Neuroblastoma bone metastasis risk prediction system, method and terminal

Abstract

This application provides a system, method, and terminal for predicting the risk of bone metastasis in neuroblastoma. By acquiring laboratory test data from multiple children with neuroblastoma and constructing internal training sets, internal validation sets, external validation sets, and external test sets, a target neuroblastoma bone metastasis risk prediction model is built. Based on the target laboratory test data of the target neuroblastoma children, the system predicts the bone metastasis risk value and risk classification results for the target neuroblastoma children. This allows for timely early warning of disease changes in the early stages, reminding children and their families to seek medical attention as soon as possible. It also assists medical staff in making medical decisions and provides diagnostic evidence, solving the technical problems of limited early disease warning, high technical requirements, poor applicability, and poor safety associated with existing imaging techniques for detecting neuroblastoma bone metastasis.

Description

Technical Field

[0001] This application relates to the field of medical data processing technology, and in particular to a system, method and terminal for predicting the risk of bone metastasis in neuroblastoma. Background Technology

[0002] Neuroblastoma (NB) is a common and prevalent extracranial solid tumor in children, characterized by early age of onset, high metastasis rate, and significant heterogeneity in its clinical course. Approximately 50% of patients already have metastases when first diagnosed with neuroblastoma. Once metastasis occurs, the patient's condition is already quite severe, with a high tumor burden and significant treatment challenges. Therefore, early disease warning and early diagnosis of neuroblastoma are crucial.

[0003] The most common sites of metastasis for neuroblastoma are bone and bone marrow. Currently, the diagnosis of neuroblastoma bone metastases mainly relies on imaging techniques such as CT, MRI, and PET / CT. Imaging techniques offer high sensitivity and specificity in cases of significant tumor-induced bone destruction. However, in the early stages of the disease, imaging techniques cannot readily identify the risk of neuroblastoma bone metastasis, limiting their diagnostic value. Furthermore, the use of imaging techniques requires high precision in the imaging equipment and skilled medical personnel, making them difficult to implement in areas with limited medical resources.

[0004] Furthermore, neuroblastoma patients are usually young children, who are difficult to cooperate with during imaging examinations, resulting in longer examination times. In addition, imaging examinations carry the risk of radiation exposure, which may have long-term health effects on children who are still growing, making them unsafe. Summary of the Invention

[0005] In view of the shortcomings of the prior art described above, the purpose of this application is to provide a neuroblastoma bone metastasis risk prediction system, method and terminal to solve the technical problems of limited early disease warning, high technical requirements, poor applicability and poor safety of existing imaging technology for detecting neuroblastoma bone metastasis.

[0006] To achieve the above and other related objectives, a first aspect of this application provides a neuroblastoma bone metastasis risk prediction system. The neuroblastoma bone metastasis risk prediction system includes: a model building module for constructing a target neuroblastoma bone metastasis risk prediction model; and a model prediction module connected to the model building module for predicting the bone metastasis risk value and bone metastasis risk classification result of the target neuroblastoma patient based on the target neuroblastoma bone metastasis risk prediction model and obtained target laboratory test data of the target neuroblastoma patient, and generating a bone metastasis risk report; wherein the construction method of the target neuroblastoma bone metastasis risk prediction model... The method includes: acquiring laboratory test data from multiple children with neuroblastoma, and constructing an internal training set, an internal validation set, an external validation set, and an external test set based on the laboratory test data; constructing multiple primary neuroblastoma bone metastasis risk prediction models based on the internal training set; comparing the diagnostic performance of each primary neuroblastoma bone metastasis risk prediction model based on the internal validation set and the external validation set, and obtaining an optimized neuroblastoma bone metastasis risk prediction model; evaluating the optimized neuroblastoma bone metastasis risk prediction model based on the external test set, and determining the model that meets the prediction requirements as the target neuroblastoma bone metastasis risk prediction model.

[0007] In some embodiments of the first aspect of this application, the method of obtaining laboratory test data includes: obtaining clinical characteristic data, peripheral blood sample test data, and bone marrow sample test data of multiple children with neuroblastoma from multiple medical institutions as laboratory test data for each child with neuroblastoma; and performing medical logic review on each laboratory test data based on a pre-constructed clinical medical knowledge graph to remove one or more abnormal laboratory test data.

[0008] In some embodiments of the first aspect of this application, the method of constructing multiple primary neuroblastoma bone metastasis risk prediction models based on the internal training set includes: using a minimum absolute contraction and selection operator to perform feature screening on each laboratory test data in the internal training set to obtain multiple key features of neuroblastoma bone metastasis risk, and generating a key feature training set accordingly; wherein, each key feature of neuroblastoma bone metastasis risk includes: bone marrow monocyte ratio, peripheral blood leukocyte count, peripheral blood eosinophil ratio, peripheral blood monocyte absolute count, hemoglobin concentration, and serum neuron-specific enolase level; the key feature training set is input into a logistic regression binary classification model, a random forest binary classification model, a support vector machine binary classification model, a gradient boosting machine binary classification model, a neural network binary classification model, a Naive Bayes binary classification model, and a K-nearest neighbor binary classification model, respectively, to train and obtain multiple converged primary neuroblastoma bone metastasis risk prediction models, and outputting the high-risk threshold of primary bone metastasis predicted by each primary neuroblastoma bone metastasis risk prediction model.

[0009] In some embodiments of the first aspect of this application, the method of comparing the model performance of various primary neuroblastoma bone metastasis risk prediction models to obtain an optimized neuroblastoma bone metastasis risk prediction model includes: generating an internal key feature validation set and an external key feature validation set based on the key features of each neuroblastoma bone metastasis risk, according to the internal validation set and the external validation set; inputting the internal key feature validation set and the external key feature validation set into each primary neuroblastoma bone metastasis risk prediction model, calculating the area under the receiver operating characteristic (ROC) curve of each primary neuroblastoma bone metastasis risk prediction model, and selecting the primary neuroblastoma bone metastasis risk prediction model with the largest ROC curve area as a candidate neuroblastoma bone metastasis risk prediction model; adjusting the model hyperparameters of the candidate neuroblastoma bone metastasis risk prediction model according to the internal key feature validation set and the external key feature validation set, and obtaining the optimized neuroblastoma bone metastasis risk prediction model and the target bone metastasis high-risk threshold predicted by the optimized neuroblastoma bone metastasis risk prediction model.

[0010] In some embodiments of the first aspect of this application, the optimized neuroblastoma bone metastasis risk prediction model is obtained by training and validating a gradient booster binary classification model.

[0011] In some embodiments of the first aspect of this application, the method of obtaining a corresponding primary neuroblastoma bone metastasis risk prediction model based on gradient boosting machine binary classification model training includes: initializing the gradient boosting machine binary classification model and initializing the model prediction value according to the sample ratio of the key feature training set; inputting the key feature training set into the gradient boosting machine binary classification model and iteratively training the gradient boosting machine binary classification model until a preset number of iterations is reached to obtain the corresponding primary neuroblastoma bone metastasis risk prediction model; wherein, the method of training the gradient boosting machine binary classification model in each iteration includes: calculating the pseudo residual of the current model using a logarithmic loss function and training a shallow decision tree for fitting the pseudo residual; updating the current model according to the prediction value of the shallow decision tree and a preset learning rate; verifying the updated current model according to the internal key feature validation set; if the model performance no longer improves, stopping the iterative training and using the current model as the primary neuroblastoma bone metastasis risk prediction model.

