Prediction method and device for thyroid dysfunction associated with immune checkpoint inhibitors

Abstract

The present invention provides a prediction method and device for abnormal thyroid function associated with immune checkpoint inhibitors, which are applied to the field of data processing technology. The present application obtains a total sample set and target user information; pre-processes the total sample set to generate a total sample set with target prediction factors; processes the total sample set with target prediction factors based on preset processing rules to generate a training set and a test set; trains an initial immune checkpoint inhibitor-related abnormal thyroid function risk prediction model based on the training set and the test set to generate a target immune checkpoint inhibitor-related abnormal thyroid function risk prediction model; processes the target user information based on a preset data integrity detection model to generate user information to be detected; processes the user information to be detected based on the target immune checkpoint inhibitor-related abnormal thyroid function risk prediction model to generate a detection result.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method and device for predicting abnormal thyroid function associated with immune checkpoint inhibitors. Background Art

[0002] Immune Checkpoint Inhibitors (ICIs) are anti-tumor drugs that have made outstanding progress in recent years. They exert anti-tumor effects by regulating the body's own immune response. Their targets include anti-cytotoxic T lymphocyte associated protein-4 (CTLA-4), anti-programmed cell death receptor 1 (PD-1), and anti-programmed cell death-ligand 1 (PD-L1). With the development and use of ICIs, their unique immune-related adverse events (irAEs) have become a major problem in clinical practice.

[0003] ICIs-related endocrinopathy is one of the most common irAEs, among which ICIs-related thyroid dysfunction (ICI-TD) is the most common. ICI-TD mainly includes hyperthyroidism, hypothyroidism and thyroiditis, and the time span of occurrence is long. ICI-TD not only has a high incidence, but also may lead to prolonged treatment intervals or treatment interruptions, increased complications, prolonged hospitalization, decreased quality of life, increased medical burden, etc., which seriously troubles medical staff and patients. Generally speaking, the clinical manifestations of early ICI-TD are complex and lack specificity, which is easy to be ignored and may be masked by cancer symptoms. In addition, it is necessary to distinguish between primary and secondary thyroid dysfunction. Therefore, cancer patients receiving ICIs treatment should undergo ICI-TD risk assessment in advance.

[0004] Although the guidelines give recommendations for the diagnosis and treatment of ICI-TD, the monitoring and management of ICI-TD in clinical practice is still passive. Therefore, it is urgent and beneficial to understand the risk factors for ICI-TD in cancer patients receiving ICIs and to predict the risk in advance to optimize the prevention and treatment of adverse events. However, there is currently a lack of overall understanding and comprehensive evidence of the risk factors for ICI-TD.

[0005] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0006] The purpose of this application is to provide a method and device for predicting immune checkpoint inhibitor-related thyroid dysfunction, which at least to a certain extent overcomes the problems existing in the prior art, by obtaining risk factors for immune checkpoint inhibitor-related thyroid dysfunction in cancer patients treated with immune checkpoint inhibitors, and predicting the risks of corresponding patients in advance based on a pre-trained immune checkpoint inhibitor-related thyroid dysfunction risk prediction model, so as to optimize the prevention and treatment of adverse events.

[0007] Other features and advantages of the present application will become apparent from the following detailed description, or may be learned in part by the practice of the present invention.

[0008] According to one aspect of the present application, a method for predicting abnormal thyroid function associated with immune checkpoint inhibitors is provided, comprising: obtaining a total sample set and target user information, wherein the total sample set includes adult hospitalized users who are pathologically diagnosed with malignant tumors and receive immune checkpoint inhibitor treatment for the first time; preprocessing the total sample set to generate a total sample set with a target prediction factor, wherein the target prediction factor is used to characterize that the user's thyroid function is in an abnormal state; processing the total sample set with the target prediction factor based on preset processing rules to generate a training set and a test set; constructing an initial immune checkpoint inhibitor-related abnormal thyroid function risk prediction model based on a lightweight gradient boosting machine algorithm; training the initial immune checkpoint inhibitor-related abnormal thyroid function risk prediction model based on the training set and the test set to generate a target immune checkpoint inhibitor-related abnormal thyroid function risk prediction model; processing the target user information based on a preset data integrity detection model to generate user information to be detected; processing the user information to be detected based on the target immune checkpoint inhibitor-related abnormal thyroid function risk prediction model to generate a detection result.

[0009] In one embodiment of the present application, the total sample set is preprocessed to generate a training sample set with a target prediction factor, including: obtaining case information and physiological state factor information of the training sample set based on clinical data of different users respectively; processing the case information based on a deep learning model to generate the physiological state factor information and the degree of influence of the case information; processing the degree of influence of the physiological state factor information and the case information based on risk classification rules to generate a risk abnormality indicator; processing the risk abnormality indicator based on a preset feature attribution method to generate a target prediction factor.

[0010] In one embodiment of the present application, the total sample set with the target prediction factor is processed based on a preset processing rule to generate a training set and a test set, including: dividing the training set based on the identification information to generate a number of number class samples; obtaining any number of number class samples in the training set; generating adjacent samples based on the distance between the any number of number class samples and other numbers of number class samples of the same category, wherein the adjacent samples include a preset number of the any number of number class samples; sampling the adjacent samples to generate a preset number of sampled samples; generating multiple data groups based on any minority class sample and each sampled sample, wherein each data group includes a preset number of data samples.

[0011] In one embodiment of the present application, the sampling of the adjacent samples to generate a preset number of sampling samples includes: determining a sampling ratio based on the number of samples of each category in the training set; determining a sampling ratio based on the sampling ratio; and sampling the adjacent samples based on the sampling ratio to generate a preset number of sampling samples.

[0012] In one embodiment of the present application, the target user information is processed based on a preset data integrity detection model to generate user information to be detected, including: the preset data integrity detection model includes multiple interpolation rules; feature extraction is performed on the target user information to determine an original feature library; the original feature library is processed based on a preset time sorting rule to generate missing information of the target user information; the missing information of the target user information is processed based on the multiple interpolation rules to generate the user information to be detected.

[0013] In one embodiment of the present application, the initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model is trained based on the training set and the test set to generate a target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model, including: performing predictive processing on the test set based on a preset algorithm to generate a test set prediction result; processing the test set based on the initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model to generate a test result; if the test result is the same as the test set prediction result, generating a target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model.

