Medulloblastoma survival rate prediction model training and application method, equipment and medium

Abstract

This application discloses a method, device, and medium for training and applying a medulloblastoma survival prediction model, relating to the field of survival prediction technology. The method comprises: obtaining a first training dataset, using the first training dataset as input to train multiple initial models, obtaining multiple first-trained models and the model performance of each first-trained model, and selecting the first-trained model with the best model performance as the medulloblastoma survival prediction model. The medulloblastoma survival prediction model is used to predict the survival rate of MB patients at a predicted time point. This application can improve prediction accuracy.

Description

Technical Field

[0001] The present application relates to the technical field of survival rate prediction, and in particular to a method, device and medium for training and applying a medulloblastoma survival rate prediction model. Background Art

[0002] Medulloblastoma (MB) is a common tumor of the central nervous system in children, most commonly occurring in children aged 5 to 10 years. It is highly malignant, has a poor prognosis, and a high mortality rate. Despite advances in the treatment and management of MB, endocrine-related dysfunction, neurocognitive deficits, and the increased risk of secondary neoplastic lesions significantly impact the quality of life of MB patients. Therefore, prognostic prediction plays a crucial role in optimizing treatment strategies for MB patients. Most researchers use Cox regression (i.e., Cox proportional hazards regression model) to develop nomograms based on clinical information, imaging data, or genomic data to predict MB patient-specific survival. However, multivariate Cox regression is a semiparametric model that assumes that the risk of death in MB patients is a linear combination of their covariates, which has certain limitations. Due to these limitations, some researchers have considered using machine learning methods to predict MB patient-specific survival based on clinical information, imaging data, or genomic data, but the prediction accuracy remains poor. Summary of the Invention

[0003] The purpose of this application is to provide a medulloblastoma survival rate prediction model training and application method, equipment and medium, which can improve the prediction accuracy.

[0004] To achieve the above objectives, this application provides the following solutions.

[0005] In a first aspect, the present application provides a method for training a medulloblastoma survival rate prediction model, the method comprising:

[0006] Obtain a first training data set; the first training data set is a data set obtained by collecting patient information of MB patients in China, the first training data set includes multiple first samples, the first samples include first feature data and label data, the first feature data includes the MB patient's molecular subtype, gender, age, metastasis status, surgical resection status, histological subtype, treatment plan, whole-brain and whole-spinal cord radiotherapy dose, and posterior fossa whole-body or tumor bed localized radiotherapy dose; the label data includes the MB patient's survival status; the MB patients all underwent surgical resection and received postoperative radiotherapy and / or chemotherapy;

[0007] Using the first training data set as input, training multiple initial models to obtain multiple first trained models and the model performance of each first trained model; the initial models are models that can perform prediction functions;

[0008] A first post-training model with the best model performance is selected as a medulloblastoma survival rate prediction model; the medulloblastoma survival rate prediction model is used to predict the survival rate of MB patients at a predicted time point.

[0009] In a second aspect, the present application provides a method for training a medulloblastoma survival rate prediction model, the method comprising:

[0010] Obtain a second training data set; the second training data set is a data set obtained by collecting patient information of MB patients in China, the second training data set includes multiple second samples, the second samples include second feature data and label data, the second feature data includes the MB patient's molecular subtype, gender, age, metastasis status, histological subtype, whole brain and spinal cord radiotherapy dose, posterior fossa whole or tumor bed local boost radiotherapy dose, MYC expression level, MYCN expression level, OTX2 expression level, and GFI1 expression level; the label data includes the MB patient's survival status; the MB patients all underwent surgical resection and received postoperative radiotherapy and / or chemotherapy;

[0011] Using the second training data set as input, training multiple initial models to obtain multiple second trained models and the model performance of each second trained model; the initial model is a model that can perform a prediction function;

[0012] The second trained model with the best model performance is selected as the medulloblastoma survival rate prediction model; the medulloblastoma survival rate prediction model is used to predict the survival rate of MB patients at the predicted time point.

[0013] In a third aspect, the present application provides a method for applying a medulloblastoma survival rate prediction model, the method comprising:

[0014] Obtaining patient information of the MB patient to be predicted; the patient information includes the molecular subtype, gender, age, metastatic status, surgical resection status, histological subtype, treatment plan, whole-brain and whole-spinal cord radiotherapy dose, and posterior fossa overall or tumor bed local boost radiotherapy dose of the MB patient to be predicted, or the patient information includes the molecular subtype, gender, age, metastatic status, histological subtype, whole-brain and whole-spinal cord radiotherapy dose, posterior fossa overall or tumor bed local boost radiotherapy dose, MYC expression level, MYCN expression level, OTX2 expression level, and GFI1 expression level of the MB patient to be predicted;

[0015] The patient information is used as input, and the medulloblastoma survival rate prediction model is used to predict the survival rate of the MB patient to be predicted at the predicted time point to obtain the predicted survival rate of the MB patient to be predicted; the medulloblastoma survival rate prediction model is a medulloblastoma survival rate prediction model trained using the above-mentioned medulloblastoma survival rate prediction model training method.

[0016] In a fourth aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned medulloblastoma survival rate prediction model training method or the above-mentioned medulloblastoma survival rate prediction model application method.

[0017] In a fifth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned medulloblastoma survival rate prediction model training method or the above-mentioned medulloblastoma survival rate prediction model application method.

[0018] According to the specific embodiments provided in this application, this application has the following technical effects.

