Prediction method and system for pathological state of lymph node after immunotherapy

Abstract

The invention provides a method and a system for predicting pathological states of lymph nodes after immunotherapy, which are applied to the technical field of data processing. The method comprises the following steps: acquiring clinical parameter information of a target user, iconography parameter information of the target user, metabolic parameter information of the target user, a training sample set and a preset metastatic lymph node pathology remission rate prediction model; processing a preset metastatic lymph node pathology remission rate prediction model based on the training sample set to generate a target metastatic lymph node pathology remission rate prediction model; processing the clinical parameter information of the target user, the iconography parameter information of the target user and the metabolic parameter information of the target user to generate risk prediction factors; processing the risk prediction factor based on a target metastatic lymph node pathology remission rate prediction model to generate lymph node residual information of the target user; and processing the lymph node residual information of the target user to generate adjuvant therapy intervention information of the target user.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method and system for predicting the pathological state of lymph nodes after immunotherapy. Background Art

[0002] In recent years, neoadjuvant immunotherapy has been used in the treatment of non-small cell lung cancer, which has increased the pathological remission rate of patients and significantly improved their survival. Although neoadjuvant immunotherapy has brought certain positive effects, more than half of the patients still fail to benefit fully from the treatment. Therefore, there is an urgent need to screen out patients with poor treatment effects.

[0003] At present, neoadjuvant therapy research mainly focuses on the evaluation of primary tumors, while lymph nodes have not received enough attention. Clinical practice shows that in addition to primary tumors, the pathological status of lymph nodes is also closely related to the patient's survival and prognosis. On the one hand, independent evaluation of lymph nodes has not yet become a clinical standard; on the other hand, existing evaluation methods are not effective in predicting lymph node pathological remission rates, and there is no effective method for predicting lymph node pathological remission status.

[0004] 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

[0005] The purpose of this application is to provide a method and system for predicting the pathological state of lymph nodes after immunotherapy, which at least overcomes the problems existing in the prior art to a certain extent, by obtaining the clinical, imaging, and metabolic parameter information of the target user, as well as the training sample set and the preset model, and processing the preset model with the training sample set, and generating a prediction model for the pathological remission rate of target metastatic lymph nodes through operations such as data feature acquisition, sampling, grouping, and missing value interpolation. Then, features are extracted from the multi-source parameter information and regression analysis is performed to obtain risk prediction factors including the diameter of the primary tumor after treatment. The risk prediction factor is input into the target model, and the predicted value of the pathological remission state of the lymph nodes is calculated using a specific formula, thereby obtaining the residual information of the lymph nodes. Finally, the residual information of the lymph nodes is quantitatively analyzed, and the information on the changes in imaging and metabolic parameters within a preset time period, as well as clinical parameters, is combined to generate the residual association information of the lymph nodes, and adjuvant treatment intervention information is formulated accordingly.

[0006] 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.

[0007] According to one aspect of the present application, a method for predicting the pathological status of lymph nodes after immunotherapy is provided, comprising: obtaining clinical parameter information of a target user, imaging parameter information of a target user, metabolic parameter information of a target user, a training sample set, and a preset metastatic lymph node pathological remission rate prediction model, wherein the target user is a user of lymphatic metastatic lung cancer, and the training sample set is a patient with lymphatic metastatic lung cancer who has completed lymph node metastasis treatment; processing the preset metastatic lymph node pathological remission rate prediction model based on the training sample set to generate a target metastatic lymph node pathological remission rate prediction model; processing the clinical parameter information of the target user, the imaging parameter information of the target user, and the metabolic parameter information of the target user to generate a risk prediction factor; processing the risk prediction factor based on the target metastatic lymph node pathological remission rate prediction model to generate lymph node residual information of the target user; processing the lymph node residual information of the target user to generate auxiliary treatment intervention information for the target user.

[0008] Another aspect of the present application is a device for predicting the pathological state of lymph nodes after immunotherapy, characterized in that it includes: an acquisition module, used to acquire clinical parameter information of a target user, imaging parameter information of a target user, metabolic parameter information of a target user, a training sample set and a preset metastatic lymph node pathological remission rate prediction model, wherein the target user is a user of lymphatic metastatic lung cancer, and the training sample set is a patient with lymphatic metastatic lung cancer who has completed lymph node metastasis treatment; a processing module, used to process the preset metastatic lymph node pathological remission rate prediction model based on the training sample set to generate a target metastatic lymph node pathological remission rate prediction model; process the clinical parameter information of the target user, the imaging parameter information of the target user and the metabolic parameter information of the target user to generate a risk prediction factor; process the risk prediction factor based on the target metastatic lymph node pathological remission rate prediction model to generate lymph node residual information of the target user; process the lymph node residual information of the target user to generate auxiliary treatment intervention information of the target user.

[0009] According to another aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a second processor, the method for predicting the pathological state of lymph nodes after immunotherapy is implemented.

[0010] The present application provides a method and system for predicting the pathological state of lymph nodes after immunotherapy. The present application obtains the clinical, imaging, and metabolic parameter information of the target user, as well as a training sample set and a preset model, processes the preset model with the training sample set, and generates a target metastatic lymph node pathological remission rate prediction model through operations such as data feature acquisition, sampling, grouping, and missing value interpolation. Next, features are extracted from the multi-source parameter information and regression analysis is performed to obtain risk prediction factors including the diameter of the primary tumor after treatment. The risk prediction factor is input into the target model, and the predicted value of the pathological remission state of the lymph nodes is calculated using a specific formula to obtain the residual lymph node information. Finally, the residual lymph node information is quantitatively analyzed, and the imaging and metabolic parameter change information within the preset time period, as well as the clinical parameters, are combined to generate lymph node residual association information, and adjuvant treatment intervention information is formulated accordingly.

[0011] 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

[0012] Figure 1 A flow chart showing a method for predicting pathological status of lymph nodes after immunotherapy provided in one embodiment of the present application; Figure 2 A schematic diagram of the structure of a device for predicting the pathological state of lymph nodes after immunotherapy provided in one embodiment of the present application is shown. DETAILED DESCRIPTION

[0013] 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.

[0014] Combine the following Figure 1 The following describes a method for predicting pathological status of lymph nodes after immunotherapy according to an exemplary embodiment of the present application. In one embodiment, the present application also proposes a method and system for predicting pathological status of lymph nodes after immunotherapy. Figure 1 The flowchart of a method for predicting the pathological state of lymph nodes after immunotherapy according to an embodiment of the present application is schematically shown. Figure 1 As shown, the method is applied to a server, comprising: S101, obtaining clinical parameter information of a target user, imaging parameter information of a target user, metabolic parameter information of a target user, a training sample set, and a preset metastatic lymph node pathological remission rate prediction model.