[0014] In one embodiment of the present application, the initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model is trained based on the training set and the test set to generate a target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model, and also includes: obtaining an external immune checkpoint inhibitor-related thyroid dysfunction verification model; processing the test set based on the external immune checkpoint inhibitor-related thyroid dysfunction verification model to generate a test result; if the test result is the same as the test result generated by the initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model, then using the initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model as the target immune checkpoint inhibitor-related thyroid function prediction model.

[0015] Another aspect of the present application is a prediction device for immune checkpoint inhibitor-related thyroid dysfunction, characterized in that it includes: an acquisition module, used to acquire a total sample set and target user information, wherein the total sample set includes adult hospitalized users who are pathologically diagnosed with malignant tumors and receive immune checkpoint inhibitor treatment for the first time; a processing module, used to pre-process the total sample set to generate a total sample set with a target prediction factor, wherein the target prediction factor is used to characterize that the user's thyroid function is in an abnormal state; the total sample set with the target prediction factor is processed based on preset processing rules to generate a training set and a test set; an initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model is constructed based on a lightweight gradient boosting machine algorithm; the initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model is trained based on the training set and the test set to generate a target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model; the target user information is processed based on a preset data integrity detection model to generate user information to be detected; the user information to be detected is processed based on the target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model to generate a detection result.

[0016] According to another aspect of the present application, an electronic device is characterized in that it includes: a first processor; and a memory for storing executable instructions of the first processor; wherein the first processor is configured to execute the above-mentioned method for predicting immune checkpoint inhibitor-related thyroid dysfunction by executing the executable instructions.

[0017] According to another aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a second processor, the above-mentioned method for predicting immune checkpoint inhibitor-related thyroid dysfunction is implemented.

[0018] According to another aspect of the present application, a computer program product is provided, comprising a computer program, wherein the computer program implements the above-mentioned method for predicting immune checkpoint inhibitor-related thyroid dysfunction when executed by a third processor.

[0019] The present application provides a prediction method and device for immune checkpoint inhibitor-related thyroid dysfunction, wherein a server obtains a total sample set and target user information, wherein the total sample set includes adult hospitalized users who are pathologically diagnosed with malignant tumors and receive immune checkpoint inhibitor treatment for the first time; the total sample set is preprocessed to generate a total sample set with a target prediction factor, wherein the target prediction factor is used to characterize that the user's thyroid function is in an abnormal state; the total sample set with the target prediction factor is processed based on a preset processing rule to generate a training set and a test set; an initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model is constructed based on a lightweight gradient boosting machine algorithm; the initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model is trained based on the training set and the test set to generate a target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model; the target user information is processed based on a preset data integrity detection model to generate user information to be detected; the user information to be detected is processed based on the target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model to generate a detection result. By obtaining the risk factors for immune checkpoint inhibitor-associated thyroid dysfunction in cancer patients receiving immune checkpoint inhibitor treatment, and predicting the risk of corresponding patients in advance based on a pre-trained immune checkpoint inhibitor-associated thyroid dysfunction risk prediction model, the prevention and treatment of adverse events can be optimized.

[0020] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A flow chart showing a method for predicting abnormal thyroid function associated with immune checkpoint inhibitors provided in one embodiment of the present application;

[0022] Figure 2 A schematic diagram showing the structure of a device for predicting abnormal thyroid function associated with immune checkpoint inhibitors provided in one embodiment of the present application is shown;

[0023] Figure 3 A schematic diagram of the structure of an electronic device provided by an embodiment of the present application is shown;

[0024] Figure 4 A schematic diagram of a storage medium provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0025] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0026] Combine the following Figure 1 To describe the prediction method of thyroid dysfunction associated with immune checkpoint inhibitors according to an exemplary embodiment of the present application. It should be noted that the following application scenarios are only shown for the sake of understanding the spirit and principles of the present application, and the embodiments of the present application are not limited in this regard. On the contrary, the embodiments of the present application can be applied to any applicable scenario.

[0027] In one embodiment, the present application also proposes a method and device for predicting immune checkpoint inhibitor-related thyroid dysfunction. Figure 1 The flowchart of a method for predicting abnormal thyroid function associated with immune checkpoint inhibitors according to an embodiment of the present application is schematically shown. Figure 1 As shown, the method is applied to a server, comprising:

[0028] S101, obtaining the total sample set and target user information.

[0029] In one embodiment, the total sample set includes adult hospitalized users who are pathologically diagnosed with malignant tumors and receive immune checkpoint inhibitors for the first time. If the patient is diagnosed with thyroid dysfunction or takes thyroid-related drugs before receiving immune checkpoint inhibitors, is pregnant or lactating, or has missing thyroid-related indicator data, they are excluded. Target user information, that is, user information for immune checkpoint inhibitor-related thyroid dysfunction testing.

[0030] S102, preprocessing the total sample set to generate a total sample set with target prediction factors.

[0031] In one embodiment, the present invention includes a total of 915 patients, excluding 78 patients with thyroid dysfunction before medication and 216 patients with missing thyroid hormone level information. Finally, 621 patients were included in model construction and validation, of which 414 (66.67%) were male, with a median age of 63 [55, 70] years old. After judging the association of adverse drug reactions, 36 patients had immune checkpoint inhibitor-related thyroid dysfunction. There were 30 positive cases and 446 negative cases in the model development set. There were 6 positive cases and 139 negative cases in the external validation set.

[0032] In another embodiment, case information and physiological status factor information of the total sample set are obtained based on the clinical data of different users respectively; the case information is processed based on the deep learning model to generate the influence degree of the physiological status factor information and the case information; the influence degree of the physiological status factor information and the case information is processed based on the risk classification rules to generate the risk abnormality index; the risk abnormality index is processed based on the preset feature attribution method to generate the target prediction factor.