[0019] The present application provides a method, device, and medium for training and applying a medulloblastoma survival prediction model. When establishing a medulloblastoma survival prediction model through training, the first feature data used includes the MB patient's molecular subtype, gender, age, metastasis status, surgical resection status, histological subtype, treatment plan, whole-brain and whole-spinal cord radiotherapy dose, and posterior cranial fossa overall or tumor bed local boost radiotherapy dose. By selecting more complete patient information as the first feature data, the medulloblastoma survival prediction model obtained at this time can consider more information and the prediction accuracy is significantly improved. Multiple initial models are trained to obtain multiple first trained models, and the model performance of the multiple first trained models is compared. The first trained model with the best model performance is selected as the medulloblastoma survival prediction model. The medulloblastoma survival prediction model obtained at this time has better model performance and significantly improved prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0021] Figure 1 This is an application environment diagram of a medulloblastoma survival rate prediction model training and application method provided in this application.

[0022] Figure 2 A flowchart of a medulloblastoma survival rate prediction model training method provided in this application.

[0023] Figure 3 Schematic diagram of the ROC curve (Receiver Operating Characteristic curve) performance comparison of six models for 5-year survival rate prediction provided in this application.

[0024] Figure 4 This application provides a schematic diagram of the ROC curve performance comparison of six models for 10-year survival rate prediction.

[0025] Figure 5 Schematic diagram of the performance comparison of clinical decision curve analysis (DCA) of six models for 5-year survival rate prediction provided in this application.

[0026] Figure 6 Schematic diagram of the comparison of clinical decision curve performance of six models for 10-year survival rate prediction provided in this application.

[0027] Figure 7 Schematic diagram showing the performance comparison of calibration curves for the 5-year survival rate prediction provided in this application, between the XGBoost (Extreme Gradient Boosting) model and the Coxph model (Cox Proportional Hazards Regression Model).

[0028] Figure 8 Schematic diagram of the comparison of calibration curve performance between the XGBoost model and the Coxph model for 10-year survival rate prediction provided in this application.

[0029] Figure 9SHAP (SHapley Additive exPlanation) variable importance bar chart for the XGBoost model provided in this application.

[0030] Figure 10 SHAP summary plot of the XGBoost model provided for this application.

[0031] Figure 11 Schematic diagram of the ROC curve for external validation of the XGBoost model provided in this application.

[0032] Figure 12 Schematic diagram of the screening of molecular features provided in this application.

[0033] Figure 13 A flowchart of a medulloblastoma survival rate prediction model training method provided in this application.

[0034] Figure 14 The survival prediction model for medulloblastoma patients of Group_3 subtype (3 subtypes) or Group_4 subtype (4 subtypes) provided in this application is aimed at predicting the 5-year survival rate, and a schematic diagram of the ROC curve performance comparison of six models.

[0035] Figure 15 The survival prediction model for medulloblastoma patients of Group_3 or Group_4 subtype provided in this application is a schematic diagram of the ROC curve performance comparison of six models for 10-year survival rate prediction.

[0036] Figure 16 A flowchart of a method for applying a medulloblastoma survival rate prediction model provided in this application.

[0037] Figure 17 A schematic diagram of the structure of a computer device provided in this application. DETAILED DESCRIPTION

[0038] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0039] Example 1.

[0040] The medulloblastoma survival rate prediction model training method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown, the terminal communicates with the server via a network. The data storage system can store data that the server needs to process. The data storage system can be set up separately, integrated on the server, or placed on the cloud or other servers. The terminal can send a pending establishment request to the server. After the server receives the pending establishment request, the server obtains a first training data set for the pending establishment request; uses the first training data set as input to train multiple initial models to obtain multiple first trained models and the model performance of each first trained model; selects the first trained model with the best model performance as the medulloblastoma survival rate prediction model. The medulloblastoma survival rate prediction model is used to predict the survival rate of MB patients at a predicted time point. In this case, the prediction model is established based on clinical data. Alternatively, the server obtains a second training data set; uses the second training data set as input to train multiple initial models to obtain multiple second trained models and the model performance of each second trained model; selects the second trained model with the best model performance as the medulloblastoma survival rate prediction model. The medulloblastoma survival rate prediction model is used to predict the survival rate of MB patients at a predicted time point. In this case, the prediction model is established based on clinical data and molecular data. The server may feed back the obtained establishment result of the medulloblastoma survival rate prediction model requested to the terminal.

[0041] In addition, in some embodiments, the medulloblastoma survival rate prediction model training method can also be implemented independently by a server or a terminal. For example, the terminal can directly process the pending establishment request, or the server can obtain the pending establishment request from the data storage system and process the pending establishment request.

[0042] The terminals may include, but are not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices may include smart speakers, smart TVs, smart air conditioners, and smart car devices. Portable wearable devices may include smart watches, smart bracelets, and head-mounted devices. The server may be implemented as a standalone server or a server cluster consisting of multiple servers, or as a cloud server.

[0043] In an exemplary embodiment, Figure 2 As shown, a method for training a medulloblastoma survival rate prediction model is provided. The method is executed by a computer device, specifically a computer device such as a terminal or a server, or a terminal and a server. In the embodiment of the present application, the method is applied to Figure 1 The following steps are used as an example to illustrate the server.

[0044] Step S101: Obtain a first training data set; the first training data set is a data set obtained by collecting patient information of MB patients in China, the first training data set includes multiple first samples, the first samples include first feature data and label data, the first feature data includes the MB patient's molecular subtype, gender, age, metastasis status, surgical resection status, histological subtype, treatment plan, whole-brain and whole-spinal cord radiotherapy dose, and posterior cranial fossa whole or tumor bed local boost radiotherapy dose, the label data includes the MB patient's survival status; the MB patients all underwent surgical resection and received postoperative radiotherapy and / or chemotherapy.

[0045] Step S102: Using the first training data set as input, multiple initial models are trained to obtain multiple first trained models and the model performance of each first trained model; the initial model is a model that can perform a prediction function.

[0046] Step S103 , selecting a first trained model with the best model performance as a medulloblastoma survival rate prediction model; the medulloblastoma survival rate prediction model is used to predict the survival rate of MB patients at a predicted time point.

