Ssepsis risk prompt detection method based on model prediction

Abstract

The invention relates to the field of medical treatment, and discloses a sepsis risk prompt detection method based on model prediction, which comprises the following steps of: collecting a clinical blood sample of an acute infection patient, matching gene expression characteristics of the clinical blood sample with a database through gene target analysis, primarily screening out gene expression data obviously related to sepsis, and comparing the gene expression data with the database; according to the gene expression data of the marked sepsis patient samples and the gene expression data of the non-sepsis patient samples, gene expression differences are obtained, differential gene data and clinical data are established, a machine learning algorithm model is trained, and a sepsis recognition model is obtained; differential gene detection data related to sepsis is input into a sepsis recognition model for prediction, a sepsis recognition result is output, the sepsis recognition result and detection data of a sepsis rehabilitation patient are compared and analyzed, a model prediction effect is evaluated, and early warning is performed on prediction and evaluation results. The accuracy of model prediction is improved, and the prediction capability of the model is enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the medical technical field, more particularly to a model prediction-based sepsis risk prompt detection method. BACKGROUND

[0002] Sepsis is a severe systemic inflammatory response syndrome, usually caused by infection, which can lead to organ failure and even death. Early identification of high-risk patients with sepsis is the key to effective treatment; with the development of precision medicine, individualized risk assessment is needed for patients to identify and develop more appropriate treatment plans as soon as possible. With the development of big data and machine learning, a large amount of clinical data is collected and analyzed to extract features that are helpful for early diagnosis of sepsis, and a mathematical model is constructed to predict the likelihood of patients developing sepsis, so as to intervene in advance.

[0003] However, the above process still has the following shortcomings:

[0004] First, the existing model prediction method for sepsis risk lacks a comparison of the gene expression difference values between the sepsis group and the control group, further screening of the differential genes related to sepsis, and improper selection of input features, resulting in reduced accuracy of model prediction.

[0005] Second, the existing model prediction method for sepsis risk lacks the combination of gene expression data and clinical data to construct a prediction model, and the evaluation and feedback of the prediction effect of the model, which may lead to misdiagnosis and missed diagnosis due to insufficient single data source, and the inability to timely discover the shortcomings of the model. SUMMARY

[0006] In order to overcome the above-mentioned defects of the prior art, the present application provides a model prediction-based sepsis risk prompt detection method to solve the problems in the above background.

[0007] The present application provides the following technical solution: a model prediction-based sepsis risk prompt detection method, comprising:

[0008] S1: Collecting clinical blood samples of patients with acute infection, including sepsis patient sample data and non-sepsis patient sample, detecting the collected clinical blood samples of patients with acute infection, obtaining standardized clinical detection data, storing the standardized sample data into a biological sample library, and transmitting the processing result to S3;

[0009] S2: By searching for gene targets related to infection progression, and performing panel testing on clinical blood sample data collected from acute infection patients, the gene set is screened and enriched, and the gene target data is screened and extracted, and the extracted gene target data is entered and stored in the database, and the results are transmitted to S3;

[0010] S3: By matching the processed sepsis patient sample clinical testing data with the gene target data in the database, the gene expression data significantly related to sepsis is initially screened, and the initial screening results are transmitted to S4;

[0011] S4: According to the gene expression data significantly related to sepsis screened, the gene expression data of sepsis patient samples and non-sepsis patient samples are marked respectively, and by analyzing and comparing the gene variation between the marked gene expression data of sepsis patient samples and non-sepsis patient samples, the difference value of gene expression between the sepsis group sample data set and the control group sample data set is calculated, thereby screening the differential genes related to sepsis, and the screening results are transmitted to S5 and S6;

[0012] S5: According to the screened differential gene data related to sepsis and clinical data, a machine learning algorithm model is established, and the machine learning model is trained to obtain a sepsis recognition model, and the obtained sepsis recognition model is transmitted to S6;

