Intelligent classification prediction method, system and equipment for tumor treatment and protection

Abstract

The invention is suitable for the technical field of tumor treatment and protection, and particularly relates to an intelligent classification prediction method, system and equipment for tumor treatment and protection, and the method comprises the steps: obtaining tumor-related molecular marker detection values and clinical diagnosis and treatment features of all target patients, and comprehensively integrating the multi-dimensional medical information of the patients; according to the relevance of molecular markers and clinical features, a scientific and quantitative multi-dimensional prognosis layering index set is constructed, and an accurate layering evaluation framework is formed; inputting the medical data into a pre-trained multilayer neural network model, and generating a high-resolution tumor characterization map through a cross-modal feature fusion technology; based on the dynamic matching of the atlas and the hierarchical index set, the hierarchical contribution degree parameter and the prognosis risk score are calculated in combination with the random forest model, the precise quantitative evaluation of the condition of the patient is completed, a high-credibility quantitative basis is provided for clinical decision making, the over-treatment risk is effectively reduced, and the prognosis management effect of the patient is improved.

Description

Technical Field

[0001] The present application belongs to the technical field of tumor treatment and care, and in particular, relates to a method, system and device for intelligent classification and prediction of tumor treatment and care. Background Art

[0002] Intelligent classification and prediction of tumor treatment and care is a key technology for clinical medical decision support. It aims to achieve accurate diagnosis and treatment plan recommendations by integrating patients' multi-dimensional medical data. By analyzing static indicators such as tumor staging, histological type, and gene mutation status, it provides a reference for the selection of treatment plans. It is widely used in scenarios such as personalized medication guidance, prognosis evaluation, and efficacy monitoring.

[0003] Traditional tumor classification and prediction methods have significant limitations, including insufficient data integration capabilities, limited model processing capabilities, low adaptability of the classification system, and lack of effective classification and prediction methods, which lead to inefficient clinical decision-making and improper allocation of medical resources. At the same time, it restricts the optimization space of personalized treatment plans and affects the patient's quality of life and prognosis improvement. Summary of the invention

[0004] The embodiments of the present application provide an intelligent classification prediction method, system and device for tumor treatment and care, which can solve the problem of low clinical decision-making efficiency, improper allocation of medical resources, and impact on patient quality of life and prognosis improvement due to the lack of effective classification prediction means in the tumor classification prediction process.

[0005] In a first aspect, the present application provides an intelligent classification prediction method for tumor treatment, including: Obtaining medical data of each target patient; wherein the medical data includes tumor-related molecular marker detection values ​​and clinical diagnosis and treatment characteristics; Constructing a multi-dimensional prognostic stratification indicator set based on the tumor-related molecular marker detection values ​​and the clinical diagnosis and treatment characteristics; Input the medical data into a pre-trained multi-layer neural network model for feature fusion to obtain a tumor characterization map of each target patient; wherein the tumor characterization map is a visual data model generated by performing feature fusion on the medical data through a deep learning model; Calculating the stratification contribution parameter and the prognostic risk score of each of the target patients according to the tumor characterization map and the multi-dimensional prognostic stratification indicator set; Based on the stratification contribution parameters and the prognostic risk score of each target patient, a personalized treatment plan classification result is generated for each target patient.

[0006] The above technical solutions in the embodiments of the present application have at least the following technical effects: The intelligent classification prediction method for tumor care provided in the embodiment of the present application obtains the tumor-related molecular marker detection values ​​and clinical diagnosis and treatment characteristics of each target patient, comprehensively integrates the multi-dimensional medical information of the patient, and provides a data basis for subsequent stratification prediction. According to the correlation between molecular markers and clinical characteristics, a scientifically quantified multi-dimensional prognostic stratification indicator set is constructed to form an accurate stratification evaluation framework. The medical data is input into the pre-trained multi-layer neural network model, and a high-resolution tumor characterization map is generated through cross-modal feature fusion technology to achieve a deep analysis of the biological characteristics of the tumor. Based on the dynamic matching of the map and the stratification indicator set, the stratification contribution parameters and prognostic risk scores are calculated in combination with the random forest model to complete the accurate quantitative evaluation of the patient's condition. Finally, according to the collaborative analysis of dynamic contribution and risk score, a personalized classification result of the treatment plan is generated, which significantly improves the objectivity and prediction efficiency of tumor stratification, optimizes the allocation of medical resources, and enhances the interpretability of the model through the dynamic weight mechanism, providing a highly reliable quantitative basis for clinical decision-making, effectively reducing the risk of overtreatment and improving the prognosis management effect of patients.

[0007] In a second aspect, the embodiment of the present application provides an intelligent classification prediction system for tumor treatment, including: An acquisition unit, used to acquire medical data of each target patient; wherein the medical data includes tumor-related molecular marker detection values ​​and clinical diagnosis and treatment characteristics; A stratification unit, used to construct a multi-dimensional prognostic stratification indicator set based on the tumor-related molecular marker detection values ​​and the clinical diagnosis and treatment characteristics; A fusion unit, used for inputting the medical data into a pre-trained multi-layer neural network model for feature fusion to obtain a tumor characterization map; wherein the tumor characterization map is a visual data model generated by performing feature fusion on the medical data through a deep learning model; A scoring unit, used to calculate the stratification contribution parameter and the prognostic risk score of each of the target patients according to the tumor characterization map and the multi-dimensional prognostic stratification indicator set; The prediction unit generates a personalized treatment plan classification result for each target patient based on the stratification contribution parameter and the prognostic risk score of each target patient.

[0008] In a third aspect, an embodiment of the present application provides an intelligent classification and prediction device for tumor treatment, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements a method as described in any one of the above aspects when executing the computer program.

[0009] In a fourth aspect, an embodiment of the present application provides a computer program product. When the computer program product is run on a tumor therapy intelligent classification prediction device, the tumor therapy intelligent classification prediction device executes the method of any one of the above aspects.

[0010] It can be understood that the beneficial effects of the second to fourth aspects mentioned above can be found in the relevant descriptions in the above aspects and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. 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 paying any creative work.

[0012] Figure 1 It is a flowchart of an intelligent classification prediction method for tumor treatment provided by an embodiment of the present application; Figure 2 Schematic diagram of the operation of the intelligent classification prediction method for tumor treatment provided by one embodiment of the present application; Figure 3 It is a structural diagram of an intelligent classification prediction system for tumor treatment provided by an embodiment of the present application; Figure 4 It is a structural diagram of an intelligent classification and prediction device for tumor treatment provided in one embodiment of the present application. DETAILED DESCRIPTION

[0013] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0014] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0015] It should also be understood that the term “and / or” used in the specification and appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0016] As used in the specification of this application and the appended claims, the term "if" can be interpreted as "when" or "uponce" or "in response to determining" or "in response to detecting" depending on the context. Similarly, the phrase "if it is determined" or "if the described condition or event is detected" can be interpreted as meaning "uponce it is determined" or "in response to determining" or "uponce the described condition or event is detected" or "in response to detecting the described condition or event" depending on the context.