[0047] In implementing steps S101 to S103 above, in this embodiment, when establishing a medulloblastoma survival prediction model through training, the first feature data used includes the MB patient's molecular subtype, gender, age, metastasis status, surgical resection status, histological subtype, treatment plan, whole-brain and whole-spinal cord radiotherapy dose, and posterior fossa whole-body or tumor bed localized boost radiotherapy dose. By selecting more complete patient information as the first feature data, the resulting medulloblastoma survival prediction model can consider more information, and the prediction accuracy is significantly improved. Multiple initial models are trained to obtain multiple first trained models. Through model performance comparison, the first trained model with the best model performance is selected as the medulloblastoma survival prediction model. The resulting medulloblastoma survival prediction model has better model performance and significantly improved prediction accuracy.

[0048] After obtaining the medulloblastoma survival rate prediction model, the medulloblastoma survival rate prediction model training method of this embodiment further includes: obtaining a first external dataset, where the first external dataset is a dataset obtained by collecting patient information of MB patients worldwide, the first external dataset including multiple first samples, the first samples including first feature data and label data, the first feature data including the MB patient's molecular subtype, gender, age, metastasis status, surgical resection status, histological subtype, treatment plan, whole-brain and whole-spinal cord radiotherapy dose, and posterior fossa whole or tumor bed local boost radiotherapy dose, and the label data including the MB patient's survival status; using the first training dataset as input, performing internal validation on the medulloblastoma survival rate prediction model to obtain internal validation results and test the model stability; and using the first external dataset as input, performing external validation on the medulloblastoma survival rate prediction model to obtain external validation results and test the model extrapolation.

[0049] The medulloblastoma survival rate prediction model training method of this embodiment specifically includes the following steps.

[0050] (1) Obtain a first training dataset and a first external dataset.

[0051] This example collects domestic clinical data (also known as clinical pathological data). Specifically, a multi-institutional cohort of 1043 MB patients diagnosed in China between September 2001 and April 2023 was collected. Among them, 729 MB patients had complete clinical data, including molecular subtype, gender, age, metastatic status, surgical resection status, histological subtype, treatment regimen, whole-brain and whole-spinal cord radiotherapy dose, posterior cranial fossa whole or tumor bed local boost radiotherapy dose, surgery date, follow-up deadline, and survival status. Molecular subtypes include WNT, SHH, Gr.3 (i.e., Group_3), Gr.4 (i.e., Group_4), and unknown molecular subtypes. Gender includes male and female, and age includes infants and young children (generally The study included 52 patients (aged 0-3 years), children (generally 3-10 years), adolescents (generally 10-17 years), and adults (generally 18 years and older). Metastasis status included no metastasis and metastasis. Surgical resection included partial resection and near-total / total resection. Histological subtypes included classic, desmoplastic / nodular, extensively nodular, large cell / anaplastic, and unknown histological subtype. Treatment options included postoperative radiotherapy alone and combined postoperative chemoradiotherapy (i.e., simultaneous postoperative radiotherapy and chemotherapy). The dose of whole-brain and whole-spinal radiotherapy was measured in Gy, and the dose of posterior fossa or localized tumor bed boost radiotherapy was measured in Gy. The survival of MB patients was defined as the time from surgery to death / follow-up cutoff (August 31, 2023). The first training dataset was subsequently constructed based on the aforementioned domestic clinical data.

[0052] This example collected global clinical data, specifically a multicenter cohort of 116 MB patients from 23 institutions worldwide. Clinical data included molecular subtype, gender, age, metastatic status, surgical resection, histological subtype, treatment regimen, whole-brain and whole-spinal radiotherapy dose, posterior fossa whole-body or localized tumor bed boost radiotherapy dose, survival duration, and survival status. This global clinical data was subsequently used as external validation data to construct the first external dataset.

[0053] It should be noted that all the above-mentioned MB patients underwent surgical resection and received radiotherapy and / or chemotherapy after surgery. MB patients who only underwent biopsy or did not receive radiotherapy were not included. The median age at diagnosis was 8 years old, and the IQR (Interquartile Range) was 6 to 11 years old.

[0054] During training, this embodiment first selects data for model training. Specifically, the first feature data is designed to include molecular subtype, gender, age, metastasis status, surgical resection status, histological subtype, treatment plan, whole-brain and whole-spinal cord radiotherapy dose, and posterior fossa overall or tumor bed local boost radiotherapy dose. Therefore, this embodiment can design a first training dataset and a first external dataset. The first training dataset is a dataset obtained by collecting patient information of MB patients in China, and the first external dataset is a dataset obtained by collecting patient information of MB patients worldwide. The first training dataset and the first external dataset both include multiple first samples. The first samples include first feature data and label data. The first feature data includes the MB patient's molecular subtype, gender, age, metastasis status, surgical resection status, histological subtype, treatment plan, whole-brain and whole-spinal cord radiotherapy dose, and posterior fossa overall or tumor bed local boost radiotherapy dose. The label data includes the MB patient's survival status. Whether the MB patient will survive at the predicted time point is determined based on the survival status. All MB patients underwent surgical resection and received radiotherapy and / or chemotherapy after surgery.

[0055] (2) Using the first training data set as input, multiple initial models are trained to obtain multiple first trained models and the model performance of each first trained model.

[0056] In this embodiment, data cleaning is performed on the first training dataset and the first external dataset. Data cleaning includes: deleting duplicate data, reviewing abnormal data (i.e., outliers) to remove logical errors, and processing numerical data units to maintain data consistency. The data in the first cleaned training dataset and the first cleaned external dataset obtained are high-quality, consistent, non-missing, and non-anomalous data after cleaning, processing, and formatting, and can be directly used for model training and verification.