[0015] In one embodiment, for target users with lymphatic metastatic lung cancer who have completed neoadjuvant immunotherapy, clinical parameter information covers multiple key aspects. Age is an important factor. There are differences in the physical function, tolerance to treatment, and tumor development speed of patients of different age groups. For example, elderly patients have poor tolerance to high-intensity treatment. Gender also affects the development of tumors and the response to treatment. Clinical studies have found that there are certain differences in the incidence, pathological type, and treatment effect of lung cancer between men and women. Smoking history is closely related to the occurrence and development of lung cancer. Tumor cells of long-term smoking patients have unique biological characteristics, which affect the treatment effect. Tumor clinical staging directly reflects the severity and spread of the tumor and is an important basis for formulating treatment strategies. PD-L1 expression level is a key biomarker. Its expression level is closely related to the efficacy of immunotherapy. Patients with high expression are more sensitive to immunotherapy. Pathological type should not be ignored. Different lung cancer pathological types (such as adenocarcinoma, squamous cell carcinoma, etc.) vary in biological behavior and treatment sensitivity.

[0016] Imaging parameter information is mainly obtained through chest CT or PET / CT scans. The diameter of the primary tumor before treatment, the diameter of the primary tumor after treatment, and the rate of change of the two can intuitively reflect the size change of the tumor during the treatment process. If the diameter of the primary tumor is significantly reduced after treatment and the rate of change is large, it usually indicates that the treatment effect is good. The lymph node diameter before treatment, the lymph node diameter after treatment, and the rate of change of lymph node diameter before and after treatment are crucial for judging lymph node metastasis and treatment effect. Changes in lymph node diameter indicate the proliferation or regression of tumor cells in the lymph nodes, which is an important indicator for evaluating treatment effect and disease progression. Metabolic parameter information mainly includes the metabolic value of the primary tumor before treatment, the metabolic value of the primary tumor after treatment, and the rate of change of the metabolic value, as well as the metabolic value of the lymph node before treatment, the metabolic value of the lymph node after treatment, and the rate of change of the metabolic value. These parameters are measured by PET / CT scanning, among which the maximum standardized uptake value (SUVmax) is a commonly used indicator. Changes in metabolic values ​​reflect changes in the metabolic activity of tumor cells. Reduced metabolic activity of tumor cells usually means that the treatment has effectively inhibited tumor growth; conversely, increased metabolic activity indicates tumor progression or poor treatment effect.

[0017] The training sample set is selected from patients with lymph node metastatic lung cancer who have completed lymph node metastasis treatment. The treatment process and outcome data of these patients are valuable resources for model training. The sample set contains various clinical characteristics, imaging and metabolic parameter information of patients before treatment, as well as the results of lymph node pathological status after treatment. By learning a large number of such samples, the model can explore the potential relationship between different parameters and lymph node pathological status. For example, the lymph node pathological remission of patients under different combinations of clinical characteristics, imaging and metabolic parameters is analyzed to provide a basis for predicting the lymph node pathological status of new target users. The diversity and representativeness of the sample set are crucial, covering patients of different ages, genders, pathological types, and treatment plans, which can improve the generalization ability of the model and enable it to have better prediction performance when facing various actual cases.

[0018] The preset metastatic lymph node pathological remission rate prediction model uses Lasso regression. Its core lies in the analysis of a large amount of relevant data, screening key features and constructing a prediction formula to achieve effective prediction of lymph node pathological remission rate. Lasso regression (Least Absolute Shrinkage and Selection Operator regression) is an extension of linear regression. It introduces the L1 regularization term on the basis of ordinary linear regression. This feature enables Lasso regression to not only select variables, but also screen out variables that have a significant impact on the dependent variable (lymph node pathological remission status) from a large number of independent variables (such as clinical parameters, imaging parameters, metabolic parameters, etc.), and compress the regression coefficients, effectively avoiding model overfitting and improving the generalization ability of the model. When predicting the pathological remission rate of metastatic lymph nodes, Lasso regression can accurately identify key predictive factors in the case of high-dimensional data (that is, data containing a large number of features) and build a concise and effective prediction model.

[0019] When constructing a preset metastatic lymph node pathological remission rate prediction model based on Lasso regression, the overall model can be divided into a data preprocessing layer, a model training layer, and a model output layer. The data preprocessing layer is responsible for cleaning, standardizing, and processing missing values ​​of the collected clinical parameter information, imaging parameter information, metabolic parameter information, and training sample sets of the target users to ensure the quality and consistency of the input data and provide reliable data for subsequent model training. The model training layer is the core layer, which uses the Lasso regression algorithm to learn the preprocessed data, and continuously adjusts the regression coefficient to find the optimal solution, so that the model can accurately capture the relationship between each parameter and the lymph node pathological remission status. The model output layer calculates the newly input target user data based on the trained model and outputs the prediction results of the lymph node pathological remission status.

[0020] Structurally, the model based on Lasso regression is mainly composed of a regression coefficient vector and a regularization parameter. The regression coefficient vector determines the weight of each independent variable in the prediction. For example, when predicting the pathological remission rate of lymph nodes, the independent variables such as the post-treatment PT diameter, PT diameter reduction rate, PT-ΔSUVmax and LN-ΔSUVmax have corresponding regression coefficients, which reflect the contribution of each variable to the prediction result. In addition to the regression coefficients and regularization parameters mentioned above, there are some important parameters related to model construction and training. In the data preprocessing stage, standardization involves calculating the mean and standard deviation of the data, which is used to normalize the data so that different features have the same scale and avoid affecting the model training effect due to large differences in feature scales. In the model training process, the number of iterations determines the number of calculations of the algorithm when searching for the optimal solution. The appropriate number of iterations can ensure that the model converges to a better result. If the number of iterations is too small, the model does not fully learn the data features, resulting in inaccurate predictions; if the number of iterations is too large, the calculation time will increase and overfitting will occur. In addition, the learning rate controls the step size of the regression coefficient update in each iteration. A suitable learning rate can speed up the convergence of the model and improve the training efficiency.

[0021] S102, processing the preset metastatic lymph node pathological remission rate prediction model based on the training sample set to generate a target metastatic lymph node pathological remission rate prediction model.

[0022] In one embodiment, any number of data features in the training sample set are obtained, and some data features are selected from the training sample set of 1,000 patients for analysis. For example, 20 data features are selected, including age, gender, smoking history, clinical tumor stage, PD-L1 expression level, pathological type, primary tumor diameter before treatment, primary tumor diameter after treatment, primary tumor diameter change rate before and after treatment, lymph node diameter before treatment, lymph node diameter after treatment, lymph node diameter change rate before and after treatment, primary tumor metabolic value before treatment, primary tumor metabolic value after treatment, primary tumor metabolic value change rate before and after treatment, lymph node metabolic value before treatment, lymph node metabolic value after treatment, lymph node metabolic value before treatment, and lymph node metabolic value change rate before and after treatment. These features cover the basic clinical information, imaging information, and metabolic information of the patient, and are of great significance for predicting the lymph node pathological remission rate.