[0033] For example, to determine whether different users have experienced abnormal thyroid function within a specific period of time, we will obtain various indicators of the users, analyze the factors that cause abnormal thyroid function, and whether it is related to other diseases of the users. Through the above considerations, we can obtain the degree of influence of the two. Based on the risk classification rules, the influence of physiological status factor information and case information is processed to generate risk abnormality indicators, wherein the risk abnormality indicators include but are not limited to demographic characteristics, such as gender, age, number of hospitalizations; vital signs, such as body mass index (BMI), body temperature, pulse; diagnostic information, such as number of diagnoses, tumor type, whether the tumor has been removed, whether the tumor has recurred, whether the tumor has metastasized, hypertension, dyslipidemia, diabetes, heart disease; treatment information, such as the name of immune checkpoint inhibitor drugs, radiotherapy history, chemotherapy history, and combination with other anti-tumor drugs; test results, such as thyroid hormones: free thyroxine (FT4), free triiodothyronine (FT3), total thyroxine (TT4), total triiodothyronine (TT3), thyroid stimulating hormone (TSH);Other types of test results, such as Absolute neutrophil counts (ANC), Alanine aminotransferase (ALT), Aspartate aminotransferase (AST), Creatine kinase (CK), Direct bilirubin (Dbil), Glucose (Glc), Haemoglobin (Hb), Lymphocyte counts (LYM), Mean hemoglobin (MCH), Mean platelet volume (MPV), Monocyte counts (MONO), Packed cell volume (PCV), Platelet count (PLT), Red blood cells (RBC), Serum creatinine (SCr), The percentage of monocytes (The percentage of monocytes (The percentage of monocytes (The percentage of monocytes (The percentage of monocytes (The percentage of monocytes (The percentage of monocytes (The percentage of monocytes (The percentage of monocytes (The percentage of monocytes (The percentage of monocytes (The percentage of monocytes (The percentage of monocytes (The percentage of monocytes (The percentage of monocytes (The percentage of monocytes (The percentage of monocytes (The percentage of monocytes (The percentage of monocytes (The percentage of monocytes (The percentage of monocytes (The percentage of monocytes ( ofmonocytes,MONO%),The percentage oflymphocytes,LYM%),The percentage ofbasophils,BASO%),The percentage ofeosinophils,EOS%),The percentage ofneutrophils,NEUT%),Total bilirubin,Total cholesterol,TC),Triglyceride,TG,Urea,Uric acid,WBC.;

[0034] Among the above-mentioned risk abnormality indicators, the most important variables were screened by the SHapleyAdditive exPlanations (SHAP) method, and finally the percentage of eosinophils, uric acid, free triiodothyronine, number of diagnoses, urea, free thyroxine, thyroid-stimulating hormone, body mass index, total thyroxine, percentage of lymphocytes, total cholesterol, aspartate aminotransferase, absolute value of lymphocytes, mean platelet volume, percentage of monocytes, triglycerides and pulse rate were used as target predictors to develop a risk prediction model for thyroid dysfunction associated with immune checkpoint inhibitors. Among them, the target predictor is used to characterize that the user's thyroid function is in an abnormal state.

[0035] In another embodiment, the present invention has found some clinical features that are significantly associated with ICI-TD, and these features need to be considered in cancer patients treated with ICIs. In terms of demographic characteristics and lifestyle, BMI should be considered. In addition, the number of diagnoses is also included in the prediction model. Clinicians should pay attention to baseline factors related to thyroid function, including FT3, FT4, TT4 and TSH. In addition, laboratory test-related factors may also affect the risk of ICI-TD, including eosinophil percentage, uric acid, urea, lymphocyte percentage, total cholesterol, aspartate aminotransferase, lymphocyte count, mean platelet volume, monocyte percentage and triglycerides.

[0036] BMI plays an important role in the final prediction model. The applicant's study found that patients with higher BMI were more likely to develop ICI-TD. Compared with thinner patients, obese patients developed thyrotoxicosis earlier, suggesting a complex interaction between obesity and immune response in cancer patients treated with ICIs. Overweight or obese patients have high serum FT3 concentrations, high FT3 / FT4 ratios, and low FT4 concentrations, while underweight patients have high serum FT4 concentrations and low FT3 concentrations. In addition, thyroid function may be a factor affecting weight and obesity complications. TSH has been reported to be positively correlated with BMI. In normal-weight euthyroid people, changes in serum FT4 concentrations may lead to changes in body weight. The clinical benefits of ICIs in the high BMI population may be limited to patients who receive personalized drug therapy based on weight. Considering efficacy and safety, future studies may need to further explore the individualization of ICIs dosage.

[0037] TSH, FT4, FT3, and TT4 at baseline are associated with the occurrence of ICI-TD. Some studies have reported that baseline ATA can be used as a biomarker to predict the occurrence of ICI-TD. The applicant's investigation found that ATA positivity is associated with significant thyroid dysfunction. For clinical practice, researchers recommend stratifying the risk of ICI-TD based on baseline thyroid-related hormones and ATA levels in order to promote individualized safety management of ICIs.

[0038] In the model, the SHAP value of the percentage of eosinophils was the highest. Eosinophils are a type of white blood cell that are classified as granulocytes because they have large cytoplasmic secretory granules. There is evidence that eosinophils can regulate the number of T cells and B cells, increase the survival of granulocytes in inflamed tissues, and propagate Th1 and Th2 CD4+ immune responses. In addition, cytokines secreted by eosinophils can affect the balance of Th1 and Th2CD4+T cell populations by interacting with CD80 / CD86 and CTLA-4 surface antigens. In the treatment of melanoma, peripheral blood eosinophil counts are associated with ICIs efficacy outcomes. Eosinophils can regulate the tumor microenvironment and play a direct role in the therapeutic response to ICIs. Future studies should focus on how eosinophils affect tumor-related prognosis and whether they can predict ICI-TD.

[0039] In the model, the SHAP value of uric acid ranked second. According to the applicant's research, reduced uric acid levels can be used as a risk factor for predicting thyroid dysfunction in Chinese people. The hypothesis is that a moderate increase in plasma uric acid levels may affect thyroid hormones due to its powerful antioxidant effect. It is worth noting that there are differences in the appropriate uric acid levels between men and women. Similarly, a systematic review and meta-analysis showed that subclinical thyroid dysfunction is associated with an increased incidence of hyperuricemia.

[0040] S103, processing the total sample set with the target prediction factor based on a preset processing rule to generate a training set and a test set.

[0041] In one embodiment, patient data of ICIs (immune checkpoint inhibitors) is used. The dataset is randomly divided into a training set and a test set (80% and 20%, respectively). Five-fold cross-validation resampling is used to divide the data into five subsets. In the first iteration, the first group is used as a test set and the remaining groups are used as training sets. This process is repeated until each of the five subsets is used as a test set.