[0057] At this time, in this embodiment, the first training data set is used as input to train multiple initial models, specifically including: data cleaning the first training data set to obtain a first cleaned training data set; using the first cleaned training data set as input to train multiple initial models, data cleaning includes: deleting duplicate data, reviewing abnormal data, and unifying the units of numerical data of the same type.

[0058] In this embodiment, the initial model is designed to be a model that can perform the prediction function. As an example, the initial models include: Cox proportional hazard regression model (abbreviated as Coxph), random survival forest model (Random Survival Forests, abbreviated as RSF), extreme gradient boosting model (abbreviated as XGBoost), elastic network model (Elastic Net Regularization, abbreviated as ENET), DeepSurv model (Deep Survival Model) and gradient boosting machine model (Gradient Boosting Machine, abbreviated as GBM). Of course, other models that can perform the prediction function can be used as the initial model, and this embodiment does not impose any restrictions on this.

[0059] During training, this embodiment uses the first training data set to train each initial model separately. During the training process, a 5-fold cross-validation method combined with a grid search algorithm is used to tune hyperparameters, and the initial model is retrained on the first training data set using the optimal hyperparameters finally obtained.

[0060] At this time, in this embodiment, the first training data set is used as input to train multiple initial models to obtain multiple first trained models and the model performance of each first trained model, specifically including: for each initial model, the first training data set is used as input, and the 5-fold cross-validation method and the grid search algorithm are used to tune the hyperparameters of the initial model to obtain the optimal hyperparameters of the initial model; the first training data set is used to train the initial model using the optimal hyperparameters to obtain the first trained model corresponding to the initial model and the model performance of each first trained model.

[0061] Multiple performance evaluation indicators are used to evaluate model performance, including the area under the receiver operating characteristic curve (AUC), decision curve analysis, and calibration curve analysis. The model performance is compared using the performance evaluation indicators, and the best prediction model (i.e., the model with the best model performance) is selected as the medulloblastoma survival rate prediction model. At this time, in this embodiment, the model performance is characterized by performance evaluation indicators, including the area under the receiver operating characteristic curve, decision curve analysis, and calibration curve analysis.

[0062] (3) The first post-training model with the best model performance is selected as the medulloblastoma survival rate prediction model. The medulloblastoma survival rate prediction model is used to predict the survival rate of MB patients at the predicted time point.

[0063] After experiments, the medulloblastoma survival rate prediction model is the XGBoost model. When tuning the hyperparameters of the XGBoost model, the grid search parameter range is: the number of iterations (nrounds) range is (200, 300), the maximum depth (max_depth) range is (2, 6), and the learning rate (eta) range is (0.01, 0.3). The optimal hyperparameters obtained are: number of iterations = 207, maximum depth = 2, and learning rate = 0.08004. After obtaining the medulloblastoma survival prediction model, an online survival prediction calculator based solely on clinical information (i.e., clinical data) was constructed based on the medulloblastoma survival prediction model and deployed on the software interface, thereby providing a medulloblastoma survival prediction method based on multimodal data and the XGBoost model to predict the survival rate of medulloblastoma patients. During the prediction, the input of the medulloblastoma survival prediction model includes molecular subtype, gender, age, metastasis status, surgical resection status, histological subtype, treatment plan, whole brain and whole spinal cord radiotherapy dose, and posterior cranial fossa overall or tumor bed local enhanced radiotherapy dose. The output includes the predicted survival rate at the predicted time points of 5 years, 10 years, and 15 years. At the same time, the online survival prediction calculator can also perform risk stratification judgment. The standard risk is defined as: age > 3 years, no metastasis, and total resection / near total resection. This situation is a low-risk population, and the rest are high-risk populations, such as Figure 3 、 Figure 4 、 Figure 5 、 Figure 6 、 Figure 7 、 Figure 8 shown. Figure 3 In the above example, CI refers to Confidence Interval, which represents the confidence interval. Figure 5 In the figure, the All line refers to the clinical decision curve obtained by assuming that active treatment intervention is taken for all patients without considering their individual predicted risks, and the None line refers to the clinical decision curve obtained by assuming that no additional active treatment intervention is taken for any patient.

[0064] It should be noted that the medulloblastoma survival rate prediction model corresponding to different prediction time points may be different. For each prediction time point, the medulloblastoma survival rate prediction model is determined according to the training method.

[0065] (4) Using the first training data set as input, the medulloblastoma survival prediction model was internally validated to obtain the internal validation results.

[0066] Since the first training dataset and the first external dataset have been cleaned in advance, in this embodiment, the first training dataset is used as input to perform internal verification of the medulloblastoma survival rate prediction model. Specifically, the verification includes: performing data cleaning on the first training dataset to obtain a first cleaned training dataset; and performing internal verification of the medulloblastoma survival rate prediction model using the first cleaned training dataset as input. The data cleaning includes: deleting duplicate data, reviewing abnormal data, and unifying the units of numerical data of the same type.

[0067] This example uses the first training dataset as input and adopts the Bootstrap method (resampling 1000 times) for internal validation. By performing sampling with replacement in the model training queue, Bootstrap resampling samples of the same sample size are constructed to evaluate model performance. This process is repeated 1000 times to obtain the stability of the model in internal validation, test the repeatability of the model development process, and prevent overestimation of model performance due to model overfitting.

[0068] (5) Using the first external dataset as input, the medulloblastoma survival prediction model was externally validated to obtain external validation results.

[0069] Using the first external dataset as input, external validation of the medulloblastoma survival rate prediction model is performed. Specifically, the method includes: performing data cleaning on the first external dataset to obtain a first cleaned external dataset; and external validation of the medulloblastoma survival rate prediction model using the first cleaned external dataset as input. The data cleaning includes: deleting duplicate data, reviewing abnormal data, and unifying the units of numerical data of the same type. By performing external validation using the first external dataset as input, the prediction performance of the optimal prediction model on the external data is tested. Specifically, the external validation using the first external dataset is used to verify the extrapolation of the model.