[0023] Based on the number of each data feature in the training sample set, a sampling ratio is generated, and the training sample set is sampled based on the sampling ratio to generate a preset number of sampling features. The sampling ratio is obtained by calculating the ratio of the number of each feature in the selected 20 data features to the total number of features. Assuming that among the 20 features, the data integrity of the age feature is relatively high, with 950 valid data, then the sampling ratio of the age feature is 950÷1000=0.95; and due to the limitation of detection technology, a certain metabolic value feature has only 800 valid data, and its sampling ratio is 800÷1000=0.8. In this way, the sampling weight is determined for each data feature to ensure that different features can be reasonably sampled according to their data richness in the subsequent sampling process. According to the calculated sampling ratio, the training sample set is sampled. The preset number of sampling features is set to 500, and data is extracted from the training sample set by using random sampling combined with weights. For example, when extracting data about age, according to the sampling ratio of 0.95, it is expected that 950×(500÷1000)=475 data will be extracted; for metabolic value features, according to the sampling ratio of 0.8, it is expected that 800×(500÷1000)=400 data will be extracted. After sampling, a subset of 500 sampling features is obtained, which contains partial data of different data features and retains the distribution characteristics of the original data features.

[0024] Based on any data feature and each sampling feature, a response group and a poor response group are generated, wherein each data group contains a preset number of data samples, and at least one data sample includes identification information. Select one of the data features, such as the diameter of the primary tumor after treatment, and compare and analyze it with each sampling feature. Set a threshold, assuming that the diameter of the primary tumor after treatment is less than 1 cm as an indicator of a good treatment response. According to this standard, the patient data corresponding to the 500 sampling features are divided into two groups. The patient data with a primary tumor diameter of less than 1 cm after treatment are classified as the response group, which contains 300 data samples; the patient data with a primary tumor diameter greater than or equal to 1 cm are classified as the poor response group, which contains 200 data samples. In these data samples, each sample contains various information about the patient, such as age, gender, etc., and at least one of the data samples contains information that can identify whether the patient has risk factors affecting lymphatic metastatic lung cancer, such as specific gene markers or special clinical symptoms.

[0025] The missing values ​​of the response group and the poor response group were interpolated to generate missing value interpolation prediction information. The preset metastatic lymph node pathological remission rate prediction model was trained based on the data samples in the response group and the poor response group to generate the trained metastatic lymph node pathological remission rate prediction model. The data of the response group and the poor response group were checked, and it was found that some data were missing. For the missing age data in the response group, the multiple imputation method was used for processing. The age, gender, tumor stage and other related information of other patients in the group were used to construct a regression model to predict the missing age value. For example, through analysis, it was found that age and tumor stage had a certain correlation. This relationship was used to establish a regression equation to predict the missing age data. Similarly, the missing data of the poor response group were processed similarly. After processing, more complete data of the response group and the poor response group were obtained, and these supplementary data were the missing value interpolation prediction information. The data of the response group and the poor response group that were processed with missing value interpolation were input into the preset metastatic lymph node pathological remission rate prediction model based on Lasso regression. The model begins to learn the relationship between the features in the data and the lymph node pathological remission rate, and adjusts the regression coefficients in the model, such as the coefficients of the parameters such as the PT diameter after treatment and the PT diameter reduction rate. After multiple iterations of training, the model gradually fits the data and obtains the trained metastatic lymph node pathological remission rate prediction model. During the training process, the model is continuously optimized to improve the prediction accuracy of the lymph node pathological remission rate.

[0026] The trained metastatic lymph node pathological remission rate prediction model is processed to generate a verification result. If the data sample containing the identification information in the verification result is a risk factor that characterizes the risk of lymph node metastatic lung cancer, the trained metastatic lymph node pathological remission rate prediction model is used as the target metastatic lymph node pathological remission rate prediction model. The trained model is verified using the data of another 500 patients in the training sample set who did not participate in the training. The relevant data of these patients are input into the model, and the model outputs the prediction results of the lymph node pathological remission rate of these patients. Then, the prediction results are compared with the actual lymph node pathological examination results of these patients, and the prediction accuracy, specificity, sensitivity and other indicators of the model are calculated. For example, the model predicts the lymph node pathological remission status of 400 patients, of which 300 are predicted correctly, then the prediction accuracy is 300÷400=75%. In the verification result, the data sample containing the identification information is checked. Assuming that the identification information is a specific gene marker, if the data sample carrying the gene marker shows that patients with the marker have a higher risk of lymph node metastasis in the model prediction, and it is consistent with the actual clinical situation, that is, the lymph node pathological remission of these patients is indeed poor, then it means that the trained model can effectively identify the risk factors affecting lymphatic metastatic lung cancer. At this time, the trained metastatic lymph node pathological remission rate prediction model is used as the target metastatic lymph node pathological remission rate prediction model for subsequent prediction of the target user's lymph node pathological remission rate.

[0027] In another embodiment, the missing sample information of the response group and the poor response group is obtained respectively, and the response group and the poor response group are processed based on the missing sample information of the response group and the poor response group respectively to generate the predicted value of the missing sample. Both the response group and the poor response group may have missing data. For example, in the response group, it was found that the PD-L1 expression level data of some patients were missing, involving a total of 50 patients; in the poor response group, the lymph node metabolism values ​​before treatment were missing, involving 30 patients. These missing data may affect the accuracy of model training, so further processing is required. For the missing PD-L1 expression level data in the response group, we used a machine learning-based method to predict the missing values. Using other complete clinical parameters in the response group (such as age, gender, smoking history, tumor clinical stage, pathological type), imaging parameters (primary tumor diameter before treatment, primary tumor diameter after treatment, primary tumor diameter change rate before and after treatment, lymph node diameter before treatment, lymph node diameter after treatment, lymph node diameter change rate before and after treatment) and metabolic parameters (primary tumor metabolic value before treatment, primary tumor metabolic value after treatment, primary tumor metabolic value change rate before and after treatment, lymph node metabolic value after treatment, lymph node metabolic value change rate before and after treatment) as features, the random forest algorithm was used to build a prediction model. After training, the 50 patients with missing PD-L1 expression levels were predicted to obtain the predicted values. For example, one of the patients was predicted by the model to have a predicted value of 50% for PD-L1 expression level.

[0028] For the missing pre-treatment lymph node metabolic values ​​in the poor response group, a prediction model was also constructed based on other complete parameters in the group. A linear regression model was used here to predict 30 patients with missing pre-treatment lymph node metabolic values, using age, clinical tumor stage, and pre-treatment primary tumor metabolic values ​​as independent variables. Assume that a patient is predicted to have a pre-treatment lymph node metabolic value of 3.5 (SUVmax).