[0012] In some embodiments of the first aspect of this application, the method for predicting the bone metastasis risk value and bone metastasis risk classification result of the target neuroblastoma patient includes: generating target neuroblastoma bone metastasis risk key feature data for the target neuroblastoma patient based on the target laboratory test data of the target neuroblastoma patient, according to each neuroblastoma bone metastasis risk key feature data; inputting the target neuroblastoma bone metastasis risk key feature data into the target neuroblastoma bone metastasis risk prediction model to obtain the bone metastasis risk value of the target neuroblastoma patient. The Shapley interpretable algorithm is used to calculate the contribution of each key feature of neuroblastoma bone metastasis risk to the bone metastasis risk value. The bone metastasis risk value is compared with the target high-risk bone metastasis threshold to determine whether the target neuroblastoma child has a low or high risk of neuroblastoma bone metastasis, and a bone metastasis risk classification result is generated for the target neuroblastoma child. Based on the bone metastasis risk value, the bone metastasis risk classification result, and the contribution values ​​of each key feature of neuroblastoma bone metastasis risk, a bone metastasis risk report for the target neuroblastoma child is generated.

[0013] In some embodiments of the first aspect of this application, the neuroblastoma bone metastasis risk prediction system further includes: an electronic medical record integration module, connected to the model prediction module and connected to an external electronic medical record system, for obtaining target laboratory test data or key feature data of target neuroblastoma bone metastasis risk from the electronic medical record system and sending it to the model prediction module; simultaneously, receiving the bone metastasis risk report generated by the model prediction module and sending it to the electronic medical record system; and a human-computer interaction module, connected to the model prediction module, for providing a visual interface to collect target laboratory test data or key feature data of target neuroblastoma bone metastasis risk from the target neuroblastoma patient and send it to the model prediction module, and receiving the bone metastasis risk report generated by the model prediction module for visual display.

[0014] To achieve the above and other related objectives, a second aspect of this application provides a method for predicting the risk of bone metastasis in neuroblastoma. The method includes: constructing a target neuroblastoma bone metastasis risk prediction model; based on the target neuroblastoma bone metastasis risk prediction model, and according to the obtained target laboratory test data of the target neuroblastoma patient, predicting the bone metastasis risk value and bone metastasis risk classification result of the target neuroblastoma patient, and generating a bone metastasis risk report; wherein, the construction of the target neuroblastoma bone metastasis risk prediction model includes: obtaining multiple neuroblastomas... Laboratory test data of the children were collected, and internal training sets, internal validation sets, external validation sets, and external test sets were constructed based on the laboratory test data. Multiple primary neuroblastoma bone metastasis risk prediction models were constructed based on the internal training set. The diagnostic performance of each primary neuroblastoma bone metastasis risk prediction model was compared based on the internal and external validation sets to obtain an optimized neuroblastoma bone metastasis risk prediction model. Based on the external test set, the optimized neuroblastoma bone metastasis risk prediction model was evaluated, and the model that met the prediction requirements was identified as the target neuroblastoma bone metastasis risk prediction model.

[0015] To achieve the above and other related objectives, a third aspect of this application provides a neuroblastoma bone metastasis risk prediction terminal, the neuroblastoma bone metastasis risk prediction terminal comprising: a memory and a processor; the memory for storing a computer program; the processor for executing the computer program stored in the memory, so that the terminal performs the functions of the neuroblastoma bone metastasis risk prediction system described in any of the above embodiments.

[0016] As described above, this application provides a system, method, and terminal for predicting the risk of bone metastasis in neuroblastoma. It acquires laboratory test data from multiple children with neuroblastoma and constructs internal training sets, internal validation sets, external validation sets, and external test sets to build a target neuroblastoma bone metastasis risk prediction model. Based on the target laboratory test data of the target neuroblastoma children, it predicts the bone metastasis risk value and risk classification results for the target neuroblastoma children. This application has the following beneficial effects: it can provide timely early warning of disease changes, reminding children and their families to seek medical attention as soon as possible. It also helps medical staff make medical decisions, provides diagnostic evidence, and solves the technical problems of limited early disease warning, high technical requirements, poor applicability, and poor safety in existing imaging techniques for detecting neuroblastoma bone metastasis. Attached Figure Description

[0017] Figure 1 The diagram shown is a structural schematic of a neuroblastoma bone metastasis risk prediction system according to an embodiment of this application.

[0018] Figure 2 The diagram shown is a flowchart illustrating the process of constructing a risk prediction model for bone metastasis of target neuroblastoma in one embodiment of this application.

[0019] Figure 3 The diagram shown is a flowchart illustrating the training process of a gradient booster binary classification model in one embodiment of this application.

[0020] Figure 4 The diagram shows a flowchart of the prediction of bone metastasis risk value and bone metastasis risk classification results in one embodiment of this application.

[0021] Figure 5 The diagram shown is a visualization of bone metastasis risk classification results in one embodiment of this application.

[0022] Figure 6 The diagram shown is a flowchart illustrating a method for predicting the risk of bone metastasis in neuroblastoma according to an embodiment of this application.

[0023] Figure 7 The diagram shown is a structural schematic of a neuroblastoma bone metastasis risk prediction terminal in one embodiment of this application. Detailed Implementation

[0024] The following specific examples illustrate the implementation of this application. Those skilled in the art can understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.

[0025] In the embodiments of this application, terms such as "first" and "second" are used to distinguish identical or similar items with essentially the same function and effect. For example, the first ratio and the second ratio are merely used to distinguish different ratio values ​​and do not limit their order. Those skilled in the art will understand that terms such as "first" and "second" do not limit the quantity or execution order, and that terms such as "first" and "second" do not necessarily imply that they are different.

[0026] To address the problems mentioned above in the background technology, this application provides a neuroblastoma bone metastasis risk prediction system, method, and terminal. The system aims to predict the bone metastasis risk value and risk classification results of target neuroblastoma children based on target laboratory test data using a pre-constructed target neuroblastoma bone metastasis risk prediction model. This solves the technical problems of existing imaging techniques for detecting neuroblastoma bone metastasis, including limited early disease warning, high technical requirements, poor applicability, and poor safety.

[0027] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this application are further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for explaining this application and are not intended to limit this application.

[0028] like Figure 1 The diagram illustrates the structure of a neuroblastoma bone metastasis risk prediction system according to an embodiment of this application. The neuroblastoma bone metastasis risk prediction system in this embodiment mainly includes a model building module and a model prediction module.

[0029] The model building module is used to construct a risk prediction model for bone metastasis of the target neuroblastoma.