[0070] This embodiment can also calculate the SHAP values ​​of all features of all first samples, analyze the importance ranking of each feature based on the SHAP value, and analyze how the features affect the prediction results of the medulloblastoma survival rate prediction model, and explain the medulloblastoma survival rate prediction model.

[0071] At this time, in this embodiment, after obtaining the medulloblastoma survival rate prediction model, the medulloblastoma survival rate prediction model training method of this embodiment further includes: using the first training data set as input, using the SHapley AdditiveexPlanation method to calculate the SHAP value of each feature in each first sample, and based on the SHAP value of each feature in each first sample, determining the importance of each feature and its impact on the prediction result of the medulloblastoma survival rate prediction model, so as to interpret the medulloblastoma survival rate prediction model, wherein the feature is a type of data in the first feature data. The SHAP variable importance bar chart and SHAP summary chart of the XGBoost model are shown in FIG. Figure 9 and Figure 10 As shown in Figure 2, the prediction performance of the XGBoost model applied to external validation is as follows: Figure 11 shown.

[0072] Through the above process, a prediction model can be established based on clinical data. This embodiment further introduces genomic data (i.e., molecular data). Considering the difficulty in distinguishing Group_3 and Group_4 tumors, Group_3 / Group_4 patients are combined to establish a prediction model that combines clinical data and molecular data. Gene expression profiles were measured in medulloblastoma tissue samples using the BGISEQ-50 platform, and the gene expression profile reads were aligned with the hg38 human reference genome. The gene expression profiles were then normalized to obtain a tpm (transcripts per million) matrix. The tpm matrix was then logarithmically transformed (log2(tpm+1)) to obtain a transformed matrix. The expression levels of each gene were determined based on the transformed matrix to obtain molecular data. At this time, when the molecular subtype was Group_3 / Group_4 subtype, the variable screening process was as follows: To improve the accuracy and predictive efficacy of the prediction model, the molecular risk stratification conditions of Group_3 / Group_4 subtypes were fully considered. Based on previous studies, the key molecular events were included: GLI2 activation, MYC activation, MYCN activation, OTX2 activation, SNCAIP activation, PRDM6 activation, GFI1B activation, GFI1 activation, and CDK6 activation. Manual feature selection was performed based on clinical expertise and previous studies, and univariate Cox analysis was used to select candidate molecular features. Specifically, molecular features with a P value (used to measure whether the association between a molecular feature and survival outcome is statistically significant) < 0.1 were selected, such as Figure 12As shown in the figure, because MYC, MYCN and OTX2 are common driver genes in medulloblastoma, the second characteristic data finally included include: molecular subtype, gender, age, metastatic status, histological subtype, whole brain and whole spinal cord radiotherapy dose, posterior fossa whole or tumor bed local boost radiotherapy dose, and the expression levels of MYC, MYCN, OTX2 and GFI1 in MB patients (specifically, the median is used to divide the expression level into high and low, greater than or equal to the median is high expression, and less than the median is low expression).

[0073] When constructing a survival prediction model for Group 3 / Group 4 medulloblastoma (194 cases), the secondary feature data included: molecular subtype, sex, age, metastatic status, histological subtype, whole-brain and whole-spinal radiotherapy dose, posterior fossa or tumor bed boost radiotherapy dose, MYC expression (median = 1.66216), MYCN expression (median = 2.885574), OTX2 expression (median = 7.725381), and GFI1 expression (median = 0.09083753). Aside from the difference in the secondary feature data, the training, internal validation, and external validation processes were identical to those for the training method based on clinical data, resulting in a prediction model based on both clinical and molecular data.

[0074] like Figure 13 As shown, the medulloblastoma survival rate prediction model training method of this embodiment includes the following steps.

[0075] Step S201: Obtain a second training data set; the second training data set is a data set obtained by collecting patient information of MB patients in China, the second training data set includes multiple second samples, the second samples include second feature data and label data, the second feature data includes the MB patient's molecular subtype, gender, age, metastasis status, histological subtype, whole-brain and whole-spinal cord radiotherapy dose, posterior fossa whole or tumor bed local boost radiotherapy dose, MYC expression level, MYCN expression level, OTX2 expression level, and GFI1 expression level; the label data includes the MB patient's survival status; the MB patients all underwent surgical resection and received radiotherapy and / or chemotherapy after surgery.

[0076] Step S202: Using the second training data set as input, multiple initial models are trained to obtain multiple second trained models and the model performance of each second trained model; the initial model is a model that can complete the prediction function.

[0077] Step S203 , selecting a second trained model with the best model performance as a medulloblastoma survival rate prediction model; the medulloblastoma survival rate prediction model is used to predict the survival rate of MB patients at a predicted time point.

[0078] In implementing steps S201 to S203 above, in this embodiment, when establishing a medulloblastoma survival rate prediction model through training, the second feature data used includes the MB patient's molecular subtype, gender, age, metastasis status, histological subtype, whole-brain and whole-spinal cord radiotherapy dose, posterior fossa overall or tumor bed local boost radiotherapy dose, MYC expression level, MYCN expression level, OTX2 expression level, and GFI1 expression level. In this case, the resulting medulloblastoma survival rate prediction model can consider more information, fully consider clinical data and molecular data, and significantly improve prediction accuracy. Multiple initial models are trained to obtain multiple second-trained models. Through model performance comparison, the second-trained model with the best model performance is selected as the medulloblastoma survival rate prediction model. In this case, the resulting medulloblastoma survival rate prediction model has better model performance and significantly improved prediction accuracy.

[0079] The medulloblastoma survival rate prediction model training method of this embodiment specifically includes the following steps.