[0042] The case data were divided into a development set (75%-80%) and an external validation set (20%-25%) according to the time of consultation. The development set data were randomly divided into a training set (80%) and a test set (20%). In the training set data, mixed sampling was used to adjust the proportion of positive cases. There were 54 candidate variables, involving basic information, past medical history, diagnosis, treatment plan, test information, etc. In the training set data, mixed sampling was used to adjust the proportion of positive cases. There were 54 candidate variables, involving basic information, past medical history, diagnosis, treatment plan, test information, etc. Logistic regression, linear discriminant analysis, Bayesian network, K nearest neighbor, decision tree, random forest (RF), support vector machine, gradient boosting decision tree (GBDT), extreme gradient boosting (eXtreme Gradient Boosting, XGBoost), Light Gradient Boosting Machine (LightGBM) and neural network methods were used to establish the prediction model. Five-fold cross-validation resampling was used. The model discrimination was evaluated by the area under the receiver operating characteristic curve (AUC), sensitivity, specificity, positive predictive value, negative predictive value, and accuracy. The model calibration was evaluated by the Brier score and calibration curve.

[0043] In another embodiment, the training set is divided based on the identification information to generate a number of number class samples; any number of number class samples in the training set are obtained; adjacent samples are generated based on the distance between any number of number class samples and other number of number class samples of the same category, wherein the adjacent samples include a preset number of any number of number class samples; adjacent samples are sampled to generate a preset number of sampled samples; multiple groups of data groups are generated based on any minority class sample and each sampled sample, wherein each group of data groups includes a preset number of data samples. The distance between any number of number class samples and each other number of number class samples of the same category can be Euclidean distance or other cosine distance, which is not specifically limited here. All other number of number class samples can be arranged in order from near to far distance from other number of number class samples, and then the data in the first few positions of the arrangement result are selected as adjacent samples. The purpose of sampling is to increase the number of minority class samples to achieve a balance between the number of minority class samples and majority class samples in the initial sample set. Here, oversampling can be achieved by a sampling algorithm, which is not specifically limited here. After oversampling, multiple data groups can be determined based on the oversampling results and the training set. At this time, the number of descriptive data samples corresponding to each attribute category label in the data group is equivalent, and there are no large number of class samples or large number of class samples. In this embodiment, the training set is preliminarily divided first, and then the large number of class samples are merged with other large number of class samples.

[0044] In another embodiment, the sampling ratio is determined based on the number of samples of each category in the training set; the sampling ratio is determined based on the sampling ratio; and the adjacent samples are sampled based on the sampling ratio to generate a preset number of sampled samples. First, obtain which data groups are included in the training set, confirm how many sample types there are in each data group, and determine the sampling ratio according to the number of sample types in each data group, and then use the sampling ratio to determine the sampling ratio. The sampling ratio can be set as needed, for example, it can be 50% or other ratios, which are not specifically limited here. Finally, the sampling ratio is used to sample a preset number of adjacent samples to obtain a specified number of sampled samples, and all other samples of smaller number of categories can be arranged in order from near to far from any sample of smaller number of categories, and then select the samples of the preset number ranked first in the arrangement result as the neighboring samples, and the preset number can be set according to the actual situation, which is not specifically limited here. Based on any sample of smaller number of categories and each sampled sample, a new sample is generated and used as a new data group, so that the balance effect of the data group with smaller number and the data group with larger number can be further improved.

[0045] S104, construct an initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model based on a lightweight gradient boosting machine algorithm.

[0046] In one embodiment, the LightGBM (AUC = 0.96), XGBoost (AUC = 0.95), RF (AUC = 0.94) and GBDT (AUC = 0.92) models performed best in the training set. Combined with comprehensive evaluation of evaluation indicators such as sensitivity, specificity, and accuracy, the prediction model based on the LightGBM algorithm has the best performance and the strongest clinical applicability. LightGBM uses weak classifiers (decision trees) for iterative training to obtain the optimal model.

[0047] In another embodiment, an initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model can also be constructed based on the following rules. They are:

[0048] An initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model was constructed based on the logistic regression rule. The model was trained using the training set to obtain the following parameters, including AUC of 0.44, sensitivity of 0.00, specificity of 1.00, positive predictive value of 0.00, negative predictive value of 0.85, accuracy of 0.85, and Brier score of 0.15.

[0049] An initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model was constructed based on the linear discriminant analysis rule. The model was trained using the training set to obtain the following parameters, including AUC of 0.72, sensitivity of 0.07, specificity of 0.99, positive predictive value of 0.60, negative predictive value of 0.86, accuracy of 0.85, and Brier score of 0.12.

[0050] An initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model was constructed based on the naive Bayes rule. The model was trained using the training set to obtain the following parameters, including an AUC of 0.73, a sensitivity of 0.05, a specificity of 0.99, a positive predictive value of 0.50, a negative predictive value of 0.85, an accuracy of 0.85, and a Brier score of 0.13.

[0051] An initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model was constructed based on the K nearest neighbor rule. The model was trained using the training set to obtain the following parameters, including AUC of 0.88, sensitivity of 0.77, specificity of 1.00, positive predictive value of 0.74, negative predictive value of 0.95, accuracy of 0.88, and Brier score of 0.09.

[0052] An initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model was constructed based on the decision tree rule. The model was trained using the training set to obtain the following parameters, including AUC of 0.92, sensitivity of 0.75, specificity of 1.00, positive predictive value of 0.91, negative predictive value of 0.95, accuracy of 0.93, and Brier score of 0.06.

[0053] An initial immune checkpoint inhibitor-related thyroid dysfunction prediction model was constructed based on the random forest rule. The model was trained using the training set to obtain the following parameters, including an AUC of 0.94, a sensitivity of 0.82, a specificity of 1.00, a positive predictive value of 0.95, a negative predictive value of 0.92, an accuracy of 0.93, and a Brier score of 0.06.

[0054] An initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model was constructed based on the support vector machine rule. The model was trained using the training set to obtain the following parameters, including an AUC of 0.94, a sensitivity of 0.84, a specificity of 0.97, a positive predictive value of 0.92, a negative predictive value of 1.00, an accuracy of 0.93, and a Brier score of 0.06.

[0055] An initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model was constructed based on the gradient boosting rule. The model was trained using the training set to obtain the following parameters, including AUC of 0.94, sensitivity of 0.82, specificity of 1.00, positive predictive value of 0.92, negative predictive value of 0.92, accuracy of 0.92, and Brier score of 0.08.