[0013] S6: By inputting the screened differential expression genes that meet the requirements as differential gene data related to sepsis into the sepsis recognition model to predict the sepsis state, output the sepsis recognition result, and display and warn the sepsis recognition result, and output the result to S7 and S8;

[0014] S7: By obtaining the treatment archive data of sepsis rehabilitation patients who have been discharged, and extracting the sepsis rehabilitation patient detection data from it, comparing and analyzing the sepsis recognition result with the sepsis rehabilitation patient detection data, and calculating the sepsis recognition result difference factor, the model prediction effect is evaluated according to the sepsis recognition result difference factor, and the analysis result is transmitted to S8;

[0015] S8: According to the sepsis recognition result and the sepsis recognition result difference factor, a prediction result monitoring report is automatically generated, the sepsis recognition result is displayed in a visual way, and warning prompt information is generated to feedback to the detection personnel.

[0016] Preferably, the S1 collects clinical blood sample data of sepsis patients and non-sepsis patients from acute infection patients, including white blood cell count, C-reactive protein, procalcitonin (PCT), lactate level and platelet count, and establishes a biological sample library, standardizes the collected clinical blood sample data of acute infection patients, stores the processed sample data in the biological sample library, and labels the sample data of sepsis patients as sepsis group sample data and the sample data of non-sepsis patients as control group sample data.

[0017] Preferably, the S2 identifies gene targets related to infection progression through literature retrieval, bioinformatics analysis and expert consultation, designs and constructs a Panel for detecting gene targets according to the obtained gene targets, detects the clinical blood sample data of acute infection patients using the constructed Panel, screens out a gene set related to infection progression according to the detection results of the Panel, performs functional enrichment and pathway analysis on the screened gene set using Gene Ontology and Kyoto Encyclopedia of Genes and Genomes databases, and extracts gene target data highly related to infection progression from the enrichment analysis results.

[0018] Preferably, the S3 maps the gene expression data obtained by processing the sample data of sepsis patients to the gene target data in the database through the following specific steps:

[0019] Step S311: Create a mapping table of gene identifiers and expression values, using the gene identifiers as keys;

[0020] Step S312: Correspond the expression values in the sample data to the target genes in the database, and match the corresponding correct gene identifiers;

[0021] Step S313: Integrate the matched expression value data into a unified data framework, thereby obtaining the gene expression data significantly related to sepsis after the first screening.

[0022] Preferably, the S4 combines the labeled gene expression data of sepsis patients into a sepsis group sample data set X = {x1, x2, …, x i , …, x n}, combines the labeled gene expression data of non-sepsis patients into a control group sample data set K = {k1, k2, …, k j , …, k m}, and then uses statistical methods to compare and analyze the gene expression data, and calculates the difference value of the gene expression amount between the sepsis group sample data set and the control group sample data set as follows:

[0023]

[0024] wherein, represents the average gene expression of the sepsis group, represents the average gene expression of the control group, X δ 2 represents the variance of the gene expression of the sepsis group, K δ 2 represents the variance of the gene expression of the control group, n represents the number of samples of the sepsis group, and m represents the number of samples of the control group;

[0025] By comparing the difference threshold value ω1 of the first gene expression and the difference threshold value ω2 of the second gene expression with the difference value t of the gene expression between the sample data set of the sepsis group and the sample data set of the control group, and ω1> ω2, if the difference value t of the gene expression between the sample data set of the sepsis group and the sample data set of the control group is greater than the difference threshold value ω1 of the first gene expression or the difference value t of the gene expression between the sample data set of the sepsis group and the sample data set of the control group is less than the difference threshold value ω2 of the second gene expression, there is a significant difference in expression between the sepsis and the control group, and the screened differential expression genes are transmitted to S5 as the differential expression genes related to sepsis.