[0029] Get the true value corresponding to the missing sample, process the predicted value of the missing sample and the true value corresponding to the missing sample, and generate missing value interpolation prediction information. Take the missing value processing of PD-L1 expression level in the response group as an example, assuming that the formula parameters obtained through a large number of sample training are: , clinical parameter related regression coefficient (taking age as an example, assuming the age range is 1-100 years old, which has been standardized) , Regression coefficients of imaging parameters (taking the diameter of the primary tumor after treatment as an example, assuming the diameter range is 1-10 cm, which has been standardized) , metabolic parameter related regression coefficient (taking the metabolic value of the primary tumor after treatment as an example, assuming that the metabolic value range is 1-10, which has been standardized) For a patient with a missing PD-L1 expression level, the standardized value for age is 0.6, the standardized value for the primary tumor diameter after treatment is 0.4, and the standardized value for the primary tumor metabolic value after treatment is 0.5. The method also includes a calculation formula for obtaining missing value interpolation prediction information, which is: ;in, represents the constant term of the formula, represents the regression coefficient related to clinical parameters, represents the value of the kth clinical parameter, represents the regression coefficient related to imaging parameters, Representative The value of the imaging parameter, represents the regression coefficient related to metabolic parameters, Representative The value of each metabolic parameter is calculated according to the formula: .

[0030] The calculated results are de-normalized (assuming the de-normalized range is 0-100%), and the missing value interpolation prediction information of the patient's PD-L1 expression level is 40.5% (the actual de-normalization calculation is determined according to the normalization method of the original data). In this way, the predicted values ​​and true values ​​of all missing samples are processed to obtain complete missing value interpolation prediction information for subsequent model training to improve the accuracy and reliability of the model.

[0031] S103, processing the clinical parameter information of the target user, the imaging parameter information of the target user, and the metabolic parameter information of the target user to generate a risk prediction factor.

[0032] In one embodiment, the clinical parameter information of the target user is subjected to feature extraction processing to generate the clinical parameter features of the target user, wherein the clinical parameter features of the target user include age, gender, smoking history, clinical stage of tumor, PD-L1 expression level and pathological type. Assume that the target user is a 58-year-old male non-small cell lung cancer patient with a 30-year smoking history, a clinical stage of tumor IIIB, a PD-L1 expression level of 50%, and a pathological type of adenocarcinoma. Feature extraction is performed on the clinical parameter information of the target user. Age is a continuous variable and is directly recorded as 58 years old. Gender is male, and a coding method can be used, such as "1" for male and "0" for female. The smoking history is 30 years, which can be used as a quantitative indicator, or it can be divided into different levels according to medical research, such as 0-10 years for light smoking history, 11-30 years for moderate smoking history, and more than 30 years for heavy smoking history. This patient has a heavy smoking history. The clinical stage of the tumor is stage IIIB, and it is recorded according to medical standard stages. The PD-L1 expression level of 50% is a key indicator reflecting the sensitivity of immunotherapy, and this value is recorded directly. The pathological type is adenocarcinoma, which is also recorded according to the medical classification standards. Through these processes, the clinical parameter characteristics of the target user are generated. These characteristics provide basic information for subsequent analysis from the aspects of patient basic conditions, disease characteristics and immune-related indicators.

[0033] The imaging parameter information of the target user is subjected to feature extraction processing to generate the imaging features of the target user, wherein the imaging features of the target user include the primary tumor diameter before treatment, the primary tumor diameter after treatment, the change rate of the primary tumor diameter before and after treatment, the lymph node diameter before treatment, the lymph node diameter after treatment, and the change rate of the lymph node diameter before and after treatment. The patient underwent chest CT scans before the first treatment and after neoadjuvant chemotherapy. The primary tumor diameter was measured to be 4 cm before treatment, and the primary tumor diameter was reduced to 2 cm after treatment. The change rate of the primary tumor diameter before and after treatment was (4-2) ÷ 4 × 100% = 50%. The lymph node diameter was measured to be 1.5 cm before treatment, and it was reduced to 1 cm after treatment. The change rate of the lymph node diameter before and after treatment was (1.5-1) ÷ 1.5 × 100% ≈ 33.3%. These data intuitively reflect the changes in the size of tumors and lymph nodes during treatment, and are an important basis for evaluating the treatment effect and disease progression. By extracting imaging parameter information, the image features of the target user are obtained. These features provide a visual quantitative basis for judging the status of tumors and lymph nodes.

[0034] The metabolic parameter information of the target user is subjected to feature extraction processing to generate the metabolic parameter features of the target user, wherein the metabolic parameter features of the target user include the metabolic value of the primary tumor before treatment, the metabolic value of the primary tumor after treatment, the change rate of the metabolic value of the primary tumor before and after treatment, the metabolic value of the lymph node before treatment, the metabolic value of the lymph node after treatment, and the change rate of the metabolic value of the lymph node before and after treatment. The metabolic parameter information of the patient is obtained by PET / CT scanning. The metabolic value of the primary tumor before treatment (expressed as SUVmax) was 8.0, which dropped to 4.0 after treatment. The change rate of the metabolic value of the primary tumor before and after treatment was (8.0-4.0) ÷ 8.0 × 100% = 50%. The metabolic value of the lymph node (SUVmax) before treatment was 6.0, which changed to 3.0 after treatment. The change rate of the metabolic value of the lymph node before and after treatment was (6.0-3.0) ÷ 6.0 × 100% = 50%. The changes in these metabolic parameters reflect the changes in the metabolic activity of tumor cells. A decrease in metabolic value usually means that the treatment effectively inhibits tumor growth. By extracting metabolic parameter information, the metabolic parameter characteristics of the target user are generated, providing an important reference for evaluating the biological behavior of tumors.

[0035] The clinical parameter characteristics, imaging characteristics and metabolic parameter characteristics of the target user are subjected to regression analysis to generate risk prediction factors, where the risk prediction factors include the primary tumor diameter after treatment, the reduction rate of the primary tumor diameter, the change in the maximum standardized uptake value of the primary tumor, and the change in the maximum standardized uptake value of the lymph node. The clinical parameter characteristics, imaging characteristics and metabolic parameter characteristics extracted above are integrated for regression analysis. The regression analysis aims to find the potential relationship between these characteristics and the pathological remission status of the lymph nodes. In this process, statistical software or machine learning algorithms are used to perform regression calculations with the primary tumor diameter after treatment, the reduction rate of the primary tumor diameter, the change in the maximum standardized uptake value of the primary tumor (i.e., the change in SUVmax corresponding to the change rate of the metabolic value of the primary tumor before and after treatment), and the change in the maximum standardized uptake value of the lymph node (i.e., the change in SUVmax corresponding to the change rate of the metabolic value of the lymph node before and after treatment) as dependent variables, and other characteristics as independent variables. For example, regression analysis found that for every 1 cm increase in the diameter of the primary tumor after treatment, the risk of lymph node pathological remission decreased by a certain percentage; for every 10% increase in the reduction rate of the primary tumor diameter, the possibility of lymph node pathological remission increased to a certain extent, etc. Through regression analysis, risk prediction factors were finally generated. These factors comprehensively reflect the degree of influence of multiple characteristics on the state of lymph node pathological remission, providing key input for the subsequent use of prediction models to evaluate lymph node residual information.