[0030] In one embodiment, such as Figure 2 As shown, the construction method of the target neuroblastoma bone metastasis risk prediction model includes the following steps.

[0031] Step S11: Obtain laboratory test data from multiple children with neuroblastoma, and construct an internal training set, an internal validation set, an external validation set, and an external test set based on the laboratory test data.

[0032] In one embodiment, the method for obtaining laboratory test data includes the following steps.

[0033] ① Clinical characteristic data, peripheral blood sample test data, and bone marrow sample test data of multiple children with neuroblastoma were obtained from multiple medical institutions to serve as laboratory test data for each child with neuroblastoma.

[0034] The patient's clinical characteristics data include at least the following: gender, age, primary tumor location, initial tumor grade, bone metastasis status, and bone marrow metastasis status of the corresponding neuroblastoma child; the peripheral blood sample test data include at least the following: total white blood cell count, neutrophil count, lymphocyte count, monocyte count, eosinophil count, basophil count, red blood cell count, hemoglobin concentration, hematocrit, mean corpuscular hematocrit, mean corpuscular hemoglobin concentration, platelet count, and serum neuron-specific enolase level; the bone marrow sample test data include at least the following: degree of bone marrow hyperplasia, proportion of granulocytes, proportion of erythroids, number of megakaryocytes, proportion of lymphocytes, proportion of monocytes, and whether metastatic cancer cells are present.

[0035] Therefore, the laboratory test data obtained based on the patient's clinical characteristics, peripheral blood sample test data, and bone marrow sample test data includes multiple feature parameters based on prior clinical knowledge, involving both textual data and structured quantitative data. This laboratory test data can be obtained through routine blood tests on peripheral blood samples and bone marrow smear cytology tests on bone marrow samples from children with neuroblastoma. This requires only blood draws and bone marrow aspirations, resulting in low acquisition costs, minimal trauma to the children, no radiation exposure risk, and low price. Repeated sampling and follow-up are also possible. Therefore, constructing a target neuroblastoma bone metastasis risk prediction model based on this laboratory test data to predict the bone metastasis risk in children with target neuroblastoma is cost-effective and suitable for widespread adoption. Even in areas with limited medical resources, lacking imaging equipment or imaging specialists, this target neuroblastoma bone metastasis risk prediction model can help identify the risk of bone metastasis in children early, enabling timely medical intervention, controlling disease progression, and improving cure rates and prognosis.

[0036] ② Based on the pre-constructed clinical medical knowledge graph, medical logic is reviewed on the laboratory test data of each laboratory, and one or more abnormal laboratory test data are removed.

[0037] To ensure the authenticity of laboratory test data, each laboratory's test data needs to undergo medical logic review to eliminate abnormal laboratory test data that does not conform to the clinical course of the disease or the laws of cell development.

[0038] The clinical medical knowledge graph is a large-scale network that organizes and represents clinical medical knowledge in a structured and semantic manner. Nodes in the knowledge graph represent clinical entities, such as diseases, symptoms, examinations, tests, and cell characteristics, while edges represent the relationships between these entities. This knowledge graph can be constructed based on medical textbooks, clinical guidelines, medical literature, standard medical terminology sets, and the clinical experience of senior physicians. This allows for the identification of abnormal data in laboratory test data, such as data that does not conform to the clinical course of the disease or the laws of cell development. It enables quality control of laboratory test data, ensuring the quality of the internal training set, internal validation set, external validation set, and external test set constructed based on the test data from each laboratory. This ensures that the target neuroblastoma bone metastasis risk prediction model is trained on high-quality data that is real, accurate, and biologically logical, resulting in more clinically valuable feature interactions captured by the model and superior diagnostic performance.

[0039] In this embodiment, the method of constructing the internal training set, the internal validation set, the external validation set, and the external test set based on the test data of each laboratory includes the following steps.

[0040] ①Based on a preset first ratio, laboratory test data from multiple children with neuroblastoma obtained from this medical institution are randomly selected to construct the internal training set and the internal validation set respectively.

[0041] ② Based on a preset second ratio, laboratory test data of multiple neuroblastoma patients obtained from multiple other medical institutions are randomly selected to construct the external validation set and the external test set respectively.

[0042] It should be noted that the first ratio and the second ratio can be set by the user according to their needs, and this application does not specifically limit them.

[0043] Step S12: Based on the internal training set, construct multiple primary neuroblastoma bone metastasis risk prediction models respectively.

[0044] Specifically, step S12 includes the following steps.

[0045] ① The minimum absolute contraction and selection operator is used to screen the laboratory test data in the internal training set to obtain multiple key features of neuroblastoma bone metastasis risk, and a key feature training set is generated accordingly.

[0046] It should be understood that the Least Absolute Shrinkage and Selection Operator (LASSO) is a feature selection method for regression analysis. Its core principle is to shrink and select feature coefficients by introducing an L1 regularization term. It has advantages such as automatic feature selection, processing of high-dimensional data, and suppression of multicollinearity.

[0047] The bone marrow sample test data in the laboratory test data can provide the cytological characteristics of the child's bone marrow. Since peripheral blood cells originate from the bone marrow, the peripheral blood sample test data and the bone marrow sample test data are multicollinear and influence each other. Based on this, this application first uses the LASSO algorithm to perform regression analysis on the laboratory test data. From the numerous features of the laboratory test data, it automatically selects several key features of neuroblastoma bone metastasis risk that are most important for the prediction target, namely, the prediction of neuroblastoma bone metastasis risk. Based on this, a training set of the key features is generated. This not only limits the model complexity of each primary neuroblastoma bone metastasis risk prediction model and improves the model's generalization ability on unknown data, but also improves the model's prediction accuracy and ensures the diagnostic performance of the final target neuroblastoma bone metastasis risk prediction model.

[0048] Key characteristics for the risk of bone metastasis in various neuroblastomas include: bone marrow mononuclear cell ratio, peripheral blood leukocyte count, peripheral blood eosinophil ratio, absolute peripheral blood mononuclear cell count, hemoglobin concentration, and serum neuron-specific enolase level.

[0049] ② Input the key feature training set into the logistic regression binary classification model, random forest binary classification model, support vector machine binary classification model, gradient boosting machine binary classification model, neural network binary classification model, Naive Bayes binary classification model and K-nearest neighbor binary classification model respectively, train to obtain multiple convergent primary neuroblastoma bone metastasis risk prediction models, and output the high-risk threshold of primary bone metastasis predicted by each primary neuroblastoma bone metastasis risk prediction model.

[0050] Specifically, multiple binary classification models are constructed using the "caret" package in the R language software. The training set of the key features is input into each binary classification model, including: Logistic Regression (LR), Random Forest (RF), Support Vector Machine (SVM), Gradient Boosting Machine (GBM), Neural Network (N-Net), Naive Bayes (NB), and K-Nearest Neighbors (KNN). This yields multiple corresponding primary neuroblastoma bone metastasis risk prediction models, which are used to perform binary classification tasks to identify the bone metastasis risk value of each neuroblastoma patient and generate low-risk or high-risk bone metastasis classification results. The models also output primary high-risk bone metastasis thresholds used to distinguish between these classifications.