[0026] Preferably, S5 collects clinical information related to the clinical blood sample of the patient with acute infection, including age, gender, disease symptoms and treatment methods, and then combines the differential gene data related to sepsis and the clinical data according to the ID of the clinical blood sample of the patient with acute infection, and forms a clinical data set by combining the differential gene data related to sepsis and the clinical data, divides the clinical data set into a training set and a test set, wherein the data amount of the training set is much larger than that of the test set, uses random forest as a machine learning model, adjusts the parameters of the random forest, uses the training set data to train the model, obtains the sepsis recognition model, and then distinguishes the sepsis patients and the non-sepsis patients, and tests the sepsis recognition model through the test set of the divided clinical data set, including accuracy, recall rate, F1 score and specificity, to determine the accuracy of the sepsis recognition model.

[0027] Preferably, S6 inputs the differential expression gene data related to sepsis screened in real time into the sepsis recognition model, outputs the sepsis recognition result, and displays and warns the sepsis recognition result in a visual manner, wherein the sepsis recognition result includes whether the patient to be treated is infected with sepsis, the probability of the patient to be treated being infected with sepsis, and the risk degree of the patient to be treated being infected with sepsis.

[0028] Preferably, the S7 is compared and analyzed by extracting the differential expression gene set D of the model predicted sepsis identification result and the differential expression gene set U in the recovery patient detection data respectively, and the formula of the sepsis identification result difference factor is as follows:

[0029]

[0030] By setting a difference threshold, the sepsis identification result difference factor F is compared with the difference threshold, if F >= phi, it indicates that the model prediction effect is acceptable, if F < phi, it indicates that the model prediction effect is not ideal, and the model needs to be optimized and adjusted.

[0031] Preferably, the prediction result monitoring report generated by the S8 includes the sepsis identification result difference factor calculation, the model performance index, and the visual chart, the data visualization tool is used to create a chart to display the sepsis identification result, and according to the sepsis identification result, it is automatically judged whether the early warning prompt information needs to be generated, and the generated early warning prompt information is sent to the detection personnel terminal.

[0032] The technical effects and advantages of the present application are as follows:

[0033] The present application matches the clinical blood samples of the acute infection patients and the extracted gene target data, preliminarily screens the gene expression data significantly related to sepsis, marks the gene expression data of the sepsis patient sample and the gene expression data of the non-sepsis patient sample according to the preliminarily screened gene expression data significantly related to sepsis, analyzes and compares the gene variation between the marked gene expression data of the sepsis patient sample and the gene expression data of the non-sepsis patient sample, calculates the difference value of the gene expression amount between the sepsis group sample data set and the control group sample data set, thereby screening the differential genes related to sepsis, which is beneficial to improve the accuracy of model prediction and enhance the prediction ability of the model. The machine learning algorithm model is established by using the screened differential gene data related to sepsis and the clinical data, the machine learning model is trained, the sepsis identification model is obtained, the screened differential expression genes are input into the sepsis identification model as the differential gene data related to sepsis to predict the sepsis state, the sepsis identification result is output, the treatment archive data of the sepsis recovery patient is obtained, the sepsis recovery patient detection data is extracted from the treatment archive data, the sepsis identification result is compared and analyzed with the sepsis recovery patient detection data, the sepsis identification result difference factor is calculated, the model prediction effect is evaluated, and finally the model prediction result is prompted. By comprehensively utilizing the gene and clinical data, the model can more accurately identify the gene variation and clinical manifestations related to the disease, thereby reducing the misdiagnosis and missed diagnosis caused by the insufficient single data source, timely discovering the shortcomings of the model, and further enhancing the credibility of the model. Attached Figure Description

[0034] Figure 1 This is a diagram illustrating the method steps of the present invention.

[0035] Figure 2 This is a system structure block diagram of the present invention. Detailed Implementation

[0036] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. In addition, the forms of the various structures described in the following embodiments are merely illustrative. The sepsis risk warning detection method based on model prediction involved in the present invention is not limited to the structures described in the following embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] like Figure 1 The embodiment shown provides a sepsis risk warning detection method based on model prediction, including:

[0038] S1: By collecting clinical blood samples from patients with acute infections, including sample data from sepsis patients and samples from non-sepsis patients, the collected clinical blood samples from patients with acute infections are tested and processed to obtain standardized clinical test data. The standardized sample data is then stored in a biobank, and the processing results are transmitted to S3.