[0017] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0018] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0019] Traditional tumor classification and prediction methods have significant limitations, including insufficient data integration capabilities, limited model processing capabilities, low adaptability of the classification system, and lack of effective classification and prediction methods, which lead to inefficient clinical decision-making and improper allocation of medical resources. At the same time, it restricts the optimization space of personalized treatment plans and affects the patient's quality of life and prognosis improvement.

[0020] To solve the above problems, the embodiments of the present application provide a method, system and device for intelligent classification prediction of tumor care. In this method, by obtaining the tumor-related molecular marker detection values ​​and clinical diagnosis and treatment characteristics of the target patient, the multi-dimensional medical information of the patient is fully integrated to provide a data basis for subsequent stratification prediction. According to the correlation between molecular markers and clinical characteristics, a scientifically quantified multi-dimensional prognosis stratification indicator set is constructed to form an accurate stratification evaluation framework. The medical data is input into the pre-trained multi-layer neural network model, and a high-resolution tumor characterization map is generated through cross-modal feature fusion technology to achieve a deep analysis of the biological characteristics of the tumor. Based on the dynamic matching of the map and the stratification indicator set, the stratification contribution parameters and prognosis risk scores are calculated in combination with the random forest model to complete the accurate quantitative evaluation of the patient's condition. Finally, according to the collaborative analysis of dynamic contribution and risk score, a personalized treatment and care plan classification result is generated, which significantly improves the objectivity and prediction efficiency of tumor stratification, optimizes the allocation of medical resources, and enhances the interpretability of the model through the dynamic weight mechanism, providing a highly reliable quantitative basis for clinical decision-making, effectively reducing the risk of overtreatment and improving the prognosis management effect of patients.

[0021] The intelligent classification and prediction method for tumor treatment provided in the embodiment of the present application can be applied to an intelligent classification and prediction device for tumor treatment. In this case, the intelligent classification and prediction device for tumor treatment is the executor of the intelligent classification and prediction method for tumor treatment provided in the embodiment of the present application. The embodiment of the present application does not impose any restrictions on the specific type of the intelligent classification and prediction device for tumor treatment.

[0022] For example, the intelligent classification and prediction device for tumor treatment can be a mobile phone, a tablet computer, a laptop computer, an ultra-mobile personal computer (UMPC), a netbook, a desktop computer, a smart large screen, a smart TV, a handheld device with wireless communication function, a computing device or other processing device connected to a wireless modem, a vehicle-mounted device, a vehicle networking terminal, a computer, a laptop computer, a communication device, a computing device, etc.

[0023] In order to better understand the intelligent classification and prediction method for tumor treatment and nursing provided in the embodiment of the present application, the specific implementation process of the intelligent classification and prediction method for tumor treatment and nursing provided in the embodiment of the present application is exemplarily introduced below.

[0024] Figure 1 A schematic flow chart of the intelligent classification prediction method for tumor treatment and nursing provided in an embodiment of the present application is shown. Figure 2 The flowchart of the operation of the intelligent classification prediction method for tumor treatment and nursing provided in the embodiment of the present application is shown. The intelligent classification prediction method for tumor treatment and nursing includes: S100, obtaining medical data of each target patient; wherein the medical data includes tumor-related molecular marker detection values ​​and clinical diagnosis and treatment characteristics.

[0025] It can be understood that medical data includes tumor-related molecular marker detection values ​​and clinical diagnosis and treatment characteristics. Tumor-related molecular marker detection values ​​can include molecular biological indicators such as circulating tumor DNA (ctDNA) concentration, PD-L1 expression level, tumor mutation burden (TMB), and microsatellite instability (MSI) status. Tumor-related molecular marker detection values ​​are obtained through laboratory tests and reflect the molecular characteristics and biological behavior of the tumor. Clinical diagnosis and treatment characteristics can include structured clinical data such as TNM staging, ECOG physical fitness score, treatment response rate, pathological grade, and complication occurrence. TNM staging is a standard for evaluating tumor size, lymph node metastasis, and distant metastasis, while ECOG physical fitness score evaluates the patient's physical function status. Medical data can be automatically captured from the hospital laboratory information system (LIS) and electronic medical record system (EMR) through the API interface. For unstructured text data, such as doctors' diagnosis and treatment records, natural language processing technology (such as the BERT model) can be used for entity recognition and structured conversion to ensure the integrity and availability of the data.

[0026] S200, based on the detection values ​​of tumor-related molecular markers and clinical diagnosis and treatment characteristics, constructs a multidimensional prognostic stratification indicator set.

[0027] It can be understood that by integrating the detection values ​​of tumor-related molecular markers and clinical diagnosis and treatment characteristics, a set of indicators that have a significant impact on the patient's prognosis can be formed, that is, a multidimensional prognostic stratification indicator set can be constructed. Factor analysis can be performed on the detection values ​​of tumor-related molecular markers to obtain a set of molecular features of each target patient after dimensionality reduction. Factor analysis is a statistical method used to identify potential structures hidden in a large number of variables, thereby reducing the dimension of the data. Spectral clustering is performed on the clinical diagnosis and treatment characteristics to generate clinical feature classification groups for each target patient. Spectral clustering is a hierarchical clustering method that classifies similar features into one category by calculating the similarity between features. Finally, the molecular feature set and clinical feature classification group of each target patient are obtained, and then the combination of strongly associated features is screened out, a hierarchical classification framework is constructed, and a multidimensional prognostic stratification indicator set is formed.

[0028] In a possible implementation, S200 constructs a multi-dimensional prognostic stratification indicator set based on tumor-related molecular marker detection values ​​and clinical diagnosis and treatment characteristics, including: S210, performing factor analysis on the tumor-related molecular marker detection values ​​to obtain a molecular feature set of each target patient after dimensionality reduction.

[0029] It can be understood that factor analysis methods can be used to reduce the dimensionality of tumor-related molecular marker detection data. Factor analysis is a statistical method that reduces the number of variables by identifying potential factors in the data and retains the most important features for prognostic analysis. For example, ctDNA concentration and TMB may be combined into a potential "tumor burden factor" through factor analysis, thereby simplifying the complexity of the data while retaining key information. The set of molecular features after dimensionality reduction provides a more efficient input for subsequent analysis.

[0030] For example, the factor analysis process can condense high-dimensional molecular data through statistical methods. First, the covariance matrix between each indicator is calculated to identify potential common factors. For example, PD-L1 expression and TMB may share the "immune activity" potential factor. The maximum variance rotation is used to optimize the factor loading matrix so that each principal component clearly represents a specific biological process. The molecular feature set generated after dimensionality reduction retains 85% of the information content of the original data and eliminates the problem of multicollinearity. For example, 12 original indicators are condensed into 5 principal components, where the first principal component reflects the characteristics related to the immune checkpoint, the second principal component represents the tumor proliferation activity, etc....

[0031] S220, perform spectral clustering on the clinical diagnosis and treatment characteristics to generate clinical characteristic classification groups for each target patient.

[0032] It can be understood that clinical diagnosis and treatment characteristics can be divided into different classification groups through pedigree clustering algorithms. Pedigree clustering is a hierarchical clustering method that generates a dendrogram to represent the grouping results of characteristics by gradually merging or splitting data points. For example, ECOG physical performance score and TNM stage may be divided into categories such as "early low risk" and "late high risk" through clustering. Pedigree clustering can group complex clinical data according to feature similarity, which is convenient for subsequent analysis and decision-making.