[0051] Step S13: Based on the internal validation set and the external validation set respectively, compare the diagnostic performance of each primary neuroblastoma bone metastasis risk prediction model to obtain the optimized neuroblastoma bone metastasis risk prediction model.

[0052] Specifically, step S13 includes the following steps.

[0053] ① Based on the key features of bone metastasis risk of each neuroblastoma, an internal key feature validation set and an external key feature validation set are generated according to the internal validation set and the external validation set, respectively.

[0054] That is, the feature values ​​of each key feature of neuroblastoma bone metastasis risk in the laboratory test data of each laboratory in the internal validation set are screened to generate multiple corresponding first key feature data of neuroblastoma bone metastasis risk, thereby constructing an internal key feature validation set and an external key feature validation set; wherein, each first key feature data of neuroblastoma bone metastasis risk includes the feature values ​​of each key feature of neuroblastoma bone metastasis risk in a corresponding neuroblastoma patient.

[0055] Feature values ​​of key features of neuroblastoma bone metastasis risk in laboratory test data from the external validation set are selected to generate multiple corresponding second neuroblastoma bone metastasis risk key feature data, thereby constructing an internal key feature validation set and an external key feature validation set respectively; wherein, each second neuroblastoma bone metastasis risk key feature data includes feature values ​​of key features of neuroblastoma bone metastasis risk for a corresponding neuroblastoma patient.

[0056] ② Input the internal key feature validation set and the external key feature validation set into each primary neuroblastoma bone metastasis risk prediction model, and calculate the area under the receiver operating characteristic curve (AUC) of each primary neuroblastoma bone metastasis risk prediction model. Select the primary neuroblastoma bone metastasis risk prediction model with the largest AUC as the candidate neuroblastoma bone metastasis risk prediction model.

[0057] It should be understood that the Area Under the Receiver Operating Characteristic (ROC) curve (AUC) is used to describe the performance of a model in a binary classification task. Specifically, AUC measures the ability of a classification model to distinguish between "positive" and "negative" examples. The AUC value describes the probability that the target model is more likely to predict a positive sample as positive than the probability that it is to predict a negative sample as positive for a randomly selected positive sample and a negative sample as negative. The higher the AUC value, the better the overall discriminative ability of the model. This application uses AUC to evaluate the diagnostic performance of various primary neuroblastoma bone metastasis risk prediction models, and then selects the optimal primary neuroblastoma bone metastasis risk prediction model as a candidate neuroblastoma bone metastasis risk prediction model.

[0058] In a preferred embodiment, the model performance metrics of each primary neuroblastoma bone metastasis risk prediction model obtained based on the key feature training set, and the model performance metrics of each primary neuroblastoma bone metastasis risk prediction model verified based on the internal key feature validation set and the external key feature validation set are shown in Table 1. As shown in Table 1, the Gradient Boosting Machine (GBM) binary classification model has an AUC of 0.931, accuracy of 0.880, sensitivity of 0.833, specificity of 0.906, and Brier Score of 0.111, demonstrating good predictive ability and diagnostic performance. Therefore, the corresponding primary neuroblastoma bone metastasis risk prediction model is selected as the candidate neuroblastoma bone metastasis risk prediction model.

[0059] The Brier Score is a scoring metric used to evaluate the accuracy of various primary neuroblastoma bone metastasis risk prediction models. It measures the mean squared error between the predicted probability and the actual occurrence. The closer the predicted probability is to the true outcome (1 or 0), the lower the score, and the better the model.

[0060] Table 1. Model performance metrics of the primary neuroblastoma bone metastasis risk prediction model

[0061]

[0062] ③ Based on the internal key feature validation set and the external key feature validation set, adjust the model hyperparameters of the candidate neuroblastoma bone metastasis risk prediction model to obtain the optimized neuroblastoma bone metastasis risk prediction model and the target high-risk threshold for bone metastasis predicted by the optimized neuroblastoma bone metastasis risk prediction model.

[0063] Preferably, the optimized neuroblastoma bone metastasis risk prediction model is obtained by training and validating a gradient booster binary classification model.

[0064] It should be understood that GBM is a powerful ensemble learning algorithm that primarily constructs a high-precision prediction model by iteratively combining multiple weak learners (such as decision trees). Each new tree is dedicated to correcting the errors of the previous tree, and the predictions of all trees are weighted and summed to form a powerful and accurate target model. Its core objective is to reduce bias and pursue extremely high prediction accuracy.

[0065] The laboratory test data used in the internal training dataset constructed in this application is relatively complex, containing both textual and structured quantitative data. The gradient boosting machine binary classification model demonstrates a highly targeted core advantage when processing this laboratory test data. It eliminates the need for dimensionality reduction of complex data, simplifies the data processing flow, and preserves the clinical information in the original data to the greatest extent possible, avoiding information loss. It is highly compatible with the complex laboratory test data of this application.

[0066] More importantly, in binary classification prediction tasks, GBM's Boosting mechanism has significant advantages, enabling it to fit the complex and subtle patterns in the laboratory test data in a "refined" manner, thus providing extremely high prediction accuracy. The gradient boosting machine binary classification model, through its unique Boosting mechanism, can iteratively combine multiple weak learners. By continuously learning the residuals, the entire ensemble model moves along the direction of the fastest descent of the preset logarithmic loss function, more proactively and deeply capturing the complex nonlinear relationships and interactions between multiple features. This is beneficial for maximizing the predicted AUC value, thereby improving the model's accuracy and robustness in predicting the risk of neuroblastoma bone metastasis.

[0067] In one embodiment, such as Figure 3 As shown, the method for obtaining the corresponding primary neuroblastoma bone metastasis risk prediction model based on gradient boosting machine binary classification model training, and then obtaining the optimized neuroblastoma bone metastasis risk prediction model includes the following steps.

[0068] Step S131: Initialize the gradient boosting machine binary classification model and initialize the model prediction value according to the sample ratio of the training set of the key features.

[0069] Among them, the model prediction value of the gradient booster binary classification model can be initialized as the log odds of the proportion of positive samples.

[0070] Step S132: Calculate the pseudo residual of the current model using the log loss function, train a shallow decision tree to fit the pseudo residual, and update the current model based on the predicted value of the shallow decision tree and the preset learning rate.

[0071] Step S133: Verify the updated current model based on the internal key feature verification set to determine whether the model performance of the current model has improved.

[0072] If the model performance improves, determine whether the preset number of iterations has been reached. If the preset number of iterations has been reached, execute step S134 to stop iterative training and use the current model as the primary neuroblastoma bone metastasis risk prediction model. However, if the preset number of iterations has not been reached, repeat steps S132 and S133 to iteratively train the gradient booster binary classification model until the preset number of iterations is reached.

[0073] If the model performance no longer improves, proceed to step S134 to stop iterative training and use the current model as a primary neuroblastoma bone metastasis risk prediction model. In this embodiment, this application employs an early stopping method to monitor model performance during iterative training. Once the model performance no longer improves or begins to decline, iterative training is immediately stopped to lock in the optimal number of iterations and prevent the model from "memorizing" the training data or overfitting.