[0039] In this embodiment, step S1 involves collecting clinical blood sample data from patients with sepsis and those without sepsis from patients seeking treatment for acute infections. This data includes white blood cell count, C-reactive protein, procalcitonin (PCT), lactate level, and platelet count. A biobank is then established. The collected clinical blood sample data from patients with acute infections is standardized and stored in the biobank. Simultaneously, the sample data from patients with sepsis is labeled as sepsis group sample data, and the sample data from those without sepsis is labeled as control group sample data.

[0040] It should be noted that the sepsis group sample data was collected from blood samples taken from patients with sepsis, while the control group sample data was collected from blood samples taken from patients who did not have sepsis but had other diseases. The control group sample data was set up to compare and identify specific gene changes associated with sepsis.

[0041] S2: By searching for gene targets related to infection progression, and collecting clinical blood sample data from acute infection patients, and screening and enrichment analysis of gene sets, screening gene target data, extracting gene target data, and entering and storing the extracted gene target data into the database, and transmitting the results to S3.

[0042] In this embodiment, S2 identifies gene targets related to infection progression through literature search, bioinformatics analysis, and expert consultation, designs and constructs a panel for detecting gene targets based on the obtained gene targets, and uses the constructed panel to detect clinical blood sample data from acute infection patients, screens gene sets related to infection progression based on the panel detection results, and uses GeneOntology and Kyoto Encyclopedia of Genes and Genomes databases to perform functional enrichment and pathway analysis on the screened gene sets, and extracts gene target data highly related to infection progression from the enrichment analysis results.

[0043] It should be noted that the Pubmed, MEDLINE databases, acute infection, sepsis, transcriptome, and RNAprofiling are used to extract gene targets related to infection progression, and bioinformatics tools and databases are used to analyze potential gene targets.

[0044] S3: By matching the processed sepsis patient sample clinical detection data with the gene target data in the database, the gene expression data significantly related to sepsis is initially screened, and the initial screening results are transmitted to S4.

[0045] In this embodiment, S3 maps the gene expression data obtained from the processed sepsis patient sample data to the gene target data in the database through the following specific steps:

[0046] Step S311: Create a mapping table of gene identifiers and expression values, using gene identifiers as keys;

[0047] Step S312: Correspond the expression values in the sample data to the target genes in the database, and match the corresponding correct gene identifiers;

[0048] Step S313: Integrate the matched expression value data into a unified data framework, thereby obtaining the gene expression data significantly related to sepsis that is initially screened.

[0049] S4: According to the gene expression data of the screened genes significantly related to sepsis, the gene expression data of the sepsis patient samples and the gene expression data of the non-sepsis patient samples are marked respectively, the difference value of the gene expression between the sepsis group sample data set and the control group sample data set is calculated by analyzing and comparing the gene variation between the marked gene expression data of the sepsis patient samples and the marked gene expression data of the non-sepsis patient samples, the differential genes related to sepsis are screened, and the screening results are transmitted to S5 and S6.

[0050] In this embodiment, the S4 combines the marked gene expression data of the sepsis patient samples into a sepsis group sample data set X = {x1, x2, …, x i , …, x n}, combines the marked gene expression data of the non-sepsis patient samples into a control group sample data set K = {k1, k2, …, k j , …, k m}, and then uses a statistical method to compare and analyze the gene expression data, and the formula for calculating the difference value of the gene expression between the sepsis group sample data set and the control group sample data set is as follows:

[0051]

[0052] Wherein, represents the average gene expression of the sepsis group, represents the average gene expression of the control group, X δ 2 represents the variance of the gene expression of the sepsis group, K δ 2 represents the variance of the gene expression of the control group, n represents the number of sepsis group samples, and m represents the number of control group samples.