[0033] Optionally, S220, spectral clustering is performed on the clinical diagnosis and treatment characteristics to generate a clinical characteristic classification group for each target patient, including: S221, standardize the clinical diagnosis and treatment characteristics to obtain a standardized clinical characteristic matrix.

[0034] It can be understood that the standardization process aims to eliminate the interference of different clinical feature dimensions on subsequent analysis. For example, age (0-100), tumor size (0-30cm) and ECOG score (0-5) need to be converted to a distribution with a mean of 0 and a standard deviation of 1 through Z-score standardization. The mean and standard deviation of each clinical diagnosis and treatment feature can be calculated, and the original value can be converted according to the formula to convert clinical features of different types and dimensions into a structured data matrix with a unified standard distribution. For missing values, the chained equation multiple imputation method (MICE) is used to predict the missing ECOG score based on other features (such as the correlation between age and tumor stage). In the final generated standardized matrix, each patient is represented as a v=[Zage,ZECOG,ZTNM,...] vector, providing comparable input for clustering.

[0035] For example, clinical feature standardization can use a robust processing method to use a scaling method based on the median and interquartile range for continuous variables that may contain outliers (such as white blood cell counts) to avoid interference from extreme values. For example, a patient's white blood cell count is 28×10^9 / L (normal range 4-10), which will not deviate excessively from other samples after processing by RobustScaler. For categorical variables (such as TNM staging), first perform one-hot encoding to generate a sparse vector, and then project it into a continuous space through an embedding layer, so that there is a computable semantic distance between stages (such as the vector distance between stage III and stage IV is less than that between stage I and stage IV) S222, calculating the mutual information matrix between each clinical diagnosis and treatment feature to obtain a hierarchical clustering distance matrix between each clinical diagnosis and treatment feature.

[0036] It can be understood that mutual information is used to quantify the nonlinear association between features. The mutual information matrix is ​​a symmetric matrix based on the difference measure between features, which is used to guide the hierarchical clustering algorithm to construct a dendrogram. The amount of information sharing for each pair of clinical diagnosis and treatment features can be calculated by the k-nearest neighbor estimation method. For example, the mutual information value of the number of chemotherapy cycles and neutropenia is high, indicating that there is a strong correlation between the two. The hierarchical clustering algorithm is used to construct a dendrogram based on the mutual information distance: in the initial stage, each feature is a class in its own right, and the classes with the closest distance are gradually merged to form a hierarchical structure. The optimal number of clusters can be automatically determined by the gradient change of the merged distance through dynamic threshold segmentation to avoid manual setting deviations. For example, when the merged distance suddenly increases by 30%, it is determined that the current classification level is reasonable.

[0037] For example, age and the incidence of chemotherapy side effects may have a U-shaped relationship rather than a linear correlation. For continuous features (such as age) and categorical features (such as cancer subtypes), the continuous variable can be discretized into 5 intervals through adaptive binning, and then the normalized mutual information (NMI) is calculated based on the joint probability distribution. Assume that the NMI value of "TNM stage" and "lymph node metastasis" is 0.8, indicating that the two are highly correlated, while the NMI of "sex" and "pathological grade" is 0.1, which is considered a weak association. According to 1-NMI, the mutual information matrix is ​​converted into a difference matrix (difference value = 1-NMI), and then the Ward variance minimization algorithm is used to calculate the hierarchical clustering distance to form the branch structure of the dendrogram.

[0038] S223, performing dynamic threshold segmentation based on the hierarchical clustering distance matrix to obtain an initial clinical feature group for each target patient.

[0039] It can be understood that dynamic threshold segmentation determines the optimal cutting point by analyzing the branch height of the hierarchical clustering dendrogram. For example, the height difference between adjacent merging steps in the dendrogram is calculated (such as the difference is 1.3 from the merging height 1.2 to 2.5), and the position corresponding to the maximum difference is selected as the cutting threshold. If the maximum difference occurs when merging into 3 groups, the dendrogram is cut into 3 initial clinical feature groups (such as "low-risk group", "medium-risk group" and "high-risk group"). For special scenarios (such as multiple differences with similar heights), the threshold corresponding to the local maximum density can be selected through density peak detection (DPC algorithm) to avoid subjective setting of the number of groups.

[0040] For example, dynamic threshold segmentation can use an adaptive mechanism to monitor the rate of change of class intervals in real time during hierarchical clustering. When a significant jump is detected (such as an increase in intervals exceeding 2 standard deviations of the historical average), the segmentation operation is triggered to generate a new cluster. For example, "liver function index" and "renal function index" are divided into independent clusters, while "thrombocytopenia" and "anemia" of the same hematological toxicity are grouped together.

[0041] S224, evaluating the clustering quality of each initial clinical feature group by using the silhouette coefficient, merging the initial clinical feature groups with overlaps higher than a preset threshold, and generating an optimized clinical feature classification group.

[0042] It can be understood that the silhouette coefficient evaluation optimizes clustering quality by quantifying the intra-cluster compactness and inter-cluster separation. For each initial clinical feature group, its similarity with the same cluster feature (a value) and the difference with the nearest different cluster feature (b value) are calculated. When (ba) / max(a,b) is close to 1, it is determined to be an ideal classification. The merging threshold is set to the average silhouette coefficient difference <0.15. For example, when the silhouette coefficients of the two groups are 0.68 and 0.63 respectively, the merge is performed. This mechanism can automatically eliminate over-segmentation, such as merging semantically similar features such as "gastrointestinal reaction" and "loss of appetite".

[0043] S230, calculating mutual information between the molecular feature set and the clinical feature classification group of each target patient, and screening out a strongly correlated feature combination; wherein the strongly correlated feature combination is used to reflect a multimodal feature set greater than a preset threshold.

[0044] It can be understood that mutual information (MI) is used to capture the nonlinear association between molecules and clinical features. For example, to calculate the MI values ​​of "tumor metabolic factors" and "postoperative recurrence", the continuous molecular factors can be first binned (such as 10 intervals) and the joint probability distribution of the factors and the binary recurrence events can be calculated, and then the MI value can be obtained through the discrete entropy formula. If the MI value of a combination exceeds the preset threshold (such as 0.3), it is considered to have a strong correlation (such as the correlation between "TP53 mutation + ECOG ≥ 2 points" and prognosis is retained). At the same time, redundant combinations are excluded (such as the MI value of "EGFR mutation + BMI" is lower than the threshold), and finally 10-15 key cross-modal features are screened out.

[0045] S240, construct a hierarchical classification framework based on the combination of strongly correlated features to obtain a multi-dimensional prognostic stratification indicator set.

[0046] It can be understood that hierarchical classification is to recursively divide the patient population to obtain a multi-dimensional prognostic stratification indicator set. For example, the first layer is divided into early (Ⅰ-Ⅱ) and late (Ⅲ-Ⅳ) according to TNM stage; the second layer subdivides the early group according to the PD-L1 expression level (≥50% vs. <50%), and the late group is subdivided according to the ctDNA concentration (≥5% vs. <5%). Each segmentation node can select the optimal feature based on the principle of minimizing Gini impurity (such as the largest decrease in TNM's Gini coefficient), and finally generate a classification framework similar to a tree structure. Each level corresponds to a specific prognostic standard (such as "late-ctDNA high positive-immunotherapy resistance group"), forming 4-6 terminal nodes for clinical decision-making.