[0074] In a preferred embodiment, the gradient boosting machine binary classification model can be trained by constructing a parameter grid, with the following configuration parameters: 100 shallow decision trees (i.e., 100 iterations), an interaction depth of 5, a learning rate of 0.1, and a minimum number of samples per leaf node of 10.

[0075] In this embodiment, a smaller learning rate is used to control the contribution of each shallow decision tree, making the learning process more robust. Simultaneously, when training each shallow decision tree to fit the pseudo-residual of the current model, 50%-80% of the samples or features are randomly selected for randomness constraints to enhance the generalization ability of the obtained primary neuroblastoma bone metastasis risk prediction model, that is, to enhance the generalization ability of the optimized neuroblastoma bone metastasis risk prediction model and the target neuroblastoma bone metastasis risk prediction model.

[0076] However, it should be noted that users can set the configuration parameters of the model according to their needs, and this application does not specifically limit them.

[0077] Step S14: Based on the external test set, evaluate the optimized neuroblastoma bone metastasis risk prediction model, and determine it as the target neuroblastoma bone metastasis risk prediction model when it meets the prediction requirements.

[0078] This application uses laboratory test data from multiple children with neuroblastoma from various other medical institutions to simulate the real world, evaluate the optimized neuroblastoma bone metastasis risk prediction model, and verify the model's generalization and universality. Preferably, the area under the receiver operating characteristic curve (AUC) is used to evaluate the optimized neuroblastoma bone metastasis risk prediction model. The AUC value of the optimized neuroblastoma bone metastasis risk prediction model is greater than 0.75, which meets the diagnostic performance and predictive requirements of the model. Therefore, the optimized neuroblastoma bone metastasis risk prediction model is determined as the target neuroblastoma bone metastasis risk prediction model.

[0079] In a preferred embodiment, the target neuroblastoma bone metastasis risk prediction model is obtained by training and validating a gradient booster binary classification model, and its high-risk threshold for target bone metastasis is 0.332.

[0080] like Figure 1As shown, the model prediction module is connected to the model construction module. The model prediction module is used to predict the bone metastasis risk value and bone metastasis risk classification result of the target neuroblastoma child based on the target neuroblastoma bone metastasis risk prediction model and the target laboratory test data of the target neuroblastoma child, and generate a bone metastasis risk report.

[0081] In one embodiment, such as Figure 4 As shown, the method for predicting the bone metastasis risk value and bone metastasis risk classification results of the target neuroblastoma children includes the following steps.

[0082] Step S21: Based on the key features of bone metastasis risk of each neuroblastoma, generate the key feature data of bone metastasis risk of the target neuroblastoma in the target neuroblastoma patient according to the target laboratory test data of the target neuroblastoma patient.

[0083] The target laboratory test data includes: clinical characteristic data of the target neuroblastoma patient, peripheral blood sample test data, and bone marrow sample test data. The target patient clinical characteristic data includes at least: gender, age, primary tumor location, initial tumor grade, bone metastasis status, and bone marrow metastasis status of the target neuroblastoma patient. The target peripheral blood sample test data includes at least: total white blood cell count, neutrophil count, lymphocyte count, monocyte count, eosinophil count, basophil count, red blood cell count, hemoglobin concentration, hematocrit, mean corpuscular hematocrit, mean corpuscular hemoglobin concentration, platelet count, and serum neuron-specific enolase level. The target bone marrow sample test data includes at least: bone marrow hyperplasia level, granulocyte percentage, erythroid percentage, megakaryocyte count, lymphocyte percentage, monocyte percentage, and whether metastatic cancer cells are present.

[0084] The method for generating key feature data on the risk of bone metastasis of the target neuroblastoma based on the target laboratory test data of the target neuroblastoma patients includes: screening the feature values ​​of each key feature of the risk of bone metastasis of neuroblastoma in the target laboratory test data to generate corresponding key feature data on the risk of bone metastasis of the target neuroblastoma. The key feature data on the risk of bone metastasis of the target neuroblastoma includes the feature values ​​of each key feature of the risk of bone metastasis of neuroblastoma in the target neuroblastoma patients. Each key feature of the risk of bone metastasis of neuroblastoma includes: bone marrow monocyte percentage, peripheral blood leukocyte count, peripheral blood eosinophil percentage, absolute peripheral blood monocyte count, hemoglobin concentration, and serum neuron-specific enolase level.

[0085] Step S22: Input the key feature data of the target neuroblastoma bone metastasis risk into the target neuroblastoma bone metastasis risk prediction model to obtain the bone metastasis risk value of the target neuroblastoma child, and use the Shapley interpretable algorithm to calculate the contribution value of each key feature of neuroblastoma bone metastasis risk to the bone metastasis risk value.

[0086] In this embodiment, to clearly interpret the predictions of the target neuroblastoma bone metastasis risk prediction model and increase transparency, after obtaining the bone metastasis risk value of the target neuroblastoma child, the target neuroblastoma bone metastasis risk prediction model is interpreted using SHAP (SHapley Additive exPlanations). This provides a consistent and interpretable numerical value for the contribution of each key neuroblastoma bone metastasis risk feature to the prediction of the bone metastasis risk value. Specifically, an interpreter can be selected to calculate the contribution value of each key neuroblastoma bone metastasis risk feature to the bone metastasis risk value, where the bone metastasis risk value is equal to the sum of the contributions.

[0087] Step S23: Compare the bone metastasis risk value with the target bone metastasis high-risk threshold to determine whether the target neuroblastoma child has a low or high risk of neuroblastoma bone metastasis, and generate the bone metastasis risk classification result of the target neuroblastoma child.

[0088] Specifically, if the bone metastasis risk value is less than the target high-risk bone metastasis threshold, the target neuroblastoma patient is determined to have a low-risk bone metastasis risk, and the bone metastasis risk classification result is a low-risk bone metastasis classification result; otherwise, the target neuroblastoma patient is determined to have a high-risk bone metastasis risk, and the bone metastasis risk classification result is a high-risk bone metastasis classification result. Preferably, the target high-risk bone metastasis threshold is 0.332. If the bone metastasis risk value is less than 0.332, the target neuroblastoma patient is determined to have a low-risk bone metastasis risk, and the bone metastasis risk classification result is a low-risk bone metastasis classification result; otherwise, the target neuroblastoma patient is determined to have a high-risk bone metastasis risk, and the bone metastasis risk classification result is a high-risk bone metastasis classification result.

[0089] Step S24: Generate a bone metastasis risk report for the target neuroblastoma child based on the bone metastasis risk value, the bone metastasis risk classification result, and the contribution value of each neuroblastoma bone metastasis risk key feature.

[0090] The bone metastasis risk report includes, but is not limited to: the bone metastasis risk value of the target neuroblastoma patient, the contribution value of each key neuroblastoma bone metastasis risk feature to the bone metastasis risk value, the bone metastasis risk classification result, and medical advice. The bone metastasis risk classification result includes: low-risk bone metastasis classification result and high-risk bone metastasis classification result.