[0056] An initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model was constructed based on the extreme gradient boosting rule. The model was trained using the training set to obtain the following parameters, including an AUC of 0.95, a sensitivity of 0.87, a specificity of 1.00, a positive predictive value of 0.97, a negative predictive value of 0.95, an accuracy of 0.95, and a Brier score of 0.05.

[0057] An initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model was constructed based on neural network rules. The model was trained using the training set to obtain the following parameters, including AUC of 0.92, sensitivity of 0.84, specificity of 0.99, positive predictive value of 0.92, negative predictive value of 0.98, accuracy of 0.95, and Brier score of 0.06.

[0058] S105, training the initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model based on the training set and the test set to generate a target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model.

[0059] In one embodiment, the test set is predicted and processed based on a preset algorithm to generate a test set prediction result; the test set is processed based on the initial thyroid dysfunction prediction model to generate a test result; if the test result is the same as the test set prediction result, a target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model is generated. Specifically, the data classification results corresponding to the training data set are obtained, because the classification results of each training data set can be determined in advance, and the data can be directly obtained from the outside world. Compare the prediction results with the data classification results to determine the first comparison result; compare the test set prediction results with the data classification results to determine the second comparison result; when judging whether the second comparison result meets the preset requirements with the first comparison result, when the preset requirements are met, it indicates that the result of the current prediction model is more accurate. At this time, the current immune checkpoint inhibitor-related thyroid dysfunction risk prediction model can be used as the target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model.

[0060] S106: Process the target user information based on a preset data integrity detection model to generate user information to be detected.

[0061] In one embodiment, the preset data integrity detection model includes multiple imputation rules; multiple imputation rules (MI) are applied to solve the problem of missing data. MI is a powerful statistical technique that can replace each missing value with a set of credible values ​​to generate multiple complete data sets. Each data set is then analyzed and the results are summarized to obtain unbiased estimates and corrected standard errors. Through MI, the applicant can reduce the impact of missing data on the analysis and provide more reliable and valid results.

[0062] Baseline characteristics are presented as mean ± SD for normally distributed continuous variables and as median and interquartile range for skewed continuous variables. Categorical variables were summarized as number and percentage. To compare baseline characteristics between groups, Student t test or Mann-Whitney U test was used. The distribution of categorical variables was compared using chi-square test and Fisher's exact test. Differences were considered statistically significant when P value was less than 0.05. All statistical analyses were performed using Python 3.6.7 (Python Software Foundation, Wilmington, DE, United States), R 4.2.0 (https: / / www.r-project.org / ), and Microsoft Office 365 (Microsoft Corp., Redmond, WA, United States).

[0063] In another embodiment, feature extraction is performed on the target user information to determine the original feature library; the original feature library is processed based on a preset time sorting rule to generate missing information of the target user information; the missing information of the target user information is processed based on a multiple interpolation rule to generate the user information to be detected. Multiple interpolation algorithm interpolation supplements data: The multiple interpolation algorithm supplements data by performing distribution operations on the labeled data, generally including the following parts: 1) The Mice function first starts with a data set containing missing data and returns an object containing multiple (default is 5) complete data sets. Each complete data set is generated by interpolating the missing data in the original data frame. Since interpolation has a random component, each complete data set is slightly different; 2) The with function can apply a statistical model (such as a linear model or a generalized linear model) to each complete data set in turn; 3) The pool function integrates these individual analysis results into a set of results. The standard error and p-value of the final model will accurately reflect the uncertainty caused by missing values ​​and multiple interpolation. The target user information and the data set supplemented by the multiple interpolation (MICE) algorithm need to be checked for statistical differences in data validity to ensure the validity of the data. The two sets of data will be subjected to one-way ANOVA to calculate the significant difference between the two sets of data. It is necessary to ensure that there is no significant difference between the two sets of data. If there is a significant difference between the two sets of data after verification, it is necessary to adjust the number of linear fitting / linear programming models in the with function to ensure that there is no significant difference between the processed data and the original data, so that the processed data remains accurate and effective.

[0064] S107, processing the user information to be detected based on the target thyroid dysfunction prediction model to generate a detection result.

[0065] In one embodiment, the test result includes a risk level. According to the risk level classification method, a risk score less than 0.01 is low risk, between 0.01 and 0.1 is medium risk, between 0.1 and 0.5 is high risk, and greater than or equal to 0.5 is very high risk. For example, a Chinese male aged 63 years old with a body mass index of 26.0 kg / m 2 , and its information has been fully input into the target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model. The final test result is that the risk score is 0.689 and the risk level is "extremely high".

[0066] In this application, a server obtains a training sample set and target user information, wherein the training sample set includes adult hospitalized users who are pathologically diagnosed with malignant tumors and receive immune checkpoint inhibitor treatment for the first time; the total sample set is preprocessed to generate a total sample set with a target prediction factor, wherein the target prediction factor is used to characterize that the user's thyroid function is in an abnormal state; the total sample set with the target prediction factor is processed based on preset processing rules to generate a training set and a test set; an initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model is constructed based on a lightweight gradient boosting machine algorithm; the initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model is trained based on the training set and the test set to generate a target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model; the target user information is processed based on a preset data integrity detection model to generate user information to be detected; the user information to be detected is processed based on the target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model to generate a detection result. By obtaining the risk factors for immune checkpoint inhibitor-associated thyroid dysfunction in cancer patients receiving immune checkpoint inhibitor treatment, and predicting the risk of corresponding patients in advance based on a pre-trained immune checkpoint inhibitor-associated thyroid dysfunction risk prediction model, the prevention and treatment of adverse events can be optimized.

[0067] Optionally, in another embodiment of the method of the present application, the initial immune checkpoint inhibitor-associated thyroid dysfunction risk prediction model is trained based on the training set and the test set to generate a target immune checkpoint inhibitor-associated thyroid dysfunction risk prediction model, further comprising:

[0068] Obtain external validation models of immune checkpoint inhibitor-related thyroid dysfunction;

[0069] Processing the test set based on the external immune checkpoint inhibitor-related thyroid dysfunction validation model to generate a test result;

[0070] If the test result is the same as the test result generated by the initial thyroid dysfunction prediction model, the initial thyroid dysfunction prediction model is used as the target thyroid function prediction model.

[0071] In one embodiment, the model based on XGBoost has the highest AUC, specificity, positive predictive value and accuracy, while the model based on LightGBM has the highest sensitivity and negative predictive value. The model using XGBoost, LightGBM and gradient boosting technology has the lowest Brier score (0.04). Taking into account the model evaluation results, the model based on LightGBM has the best prediction performance.