[0047] S300, inputting the medical data into a pre-trained multi-layer neural network model for feature fusion to obtain a tumor characterization map; wherein the tumor characterization map is a visual data model generated after feature fusion of medical data through a deep learning model.

[0048] It can be understood that molecular and clinical diagnosis and treatment features can be integrated through deep learning models (such as Graph Convolutional Network, GCN). The molecular feature data is normalized and input into the fully connected layer, and the text-based diagnosis and treatment features are encoded into semantic vectors through the long short-term memory network (LSTM). The cross-attention mechanism is used to capture multimodal interactions (such as "tumor mutation load is associated with necrotic areas in CT images"), and then t-SNE is used to project high-order features into two-dimensional space to form a tumor representation map. The tumor representation map is a visual data model generated by feature fusion of medical data through a deep learning model. Adjacent points in the map represent similar biological behaviors (such as the tendency to metastasize), and high-risk clusters can be identified through density clustering (DBSCAN).

[0049] In a possible implementation, S300, medical data is input into a pre-trained multi-layer neural network model for feature fusion to obtain a tumor characterization map, including: S310, standardizing the tumor-related molecular marker detection values ​​to obtain standardized molecular marker data for each target patient.

[0050] It is understandable that the tumor-related molecular marker detection values ​​can be standardized to eliminate the dimension differences in the data and ensure the accuracy and stability of subsequent feature fusion. For example, the ctDNA concentration and TMB are standardized to the same value range.

[0051] For example, the skewed distribution index (such as PD-L1 expression) can be transformed by log2(x+1) to make it approximate normal distribution; binary encoding (MSS=0 / MSI-H=1) is used for categorical data (such as microsatellite status). The standardized data is stored in HDF5 files, retaining the original data, processing parameters and version information to ensure traceability of the results.

[0052] S320, embedding and compensating the clinical diagnosis and treatment features to obtain structured diagnosis and treatment data for each target patient.

[0053] It is understandable that clinical diagnosis and treatment features can be converted into low-dimensional vector representations through feature embedding, such as using Word2Vec or Transformer models to convert text data into numerical vectors. At the same time, missing values ​​are compensated, and null values ​​are processed using mean filling, interpolation, or machine learning methods (such as KNN interpolation) to ensure data integrity.

[0054] For example, a 768-dimensional context vector can be generated for a text description (such as "partial remission") to capture subtle semantic differences (such as the vector distance between "PR" and "CR" is greater than that between "PR" and "SD"). Missing value filling uses a conditional generative adversarial network: the generator receives known features (such as age, gender) to generate reasonable inferences (such as ECOG scores), and the discriminator evaluates the authenticity of the generated data through the Wasserstein distance. The Transformer encoder can use masked language modeling (MLM) + contrastive learning and fine-tune through layer-wise learning rates (top layer lr=5e-5, bottom layer lr=1e-6). The conditional generative adversarial network uses a 5-layer fully connected layer generator architecture (1024-512-256-128-64), and the discriminator is designed through spectral normalization + gradient penalty (λ=10). The training parameters are Adam (β1=0.5, β2=0.999), batch=64.

[0055] S330, combining the normalized molecular marker data and the structured diagnosis and treatment data of each target patient into a joint input tensor.

[0056] It can be understood that the joint input tensor can fuse multimodal data through channel splicing. Assuming that the normalized molecular marker data is a 100-dimensional vector (such as 20 gene expression values, 5 protein markers, and 3 factor analysis results), and the structured diagnosis and treatment data is a 50-dimensional vector (such as age, stage, treatment response markers), it is spliced ​​into a 150-dimensional 2D tensor (shape [number of patients, 150]) in the channel dimension. In order to maintain time series features (such as multiple ctDNA tests), it can be expanded to a 3D tensor ([number of patients, time step, number of features]) for recurrent neural network processing. When the time dimension is missing, it is filled with a single time step input by default to ensure that the model is compatible with heterogeneous data.

[0057] S340, extracts local features of the joint input tensor through the convolution layer of the multi-layer neural network model, and performs weighted fusion on the local features of the joint input tensor through the attention mechanism to obtain a fused cross-modal fusion feature tensor.

[0058] It can be understood that the local features of the joint input tensor are extracted through the convolution layer of the multi-layer neural network model, and a 1D convolution kernel (width = 3, step size = 1) is used on the joint input tensor to extract local patterns (such as the gain effect of "EGFR mutation + synchronous radiotherapy" on prognosis) from adjacent molecular features (such as EGFR mutation status) and clinical features (such as concomitant treatment). The attention mechanism assigns higher weights to key features (such as driver gene mutations) by calculating the feature weight matrix, and performs scaled dot-product attention on the fused features by generating query (Q), key (K), and value (V) matrices and calculating the attention score through Softmax. The final output cross-modal fusion feature tensor has high-weight features that are enhanced and noise features that are suppressed.

[0059] Optionally, in step S340, weighted fusion is performed on the local features of the joint input tensor through an attention mechanism to obtain a fused cross-modal fusion feature tensor, including: S341, generating a first feature projection based on the normalized molecular marker data, and constructing a molecular marker query matrix.

[0060] It can be understood that the first feature projection is a feature conversion process based on molecular marker data. The original data can be projected into a 32-dimensional semantic space through two fully connected layers. The first layer uses ReLU activation to extract nonlinear features, and the second layer uses linear transformation to generate the final query vector. The first fully connected layer maps the input normalized molecular marker data to a high-dimensional space through the ReLU activation function to capture nonlinear features. The nonlinear characteristics of ReLU enable the model to better fit complex biomarker data during the learning process. The second fully connected layer maps the features after ReLU activation to a 32-dimensional semantic space through a linear transformation to form a query vector. A contrastive learning strategy can be adopted in the pre-training stage: positive samples are enhanced from time series data of the same patient, and negative samples are randomly mixed with different patient features, so that query vectors of similar patients are clustered in space.

[0061] S342, generating a second feature projection based on the structured diagnosis and treatment data, and constructing a diagnosis and treatment key matrix and a diagnosis and treatment value matrix.

[0062] It can be understood that the second feature projection is a feature conversion process based on structured diagnosis and treatment data. Feature extraction of structured data can be performed through methods such as convolutional neural networks (CNN) or recurrent neural networks (RNN). For example, a one-dimensional convolutional network is used to process time series data to extract key information in the diagnosis and treatment process. By encoding the extracted features (such as using a transformation layer), a "key matrix" is obtained, which plays a key role in the subsequent cross-modal attention mechanism and acts as a pairing object for "query" data. The diagnosis and treatment value matrix contains the actual values ​​in the diagnosis and treatment data (such as treatment plans, treatment effect evaluation, etc.), which will be spliced ​​with the weighted features during fusion.

[0063] S343, calculating a cross-modal association weight matrix through dual-line attention according to the query matrix of the normalized molecular marker data and the diagnosis and treatment key matrix of the structured diagnosis and treatment data.