[0091] This application utilizes the target neuroblastoma bone metastasis risk prediction model to predict the risk of bone metastasis in the early stages of the disease based on multiple indicators in the target laboratory test data. This provides early warning of disease changes and reminds children and their families to seek medical attention as soon as possible when there is a high risk of bone metastasis, and to conduct further testing and diagnosis. At the same time, it also helps medical staff make medical decisions and provides them with diagnostic basis, including examination plans, treatment plans, and prognostic assessments.

[0092] In one embodiment, such as Figure 1 As shown, the neuroblastoma bone metastasis risk prediction system also includes an electronic medical record integration module.

[0093] Specifically, such as Figure 1 As shown, the electronic medical record integration module is connected to the model prediction module and to an external electronic medical record system.

[0094] The electronic medical record integration module is used to obtain the target laboratory test data or key feature data of target neuroblastoma bone metastasis risk of the target neuroblastoma child from the electronic medical record system, and send them to the model prediction module.

[0095] Meanwhile, the electronic medical record integration module is also used to receive the bone metastasis risk report generated by the model prediction module and send it to the electronic medical record system to update the electronic medical record of the target neuroblastoma child, which facilitates the management of the target neuroblastoma child's condition and helps medical staff make more accurate medical decisions based on comprehensive medical record data.

[0096] In one embodiment, such as Figure 1 As shown, the neuroblastoma bone metastasis risk prediction system also includes a human-computer interaction module.

[0097] Specifically, such as Figure 1 As shown, the human-computer interaction module is connected to the model prediction module. The human-computer interaction module provides a visual interface to collect target laboratory test data or key characteristic data of target neuroblastoma bone metastasis risk from the target neuroblastoma child and send them to the model prediction module. It also receives the bone metastasis risk report generated by the model prediction module and displays it visually.

[0098] In a preferred embodiment, an interactive webpage is constructed using the "shiny" package in the R language software as a human-computer interaction module. This allows users, such as medical staff or family members of the target neuroblastoma patient, to directly upload the target laboratory test data, including the target peripheral blood sample test data and the bone marrow sample test data, or directly input the key characteristic data of the target neuroblastoma bone metastasis risk (including the proportion of bone marrow mononuclear cells, peripheral blood white blood cell count, peripheral blood eosinophil proportion, absolute peripheral blood mononuclear cell count, hemoglobin concentration, and characteristic values ​​of serum neuron-specific enolase level), thereby obtaining the bone metastasis risk report of the target neuroblastoma patient in real time.

[0099] The human-computer interaction module is also used to visually display the bone metastasis risk report of the target neuroblastoma child. In a preferred embodiment, the bone metastasis risk report, including the bone metastasis risk value, the contribution of each key neuroblastoma bone metastasis risk feature to the bone metastasis risk value, the bone metastasis risk classification results, and medical advice, can be displayed on an interactive webpage.

[0100] For example, an interactive webpage might display a pie chart on the right side of the page, such as... Figure 5 As shown, this pie chart represents the risk of bone metastasis in patients with neuroblastoma. The pie chart can be composed of two colors, such as blue or red, to represent low-risk and high-risk bone metastasis classifications, respectively. The bone metastasis risk value can be displayed in the center of the pie chart. Figure 5 As shown, the Risk-filled blocks in the pie chart represent the high-risk classification results for bone metastasis, the non-Risk-filled blocks represent the low-risk classification results for bone metastasis, and the Risk text in the middle of the pie chart represents the specific bone metastasis risk value.

[0101] The interactive webpage can also generate feature importance bar charts and bee colony diagrams to provide a global explanation of the decision-making logic of the target neuroblastoma bone metastasis risk prediction model. This allows users to intuitively understand the importance ranking and overall distribution of each key feature of neuroblastoma bone metastasis risk when the model performs its prediction task. The feature importance bar chart is drawn based on the average contribution value of each key feature of neuroblastoma bone metastasis risk in each iteration of the training task. The length of the bar visually displays the average importance ranking of each key feature of neuroblastoma bone metastasis risk on the overall training set, providing a global and comprehensive perspective and quickly answering the core question of "which features are most important". In a bee swarm plot, the horizontal position of a point (i.e., the X-axis) represents the specific contribution of a key feature of neuroblastoma bone metastasis risk to the prediction of a single sample. A positive contribution increases the prediction, while a negative contribution decreases it. The vertical packing density of points shows the distribution of affected samples. Points can also be displayed using different colors, such as yellow for high feature values ​​and purple for low feature values, and these colors are correlated with the horizontal position of the point, further revealing the inherent pattern of how key features of neuroblastoma bone metastasis risk affect the prediction of a single sample. As a data visualization chart, the bee swarm plot clearly shows the distribution of data by slightly dispersing data points along the axis, effectively avoiding the problem of point overlap in traditional scatter plots. This allows it to simultaneously display the overall distribution pattern and individual point information of key features of neuroblastoma bone metastasis risk.

[0102] The interactive webpage can also generate force graphs to interpret the prediction results of individual samples, specifically providing a local interpretation of the contribution values ​​of each key feature of neuroblastoma bone metastasis risk in the bone metastasis risk prediction task for the target neuroblastoma child. It visually demonstrates how the contribution of each key feature of neuroblastoma bone metastasis risk is superimposed on the target high-risk bone metastasis threshold to ultimately obtain the bone metastasis risk value for the target neuroblastoma child. For a specific prediction sample, the contribution values ​​of each key feature of neuroblastoma bone metastasis risk are superimposed on the target high-risk bone metastasis threshold. A positive SHAP value (yellow feature bar to the right) represents an increased risk of bone metastasis, while a negative SHAP value (purple-red feature bar to the left) represents a decreased risk. The final accumulated bone metastasis risk value being greater than the target high-risk bone metastasis threshold indicates a high-risk bone metastasis classification result.

[0103] This application, through its interactive webpage of the human-computer interaction module, enables early warning of disease changes and access to medical advice even in situations with limited medical resources, such as the absence of high-precision imaging equipment or professional medical staff. This helps avoid delays in treatment and missing the optimal window for intervention. Furthermore, the interactive webpage of the human-computer interaction module supports user self-operation, with a simple and easy-to-understand interface. It is convenient for families of children with neuroblastoma or medical staff to operate, significantly reducing medical costs and supporting its widespread adoption in areas with scarce medical resources.

[0104] It should be understood that the module division in the embodiments of this application is illustrative and only represents a logical functional division. In actual implementation, there may be other division methods. Furthermore, the functional modules in the various embodiments of this application can be integrated into a single processor or functional module, exist as separate physical entities, or be divided into more functional modules. The integrated modules or units described above can be implemented in hardware or as software functional modules.

[0105] like Figure 6 The diagram illustrates a flowchart of a method for predicting the risk of bone metastasis in neuroblastoma according to an embodiment of this application. The method for predicting the risk of bone metastasis in neuroblastoma according to this embodiment mainly includes the following steps.