[0072] In another embodiment, the test set is processed based on several external immune checkpoint inhibitor-related thyroid dysfunction validation models. If the error between the test results obtained by the above-mentioned external immune checkpoint inhibitor-related thyroid dysfunction validation models and the test results based on the immune checkpoint inhibitor-related thyroid dysfunction risk prediction model generated based on LightGBM is within a preset range, then the results obtained by the immune checkpoint inhibitor-related thyroid dysfunction risk prediction model generated based on LightGBM are confirmed to be accurate.

[0073] Optionally, in another embodiment of the above method of the present application, the total sample set with the target prediction factor is processed based on a preset processing rule to generate a training set and a test set, and further includes:

[0074] Labeling the samples in the training sample set to obtain a first label for each sample;

[0075] Extracting a single sample from the training sample set, forming a first training set according to the sample and the corresponding first label, and extracting multiple times to obtain multiple first training sets;

[0076] Extract two samples from the training sample set, generate second labels for the two samples according to first labels respectively corresponding to the two samples, form a second training set according to the two samples and the corresponding second labels, and extract multiple times to obtain multiple second training sets;

[0077] A plurality of training sets are constructed according to the plurality of first training sets and the second training set.

[0078] In one implementation, a preset number of feature information can be randomly extracted with replacement from the multidimensional feature information of each sample each time to form a corresponding sub-sample. Multiple sub-samples constitute a training set. After multiple extractions, multiple training sets are constructed. The preset number can be customized according to actual needs. The training set can be divided into two parts. One part is a single sample x, and it is marked whether the target application will use it next. If so, it can be marked as 1, otherwise it can be marked as 0. The form can be (xi, yi), where yi∈{0, 1}. The other part is a triple, that is, by sampling two samples (xi, xj), if the two sample labels are consistent, they are recorded as 1, and if the labels are inconsistent, they are recorded as -1. The form is (xi, xj, γ), where γ∈{1, -1}.

[0079] Optionally, in another embodiment of the above method of the present application, the processing of the training sample set with the target prediction factor based on a preset processing rule to generate a training set and a test set further includes:

[0080] Performing feature extraction on the training data set with the target prediction factor to determine an original feature library;

[0081] Processing the original feature library based on a neural network model to generate forged feature data, wherein the forged feature data is different from the feature data in the original feature library;

[0082] The original feature library is divided to generate a training data set and a test data set, wherein the number of the training data set and the test data set is at least one, and the test data set includes forged feature data.

[0083] In one embodiment, forged feature data is generated to prepare for subsequent testing of the initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model, that is, the forged feature data is actually abnormal data, but if the initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model does not identify the current data as abnormal, it indicates that the detection results of the current model are inaccurate.

[0084] By applying the above technical solution, the server obtains the total sample set and target user information, wherein the total sample set includes adult hospitalized users who are pathologically diagnosed with malignant tumors and receive immune checkpoint inhibitors for the first time; the case information and physiological state factor information of the total sample set are obtained based on the clinical data of different users; the case information is processed based on the deep learning model to generate the influence degree of the physiological state factor information and the case information; the influence degree of the physiological state factor information and the case information is processed based on the risk classification rules to generate a risk abnormality indicator; the risk abnormality indicator is processed based on a preset feature attribution method to generate a target prediction factor, wherein the target prediction factor is used to characterize that the user's thyroid function is in an abnormal state.

[0085] The training set is divided based on identification information to generate a number of number class samples; any number of number class samples in the training set are obtained; adjacent samples are generated based on the distance between any number of number class samples and other number of number class samples of the same category, wherein the adjacent samples include a preset number of any number of number class samples; the sampling ratio is determined based on the number of each number class sample in the training set; the sampling ratio is determined based on the sampling ratio; the adjacent samples are sampled based on the sampling ratio to generate a preset number of sampling samples; multiple groups of data are generated based on any minority class sample and each sampling sample, wherein each group of data contains a preset number of data samples; an initial thyroid dysfunction prediction model is constructed based on a lightweight gradient boosting machine algorithm; the initial thyroid dysfunction prediction model is trained based on a training set and a test set to generate a target thyroid dysfunction prediction model.

[0086] The samples in the training sample set are labeled to obtain the first label of each sample; a single sample is extracted from the training sample set, a first training set is formed according to the sample and the corresponding first label, and multiple extractions are performed to obtain multiple first training sets; two samples are extracted from the training sample set, the second labels of the two samples are generated according to the first labels corresponding to the two samples, a second training set is formed according to the two samples and the corresponding second labels, and multiple extractions are performed to obtain multiple second training sets; multiple training sets are constructed according to multiple first training sets and second training sets. The preset data integrity detection model includes multiple interpolation rules; feature extraction of target user information is performed to determine the original feature library; the original feature library is processed based on the preset time sorting rules to generate missing information of target user information; the missing information of target user information is processed based on the multiple interpolation rules to generate user information to be detected; the test set is predicted based on the preset algorithm to generate the test set prediction result; the test set is processed based on the initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model to generate the test result; if the test result is the same as the test set prediction result, the target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model is generated.

[0087] Obtain an external immune checkpoint inhibitor-related thyroid dysfunction validation model; process the test set based on the external immune checkpoint inhibitor-related thyroid dysfunction validation model to generate test results; if the test results are the same as the test results generated by the initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model, the initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model is used as the target immune checkpoint inhibitor-related thyroid dysfunction prediction model. Obtain an external immune checkpoint inhibitor-related thyroid dysfunction validation model; process the test set based on the external immune checkpoint inhibitor-related thyroid dysfunction validation model to generate test results; if the test results are the same as the test results generated by the initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model, the initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model is used as the target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model. By obtaining the risk factors for immune checkpoint inhibitor-related thyroid dysfunction in cancer patients receiving immune checkpoint inhibitor treatment, and predicting the risks of the corresponding patients in advance based on the pre-trained immune checkpoint inhibitor-related thyroid dysfunction risk prediction model, the prevention and treatment of adverse events can be optimized.