[0053] By comparing the difference value t of the gene expression between the sepsis group sample data set and the control group sample data set with the first gene expression difference threshold ω1 and the second gene expression difference threshold ω2 respectively, and ω1 > ω2, if the difference value t of the gene expression between the sepsis group sample data set and the control group sample data set is greater than the first gene expression difference threshold ω1 or less than the second gene expression difference threshold ω2, the gene has a significant difference expression between sepsis and the control group, and the differential expression gene meeting the requirement is screened as the differential expression gene related to sepsis and transmitted to S5.

[0054] It should be specifically pointed out that through the sepsis group sample data set X = {x1, x2, …, x i , …, x n} respectively calculate the average gene expression of the sepsis group and the variance X of the gene expression of the sepsis group δ 2 The formula is as follows:

[0055]

[0056] Wherein, x i represents the gene expression of the i-th sepsis group sample, and n represents the number of samples of the sepsis group;

[0057] By comparing the control group sample data set K = {k1, k2, …, k j , …, k m} respectively calculate the average gene expression of the control group and the variance K of the gene expression of the control group δ 2 The formula is as follows:

[0058]

[0059] Wherein, k j represents the gene expression of the j-th control group sample, and m represents the number of samples of the control group.

[0060] S5: According to the screened sepsis related differential gene data and clinical data, a machine learning algorithm model is established, and the machine learning model is trained to obtain a sepsis recognition model, and the obtained sepsis recognition model is transmitted to S6.

[0061] In this embodiment, the S5 collects clinical information related to the clinical blood samples of the acute infection patients, including age, gender, disease symptoms and treatment methods, and then combines the sepsis related differential gene data and the clinical data according to the clinical blood sample ID of the acute infection patients, and combines the sepsis related differential gene data and the clinical data into a clinical data set, and divides the clinical data set into a training set and a test set, wherein the data amount of the training set is much larger than that of the test set, uses random forest as a machine learning model, adjusts the parameters of the random forest, uses the training set data to train the model, obtains the sepsis recognition model, and then distinguishes the sepsis patients and the non-sepsis patients, and then tests the sepsis recognition model through the test set of the divided clinical data set, including accuracy, recall rate, F1 score and specificity, and determines the accuracy of the sepsis recognition model.

[0062] Specifically, the sample ID merging process is as follows: the patient number is set as the clinical blood sample ID of the acute infection patient, the sample ID is used as a key field to match the gene expression data with the clinical data, so that each sample ID can find a corresponding record in the two data sets, after successful matching, the gene expression data and the clinical data are merged into one data set, in the merged data set, the record of each sample contains gene expression data and clinical data, and the gene expression level in the training set is taken as a feature, and the sepsis state is taken as a label.

[0063] S6: predicting the sepsis state by inputting the screened sepsis-related differentially expressed genes into the sepsis recognition model, outputting the sepsis recognition result, and displaying and warning the sepsis recognition result, and transmitting the output result to S7 and S8.

[0064] In this embodiment, S6 inputs the real-time screened sepsis-related differentially expressed gene data into the sepsis recognition model, outputs the sepsis recognition result, and displays and warns the sepsis recognition result in a visual manner, wherein the sepsis recognition result includes whether the patient to be treated is infected with sepsis, the probability of the patient to be treated being infected with sepsis, and the risk degree of the patient to be treated being infected with sepsis.

[0065] S7: obtaining the treatment archive data of the sepsis-recovered patients who have been discharged, extracting the sepsis-recovered patient detection data therefrom, comparing and analyzing the sepsis recognition result with the sepsis-recovered patient detection data, calculating the sepsis recognition result difference factor, evaluating the model prediction effect according to the sepsis recognition result difference factor, and transmitting the analysis result to S8.