[0064] It can be understood that the query matrix comes from molecular marker data and the bond matrix comes from structured diagnosis and treatment data. The two-line attention mechanism performs interactive calculations through the query matrix and the bond matrix. To calculate the similarity between the query and the bond, the dot product can be used to calculate score(q,k)=qTk, where q is the query vector and k is the bond vector. Based on the calculated similarity score, the similarity is converted into a probability distribution through the softmax function to obtain the attention weight. For each query vector, its attention weight with all bonds is calculated, and the weighted sum is used to generate a weighted feature. Finally, the cross-modal association matrix is ​​obtained through the two-line attention mechanism. Each element represents the correlation between the query and the bond, which helps to find the most relevant cross-modal features in subsequent steps.

[0065] S344, performing multi-level attention operations in parallel through a multi-scale attention mechanism according to the modal association weight matrix, and weighted concatenating the cross-modal association weight matrix with the diagnosis and treatment value matrix to obtain a weighted concatenation result.

[0066] It can be understood that at different scales, convolution kernels of different sizes or different context windows can be used to process data. For example, different time windows or spatial scales are used to capture features at different levels. The attention mechanism at each scale is calculated independently and outputs multiple weighted results. These results are executed in parallel and then summarized. The results obtained from the multi-scale attention calculation can be spliced ​​with the diagnosis and treatment value matrix. After splicing, a composite feature vector is formed, which contains weighted information from molecular markers, structured diagnosis and treatment data, and multi-scale attention. The form of weighted splicing is to splice the two matrices (modality association weight matrix and diagnosis and treatment value matrix) by column or row, and then merge them into a large feature tensor.

[0067] S345, performing skip connection and batch normalization on the weighted concatenation result after concatenation to obtain a fused cross-modal fusion feature tensor.

[0068] It can be understood that the jump connection will pass the information of the input layer directly to the output layer, skipping the intermediate layer, which helps to maintain the flow of information and speed up the training process. Batch normalization can be applied to the weighted results after splicing, the purpose of which is to standardize the data output by each layer so that its mean is 0 and the variance is 1. Batch normalization can make the training process more stable, accelerate convergence, and reduce the risk of overfitting.

[0069] S350, projecting the fused cross-modal fusion features into a two-dimensional space to obtain a tumor characterization map.

[0070] It can be understood that the fused cross-modal fusion features can be projected into a two-dimensional space feature projection, and the t-SNE (t-distributed random neighbor embedding) algorithm can be used to retain the local structure of high-dimensional data. Specific parameters can be set to perplexity (perplexity=30), learning rate (200), and number of iterations (1000). Input the cross-modal fusion tensor (such as 1024-dimensional features), and optimize the KL divergence of the target distribution and the low-dimensional embedding through gradient descent, so that adjacent points in the high-dimensional space (such as patients with high PD-L1 expression) are still closely distributed in the two-dimensional map. To enhance visualization, the results of K-means clustering (k=5) are mapped into clusters of different colors, and the distribution areas of patients with known good prognosis are marked (such as the lower left quadrant as the "immunotherapy sensitive area") to assist doctors in intuitively evaluating tumor heterogeneity.

[0071] S400, based on the tumor characterization map and the multidimensional prognostic stratification indicator set, calculate the stratification contribution parameters and prognostic risk score of each target patient.

[0072] It can be understood that the stratification contribution parameter uses SHAP value (Shapley Additive exPlanations) to quantify the impact of each stratification indicator on individual patients. For example, for a certain patient, the SHAP value of TNM stage is 0.3 and PD-L1 is 0.5, indicating that the latter contributes more to its high-risk classification. The prognostic risk score is a weighted sum of features through logistic regression (such as Score = 0.3 × TNM_stage III + 0.5 × PD-L1_high expression + 0.2 × ctDNA_positive), normalized to a threshold of 0-100 points (such as >70 points for high risk). At the same time, the score is calibrated based on the topological features of the tumor characterization map (such as the median survival rate of the patient's cluster) to make it consistent with the actual clinical outcome.

[0073] In a possible implementation, S400 calculates the stratification contribution parameter and the prognostic risk score of each target patient according to the tumor characterization map and the multi-dimensional prognostic stratification indicator set, including: S410, extracting topological connection features and probability distribution features from the tumor representation map to form a composite feature vector.

[0074] It can be understood that the tumor representation map can be converted into a graph structure through topological feature extraction, with each pixel as a node, and adjacent pixels and similar intensity regions are connected by edges. The neighborhood information is aggregated through a three-layer graph convolutional network: the first layer captures local metabolic hotspots, the second layer identifies sub-regional structures, and the third layer integrates global topological patterns. The mean of the output node embedding vector is used as the final topological feature to reflect the spatial heterogeneity of the tumor. Two types of features can be extracted from the tumor representation map: one is the topological connection feature, including the degree centrality and betweenness centrality of the patient node in the map, and the non-Euclidean spatial association of the neighborhood information is captured through the graph convolutional network (GCN); the second is the probability distribution feature, and the kernel density estimation method (KDE) is used to quantify the cumulative probability density of patient features in the global space, where h is the window width coefficient, K selects the Epanechnikov kernel function to reduce edge smoothing errors, and the two are fused to form a composite feature vector.

[0075] Exemplarily, in order to extract the spatial heterogeneity and local metabolic hotspots of the tumor, the tumor atlas is converted into a graph structure. The pixels of each image are regarded as nodes of the graph, and edges are formed between adjacent pixels or areas with similar intensities. Next, a convolutional network (GCN) is used to process these graph data. GCN gradually aggregates the neighborhood information of the nodes through multiple layers of convolutional layers. By embedding each aggregated node (the output of each layer of convolution), a low-dimensional feature representation is obtained, which reflects the topological structure and spatial heterogeneity of the tumor atlas. By taking the average of the embedding vector of each node, the global topological characteristics of the entire tumor area can be obtained. The intensity distribution of the tumor in different regions can be calculated to obtain the intensity probability distribution characteristics of the tumor region, further enriching the feature vector.

[0076] S420, constructing a stratified weight allocation tensor according to the multi-dimensional prognostic stratification indicator set.

[0077] It can be understood that the hierarchical weight tensor constructs a three-dimensional dynamic structure: the first dimension corresponds to the molecular principal components (such as PC1-PC5), the second dimension is associated with the clinical feature cluster (such as the treatment response group), and the third dimension divides the time window (0-3 months, 3-6 months, etc.). The initial weight is set according to the distribution of historical data. For example, the chemotherapy sensitivity has a higher weight in the early stage of treatment (0-3 months) and gradually decreases in the later stage. The weight is updated weekly through Kalman filtering to track the changes in treatment effects. The potential correlation between hierarchical indicators is modeled through a multi-task learning framework.

[0078] For example, a weight tensor w∈ℝ^{K×D} is defined (K is the number of layer dimensions, D is the composite feature dimension), and each column corresponds to the dynamic weight channel of a certain layer indicator (such as T stage or gene mutation type). It is initialized using the Xavier method to keep the variance consistent, and then optimized through multiple rounds of back propagation: when the gradient is updated, an L2 sparse constraint is added to the loss function of each indicator to learn the differential feature weights of each layer.