[0088] In one embodiment, if Figure 2 As shown, the present application also provides a prediction device for immune checkpoint inhibitor-related thyroid dysfunction, comprising:

[0089] An acquisition module 201 is used to acquire a total sample set and target user information, wherein the total sample set includes adult hospitalized users who are pathologically diagnosed with malignant tumors and receive immune checkpoint inhibitor treatment for the first time;

[0090] The processing module 202 is used to preprocess the total sample set to generate a total sample set with a target prediction factor, wherein the target prediction factor is used to characterize that the user's thyroid function is in an abnormal state; based on a preset processing rule, the training sample set with the target prediction factor is processed to generate a training set and a test set; based on a lightweight gradient boosting algorithm, an initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model is constructed; based on the training set and the test set, the initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model is trained to generate a target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model; based on a preset data integrity detection model, the target user information is processed to generate user information to be detected; based on the target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model, the user information to be detected is processed to generate a detection result.

[0091] In another embodiment of the present application, the processing module 202 is configured to pre-process the total sample set to generate a total sample set with a target prediction factor, including:

[0092] Obtain case information and physiological status factor information of the total sample set based on clinical data of different users;

[0093] Processing the case information based on a deep learning model to generate the physiological state factor information and the degree of influence of the case information;

[0094] Processing the influence degree of the physiological state factor information and the case information based on the risk classification rules to generate a risk abnormality indicator;

[0095] The risk abnormality indicator is processed based on a preset feature attribution method to generate a target prediction factor.

[0096] In another embodiment of the present application, the processing module 202 is configured to process the total sample set with the target prediction factor based on the preset processing rule to generate a training set and a test set, including:

[0097] Dividing the training set based on the identification information to generate a number of samples of different categories;

[0098] Obtain any number of samples of several categories in the training set;

[0099] Generate adjacent samples based on the distance between the any number of number-class samples and other number-class samples of the same category, wherein the adjacent samples include a preset number of the any number of number-class samples;

[0100] Sampling the adjacent samples to generate a preset number of sampling samples;

[0101] Based on any minority class sample and each sampling sample, multiple data groups are generated, wherein each data group includes a preset number of data samples.

[0102] In another embodiment of the present application, the processing module 202 is configured to sample the adjacent samples to generate a preset number of sample samples, including:

[0103] Determining a sampling ratio based on the number of samples of each category in the training set;

[0104] Based on the sampling ratio, determining a sampling ratio;

[0105] The adjacent samples are sampled based on the sampling ratio to generate a preset number of sampling samples.

[0106] In another embodiment of the present application, the processing module 202 is configured to process the target user information based on the preset data integrity detection model to generate the user information to be detected, including:

[0107] The preset data integrity detection model includes multiple imputation rules;

[0108] Extracting features from the target user information to determine an original feature library;

[0109] Processing the original feature library based on a preset time sorting rule to generate missing information of target user information;

[0110] The missing information of the target user information is processed based on the multiple interpolation rules to generate the user information to be detected.

[0111] In another embodiment of the present application, the processing module 202 is configured to train the initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model based on the training set and the test set to generate a target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model, including:

[0112] Performing prediction processing on the test set based on a preset algorithm to generate a test set prediction result;

[0113] Processing the test set based on the initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model to generate a test result;

[0114] If the test result is the same as the prediction result of the test set, a target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model is generated.

[0115] In another embodiment of the present application, the processing module 202 is configured to train the initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model based on the training set and the test set to generate a target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model, and further includes:

[0116] Obtain external validation models of immune checkpoint inhibitor-related thyroid dysfunction;

[0117] Processing the test set based on the external immune checkpoint inhibitor-related thyroid dysfunction validation model to generate a test result;

[0118] If the test result is the same as the test result generated by the initial immune checkpoint inhibitor-associated thyroid dysfunction risk prediction model, the initial immune checkpoint inhibitor-associated thyroid dysfunction risk prediction model is used as the target immune checkpoint inhibitor-associated thyroid function prediction model.

[0119] The present application embodiment provides an electronic device, such as Figure 3 As shown, the electronic device 3 includes a first processor 300, a memory 301, a bus 302 and a communication interface 303, and the first processor 300, the communication interface 303 and the memory 301 are connected via the bus 302; the memory 301 stores a computer program that can be run on the first processor 300, and when the first processor 300 runs the computer program, the prediction method for abnormal thyroid function associated with immune checkpoint inhibitors provided in any of the aforementioned embodiments of the present application is executed.

[0120] Among them, the memory 301 may include a high-speed random access memory (RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk storage. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 303 (which may be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. can be used. Bus 302 can be an ISA bus, a PCI bus, or an EISA bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. Among them, the memory 301 is used to store programs, and the first processor 300 executes the program after receiving the execution instruction. The prediction method of thyroid dysfunction associated with immune checkpoint inhibitors disclosed in any embodiment of the above-mentioned embodiment of the present application can be applied to the first processor 300, or implemented by the first processor 300.

[0121] The first processor 300 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the first processor 300. The above-mentioned first processor 300 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a readily available programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present application can be embodied as a hardware decoding processor to be executed, or a combination of hardware and software modules in the decoding processor to be executed. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 301, and the first processor 300 reads the information in the memory 301 and completes the steps of the above method in combination with its hardware.

[0122] The electronic device provided in the above-mentioned embodiments of the present application and the method for predicting abnormal thyroid function associated with immune checkpoint inhibitors provided in the embodiments of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0123] The present application embodiment provides a computer-readable storage medium, such as Figure 4 As shown, the computer-readable storage medium 401 stores a computer program, and when the computer program is read and executed by the second processor 402, the aforementioned method for predicting abnormal thyroid function associated with immune checkpoint inhibitors is implemented.

[0124] The technical solution of the embodiment of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling an electronic device (which may be an air conditioner, a refrigeration device, a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the method described in the embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a ROM, a RAM, a magnetic disk, or an optical disk.

[0125] The computer-readable storage medium provided in the above-mentioned embodiments of the present application and the method for predicting abnormal thyroid function associated with immune checkpoint inhibitors provided in the embodiments of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0126] An embodiment of the present application provides a computer program product, including a computer program, wherein the computer program is executed by a third processor to implement the method described above.

[0127] The computer program product provided in the above-mentioned embodiments of the present application and the method for predicting abnormal thyroid function associated with immune checkpoint inhibitors provided in the embodiments of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0128] It should be noted that, in this application, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or still includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise one..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0129] Each embodiment in this application is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the prediction method, electronic device, electronic device, and readable storage medium embodiment for evaluating immune checkpoint inhibitor-related thyroid dysfunction, since it is basically similar to the above-mentioned immune checkpoint inhibitor-related thyroid dysfunction prediction method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the above-mentioned immune checkpoint inhibitor-related thyroid dysfunction prediction method embodiment.