[0106] Step S1: Construct a risk prediction model for bone metastasis of the target neuroblastoma.

[0107] Wherein, step S1 is as follows Figure 2 As shown, it includes the following steps.

[0108] Step S11: Obtain laboratory test data from multiple children with neuroblastoma, and construct an internal training set, an internal validation set, an external validation set, and an external test set based on the laboratory test data.

[0109] Step S12: Based on the internal training set, construct multiple primary neuroblastoma bone metastasis risk prediction models respectively.

[0110] Step S13: Based on the internal validation set and the external validation set respectively, compare the diagnostic performance of each primary neuroblastoma bone metastasis risk prediction model to obtain the optimized neuroblastoma bone metastasis risk prediction model.

[0111] Step S14: Based on the external test set, evaluate the optimized neuroblastoma bone metastasis risk prediction model, and determine it as the target neuroblastoma bone metastasis risk prediction model when it meets the prediction requirements.

[0112] Step S2: Based on the target neuroblastoma bone metastasis risk prediction model, and according to the target laboratory test data of the target neuroblastoma patient, predict the bone metastasis risk value and bone metastasis risk classification result of the target neuroblastoma patient, and generate a bone metastasis risk report.

[0113] It should be noted that the neuroblastoma bone metastasis risk prediction method and the neuroblastoma bone metastasis risk prediction system provided in the above embodiments belong to the same concept. The specific method steps are detailed in the system embodiments and will not be repeated here.

[0114] Figure 7 This is a schematic diagram of the structure of the neuroblastoma bone metastasis risk prediction terminal 700 provided in an embodiment of this application. Figure 7 As shown, the neuroblastoma bone metastasis risk prediction terminal 700 includes: at least one processor 701, a memory 702, at least one network interface 703, and a user interface 705. The various components in the terminal are coupled together via a bus system 704. It is understood that the bus system 704 is used to implement communication between these components. In addition to a data bus, the bus system 704 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in... Figure 7 The general will label all buses as bus systems.

[0115] The user interface 705 may include a monitor, keyboard, mouse, trackball, clicker, button, touchpad, or touch screen.

[0116] It is understood that memory 702 can be volatile memory or non-volatile memory, or both. Non-volatile memory can be read-only memory (ROM) or programmable read-only memory (PROM), which serves as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM) and synchronous static random access memory (SSRAM). The memories described in the embodiments of this application are intended to include, but are not limited to, these and any other suitable categories of memory.

[0117] In this embodiment, the memory 702 is used to store various types of data to support the operation of the neuroblastoma bone metastasis risk prediction terminal 700. Examples of this data include any executable program that operates on the neuroblastoma bone metastasis risk prediction terminal 700, such as operating system 7021 and application program 7022; operating system 7021 includes various system programs, such as framework layer, core library layer, driver layer, etc., for implementing various basic services and handling hardware-based tasks. Application program 7022 may include various applications, such as media player, browser, etc., for implementing various application services. The implementation of the neuroblastoma bone metastasis risk prediction method provided in this embodiment can be included in application program 7022.

[0118] The neuroblastoma bone metastasis risk prediction method disclosed in the above embodiments of this application can be applied to, or implemented by, processor 701. Processor 701 may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the neuroblastoma bone metastasis risk prediction method can be completed by the integrated logic circuitry of the hardware in processor 701 or by instructions in software form. The processor 701 may be a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 701 can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor 701 may be a microprocessor or any conventional processor, etc. The steps of the neuroblastoma bone metastasis risk prediction method provided in the embodiments of this application can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor. The software modules may be located in a storage medium, which is located in a memory. The processor reads information from the memory and, in conjunction with its hardware, completes the steps of the aforementioned method.

[0119] In an exemplary embodiment, the neuroblastoma bone metastasis risk prediction terminal 700 may be used by one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), or complex programmable logic devices (CPLDs) to perform the aforementioned method.

[0120] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented using computer program-related hardware. The aforementioned computer program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0121] In the embodiments provided in this application, the computer-readable and writable storage medium may include read-only memory, random access memory, EEPROM, CD-ROM or other optical disc storage devices, disk storage devices or other magnetic storage devices, flash memory, USB flash drive, portable hard drive, or any other medium capable of storing desired program code in the form of instructions or data structures and accessible by a computer. Additionally, any connection may be appropriately referred to as a computer-readable medium. For example, if instructions are transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave, then the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of medium. However, it should be understood that computer-readable and writable storage media and data storage media do not include connections, carrier waves, signals, or other transient media, but are intended for non-transient, tangible storage media. The disks and optical discs used in the application include compact optical discs (CDs), laser optical discs, optical discs, digital multifunction optical discs (DVDs), floppy disks, and Blu-ray discs, where disks typically copy data magnetically, while optical discs use lasers to copy data optically.

[0122] In summary, this application provides a system, method, and terminal for predicting the risk of bone metastasis in neuroblastoma. By acquiring laboratory test data from multiple children with neuroblastoma and constructing internal training sets, internal validation sets, external validation sets, and external test sets, a target neuroblastoma bone metastasis risk prediction model is built. Based on the target laboratory test data of the target neuroblastoma children, the system predicts the bone metastasis risk value and risk classification results for the target neuroblastoma children. This allows for timely early warning of disease changes, reminding children and their families to seek medical attention as soon as possible. It also assists medical staff in making medical decisions and provides diagnostic evidence, solving the technical problems of limited early disease warning, high technical requirements, poor applicability, and poor safety in existing imaging techniques for detecting neuroblastoma bone metastasis. Therefore, this application effectively overcomes the various shortcomings of existing technologies and has easy promotion and high industrial application value.

[0123] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the claims of this application.

Claims

1. A neuroblastoma bone metastasis risk prediction system, characterized in that, include: The model building module is used to build a model for predicting the risk of bone metastasis in target neuroblastoma. The model prediction module, connected to the model construction module, is used to predict the bone metastasis risk value and bone metastasis risk classification result of the target neuroblastoma child based on the target neuroblastoma bone metastasis risk prediction model and the target laboratory test data of the target neuroblastoma child, and generate a bone metastasis risk report. The construction method of the target neuroblastoma bone metastasis risk prediction model includes: We obtained laboratory test data from multiple children with neuroblastoma and constructed internal training sets, internal validation sets, external validation sets, and external test sets based on the test data from each laboratory. Based on the internal training set, multiple primary neuroblastoma bone metastasis risk prediction models were constructed. Based on the internal validation set and the external validation set respectively, the diagnostic performance of each primary neuroblastoma bone metastasis risk prediction model is compared to obtain the optimized neuroblastoma bone metastasis risk prediction model. Based on the external test set, the optimized neuroblastoma bone metastasis risk prediction model is evaluated, and when it meets the prediction requirements, it is determined as the target neuroblastoma bone metastasis risk prediction model.