[0066] In this embodiment, S7 compares and analyzes the differentially expressed gene set D of the model prediction sepsis recognition result and the differentially expressed gene set U of the recovered patient detection data respectively, and the formula for calculating the sepsis recognition result difference factor is as follows:

[0067]

[0068] By setting a difference threshold, the sepsis recognition result difference factor F is compared with the difference threshold, that is, if F≥φ, it means that the model prediction effect is acceptable, and if F<φ, it means that the model prediction effect is not ideal, and the model needs to be re-optimized and adjusted.

[0069] S8: automatically generating a prediction result monitoring report according to the sepsis recognition result and the sepsis recognition result difference factor, displaying the sepsis recognition result in a visual manner, and generating a warning prompt information to feedback to the detection personnel.

[0070] In this embodiment, the prediction result monitoring report generated by S8 includes sepsis identification result difference factor calculation, model performance index, and visual chart. A data visualization tool is used to create a chart to display the sepsis identification result. According to the sepsis identification result, it is automatically judged whether an early warning prompt information needs to be generated, and the generated early warning prompt information is sent to the detection personnel terminal.

[0071] As Figure 2 The embodiment provides a sepsis risk prompt detection method based on model prediction, and an implementation system thereof. The implementation system comprises a sample data collection module, a gene target data extraction module, a data preliminary screening module, a difference analysis module, a model construction module, a model prediction module, a prediction effect evaluation module, an early warning prompt module and a database. The sample data collection module is connected with the data preliminary screening module, the gene target data extraction module is connected with the data preliminary screening module, the data preliminary screening module is connected with the difference analysis module, the difference analysis module is connected with the model construction module, the difference analysis module is connected with the model prediction module, the model construction module is connected with the model prediction module, the model prediction module is connected with the prediction effect evaluation module, the model prediction module is connected with the early warning prompt module, the prediction effect evaluation module is connected with the early warning prompt module, and the database is connected with all the modules.

[0072] The sample data collection module collects clinical blood sample data of acute infection patients, including sepsis patient sample data and non-sepsis patient sample data, standardizes the collected clinical blood sample data of acute infection patients, obtains gene expression data, and stores the processed sepsis patient sample data into a biological sample library.

[0073] The gene target data extraction module searches for gene targets related to infection progress, performs panel detection on the collected clinical blood sample data of acute infection patients, performs gene set screening and enrichment analysis, screens out gene target data, extracts the gene target data, and records and stores the extracted gene target data into the database.

[0074] The data preliminary screening module matches the processed sepsis patient sample data with the gene target data in the database, and preliminarily screens out gene expression data significantly related to sepsis.

[0075] The difference analysis module labels the gene expression data of the sepsis patient samples and the gene expression data of the non-sepsis patient samples according to the gene expression data of the screened genes significantly related to sepsis, calculates the difference value of the gene expression between the sample data set of the sepsis group and the sample data set of the control group by analyzing and comparing the gene variation between the labeled gene expression data of the sepsis patient samples and the gene expression data of the non-sepsis patient samples, and screens the differential genes related to sepsis.

[0076] The model construction module establishes a machine learning algorithm model according to the screened differential gene data related to sepsis and the clinical data, trains the machine learning model, and obtains a sepsis recognition model.

[0077] The model prediction module predicts the sepsis state by inputting the screened differential expression genes as the differential gene data related to sepsis into the sepsis recognition model, and outputs the sepsis recognition result.

[0078] The prediction effect evaluation module obtains the treatment archive data of the sepsis recovered patients who have been discharged, extracts the sepsis recovered patient detection data therefrom, compares and analyzes the sepsis recognition result with the sepsis recovered patient detection data, calculates the difference factor of the sepsis recognition result, and evaluates the prediction effect of the model according to the difference factor of the sepsis recognition result.

[0079] The early warning prompt module automatically generates a prediction result monitoring report according to the sepsis recognition result and the difference factor of the sepsis recognition result, displays the sepsis recognition result in a visual manner, and generates early warning prompt information to feedback to the detection personnel.

[0080] The database classifies and stores the collected clinical blood sample data of the acute infection patients, establishes a gene target management data table, stores and manages the gene target data, is connected with all the modules for storing the output data of all the modules, and updates and queries the sepsis prediction result in real time.