[0036] S104: Process the risk prediction factor based on the target metastatic lymph node pathological remission rate prediction model to generate lymph node residual information of the target user.

[0037] In one embodiment, the risk prediction factors are processed based on the target metastatic lymph node pathological remission rate prediction model to generate a predicted value of the lymph node pathological remission status. Known risk prediction factors include the primary tumor diameter after treatment, the primary tumor diameter reduction rate, the primary tumor maximum standardized uptake value change, and the lymph node maximum standardized uptake value change. Assume that through previous examinations and calculations, the patient's primary tumor diameter after treatment is 2 cm, the primary tumor diameter reduction rate is 50%, the primary tumor maximum standardized uptake value change (i.e., the SUVmax change corresponding to the primary tumor metabolic value change rate before and after treatment) is 4.0 (assuming that the SUVmax before treatment is 8.0 and 4.0 after treatment), and the lymph node maximum standardized uptake value change (i.e., the SUVmax change corresponding to the lymph node metabolic value change rate before and after treatment) is 3.0 (assuming that the lymph node SUVmax before treatment is 6.0 and 3.0 after treatment).

[0038] According to the formula for calculating the predicted value of lymph node pathological remission status: The predicted value of lymph node pathological remission status = 0.0253 × primary tumor diameter after treatment - 0.0221 × primary tumor diameter reduction rate - 0.2838 × primary tumor maximum standardized uptake value change - 0.0359 × lymph node maximum standardized uptake value change + 0.2093. Substituting the patient's data into the formula, we get: . This calculation result -1.0 is the predicted value of the patient's lymph node pathological remission status based on the target metastatic lymph node pathological remission rate prediction model. The meaning of this value is a quantitative evaluation result calculated based on the model's comprehensive risk prediction factors, which is used to reflect the possibility of lymph node pathological remission. Here it is a negative value, which preliminarily indicates that the patient's lymph node pathological remission status may not be ideal.

[0039] The predicted value of the pathological remission state of the lymph nodes is processed to generate the target user's lymph node residual information. After obtaining the predicted value of the pathological remission state of the lymph nodes, it is necessary to further process and generate the target user's lymph node residual information. A threshold can be set based on clinical experience and a large amount of research data to divide the lymph node residual situation. For example, when the predicted value is greater than 0, it is considered that there are fewer lymph node residuals and the pathological remission state is better; when the predicted value is less than or equal to 0, it is considered that there is a certain degree of residual lymph nodes and the pathological remission state is poor. Since the predicted value of the patient is -1.0, which is less than 0, it can be judged that the patient's lymph nodes have a certain degree of tumor residual. Specifically, it can be further described as: According to the model prediction, the 58-year-old male non-small cell lung cancer patient has a high probability of tumor cell residual in the lymph nodes after receiving neoadjuvant immunotherapy, and his lymph node pathological remission state is not ideal, which may affect the patient's subsequent treatment and prognosis. Based on this result, combined with other clinical information of the patient, the doctor can consider whether to adjust the treatment plan, such as strengthening subsequent adjuvant treatment measures.

[0040] S105: Process the target user's lymph node residual information to generate auxiliary treatment intervention information for the target user.

[0041] In one embodiment, the target user's lymph node residual information is quantitatively analyzed and processed to generate lymph node residual degree information. Assuming that the patient's lymph node pathological remission state prediction value is -1.0 through the previous calculation, it is judged that there is lymph node residual. Quantitative standards can be set based on clinical experience and a large amount of research data, such as predicted values ​​of -1.5 and below for high residual, -1.5 to -0.5 for moderate residual, and above -0.5 for low residual. The patient's predicted value of -1.0 is between -1.5 and -0.5, so it is determined that the degree of lymph node residual is moderate.

[0042] The imaging parameter information and metabolic parameter information of the target user within the preset time period are obtained, and the imaging parameter information and metabolic parameter information of the target user within the preset time period are processed to generate imaging feature change information and metabolic parameter feature change information. One month and three months after the patient completed neoadjuvant immunotherapy, a chest CT and PET / CT scan were performed on the patient, respectively. At one month, the diameter of the primary tumor after treatment was measured to be 2 cm, and the diameter of the lymph node after treatment was 1 cm; at three months, the diameter of the primary tumor was reduced to 1.5 cm, and the diameter of the lymph node was reduced to 0.8 cm. The metabolic parameters were obtained from the PET / CT scan. At one month, the metabolic value (SUVmax) of the primary tumor was 3.5, and the metabolic value of the lymph node was 2.5; at three months, the metabolic value of the primary tumor dropped to 2.8, and the metabolic value of the lymph node dropped to 2.0. The image feature change information was calculated. The change in the primary tumor diameter within 1-3 months was 1.5-2 = -0.5 cm, and the change rate was (1.5-2) ÷ 2 × 100% = -25%; the change in the lymph node diameter was 0.8-1 = -0.2 cm, and the change rate was (0.8-1) ÷ 1 × 100% = -20%. In terms of the metabolic parameter feature change information, the change in the primary tumor metabolic value was 2.8-3.5 = -0.7, and the change rate was (2.8-3.5) ÷ 3.5 × 100% = -20%; the change in the lymph node metabolic value was 2.0-2.5 = -0.5, and the change rate was (2.0-2.5) ÷ 2.5 × 100% = -20%. These change information reflects the changes in the size and metabolic activity of tumors and lymph nodes within the preset time period.

[0043] The target user's clinical parameter information, lymph node residual degree information, image feature change information, and metabolic parameter feature change information are processed to generate the target user's lymph node residual association information. The patient's clinical parameter information (age 58 years old, male, 30-year smoking history, tumor clinical stage IIIB, PD-L1 expression level 50%, pathological type adenocarcinoma), lymph node residual degree information (moderate residual), image feature change information (primary tumor and lymph node diameter change and change rate), and metabolic parameter feature change information (primary tumor and lymph node metabolic value change and change rate) are integrated. The data analysis algorithm is used to comprehensively consider the association between various factors. For example, it is found that for this patient with a long history of smoking and a late clinical stage of the tumor, the continuous decline in lymph node diameter and metabolic value has certain positive significance, but combined with the current status of moderate residual, it still indicates that the risk of tumor recurrence cannot be ignored. Through these comprehensive analyses, lymph node residual association information containing multiple factors is generated.

[0044] The target user's lymph node residual association information is processed to generate the target user's adjuvant treatment intervention information. Based on the above lymph node residual association information, if the patient's lymph node residual degree is moderate, and the size and metabolic activity of the tumor and lymph nodes have improved within the preset time period, there is still a risk. Doctors may consider formulating an enhanced adjuvant chemotherapy regimen for patients, such as appropriately increasing the dose of paclitaxel by 20% on the basis of the original paclitaxel and carboplatin chemotherapy regimen, and extending the chemotherapy cycle from the original 4 cycles to 6 cycles to further eliminate possible residual tumor cells. At the same time, combined with the patient's PD-L1 expression level of 50%, combined immunotherapy drugs, such as pembrolizumab, can be considered to enhance the patient's own immune system's killing effect on tumor cells, improve the treatment effect, and reduce the risk of tumor recurrence.