[0079] S430, input the composite feature vector into the pre-trained random forest model, calculate the dynamic contribution of each stratification indicator based on the dynamic weight channel in the weight allocation tensor, and obtain the stratification contribution parameter.

[0080] It can be understood that the composite feature vector extracted from the tumor characterization map and the multi-dimensional prognostic stratification indicator set (such as gene expression, clinical characteristics, etc.) can be combined as the input of the model and input into the pre-trained random forest model. Random forest is an ensemble learning method that can effectively process high-dimensional data and prevent overfitting by training multiple decision trees. Random forest calculates the contribution of each feature based on the training data. The degree of influence of each feature on the patient's prognostic risk score is measured by the feature gain value. The contribution of each stratification indicator (such as treatment response, gene expression, etc.) is calculated based on the dynamic weights in the weight allocation tensor.

[0081] Exemplarily, the significance of composite features is parsed based on a pre-trained random forest model, and the specific process includes: ① inputting a pre-constructed training data set (including patient characteristics and survival status labels), and defining a cross-entropy loss function for initial training; ② generating a set of decision trees through multiple rounds of iterations, calculating the feature gain value during the construction of each tree and accumulating the gain value along the iteration rounds as the total gain of each indicator; ③ setting a threshold according to the gain ranking (such as retaining the Top-15 indicators) to eliminate low-contribution features; ④ performing probability normalization (Softmax function) on the screened gain values, and outputting the stratified contribution parameters.

[0082] Optionally, S430, the composite feature vector is input into a pre-trained random forest model, and the dynamic contribution of each stratification index is calculated based on the dynamic weight channel in the weight allocation tensor to obtain the stratification contribution parameter, including: S431, obtaining a pre-constructed training data set based on patient stratification, and defining a classification error objective function based on the training data set.

[0083] It can be understood that the training data set is constructed based on the patient's stratification (such as disease stage, treatment plan, etc.). The label of each patient can be its final prognosis (such as survival, recurrence rate, etc.). The objective function is usually the classification error (such as cross entropy loss), which is used to measure the difference between the model prediction and the actual result. By optimizing the objective function, the model can learn the contribution of each stratification indicator.

[0084] S432, generating a decision tree classifier through multiple rounds of iterations, and calculating the feature gain value of each classifier, updating the contribution score of each hierarchical indicator according to the feature gain value in each round of iteration, and obtaining the cumulative contribution score of each hierarchical indicator.

[0085] It can be understood that the model can generate decision tree classifiers through multiple rounds of iterations and calculate the gain value of each feature in each classifier. In each iteration, the decision tree will update the contribution score of the stratification indicator according to the gain value of the current feature. Through multiple rounds of iterations, the contribution score of the stratification indicator can be gradually optimized and the cumulative contribution can be obtained to help understand the impact of each stratification indicator on the patient's prognosis.

[0086] S433, ranking the multidimensional prognostic stratification indicators by their feature contributions based on the cumulative contribution scores, and performing feature screening on each stratification indicator in the multidimensional prognostic stratification indicator set by filtering low-contribution indicators through a preset pruning threshold according to the feature contribution ranking.

[0087] It can be understood that each stratified indicator can be ranked by the cumulative contribution score. Indicators with higher contribution scores indicate greater impact on patient prognosis. According to the preset pruning threshold (such as contribution below a certain percentile), indicators with less contribution to the prognostic risk score are screened out, and indicators with greater impact on prediction are retained.

[0088] S434, performing probability normalization processing on the cumulative contribution scores of each stratified indicator after screening to generate stratified contribution parameters.

[0089] It can be understood that the contribution scores of the selected features can be probabilistically normalized so that the sum of the contributions of all indicators is 1. This can make the contributions of different indicators comparable and can be used for subsequent weighted calculations. Finally, a normalized stratified contribution parameter is obtained, which reflects the importance of each stratified indicator in predicting the patient's prognostic risk.

[0090] S440, performing weighted aggregation on the composite feature vector and the stratified contribution parameter, and converting them into a prognostic risk score for each target patient through a logistic function.

[0091] It can be understood that the composite feature vector and the stratified contribution parameter can be weighted and aggregated to form a new comprehensive feature vector. The weighted features are transformed by a logistic function (such as logistic regression or sigmoid function) to obtain the prognostic risk score of each patient. The prognostic risk score indicates the prognostic risk level of the patient, and usually ranges from 0 to 1, with a larger value indicating a higher risk.

[0092] Exemplarily, the composite feature vector and the contribution parameter are weighted by tensor product, which is calculated as: , and then the linear combination value is mapped to the 0-100 interval through the logistic function. The final score is based on topological calibration adjustment, and the cluster center of the target patient is retrieved from the atlas, and the actual median survival rate of the patient to whom it belongs (such as the median survival time of a cluster is 18 months) is compared with the model prediction value. The risk score is strictly consistent with the clinical observation outcome through the weighted moving average method (α=0.7 determined by learning from the validation set).

[0093] S500, generating a classification result of a personalized treatment plan for each target patient based on the stratified contribution parameter and prognostic risk score of each target patient.

[0094] It can be understood that treatment plan generation can combine evidence-based medicine rules with resource constraint optimization. For example, high-risk patients (prognostic risk>70) are given priority to match the "immunity + chemotherapy" combination recommended by the guidelines, but they need to be adjusted according to the hospital's drug inventory (such as the availability of PD-1 inhibitors). The particle swarm optimization algorithm (PSO) uses maximizing the treatment effect (based on the survival gain model in historical data) and minimizing the cost (number of chemotherapy cycles, bed occupancy time) as the objective function, and iteratively searches for the optimal combination of solutions. Finally, structured treatment plan labels (such as "high risk-priority Keytruda+platinum-need to monitor liver toxicity") and resource scheduling recommendations (such as "ICU beds need to be reserved for 3 days") are output, and doctors are assisted in decision-making through a visual interface.

[0095] In a possible implementation, S500 generates a classification result of a personalized treatment plan for each target patient based on the stratified contribution parameter and the prognostic risk score of each target patient, including: S510, based on the prognostic risk score of each target patient, obtain the tumor risk level of each target patient.

[0096] It can be understood that patients can be classified into different risk levels based on the prognostic risk score of each patient. Generally, these levels can be divided into low risk, medium risk and high risk. For example, if the prognostic risk score is greater than 70, the patient is classified as a high-risk group. The division of risk levels is the basis for the formulation of treatment plans and provides guidance for subsequent treatment decisions. Patients in different risk levels require different intensities and types of therapeutic interventions.

[0097] S520, matching a preset treatment pathway template according to the tumor risk level of each target patient to obtain a basic treatment decision for each target patient; wherein the basic treatment decision includes multiple intervention measures.

[0098] It can be understood that based on the patient's tumor risk level, the patient's basic treatment decision is generated by matching the preset treatment pathway template. For example, for high-risk patients, the treatment pathway template may recommend a combination of immunotherapy and chemotherapy; for low-risk patients, a single targeted therapy or monitoring program may be recommended. These treatment pathway templates take into account evidence-based medical guidelines and are personalized based on the patient's specific circumstances (such as age, comorbidities, etc.). Each treatment pathway contains multiple interventions, such as drug therapy, surgery, radiotherapy, nursing intervention, etc., to form a preliminary treatment plan.