[0081] Finally, the above is only the preferred embodiment of the present application and is not used to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0082] The above is only the specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any modification, equivalent replacement, improvement, etc. made within the technical range disclosed by the present application can be easily thought of by any person skilled in the art, and shall be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A sepsis risk warning detection method based on model prediction, characterized in that: include: S1: By collecting clinical blood samples from patients with acute infections, including sample data from sepsis patients and samples from non-sepsis patients, the collected clinical blood samples from patients with acute infections are tested and processed to obtain standardized clinical test data. The standardized sample data is then stored in the biobank, and the processing results are transmitted to S3. S2: By searching for gene targets related to infection progression, and performing panel testing on clinical blood sample data collected from acutely infected patients, gene set screening and enrichment analysis are performed to screen out gene target data. The extracted gene target data is then entered and stored in the database, and the results are transmitted to S3. S3: By matching the processed clinical test data of sepsis patient samples with the gene target data in the database, gene expression data that are significantly related to sepsis are initially screened out, and the initial screening results are transmitted to S4; S4: Based on the gene expression data that are significantly associated with sepsis identified in the initial screening, the gene expression data of sepsis patients and non-sepsis patients are labeled. By analyzing and comparing the gene expression data of labeled sepsis patients and non-sepsis patients, the difference in gene expression levels between the sepsis group sample dataset and the control group sample dataset is calculated, thereby screening out differentially expressed genes associated with sepsis. The screening results are then transmitted to S5 and S6. S5: Based on the selected differential gene data related to sepsis and clinical data, establish a machine learning algorithm model, train the machine learning model to obtain a sepsis identification model, and transfer the obtained sepsis identification model to S6. S6: By selecting differentially expressed genes that meet the criteria as differentially expressed genes related to sepsis and inputting them into the sepsis identification model, the sepsis status is predicted, the sepsis identification results are output, and the sepsis identification results are displayed and alerted. The output results are then transmitted to S7 and S8. S7: By acquiring the treatment record data of discharged sepsis recovered patients, extracting the sepsis recovered patients' test data from it, comparing and analyzing the sepsis identification results with the sepsis recovered patients' test data, calculating the difference factor of the sepsis identification results, evaluating the model prediction effect based on the difference factor of the sepsis identification results, and transmitting the analysis results to S8; S8: Automatically generate a prediction result monitoring report based on the sepsis identification results and the difference factors of the sepsis identification results, display the sepsis identification results in a visual way, and generate early warning information to provide feedback to the testing personnel.

2. The sepsis risk warning detection method based on model prediction according to claim 1, characterized in that: S1 involves collecting clinical blood sample data from patients with sepsis and non-sepsis from those seeking treatment for acute infections. This includes data on white blood cell count, C-reactive protein, procalcitonin (PCT), lactate level, and platelet count. A biobank is then established. The collected clinical blood sample data from patients with acute infections is standardized and stored in the biobank. Simultaneously, the sample data from patients with sepsis is labeled as sepsis group sample data, and the sample data from non-sepsis patients is labeled as control group sample data.

3. The sepsis risk warning detection method based on model prediction according to claim 1, characterized in that: S2 identifies gene targets related to infection progression through literature retrieval, bioinformatics analysis, and expert consultation. Based on the obtained gene targets, a panel for detecting gene targets is designed and constructed. The constructed panel is then used to detect clinical blood sample data from patients with acute infection. Based on the panel detection results, a set of genes related to infection progression is screened out. The GeneOntology and Kyoto Encyclopedia of Genes and Genomes databases are used to perform functional enrichment and pathway analysis on the screened gene set. Finally, gene target data highly related to infection progression are extracted from the enrichment analysis results.