2. The neuroblastoma bone metastasis risk prediction system according to claim 1, characterized in that, Methods for obtaining laboratory test data include: Clinical characteristic data, peripheral blood sample test data, and bone marrow sample test data of multiple children with neuroblastoma were obtained from multiple medical institutions to serve as laboratory test data for each child with neuroblastoma. Based on a pre-constructed clinical medical knowledge graph, medical logic is reviewed on the laboratory test data of each laboratory, and one or more abnormal laboratory test data are removed.

3. The neuroblastoma bone metastasis risk prediction system according to claim 1, characterized in that, Based on the aforementioned internal training set, the methods for constructing multiple primary neuroblastoma bone metastasis risk prediction models include: The minimum absolute contraction and selection operator were used to screen the laboratory test data in the internal training set to obtain multiple key features of neuroblastoma bone metastasis risk, and a key feature training set was generated accordingly. Among them, each key feature of neuroblastoma bone metastasis risk includes: bone marrow monocyte ratio, peripheral blood leukocyte count, peripheral blood eosinophil ratio, absolute peripheral blood monocyte count, hemoglobin concentration, and serum neuron-specific enolase level. The key feature training set is input into logistic regression binary classification model, random forest binary classification model, support vector machine binary classification model, gradient boosting machine binary classification model, neural network binary classification model, Naive Bayes binary classification model and K-nearest neighbor binary classification model respectively, and multiple convergent primary neuroblastoma bone metastasis risk prediction models are obtained. The high-risk threshold of primary bone metastasis predicted by each primary neuroblastoma bone metastasis risk prediction model is output.

4. The neuroblastoma bone metastasis risk prediction system according to claim 3, characterized in that, Comparing the model performance of various primary neuroblastoma bone metastasis risk prediction models, the methods for obtaining and optimizing the neuroblastoma bone metastasis risk prediction model include: Based on the key features of bone metastasis risk of each neuroblastoma, an internal key feature validation set and an external key feature validation set are generated according to the internal validation set and the external validation set, respectively. The internal key feature validation set and the external key feature validation set are respectively input into each primary neuroblastoma bone metastasis risk prediction model, and the area under the receiver operating characteristic curve of each primary neuroblastoma bone metastasis risk prediction model is calculated. The primary neuroblastoma bone metastasis risk prediction model with the largest area under the receiver operating characteristic curve is selected as the candidate neuroblastoma bone metastasis risk prediction model. Based on the internal key feature validation set and the external key feature validation set, the model hyperparameters of the candidate neuroblastoma bone metastasis risk prediction model are adjusted to obtain the optimized neuroblastoma bone metastasis risk prediction model and the target high-risk threshold for bone metastasis predicted by the optimized neuroblastoma bone metastasis risk prediction model.

5. The neuroblastoma bone metastasis risk prediction system according to claim 4, characterized in that, The optimized neuroblastoma bone metastasis risk prediction model was obtained by training and validating a gradient booster binary classification model.

6. The neuroblastoma bone metastasis risk prediction system according to claim 5, characterized in that, Methods for obtaining a primary neuroblastoma bone metastasis risk prediction model based on gradient boosting machine binary classification model training include: Initialize the gradient boosting machine binary classification model, and initialize the model prediction values ​​according to the sample ratio of the training set of the key features; The key feature training set is input into the gradient booster binary classification model, and the gradient booster binary classification model is iteratively trained until the preset number of iterations is reached to obtain the corresponding primary neuroblastoma bone metastasis risk prediction model. The method for training the gradient boosting machine binary classification model in each iteration includes: The pseudo residual of the current model is calculated using the log loss function, and a shallow decision tree is trained to fit the pseudo residual. The current model is then updated based on the predictions of the shallow decision tree and the preset learning rate. The updated current model is validated based on the internal key feature validation set. If the model performance no longer improves, iterative training is stopped, and the current model is used as the primary neuroblastoma bone metastasis risk prediction model.

7. The neuroblastoma bone metastasis risk prediction system according to claim 4, characterized in that, The methods for predicting the bone metastasis risk value and bone metastasis risk classification results for the target neuroblastoma children include: Based on the key characteristics of bone metastasis risk of each neuroblastoma, and according to the target laboratory test data of the target neuroblastoma patients, the key characteristics data of bone metastasis risk of the target neuroblastoma patients are generated. The key feature data of the target neuroblastoma bone metastasis risk are input into the target neuroblastoma bone metastasis risk prediction model to obtain the bone metastasis risk value of the target neuroblastoma child, and the contribution value of each key feature of neuroblastoma bone metastasis risk to the bone metastasis risk value is calculated by the Shapley interpretable algorithm. The bone metastasis risk value is compared with the target bone metastasis high-risk threshold to determine whether the target neuroblastoma child has a low or high risk of neuroblastoma bone metastasis, and to generate a bone metastasis risk classification result for the target neuroblastoma child. Based on the bone metastasis risk value, the bone metastasis risk classification results, and the contribution values ​​of each key feature of neuroblastoma bone metastasis risk, a bone metastasis risk report for the target neuroblastoma child is generated.

8. The neuroblastoma bone metastasis risk prediction system according to claim 6, characterized in that, Also includes: The electronic medical record integration module is connected to the model prediction module and to an external electronic medical record system. It is used to obtain the target laboratory test data or key feature data of target neuroblastoma bone metastasis risk of the target child from the electronic medical record system and send them to the model prediction module; at the same time, it receives the bone metastasis risk report generated by the model prediction module and sends it to the electronic medical record system. The human-computer interaction module, connected to the model prediction module, is used to provide a visualization interface to collect target laboratory test data or key feature data of target neuroblastoma bone metastasis risk of the target neuroblastoma child and send them to the model prediction module, and to receive the bone metastasis risk report generated by the model prediction module for visualization display.

9. A method for predicting the risk of bone metastasis in neuroblastoma, characterized in that, include: Construct a predictive model for the risk of bone metastasis in target neuroblastoma; Based on the target neuroblastoma bone metastasis risk prediction model, and according to the target laboratory test data of the target neuroblastoma patients, the bone metastasis risk value and bone metastasis risk classification results of the target neuroblastoma patients are predicted, and a bone metastasis risk report is generated. The construction method of the target neuroblastoma bone metastasis risk prediction model includes: We obtained laboratory test data from multiple children with neuroblastoma and constructed internal training sets, internal validation sets, external validation sets, and external test sets based on the test data from each laboratory. Based on the internal training set, multiple primary neuroblastoma bone metastasis risk prediction models were constructed. Based on the internal validation set and the external validation set respectively, the diagnostic performance of each primary neuroblastoma bone metastasis risk prediction model is compared to obtain the optimized neuroblastoma bone metastasis risk prediction model. Based on the external test set, the optimized neuroblastoma bone metastasis risk prediction model is evaluated, and when it meets the prediction requirements, it is determined as the target neuroblastoma bone metastasis risk prediction model.

10. A terminal for predicting the risk of bone metastasis in neuroblastoma, characterized in that, include: Memory and processor; The memory is used to store computer programs; The processor is used to execute the computer program stored in the memory to enable the terminal to perform the functions of the neuroblastoma bone metastasis risk prediction system as described in any one of claims 1 to 8.

Images