[0045] In another embodiment, the imaging parameter information and metabolic parameter information of the target user within a preset time period are processed to generate a first frequency access value and a second frequency moment state vector. Within a preset time period (1-3 months), the patient is subjected to multiple imaging and metabolic parameter tests. Assume that during the first test (representing the first frequency), data is obtained through chest CT and PET / CT scans. The diameter of the primary tumor after treatment is measured to be 2 cm (which can be used as a tumor size quantification indicator, corresponding to the formula ), at this time the primary tumor metabolic value (SUVmax) is 3.5 (corresponding to the quantitative index of tumor metabolic activity, which can be regarded as the This data is used as the first frequency access value.

[0046] For the moment state vector of the second frequency, we consider associating the data of the second test (assuming it is performed in the second month) with the data of the first test. For example, in the second test, the diameter of the primary tumor changes to 1.8 cm, and the metabolic value (SUVmax) changes to 3.2. In order to generate the moment state vector, we calculate the change between the two tests and some related statistical features (such as the rate of change, etc.). The change in the diameter of the primary tumor is 1.8-2=-0.2 cm, and the rate of change is (1.8-2) ÷ 2 × 100% = -10%; the change in the metabolic value is 3.2-3.5=-0.3, and the rate of change is (3.2-3.5) ÷ 3.5 × 100% ≈ -8.6%. These changes and change rates are integrated to form the moment state vector of the second frequency.

[0047] The access value of the first frequency and the moment state vector of the second frequency are processed to generate the moment state vector of the first frequency. The method includes a calculation formula for obtaining the moment state vector of the first frequency. The calculation formula is: ;in, represents the moment state vector of the first frequency calculated at time point t, Represents the first frequency access value corresponding to The value of a tumor size quantitative index at time point t, Represents the first frequency access value corresponding to The value of a tumor size quantitative index at time point t, The state vector corresponding to the second frequency The value of a quantitative index of tumor metabolic activity at time point t, The state vector corresponding to the second frequency The value of a quantitative index of tumor metabolic activity at time point t, Represents the adjustment The weight coefficient of Represents the adjustment The weight coefficient of Represents the adjustment The weight coefficient of Represents the adjustment The weight coefficient of represents the bias term in the first half of the formula, Represents the bias term in the second half of the formula.

[0048] Assume that the parameters obtained through training with a large number of samples are: .

[0049] Taking the quantitative index of tumor size (such as primary tumor diameter) and the quantitative index of tumor metabolic activity (such as primary tumor metabolic value) as examples, the moment state vector of the first frequency is calculated. Here n and m are both assumed to be 1, i=1, j=1. is 2 cm, is 3.5, is 2, is the calculated value related to the corresponding primary tumor metabolic value in the state vector at the second frequency moment (assuming the above calculated rate of change -8.6% converted to a decimal -0.086).

[0050] First calculate the exponential part of the formula: .

[0051] but Part of the calculation is: .

[0052] .

[0053] Therefore, the state vector at the first frequency is for: .

[0054] Based on the moment state vector of the first frequency and the moment state vector of the second frequency, the image feature change information and metabolic parameter feature change information of the target user are generated. Combined with the moment state vector of the first frequency ( ) and the second frequency moment state vector, generate image feature change information and metabolic parameter feature change information. Suppose we set a rule that if the value of the first frequency moment state vector is greater than a certain threshold (such as 0.5, this threshold is determined based on a large number of sample analyses), and the change rates of tumor diameter and metabolic value in the second frequency moment state vector are both negative (indicating that the tumor is shrinking and metabolic activity is decreasing), then it is considered that the image feature and metabolic parameter feature changes show a positive trend.

[0055] For this patient, The primary tumor diameter change rate was -10%, and the metabolic value change rate was -8.6%, both of which were negative numbers. From this, it can be judged that within the preset time period, the patient's imaging feature change information showed that the tumor size had a trend of shrinking, and the metabolic parameter feature change information showed that the tumor metabolic activity also decreased, which indicated that it played a role in inhibiting tumor growth to a certain extent.

[0056] The server first obtains the target user's clinical, imaging, and metabolic parameter information, as well as the training sample set and preset model. The preset model is processed with the training sample set, and the target metastatic lymph node pathological remission rate prediction model is generated through operations such as data feature acquisition, sampling, grouping, and missing value interpolation. Next, features are extracted from the multi-source parameter information and regression analysis is performed to obtain risk prediction factors including the primary tumor diameter after treatment. The risk prediction factors are input into the target model, and the predicted value of the lymph node pathological remission status is calculated using a specific formula to obtain the lymph node residual information. Finally, the lymph node residual information is quantitatively analyzed, and the imaging and metabolic parameter change information within the preset time period, as well as the clinical parameters, are combined to generate lymph node residual association information, based on which auxiliary treatment intervention information is formulated.

[0057] In one embodiment, if Figure 2 As shown, the present application also provides a device for predicting the pathological state of lymph nodes after immunotherapy, comprising: The acquisition module 201 is used to acquire clinical parameter information of a target user, imaging parameter information of a target user, metabolic parameter information of a target user, a training sample set, and a preset metastatic lymph node pathological remission rate prediction model, wherein the target user is a user of lymphatic metastatic lung cancer, and the training sample set is a patient of lymphatic metastatic lung cancer who has completed lymph node metastasis treatment; The processing module 202 is used to process the preset metastatic lymph node pathological remission rate prediction model based on the training sample set to generate a target metastatic lymph node pathological remission rate prediction model; process the clinical parameter information of the target user, the imaging parameter information of the target user and the metabolic parameter information of the target user to generate a risk prediction factor; process the risk prediction factor based on the target metastatic lymph node pathological remission rate prediction model to generate the target user's lymph node residual information; process the target user's lymph node residual information to generate the target user's auxiliary treatment intervention information.

[0058] In another embodiment of the present application, the processing module 202 is configured to process the preset metastatic lymph node pathological response rate prediction model based on the training sample set to generate a target metastatic lymph node pathological response rate prediction model, including: Get any number of data features in the training sample set; Generating a sampling ratio based on the number of each data feature in the training sample set; Sampling the training sample set based on the sampling ratio to generate a preset number of sampling features; Based on any data feature and each sampling feature, processing is performed to generate a response group and a poor response group, wherein each data group includes a preset number of data samples and at least one data sample includes identification information; Performing missing value interpolation processing on the response group and the poor response group respectively to generate missing value interpolation prediction information; Training the preset metastatic lymph node pathological response rate prediction model based on the data samples in the response group and the poor response group to generate a trained metastatic lymph node pathological response rate prediction model; Processing the trained metastatic lymph node pathological response rate prediction model to generate a verification result; If the data sample containing identification information in the verification result is a risk factor that characterizes the risk factors affecting lymph node metastatic lung cancer, the trained metastatic lymph node pathological remission rate prediction model is used as the target metastatic lymph node pathological remission rate prediction model.

