can intuitively show the fluctuation of a parameter at continuous time points, making it easier for researchers to quickly grasp the overall trend.

[0060] Bar charts are used to compare experimental results of different samples. Under the same experimental conditions, the performance of multiple samples often needs to be compared horizontally. Bar charts can clearly show the size of the differences and the relative relationships, thereby helping to identify the patterns between samples.

[0061] Scatter plots are used to present the relationship between experimental parameters and results. By plotting the distribution of points in a two-dimensional coordinate system, the correlation or distribution characteristics between parameter changes and experimental results can be observed intuitively, making it easier to judge potential causal relationships or patterns.

[0062] Heat maps are used to display data fluctuations of different devices. Since laboratories usually have multiple devices running at the same time, heat maps use color variations to visually represent the operating data of different devices within the same time period, which can quickly identify any abnormal differences between devices.

[0063] By combining the above four visualization methods, normal data can be comprehensively presented from different perspectives such as time dimension, sample comparison, parameter correlation and equipment differences. This allows the data to not only be stored and retrieved, but also to be presented in an intuitive graphical way, making it easier for managers and researchers to quickly understand and analyze the data content in subsequent use, thereby better serving the overall purpose of experimental data process management.

[0064] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the present invention should still fall within the scope of the present invention.

Claims

1. A laboratory data workflow management method based on artificial intelligence, characterized in that, Includes the following steps: Collect laboratory data, including experimental parameters and results. Based on a pre-trained artificial intelligence model and experimental parameters, determine whether the experimental results meet expectations. If not, mark them as abnormal data; if so, mark them as normal data. Establish a database to store normal data. Starting from the moment a user accesses the database, collect the normal data that the user browses during this access process at preset intervals, and use the normal data that the user browses during this access process collected at the current moment as the target data. The coordinates of the target data are generated, and the coordinates are clustered based on the DBSCAN algorithm to obtain clusters. The number of clusters N is obtained. If the number N is greater than the preset threshold, the corresponding user is marked as an abnormal user and a prompt message is sent to the preset administrator. The process of determining whether experimental results meet expectations includes: Laboratory data with the same experimental parameters as the latest collected laboratory data within the past n days are recorded as evaluation data. The mean and standard deviation of the experimental results of the evaluation data are calculated, where n is the preset number. The latest collected experimental results, the experimental results of the evaluation data, and the mean and standard deviation of the experimental results of the evaluation data are input into the pre-trained artificial intelligence model. The artificial intelligence model outputs the judgment result, which includes whether it meets the expectations or does not meet the expectations. If the judgment result is in line with expectations, then the experimental results in the latest collected laboratory data are judged to be in line with expectations. The process of training an artificial intelligence model includes: Historical laboratory data is collected. For each historical laboratory data point, the experimental results of the corresponding evaluation data, as well as the mean and standard deviation of the evaluation data, are collected. Each historical laboratory data point, the experimental results of the corresponding evaluation data, and the mean and standard deviation of the evaluation data are used as a sample. Based on manual labeling, labels are added to the sample, including those that meet expectations and those that do not. An artificial intelligence model is built based on the random forest algorithm, and the model is trained and validated based on labeled samples. The coordinates of the points used to generate the target data include: Generate the coordinates (X1, X2, ..., Xm) of the target data, where Xm represents the feature value of the m-th experimental parameter of the target data, and m represents the total number of types of experimental parameters; If the m-th experimental parameter does not exist in the target data, then the eigenvalue Xm = 0; otherwise, the eigenvalue Xm = 1.

2. The laboratory data flow management method based on artificial intelligence according to claim 1, characterized in that, After obtaining normal data, the process also includes visualization processing of the normal data. Visualization methods include line charts, scatter plots, heatmaps, and bar charts.

3. The laboratory data flow management method based on artificial intelligence according to claim 1, characterized in that, Laboratory data also includes equipment serial numbers and sample serial numbers.

4. A laboratory data workflow management system based on artificial intelligence, characterized in that, include: Data Judgment Module: Collects laboratory data, including experimental parameters and results. Based on a pre-trained artificial intelligence model and experimental parameters, it judges whether the experimental results meet expectations. If not, it marks them as abnormal data; if so, it marks them as normal data. Behavior collection module: Establishes a database to store normal data. Starting from the moment the user accesses the database, it collects the normal data browsed by the user during this visit at preset intervals, and uses the normal data browsed by the user during this visit collected at the current moment as the target data. Behavior judgment module: Generates the coordinate points of the target data, clusters the coordinate points based on the DBSCAN algorithm to obtain clusters, obtains the number of clusters N, and if the number N is greater than the preset number threshold, marks the corresponding user as an abnormal user and sends a prompt message to the preset management personnel. The process of determining whether experimental results meet expectations includes: Laboratory data with the same experimental parameters as the latest collected laboratory data within the past n days are recorded as evaluation data. The mean and standard deviation of the experimental results of the evaluation data are calculated, where n is the preset number. The latest collected experimental results, the experimental results of the evaluation data, and the mean and standard deviation of the experimental results of the evaluation data are input into the pre-trained artificial intelligence model. The artificial intelligence model outputs the judgment result, which includes whether it meets the expectations or does not meet the expectations. If the judgment result is in line with expectations, then the experimental results in the latest collected laboratory data are judged to be in line with expectations. The process of training an artificial intelligence model includes: Historical laboratory data is collected. For each historical laboratory data point, the experimental results of the corresponding evaluation data, as well as the mean and standard deviation of the evaluation data, are collected. Each historical laboratory data point, the experimental results of the corresponding evaluation data, and the mean and standard deviation of the evaluation data are used as a sample. Based on manual labeling, labels are added to the sample, including those that meet expectations and those that do not. An artificial intelligence model is built based on the random forest algorithm, and the model is trained and validated based on labeled samples. The coordinates of the points used to generate the target data include: Generate the coordinates (X1, X2, ..., Xm) of the target data, where Xm represents the feature value of the m-th experimental parameter of the target data, and m represents the total number of types of experimental parameters; If the m-th experimental parameter does not exist in the target data, then the eigenvalue Xm = 0; otherwise, the eigenvalue Xm = 1.

Images