4. The sepsis risk warning detection method based on model prediction according to claim 1, characterized in that: The specific steps in S3 involving mapping gene expression data obtained from processed sepsis patient sample data with gene target data in the database are as follows: Step S311: Create a mapping table between gene identifiers and expression values, using gene identifiers as keys; Step S312: Match the expression values ​​in the sample data with the target genes in the database and match the correct gene identifiers. Step S313: Integrate the matched expression data into a unified data framework to obtain the gene expression data that are significantly associated with sepsis from the initial screening.

5. The sepsis risk warning detection method based on model prediction according to claim 1, characterized in that: S4 involves combining the gene expression data of labeled sepsis patient samples into a sepsis group sample dataset X = {x1, x2, ..., x}. i , ..., x n The gene expression data of the labeled non-sepsis patient samples were combined to form the control group sample dataset K = {k1, k2, ..., k}. j , ..., k m Then, statistical methods are used to compare and analyze gene expression data. The formula for calculating the difference in gene expression levels between the sepsis group sample dataset and the control group sample dataset is as follows: in, This represents the average gene expression level in the sepsis group. X represents the average gene expression level in the control group. δ 2 K represents the variance of gene expression levels in the sepsis group. δ 2 The variance of gene expression levels in the control group is represented by n, where n represents the number of samples in the sepsis group and m represents the number of samples in the control group. By comparing the difference threshold ω1 of the first gene expression level and the difference threshold ω2 of the second gene expression level with the difference value t of gene expression level between the sepsis group sample dataset and the control group sample dataset, and ω1>ω2, if the difference value t of gene expression level between the sepsis group sample dataset and the control group sample dataset is greater than the first gene expression threshold ω1 or less than the second gene expression threshold ω2, then the gene has significant differential expression between the sepsis group and the control group. The differentially expressed genes that meet the criteria are selected as differentially expressed genes related to sepsis and transferred to S5.

6. The sepsis risk warning detection method based on model prediction according to claim 1, characterized in that: S5 collects clinical information related to clinical blood samples from patients with acute infections, including age, gender, disease symptoms, and treatment methods. It then merges differentially expressed genetic data related to sepsis with the clinical data according to the clinical blood sample IDs of the patients with acute infections. The merged differentially expressed genetic data and clinical data form a clinical dataset. This clinical dataset is divided into a training set and a test set, with the training set containing significantly more data than the test set. A random forest is used as the machine learning model, and its parameters are adjusted. The model is then ensemble-trained using the training set data to obtain the sepsis identification model, which distinguishes between sepsis patients and non-sepsis patients. The sepsis identification model is then tested using the test set of the divided clinical dataset, including accuracy, recall, F1 score, and specificity, to determine the accuracy of the sepsis identification model.

7. The sepsis risk warning detection method based on model prediction according to claim 1, characterized in that: S6 inputs differentially expressed gene data related to sepsis into the sepsis identification model in real time, and outputs sepsis identification results. The sepsis identification results include whether the patient to be treated is infected with sepsis, the probability of the patient to be treated being infected with sepsis, and the risk level of the patient to be treated being infected with sepsis. The sepsis identification results are displayed and warnings are given in a visual manner.

8. The sepsis risk warning detection method based on model prediction according to claim 1, characterized in that: S7 involves extracting and comparing the differentially expressed gene set D from the model's predicted sepsis identification results with the differentially expressed gene set U from the recovered patient detection data. The formula for calculating the difference factor in sepsis identification results is as follows: By setting a difference threshold, the difference factor F of the sepsis identification results is compared with the difference threshold φ. If F ≥ φ, it means that the model prediction effect is acceptable. If F < φ, it means that the model prediction effect is not ideal and the model needs to be optimized and adjusted.

9. The sepsis risk warning detection method based on model prediction according to claim 1, characterized in that: The prediction result monitoring report generated by S8 includes the calculation of the difference factor of sepsis identification results, model performance indicators, and visualization charts. Data visualization tools are used to create charts to display the sepsis identification results. Based on the sepsis identification results, it automatically determines whether to generate early warning information and sends the generated early warning information to the terminal of the testing personnel.

Images