[0059] In another embodiment of the present application, the processing module 202 is configured to process the initial metastatic lymph node pathological response rate prediction model based on the training sample set and the validation sample set to generate a target metastatic lymph node pathological response rate prediction model, and further includes: Obtain missing sample information for the response group and the poor response group respectively; The response group and the poor response group are processed based on their missing sample information respectively to generate prediction values ​​of the missing samples; Get the true value corresponding to the missing sample; Process the predicted values ​​of missing samples and the true values ​​corresponding to the missing samples to generate missing value interpolation prediction information; The method also includes a calculation formula for obtaining missing value interpolation prediction information, and the calculation formula is: .

[0060] in, represents the constant term of the formula, represents the regression coefficient related to clinical parameters, represents the value of the kth clinical parameter, represents the regression coefficient related to imaging parameters, Representative The value of the imaging parameter, represents the regression coefficient related to metabolic parameters, Representative The value of a metabolic parameter.

[0061] In another embodiment of the present application, the processing module 202 is configured to process the clinical parameter information of the target user, the imaging parameter information of the target user, and the metabolic parameter information of the target user to generate a risk prediction factor, including: Performing feature extraction processing on the clinical parameter information of the target user to generate clinical parameter features of the target user, wherein the clinical parameter features of the target user include age, gender, smoking history, tumor clinical stage, PD-L1 expression level and pathological type; Performing feature extraction processing on the imaging parameter information of the target user to generate imaging features of the target user, wherein the imaging features of the target user include the primary tumor diameter before treatment, the primary tumor diameter after treatment, the rate of change of the primary tumor diameter before and after treatment, the lymph node diameter before treatment, the lymph node diameter after treatment, and the rate of change of the lymph node diameter before and after treatment; The metabolic parameter information of the target user is subjected to feature extraction processing to generate metabolic parameter features of the target user, wherein the metabolic parameter features of the target user include the metabolic value of the primary tumor before treatment, the metabolic value of the primary tumor after treatment, the rate of change of the metabolic value of the primary tumor before and after treatment, the metabolic value of the lymph node before treatment, the metabolic value of the lymph node after treatment, and the rate of change of the metabolic value of the lymph node before and after treatment; The clinical parameter characteristics of the target user, the imaging characteristics of the target user, and the metabolic parameter characteristics of the target user are subjected to regression analysis to generate risk prediction factors, wherein the risk prediction factors include the primary tumor diameter after treatment, the primary tumor diameter reduction rate, the change in the maximum standardized uptake value of the primary tumor, and the change in the maximum standardized uptake value of the lymph nodes.

[0062] In another embodiment of the present application, the processing module 202 is configured to process the risk prediction factor based on the target metastatic lymph node pathological remission rate prediction model to generate lymph node residual information of the target user, including: Processing the risk prediction factor based on the target metastatic lymph node pathological remission rate prediction model to generate a predicted value of the lymph node pathological remission status; Processing the predicted value of the lymph node pathological remission status to generate lymph node residual information of the target user; The method also includes a calculation formula for obtaining a predicted value of the lymph node pathological remission state, and the calculation formula is: .

[0063] In another embodiment of the present application, the processing module 202 is configured to process the target user's lymph node residual information to generate auxiliary treatment intervention information of the target user, including: Performing quantitative analysis on the target user's lymph node residual information to generate lymph node residual degree information; Obtain imaging parameter information and metabolic parameter information of the target user within a preset time period; Processing the imaging parameter information and metabolic parameter information of the target user within a preset time period to generate imaging feature change information and metabolic parameter feature change information; The target user's clinical parameter information, lymph node residual degree information, image feature change information, and metabolic parameter feature change information are processed to generate lymph node residual association information of the target user; The target user's lymph node residual association information is processed to generate the target user's auxiliary treatment intervention information.

[0064] In another embodiment of the present application, the processing module 202 is configured to process the imaging parameter information and metabolic parameter information of the target user within a preset time period to generate imaging feature change information and metabolic parameter feature change information, including: Processing the imaging parameter information and metabolic parameter information of the target user within a preset time period to generate a first frequency access value and a second frequency moment state vector; Processing the access value of the first frequency and the moment state vector of the second frequency to generate the moment state vector of the first frequency; Generate image feature change information and metabolic parameter feature change information of the target user based on the moment state vector of the first frequency and the moment state vector of the second frequency; The method includes a calculation formula for obtaining a moment state vector of a first frequency, and the calculation formula is: ; in, represents the moment state vector of the first frequency calculated at time point t, Represents the first frequency access value corresponding to The value of a tumor size quantitative index at time point t, Represents the first frequency access value corresponding to The value of a tumor size quantitative index at time point t, The state vector corresponding to the second frequency The value of a quantitative index of tumor metabolic activity at time point t, The state vector corresponding to the second frequency The value of a quantitative index of tumor metabolic activity at time point t, Represents the adjustment The weight coefficient of Represents the adjustment The weight coefficient of Represents the adjustment The weight coefficient of Represents the adjustment The weight coefficient of represents the bias term in the first half of the formula, Represents the bias term in the second half of the formula.

[0065] Each embodiment in the present 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 the pathological state of lymph nodes after immunotherapy, since it is basically similar to the above-mentioned prediction method embodiment for the pathological state of lymph nodes after immunotherapy, the description is relatively simple, and the relevant parts can be referred to the partial description of the above-mentioned prediction method embodiment for the pathological state of lymph nodes after immunotherapy.

Claims

1. A method for predicting the pathological state of lymph nodes after immunotherapy, characterized in that: include: Obtaining clinical parameter information of the target user, imaging parameter information of the target user, metabolic parameter information of the target user, a training sample set, and a preset metastatic lymph node pathological response rate prediction model, wherein the target user is a user of lymphatic metastatic lung cancer, and the training sample set is a patient with lymphatic metastatic lung cancer who has completed lymph node metastasis treatment; Processing the preset metastatic lymph node pathological remission rate prediction model based on the training sample set to generate a target metastatic lymph node pathological remission rate prediction model; Processing the clinical parameter information of the target user, the imaging parameter information of the target user, and the metabolic parameter information of the target user to generate a risk prediction factor; Processing the risk prediction factor based on the target metastatic lymph node pathological remission rate prediction model to generate lymph node residual information of the target user; The lymph node residual information of the target user is processed to generate auxiliary treatment intervention information of the target user.