[0080] (1) Obtain a second training dataset and a second external dataset.

[0081] On the basis of the first training data set, molecular data (i.e., the high and low expression levels of MYC, the high and low expression levels of MYCN, the high and low expression levels of OTX2, and the high and low expression levels of GFI1) are further added to obtain the second training data set. On the basis of the first external data set, molecular data are further added to obtain the second external data set.

[0082] (2) Using the second training data set as input, multiple initial models are trained to obtain multiple second trained models and the model performance of each second trained model.

[0083] The second training data set is used as input to train multiple initial models, specifically including: performing data cleaning on the second training data set to obtain a second cleaned training data set; and using the second cleaned training data set as input to train multiple initial models, wherein data cleaning includes: deleting duplicate data, reviewing abnormal data, and unifying units of numerical data of the same type.

[0084] The second training data set is used as input to train multiple initial models to obtain multiple second trained models and the model performance of each second trained model, specifically including: for each initial model, using the second training data set as input, using a 5-fold cross-validation method and a grid search algorithm to tune the hyperparameters of the initial model to obtain the optimal hyperparameters of the initial model; using the second training data set to train the initial model using the optimal hyperparameters to obtain the second trained model corresponding to the initial model and the model performance of each second trained model.

[0085] (3) The second trained model with the best model performance is selected as the medulloblastoma survival rate prediction model. The medulloblastoma survival rate prediction model is used to predict the survival rate of MB patients at the predicted time point.

[0086] (4) Using the second training data set as input, the medulloblastoma survival prediction model was internally validated to obtain internal validation results.

[0087] The second training data set is used as input to perform internal validation on the medulloblastoma survival rate prediction model, specifically including: performing data cleaning on the second training data set to obtain a second cleaned training data set; and the second cleaned training data set is used as input to perform internal validation on the medulloblastoma survival rate prediction model, where data cleaning includes: deleting duplicate data, reviewing abnormal data, and unifying the units of numerical data of the same type.

[0088] (5) Using the second external dataset as input, the medulloblastoma survival prediction model was externally validated to obtain external validation results.

[0089] The second external data set is used as input to externally validate the medulloblastoma survival rate prediction model, specifically including: performing data cleaning on the second external data set to obtain a second cleaned external data set; and the second cleaned external data set is used as input to externally validate the medulloblastoma survival rate prediction model, and data cleaning includes: deleting duplicate data, reviewing abnormal data, and unifying the units of numerical data of the same type.

[0090] The second training dataset was used for training and internal validation, the second external dataset was used for external validation, and the receiver operating characteristic curve was used to compare the model performance, such as Figure 14 and Figure 15 shown.

[0091] After obtaining the medulloblastoma survival prediction model, the SHAP values ​​of all features of all second samples were calculated, and the importance ranking of each feature was analyzed based on the SHAP value, as well as how the features affected the prediction results, to explain the medulloblastoma survival prediction model.

[0092] After testing, the XGBoost model was used to predict the survival rate of medulloblastoma of the Group_3 / Group_4 subtype. When performing hyperparameter tuning, the grid search parameter range was: the number of iterations (nrounds) range was (20, 28), the maximum depth (max_depth) range was (2, 6), and the learning rate (eta) range was (0.01, 0.3). The optimal hyperparameters obtained were: number of iterations = 21, maximum depth = 4, and learning rate = 0.189. An online survival prediction calculator integrating clinical information (i.e., clinical data) and molecular event information (i.e., molecular data) was constructed and deployed on the software interface. During the prediction, the inputs of the medulloblastoma survival prediction model included molecular subtype, gender, age, metastatic status, histological subtype, whole-brain and whole-spinal cord radiotherapy dose, posterior fossa overall or tumor bed local enhanced radiotherapy dose, MYC expression, MYCN expression, OTX2 expression, and GFI1 expression. The outputs included predicted survival rates at predicted time points such as 5 years, 10 years, and 15 years, and further risk stratification was performed.

[0093] This embodiment proposes a method for predicting the survival rate of medulloblastoma patients based on multimodal data, comprising the following steps: cleaning clinical data and molecular data, using clinical data alone or integrating clinical data and molecular data, constructing and training six initial models (including a Cox proportional hazards regression model, a random survival forest model, an extreme gradient boosting model, an elastic network model, a DeepSurv model, and a gradient boosting machine model), training the models in a training dataset, using a 5-fold cross-validation method combined with a grid search algorithm for hyperparameter tuning, retraining the initial model on the training dataset using the final optimal hyperparameters, comparing model performance using the area under the receiver operating characteristic curve, decision curve analysis, and calibration curve analysis, determining the final medulloblastoma survival rate prediction model, performing internal validation using the Bootstrap method, performing external validation using an external dataset from a multicenter cohort of 23 institutions worldwide, and performing SHapley The AdditiveexPlanation method ranks feature importance and interprets the final medulloblastoma survival prediction model. The final medulloblastoma survival prediction model is used to construct an online survival prediction calculator and deployed in the software interface. The observation data of the individual patient to be analyzed is input into the online survival prediction calculator to perform survival prediction and obtain prediction results. This embodiment can use complete patient information for survival analysis and prediction, with good performance and accurate prediction results.

[0094] The online survival prediction calculator is deployed in a public and free software interface. By selecting individual information and clicking the predict button, 5- and 10-year survival rate predictions are made. The right-hand interface first displays the risk group, followed by the predicted 5- and 10-year survival rates. At the bottom, a line graph of the changes in the 5- and 10-year predicted survival rates is displayed. Currently, no clinical data has been studied that covers radiotherapy doses or integrates clinical and molecular data to predict survival rates. Compared with existing survival prediction models, this method has higher predictive performance. The specific advantages come from the following technical points.