[0099] S530, rank the sensitivity of the intervention measures in the basic treatment decision according to the stratified contribution parameters of each target patient, balance the medical resource constraints through the particle swarm optimization algorithm, and generate the classification results of the personalized treatment plan for each target patient.

[0100] It can be understood that each intervention in the basic treatment decision can be ranked by sensitivity based on the stratified contribution parameters. The purpose of sensitivity ranking is to evaluate the contribution of each intervention to the treatment effect. For example, for high-risk patients, immunotherapy may be more critical than chemotherapy, so immunotherapy interventions will be ranked first. The particle swarm optimization algorithm (PSO) can be used for further optimization to generate personalized treatment plan classification results, which significantly improves the objectivity and prediction efficiency of tumor stratification, optimizes the allocation of medical resources, and enhances the interpretability of the model through a dynamic weight mechanism, providing a highly reliable quantitative basis for clinical decision-making, effectively reducing the risk of overtreatment and improving patient prognosis management.

[0101] Exemplarily, PSO is an optimization algorithm that simulates the movement of particle swarms in nature, aiming to find the optimal solution. In the generation of treatment plans, the goal of particle swarm optimization is to balance the treatment effect and medical resource constraints. Specifically, the algorithm is optimized under the following goals: According to the survival gain model in historical data, a treatment combination is selected to maximize the patient's survival and quality of life. According to the resource constraints of the hospital (such as drug inventory, beds, medical equipment, etc.), the algorithm needs to control the cost of treatment and resource consumption when optimizing the treatment plan. For example, if the drug inventory is limited, the treatment plan will be adjusted to select available drugs or treatment plans. Through iterative optimization, the particle swarm algorithm can eventually generate personalized treatment plans and resource scheduling recommendations. For example, for high-risk patients, the treatment plan that may be output is: "High risk-priority Keytruda + platinum-need to monitor liver toxicity", and the recommended resource scheduling is: "ICU beds need to be reserved for 3 days".

[0102] Corresponding to the intelligent classification prediction method for tumor treatment in the above embodiment, the embodiment of the present application also provides an intelligent classification prediction system for tumor treatment, and each unit of the system can implement each step of the intelligent classification prediction method for tumor treatment. Figure 3 The structural block diagram of the intelligent classification and prediction system for tumor treatment provided in an embodiment of the present application is shown. For the sake of ease of explanation, only the parts related to the embodiment of the present application are shown.

[0103] Reference Figure 3 , the tumor care intelligent classification prediction system includes: An acquisition unit, used to acquire medical data of each target patient; wherein the medical data includes tumor-related molecular marker detection values ​​and clinical diagnosis and treatment characteristics; A stratification unit, used to construct a multi-dimensional prognostic stratification indicator set based on the tumor-related molecular marker detection values ​​and the clinical diagnosis and treatment characteristics; A fusion unit, used for inputting the medical data into a pre-trained multi-layer neural network model for feature fusion to obtain a tumor characterization map; wherein the tumor characterization map is a visual data model generated by performing feature fusion on the medical data through a deep learning model; A scoring unit, used to calculate the stratification contribution parameter and the prognostic risk score of each of the target patients according to the tumor characterization map and the multi-dimensional prognostic stratification indicator set; The prediction unit generates a personalized treatment plan classification result for each target patient based on the stratification contribution parameter and the prognostic risk score of each target patient.

[0104] It should be noted that the information interaction, execution process, etc. between the above-mentioned systems / units are based on the same concept as the method embodiment of the present application. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0105] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit module can exist physically separately, or two or more unit modules can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0106] The present application also provides an intelligent classification prediction device for tumor treatment. Figure 4 This is a schematic diagram of the structure of the intelligent classification prediction device for tumor care provided in one embodiment of the present application. Figure 4 As shown, the tumor care intelligent classification prediction device 6 of this embodiment includes: at least one processor 60 ( Figure 4 Only one is shown), at least one memory 61 ( Figure 4 Only one is shown in the figure) and a computer program 62 stored in the at least one memory 61 and executable on the at least one processor 60. When the processor 60 executes the computer program 62, the intelligent classification and prediction device for tumor therapy 6 implements the steps in any of the above-mentioned embodiments of the intelligent classification and prediction method for tumor therapy, or implements the functions of each unit in the above-mentioned system embodiments.

[0107] Exemplarily, the computer program 62 may be divided into one or more units, which are stored in the memory 61 and executed by the processor 60 to complete the present application. The one or more units may be a series of computer program instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program 62 in the tumor care intelligent classification prediction device 6.

[0108] The intelligent classification prediction device 6 for tumor care can be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The intelligent classification prediction device for tumor care can include, but is not limited to, a processor 60 and a memory 61. Those skilled in the art can understand that Figure 4 It is only an example of the intelligent classification and prediction device 6 for tumor treatment and care, and does not constitute a limitation of the intelligent classification and prediction device 6 for tumor treatment and care. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components, for example, it may also include input and output devices, network access devices, buses, etc.

[0109] The processor 60 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0110] In some embodiments, the memory 61 may be an internal storage unit of the intelligent classification prediction device for tumor therapy 6, such as a hard disk or memory of the intelligent classification prediction device for tumor therapy 6. In other embodiments, the memory 61 may also be an external storage device of the intelligent classification prediction device for tumor therapy 6, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the intelligent classification prediction device for tumor therapy 6. Further, the memory 61 may also include both an internal storage unit and an external storage device of the intelligent classification prediction device for tumor therapy 6. The memory 61 is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as the program code of the computer program, etc. The memory 61 may also be used to temporarily store data that has been output or is to be output.

[0111] An embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.

[0112] An embodiment of the present application provides a computer program product. When the computer program product is run on a tumor therapy intelligent classification prediction device, the tumor therapy intelligent classification prediction device implements the steps in any of the above-mentioned method embodiments.

[0113] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, which can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may at least include: any entity or device that can carry the computer program code to the intelligent classification prediction device for tumor care, recording medium, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. For example, a USB flash drive, a mobile hard disk, a disk or an optical disk. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electric carrier signals and telecommunication signals.

[0114] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0115] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0116] In the embodiments provided in the present application, it should be understood that the disclosed intelligent classification prediction system for tumor therapy / intelligent classification prediction device and method for tumor therapy can be implemented in other ways. For example, the embodiments of the intelligent classification prediction system for tumor therapy / intelligent classification prediction device for tumor therapy described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0117] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0118] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A tumor treatment intelligent classification prediction method, characterized in that: include: Obtaining medical data of each target patient; wherein the medical data includes tumor-related molecular marker detection values ​​and clinical diagnosis and treatment characteristics; Constructing a multi-dimensional prognostic stratification indicator set based on the tumor-related molecular marker detection values ​​and the clinical diagnosis and treatment characteristics; Inputting the medical data into a pre-trained multi-layer neural network model for feature fusion to obtain a tumor characterization map; wherein the tumor characterization map is a visual data model generated by performing feature fusion on the medical data through a deep learning model; Calculating the stratification contribution parameter and the prognostic risk score of each of the target patients according to the tumor characterization map and the multi-dimensional prognostic stratification indicator set; Based on the stratification contribution parameters and the prognostic risk score of each target patient, a personalized treatment plan classification result is generated for each target patient.