2. The method according to claim 1, characterized in that The preset metastatic lymph node pathological response rate prediction model is processed based on the training sample set to generate a target metastatic lymph node pathological response rate prediction model, including: Get any number of data features in the training sample set; Generating a sampling ratio based on the number of each data feature in the training sample set; Sampling the training sample set based on the sampling ratio to generate a preset number of sampling features; Based on any data feature and each sampling feature, processing is performed to generate a response group and a poor response group, wherein each data group includes a preset number of data samples and at least one data sample includes identification information; Performing missing value interpolation processing on the response group and the poor response group respectively to generate missing value interpolation prediction information; Training the preset metastatic lymph node pathological response rate prediction model based on the data samples in the response group and the poor response group to generate a trained metastatic lymph node pathological response rate prediction model; Processing the trained metastatic lymph node pathological response rate prediction model to generate a verification result; If the data sample containing identification information in the verification result is a risk factor that characterizes the risk factors affecting lymph node metastatic lung cancer, the trained metastatic lymph node pathological remission rate prediction model is used as the target metastatic lymph node pathological remission rate prediction model.

3. The method according to claim 2, characterized in that The initial metastatic lymph node pathological response rate prediction model is processed based on the training sample set and the validation sample set to generate a target metastatic lymph node pathological response rate prediction model, further comprising: Obtain missing sample information for the response group and the poor response group respectively; The response group and the poor response group are processed based on their missing sample information respectively to generate prediction values ​​of the missing samples; Get the true value corresponding to the missing sample; Process the predicted values ​​of missing samples and the true values ​​corresponding to the missing samples to generate missing value interpolation prediction information; The method also includes a calculation formula for obtaining missing value interpolation prediction information, and the calculation formula is: ； in, represents the constant term of the formula, represents the regression coefficient related to clinical parameters, represents the value of the kth clinical parameter, represents the regression coefficient related to imaging parameters, Representative The value of the imaging parameter, represents the regression coefficient related to metabolic parameters, Representative The value of a metabolic parameter.

4. The method according to claim 1, characterized in that Processing the clinical parameter information of the target user, the imaging parameter information of the target user, and the metabolic parameter information of the target user to generate a risk prediction factor includes: Performing feature extraction processing on the clinical parameter information of the target user to generate clinical parameter features of the target user, wherein the clinical parameter features of the target user include age, gender, smoking history, tumor clinical stage, PD-L1 expression level and pathological type; Performing feature extraction processing on the imaging parameter information of the target user to generate imaging features of the target user, wherein the imaging features of the target user include the primary tumor diameter before treatment, the primary tumor diameter after treatment, the rate of change of the primary tumor diameter before and after treatment, the lymph node diameter before treatment, the lymph node diameter after treatment, and the rate of change of the lymph node diameter before and after treatment; The metabolic parameter information of the target user is subjected to feature extraction processing to generate metabolic parameter features of the target user, wherein the metabolic parameter features of the target user include the metabolic value of the primary tumor before treatment, the metabolic value of the primary tumor after treatment, the rate of change of the metabolic value of the primary tumor before and after treatment, the metabolic value of the lymph node before treatment, the metabolic value of the lymph node after treatment, and the rate of change of the metabolic value of the lymph node before and after treatment; The clinical parameter characteristics of the target user, the imaging characteristics of the target user, and the metabolic parameter characteristics of the target user are subjected to regression analysis to generate risk prediction factors, wherein the risk prediction factors include the primary tumor diameter after treatment, the primary tumor diameter reduction rate, the change in the maximum standardized uptake value of the primary tumor, and the change in the maximum standardized uptake value of the lymph nodes.

5. The method according to claim 4, characterized in that The risk prediction factor is processed based on the target metastatic lymph node pathological remission rate prediction model to generate lymph node residual information of the target user, including: Processing the risk prediction factor based on the target metastatic lymph node pathological remission rate prediction model to generate a predicted value of the lymph node pathological remission status; Processing the predicted value of the lymph node pathological remission status to generate lymph node residual information of the target user; The method further includes a calculation formula for obtaining a predicted value of the lymph node pathological remission state, the calculation formula being: 。 6. The method according to claim 5, characterized in that The target user's lymph node residual information is processed to generate auxiliary treatment intervention information for the target user, including: Performing quantitative analysis on the target user's lymph node residual information to generate lymph node residual degree information; Obtain imaging parameter information and metabolic parameter information of the target user within a preset time period; Processing the imaging parameter information and metabolic parameter information of the target user within a preset time period to generate imaging feature change information and metabolic parameter feature change information; The target user's clinical parameter information, lymph node residual degree information, image feature change information, and metabolic parameter feature change information are processed to generate the target user's lymph node residual association information; The target user's lymph node residual association information is processed to generate the target user's auxiliary treatment intervention information.

7. The method according to claim 6, characterized in that The imaging parameter information and metabolic parameter information of the target user within a preset time period are processed to generate imaging feature change information and metabolic parameter feature change information, including: Processing the imaging parameter information and metabolic parameter information of the target user within a preset time period to generate a first frequency access value and a second frequency moment state vector; Processing the access value of the first frequency and the moment state vector of the second frequency to generate the moment state vector of the first frequency; Generate image feature change information and metabolic parameter feature change information of the target user based on the moment state vector of the first frequency and the moment state vector of the second frequency; The method includes a calculation formula for obtaining a moment state vector of a first frequency, and the calculation formula is: ； in, Indicates at a point in time The calculated state vector at the first frequency is: Represents the first frequency access value corresponding to The value of a tumor size quantitative index at time point t, Represents the first frequency access value corresponding to The value of a tumor size quantitative index at time point t, The state vector corresponding to the second frequency The value of a quantitative index of tumor metabolic activity at time point t, The state vector corresponding to the second frequency The value of a quantitative index of tumor metabolic activity at time point t, Represents the adjustment The weight coefficient of Represents the adjustment The weight coefficient of Represents the adjustment The weight coefficient of Represents the adjustment The weight coefficient of represents the bias term in the first half of the formula, Represents the bias term in the second half of the formula.

8. A device for predicting pathological status of lymph nodes after immunotherapy, characterized in that: The device comprises: An acquisition module is used to acquire clinical parameter information of a target user, imaging parameter information of a target user, metabolic parameter information of a target user, a training sample set, and a preset metastatic lymph node pathological remission rate prediction model, wherein the target user is a user of lymphatic metastatic lung cancer, and the training sample set is a patient of lymphatic metastatic lung cancer who has completed lymph node metastasis treatment; A processing module is used to process the preset metastatic lymph node pathological remission rate prediction model based on the training sample set to generate a target metastatic lymph node pathological remission rate prediction model; process the clinical parameter information of the target user, the imaging parameter information of the target user and the metabolic parameter information of the target user to generate a risk prediction factor; process the risk prediction factor based on the target metastatic lymph node pathological remission rate prediction model to generate the target user's lymph node residual information; process the target user's lymph node residual information to generate the target user's auxiliary treatment intervention information.

9. 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 the pathological state of lymph nodes after immunotherapy as described in any one of claims 1 to 7 by executing the executable instructions.

10. 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 the pathological state of lymph nodes after immunotherapy according to any one of claims 1 to 7 is implemented.

Images