[0130] Although the present application is disclosed as above, the present application is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present application. Therefore, the protection scope of the present application shall be subject to the scope defined by the claims.

Claims

1. A method for predicting thyroid dysfunction associated with immune checkpoint inhibitors, characterized in that: include: Obtaining the total sample set and target user information, wherein the total sample set includes adult hospitalized users who are pathologically diagnosed with malignant tumors and receive immune checkpoint inhibitor treatment for the first time; Preprocessing the total sample set to generate a training sample set with a target prediction factor, wherein the target prediction factor is used to characterize that the user's thyroid function is in an abnormal state; Processing the total sample set with the target prediction factor based on preset processing rules to generate a training set and a test set; An initial risk prediction model for thyroid dysfunction associated with immune checkpoint inhibitors was constructed based on a lightweight gradient boosting machine algorithm; Training the initial immune checkpoint inhibitor-associated thyroid dysfunction risk prediction model based on the training set and the test set to generate a target immune checkpoint inhibitor-associated thyroid dysfunction risk prediction model; Processing the target user information based on a preset data integrity detection model to generate user information to be detected; Processing the user information to be tested based on the target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model to generate a test result; The preprocessing of the total sample set to generate a total sample set with a target prediction factor includes: obtaining case information and physiological state factor information of the total sample set based on clinical data of different users respectively; processing the case information based on a deep learning model to generate the physiological state factor information and the influence degree of the case information; processing the influence degree of the physiological state factor information and the case information based on a risk classification rule to generate a risk abnormality index; processing the risk abnormality index based on a preset feature attribution method to generate a target prediction factor; The method processes the total sample set with the target prediction factor based on the preset processing rules to generate a training set and a test set, including: dividing the training set based on the identification information to generate a number of number class samples; obtaining any number of number class samples in the training set; generating adjacent samples based on the distance between the any number of number class samples and other number of number class samples of the same category, wherein the adjacent samples include a preset number of the any number of number class samples; sampling the adjacent samples to generate a preset number of sampled samples; generating multiple data groups based on any minority class sample and each sampled sample, wherein each data group includes a preset number of data samples; The sampling of the adjacent samples to generate a preset number of sampling samples includes: determining a sampling ratio based on the number of samples of each category in the training set; determining a sampling ratio based on the sampling ratio; and sampling the adjacent samples based on the sampling ratio to generate a preset number of sampling samples.

2. The method according to claim 1, characterized in that The processing of the target user information based on a preset data integrity detection model to generate user information to be detected includes: The preset data integrity detection model includes multiple imputation rules; Extracting features from the target user information to determine an original feature library; Processing the original feature library based on a preset time sorting rule to generate missing information of target user information; The missing information of the target user information is processed based on the multiple interpolation rules to generate the user information to be detected.

3. The method according to claim 1, characterized in that The step of training the initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model based on the training set and the test set to generate a target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model comprises: Performing prediction processing on the test set based on a preset algorithm to generate a test set prediction result; Processing the test set based on the initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model to generate a test result; If the test result is the same as the prediction result of the test set, a target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model is generated.

4. The method according to claim 3, characterized in that The training of the initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model based on the training set and the test set to generate a target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model further includes: Obtain external validation models of immune checkpoint inhibitor-related thyroid dysfunction; Processing the test set based on the external thyroid dysfunction verification model to generate a test result; If the test result is the same as the test result generated by the initial immune checkpoint inhibitor-associated thyroid dysfunction risk prediction model, the initial immune checkpoint inhibitor-associated thyroid dysfunction risk prediction model is used as the target immune checkpoint inhibitor-associated thyroid function prediction model.

5. A device for predicting abnormal thyroid function, characterized in that: The device comprises: An acquisition module, used to acquire the total sample set and target user information, wherein the total sample set includes adult hospitalized users who are pathologically diagnosed with malignant tumors and receive immune checkpoint inhibitor treatment for the first time; A processing module, used to preprocess the total sample set to generate a total sample set with a target prediction factor, wherein the target prediction factor is used to characterize that the user's thyroid function is in an abnormal state; based on a preset processing rule, the total sample set with the target prediction factor is processed to generate a training set and a test set; based on a lightweight gradient boosting algorithm, an initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model is constructed; based on the training set and the test set, the initial immune checkpoint inhibitor-related thyroid dysfunction risk prediction model is trained to generate a target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model; based on a preset data integrity detection model, the target user information is processed to generate user information to be detected; based on the target immune checkpoint inhibitor-related thyroid dysfunction risk prediction model, the user information to be detected is processed to generate a detection result; The preprocessing of the total sample set to generate a total sample set with a target prediction factor includes: obtaining case information and physiological state factor information of the total sample set based on clinical data of different users respectively; processing the case information based on a deep learning model to generate the physiological state factor information and the influence degree of the case information; processing the influence degree of the physiological state factor information and the case information based on a risk classification rule to generate a risk abnormality index; processing the risk abnormality index based on a preset feature attribution method to generate a target prediction factor; The method processes the total sample set with the target prediction factor based on the preset processing rules to generate a training set and a test set, including: dividing the training set based on the identification information to generate a number of number class samples; obtaining any number of number class samples in the training set; generating adjacent samples based on the distance between the any number of number class samples and other number of number class samples of the same category, wherein the adjacent samples include a preset number of the any number of number class samples; sampling the adjacent samples to generate a preset number of sampled samples; generating multiple data groups based on any minority class sample and each sampled sample, wherein each data group includes a preset number of data samples; The sampling of the adjacent samples to generate a preset number of sampling samples includes: determining a sampling ratio based on the number of samples of each category in the training set; determining a sampling ratio based on the sampling ratio; and sampling the adjacent samples based on the sampling ratio to generate a preset number of sampling samples.

6. An electronic device, characterized in that: include: a first processor; and a memory for storing executable instructions of the first processor; Wherein, the first processor is configured to execute the method for predicting immune checkpoint inhibitor-related thyroid dysfunction as described in any one of claims 1 to 4 by executing the executable instructions.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the second processor, the method for predicting immune checkpoint inhibitor-related thyroid dysfunction according to any one of claims 1 to 4 is implemented.

Images