2. The intelligent classification prediction method for tumor treatment as claimed in claim 1, characterized in that: The multi-dimensional prognostic stratification indicator set is constructed based on the tumor-related molecular marker detection values ​​and the clinical diagnosis and treatment characteristics, including: Performing factor analysis on the tumor-related molecular marker detection values ​​to obtain a set of molecular features of each target patient after dimensionality reduction; Performing spectral clustering on the clinical diagnosis and treatment characteristics to generate clinical characteristic classification groups for each of the target patients; Perform mutual information calculation on the molecular feature set and the clinical feature classification group of each target patient to screen out a strongly correlated feature combination; wherein the strongly correlated feature combination is used to reflect a multimodal feature set greater than a preset threshold; A hierarchical classification framework is constructed based on the combination of the strongly correlated features to obtain a multi-dimensional prognostic stratification indicator set.

3. The intelligent classification prediction method for tumor treatment as claimed in claim 2, characterized in that: The step of performing pedigree clustering on the clinical diagnosis and treatment characteristics to generate a clinical characteristic classification group for each target patient includes: Standardizing the clinical diagnosis and treatment characteristics to obtain a standardized clinical characteristic matrix; Calculating the mutual information matrix between the clinical diagnosis and treatment characteristics to obtain a hierarchical clustering distance matrix between the clinical diagnosis and treatment characteristics; Performing dynamic threshold segmentation based on the hierarchical clustering distance matrix to obtain an initial clinical feature group for each target patient; The clustering quality of each of the initial clinical feature groups is evaluated by the silhouette coefficient, and the initial clinical feature groups with overlaps higher than a preset threshold are merged to generate optimized clinical feature classification groups.

4. The intelligent classification prediction method for tumor treatment as claimed in claim 1, characterized in that: The medical data is input into a pre-trained multi-layer neural network model for feature fusion to obtain a tumor characterization map of each target patient, including: Standardizing the tumor-related molecular marker detection values ​​to obtain standardized molecular marker data for each target patient; Performing feature embedding and null value compensation on the clinical diagnosis and treatment features to obtain structured diagnosis and treatment data of each target patient; Combining the normalized molecular marker data and the structured diagnosis and treatment data of each target patient into a joint input tensor; Extracting local features of the joint input tensor through a convolutional layer of a multi-layer neural network model, and weighted fusion of the local features of the joint input tensor through an attention mechanism to obtain a fused cross-modal fusion feature tensor; The fused cross-modal fusion features are projected into a two-dimensional space to obtain a tumor characterization map.

5. The intelligent classification prediction method for tumor treatment as claimed in claim 4, characterized in that: The local features of the joint input tensor are weightedly fused through an attention mechanism to obtain a fused cross-modal fusion feature tensor, including: Generate a first feature projection based on the normalized molecular marker data and construct a molecular marker query matrix; Generate a second feature projection based on the structured diagnosis and treatment data, and construct a diagnosis and treatment key matrix and a diagnosis and treatment value matrix; Calculating a cross-modal association weight matrix through two-line attention according to the query matrix of the normalized molecular marker data and the diagnosis and treatment key matrix of the structured diagnosis and treatment data; Performing multi-level attention operations in parallel through a multi-scale attention mechanism according to the modality association weight matrix, and weighted splicing the cross-modality association weight matrix with the diagnosis and treatment value matrix to obtain a weighted splicing result; The weighted splicing results after splicing are subjected to skip connection and batch normalization to obtain a fused cross-modal fusion feature tensor.

6. The intelligent classification prediction method for tumor treatment as claimed in claim 1, characterized in that: Calculating the stratification contribution parameter and the prognostic risk score of each target patient according to the tumor characterization map and the multi-dimensional prognostic stratification indicator set includes: Extracting topological connection features and probability distribution features from the tumor representation map to form a composite feature vector; Constructing a stratified weight allocation tensor according to the multi-dimensional prognostic stratification indicator set; Inputting the composite feature vector into a pre-trained random forest model, calculating the dynamic contribution of each stratification index based on the dynamic weight channel in the weight allocation tensor, and obtaining a stratification contribution parameter; The composite feature vector and the stratification contribution parameter are weighted and aggregated, and converted into a prognostic risk score for each target patient through a logistic function.

7. The intelligent classification prediction method for tumor treatment as claimed in claim 6, characterized in that: The step of inputting the composite feature vector into the pre-trained random forest model, calculating the dynamic contribution of each stratification index based on the dynamic weight channel in the weight allocation tensor, and obtaining the stratification contribution parameter includes: Obtaining a pre-constructed training data set based on patient stratification, and defining a classification error objective function based on the training data set; A decision tree classifier is generated through multiple rounds of iterations, and the feature gain value of each classifier is calculated. In each round of iteration, the contribution score of each stratification indicator is updated according to the feature gain value to obtain the cumulative contribution score of each stratification indicator; The multidimensional prognostic stratification indicators are sorted by feature contribution based on the cumulative contribution score, and low-contribution indicators are filtered through a preset pruning threshold according to the feature contribution sorting to perform feature screening on each stratification indicator in the multidimensional prognostic stratification indicator set; The cumulative contribution scores of the screened stratification indicators are probabilistically normalized to generate stratification contribution parameters.

8. The intelligent classification prediction method for tumor treatment as claimed in claim 1, characterized in that: Generating a personalized treatment plan classification result for each target patient based on the stratified contribution parameter and the prognostic risk score of each target patient includes: Based on the prognostic risk score of each of the target patients, obtaining the tumor risk level of each of the target patients; According to the tumor risk level of each target patient, a preset treatment pathway template is matched to obtain a basic treatment decision for each target patient; wherein the basic treatment decision includes multiple intervention measures; The intervention measures in the basic treatment decision are ranked by sensitivity according to the stratified contribution parameters of each target patient, and the medical resource constraints are balanced through a particle swarm optimization algorithm to generate a personalized treatment plan classification result for each target patient.

9. An intelligent classification prediction system for tumor care, characterized in that: include: An acquisition unit, used to acquire medical data of each target patient; wherein the medical data includes tumor-related molecular marker detection values ​​and clinical diagnosis and treatment characteristics; A stratification unit, used to construct a multi-dimensional prognostic stratification indicator set based on the tumor-related molecular marker detection values ​​and the clinical diagnosis and treatment characteristics; A fusion unit, used for inputting the medical data into a pre-trained multi-layer neural network model for feature fusion to obtain a tumor characterization map; wherein the tumor characterization map is a visual data model generated by performing feature fusion on the medical data through a deep learning model; A scoring unit, used to calculate the stratification contribution parameter and the prognostic risk score of each of the target patients according to the tumor characterization map and the multi-dimensional prognostic stratification indicator set; The prediction unit generates a personalized treatment plan classification result for each target patient based on the stratification contribution parameter and the prognostic risk score of each target patient.

10. An intelligent classification prediction device for tumor care, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 9 is implemented.

Images