[0095] (1) Multimodal data fusion and high-precision prediction: This method introduces radiotherapy dose data not included in other models, as well as molecular data, achieving deep fusion of multi-source data. By using machine learning algorithms, the precision and accuracy of survival prediction were significantly improved. The stability of the prediction model was verified through internal validation, and the generalization of the model was improved through external validation.

[0096] (2) Personalized treatment plan optimization: By using the online version of the online survival prediction calculator based on the XGBoost model, clinicians can accurately predict patient survival and provide personalized treatment plans, which can help clinicians make accurate treatment decisions and reduce patient risks.

[0097] Example 2.

[0098] The application method of the medulloblastoma survival rate prediction model provided in the embodiments of the present application can be applied to Figure 1 In the application environment shown. The terminal communicates with the server through the network. The data storage system can store data that the server needs to process. The data storage system can be set up separately, integrated on the server, or placed on the cloud or other servers. The terminal can send a prediction request to be processed to the server. After the server receives the prediction request to be processed, the server obtains the patient information of the MB patient to be predicted for the prediction request; using the patient information as input, the medulloblastoma survival rate prediction model is used to predict the survival rate of the MB patient to be predicted at the prediction time point, and obtain the predicted survival rate of the MB patient to be predicted. The server can feed back the predicted survival rate obtained for the prediction request to the terminal.

[0099] In addition, in some embodiments, the medulloblastoma survival rate prediction model application method can also be implemented independently by a server or a terminal. For example, the terminal can directly process the pending prediction request, or the server can obtain the pending prediction request from the data storage system and process the pending prediction request.

[0100] In an exemplary embodiment, Figure 16 As shown, a method for applying a medulloblastoma survival rate prediction model is provided. The method is executed by a computer device, specifically a computer device such as a terminal or a server, or a terminal and a server. In the embodiment of the present application, the method is applied to Figure 1 The following steps are used as an example to illustrate the server.

[0101] Step T1, obtaining patient information of the MB patient to be predicted; the patient information includes the molecular subtype, gender, age, metastasis status, surgical resection status, histological subtype, treatment plan, whole-brain and whole-spinal cord radiotherapy dose, and posterior fossa overall or tumor bed local boost radiotherapy dose, or the patient information includes the molecular subtype, gender, age, metastasis status, histological subtype, whole-brain and whole-spinal cord radiotherapy dose, posterior fossa overall or tumor bed local boost radiotherapy dose, MYC expression level, MYCN expression level, OTX2 expression level, and GFI1 expression level of the MB patient to be predicted.

[0102] Step T2: using the patient information as input, predicting the survival rate of the MB patient to be predicted at the predicted time point using a medulloblastoma survival rate prediction model to obtain a predicted survival rate of the MB patient to be predicted; the medulloblastoma survival rate prediction model is a medulloblastoma survival rate prediction model trained using the medulloblastoma survival rate prediction model training method described in Example 1.

[0103] When the patient information includes the molecular subtype, gender, age, metastasis status, surgical resection status, histological subtype, treatment plan, whole-brain and whole-spinal cord radiotherapy dose, and posterior fossa overall or tumor bed local boost radiotherapy dose of the MB patient to be predicted, the medulloblastoma survival rate prediction model is a medulloblastoma survival rate prediction model trained using clinical data. When the patient information includes the molecular subtype, gender, age, metastasis status, histological subtype, whole-brain and whole-spinal cord radiotherapy dose, posterior fossa overall or tumor bed local boost radiotherapy dose, MYC expression level, MYCN expression level, OTX2 expression level, and GFI1 expression level of the MB patient to be predicted, the medulloblastoma survival rate prediction model is a medulloblastoma survival rate prediction model trained using clinical data and molecular data.

[0104] This application also provides an application scenario that applies the above-mentioned medulloblastoma survival rate prediction model application method. Specifically, the medulloblastoma survival rate prediction model application method provided in this embodiment can be applied in a survival prediction scenario. The survival prediction scenario includes a prediction phase and a display phase. The prediction phase is used to predict the survival rate of medulloblastoma patients at a predicted time point to obtain a predicted survival rate, and the display phase is used to display the predicted survival rate to the user. The medulloblastoma survival rate prediction model application method provided in this embodiment belongs to the prediction phase.

[0105] Example 3.

[0106] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 17 As shown. The computer device includes a processor, memory, an input / output (I / O) interface, and a communication interface. The processor, memory, and I / O interface are connected via a system bus, and the communication interface is connected to the system bus via the I / O interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store data. The I / O interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals via a network connection. When executed by the processor, the computer program implements a method for training a medulloblastoma survival rate prediction model or a method for applying a medulloblastoma survival rate prediction model.

[0107] Those skilled in the art will understand that Figure 17 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0108] In an exemplary embodiment, a computer device is provided, comprising a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the medulloblastoma survival rate prediction model training method in Example 1 or the medulloblastoma survival rate prediction model application method in Example 2 is implemented.

[0109] Example 4.

[0110] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, which, when executed by a processor, implements the medulloblastoma survival rate prediction model training method in Example 1 or the medulloblastoma survival rate prediction model application method in Example 2.

[0111] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0112] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0113] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for training a medulloblastoma survival rate prediction model, characterized in that: The medulloblastoma survival rate prediction model training method comprises: Obtain a second training data set; the second training data set is a data set obtained by collecting patient information of domestic MB patients, the second training data set includes multiple second samples, the second samples include second feature data and label data, the second feature data includes the molecular subtype, gender, age, metastasis status, histological subtype, whole brain and spinal cord radiotherapy dose, posterior cranial fossa overall or tumor bed local enhancement radiotherapy dose, MYC expression level, MYCN expression level, OTX2 expression level and GFI1 expression level of MB patients, if the MYC expression level is greater than or equal to the median of the MYC expression level, the MYC expression level is high; otherwise, the MYC expression level is low; if the MYCN expression level is greater than or equal to the median of the MYC expression level, the MYC expression level is high; otherwise, the MYC expression level is low; If the expression level of MYCN is greater than or equal to the median of the expression level of OTX2, the expression level of OTX2 is high; otherwise, the expression level of OTX2 is low; if the expression level of GFI1 is greater than or equal to the median of the expression level of GFI1, the expression level of GFI1 is high; otherwise, the expression level of GFI1 is low. The label data includes the survival status of MB patients; the MB patients all underwent surgical resection and received radiotherapy and / or chemotherapy after surgery; wherein, the high and low expression levels of MYC, MYCN, OTX2, and GFI1 are selected based on clinical expertise and univariate Cox analysis; Using the second training data set as input, training multiple initial models to obtain multiple second trained models and the model performance of each second trained model; the initial model is a model that can perform a prediction function; The second trained model with the best model performance is selected as the medulloblastoma survival rate prediction model; the medulloblastoma survival rate prediction model is used to predict the survival rate of MB patients at the predicted time point.

2. The medulloblastoma survival rate prediction model training method according to claim 1, characterized in that: After obtaining the medulloblastoma survival rate prediction model, the medulloblastoma survival rate prediction model training method further includes: Obtaining a second external dataset; the second external dataset is a dataset obtained by adding molecular data to the first external dataset, the first external dataset is a dataset obtained by collecting patient information of MB patients worldwide, and the second external dataset includes a plurality of second samples; Using the second training data set as input, internally validating the medulloblastoma survival rate prediction model to obtain internal validation results and test the model stability; The second external data set is used as input to perform external validation on the medulloblastoma survival rate prediction model, obtain external validation results, and test the model's extrapolation ability.

3. The medulloblastoma survival rate prediction model training method according to claim 2, characterized in that: Training the plurality of initial models using the second training data set as input specifically includes: performing data cleaning on the second training data set to obtain a second cleaned training data set; and training the plurality of initial models using the second cleaned training data set as input; Using the second training data set as input, internally validating the medulloblastoma survival rate prediction model, specifically comprising: performing data cleaning on the second training data set to obtain a second cleaned training data set; and using the second cleaned training data set as input, internally validating the medulloblastoma survival rate prediction model; Using the second external data set as input, externally validating the medulloblastoma survival rate prediction model, specifically comprising: performing data cleaning on the second external data set to obtain a second cleaned external data set; and using the second cleaned external data set as input, externally validating the medulloblastoma survival rate prediction model; The data cleaning includes: deleting duplicate data, reviewing abnormal data, and unifying the units of numerical data of the same type.

4. The medulloblastoma survival rate prediction model training method according to claim 1, characterized in that: The initial models include: Cox proportional hazards regression model, random survival forest model, extreme gradient boosting model, elastic network model, DeepSurv model and gradient boosting machine model; The model performance is characterized by performance evaluation indicators, including area under the receiver operating characteristic curve, decision curve analysis and calibration curve analysis.

5. The medulloblastoma survival rate prediction model training method according to claim 1, characterized in that: Using the second training data set as input, multiple initial models are trained to obtain multiple second trained models and the model performance of each second trained model, specifically including: For each initial model, the second training data set is used as input, and the hyperparameters of the initial model are tuned using the 5-fold cross-validation method and the grid search algorithm to obtain the optimal hyperparameters of the initial model; the initial model using the optimal hyperparameters is trained using the second training data set to obtain the second trained model corresponding to the initial model and the model performance of the second trained model.

6. The medulloblastoma survival rate prediction model training method according to claim 1, characterized in that: After obtaining the medulloblastoma survival rate prediction model, the medulloblastoma survival rate prediction model training method further includes: Taking the second training data set as input, the SHapley Additive exPlanation method is used to calculate the SHAP value of each feature in each second sample, and based on the SHAP value of each feature in each second sample, the importance of each feature and its impact on the prediction results of the medulloblastoma survival rate prediction model are determined to interpret the medulloblastoma survival rate prediction model; wherein the feature is a type of data in the second feature data.

7. A method for applying a medulloblastoma survival rate prediction model, characterized in that: The medulloblastoma survival rate prediction model application method comprises: Obtaining patient information of the MB patient to be predicted; the patient information includes the molecular subtype, gender, age, metastasis status, histological subtype, whole brain and spinal cord radiotherapy dose, posterior fossa whole or tumor bed local enhanced radiotherapy dose, MYC expression level, MYCN expression level, OTX2 expression level, and GFI1 expression level of the MB patient to be predicted; if the MYC expression level is greater than or equal to the median of the MYC expression level, the MYC expression level is high; otherwise, the MYC expression level is low; if the MYCN expression level is greater than or equal to the median of the MYCN expression level, the MYCN expression level is high; otherwise, the MYCN expression level is low; if the OTX2 expression level is greater than or equal to the median of the OTX2 expression level, the OTX2 expression level is high; otherwise, the OTX2 expression level is low; if the GFI1 expression level is greater than or equal to the median of the GFI1 expression level, the GFI1 expression level is high; otherwise, the GFI1 expression level is low; The patient information is used as input, and the survival rate of the MB patient to be predicted at the predicted time point is predicted using a medulloblastoma survival rate prediction model to obtain the predicted survival rate of the MB patient to be predicted; the medulloblastoma survival rate prediction model is a medulloblastoma survival rate prediction model trained using the medulloblastoma survival rate prediction model training method according to any one of claims 1 to 6.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that the processor executes the computer program to implement the medulloblastoma survival rate prediction model training method described in any one of claims 1 to 6 or the medulloblastoma survival rate prediction model application method described in claim 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for training a medulloblastoma survival rate prediction model according to any one of claims 1 to 6 or the method for applying a medulloblastoma survival rate prediction model according to claim 7 is implemented.

Images