Gastric cancer chemotherapy benefit prediction analysis method and system

Abstract

The invention relates to a gastric cancer chemotherapy benefit prediction analysis method and system, and the method comprises the steps: obtaining clinical chemotherapy index data, imaging data and basic index data, and carrying out the standard integration processing, and obtaining a clinical integrated data set; carrying out image structure and basic target extraction on the clinical integrated data set to obtain an image structured feature vector and target feature data; obtaining knowledge graph data, and performing association analysis on the knowledge graph data and the target feature data to obtain a corresponding initial scheme group; performing trend prediction on the clinical integration data set according to the dynamic monitoring data to obtain corresponding dynamic trend information; inputting a preset four-factor joint prediction model to perform effect trend prediction to obtain a corresponding benefit prediction result; and performing overall analysis on the initial scheme group and the benefit prediction result to obtain a corresponding overall benefit information report. According to the method, the data basic integrity of benefit prediction can be improved, and more reliable data support is provided for subsequent analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence, and in particular to a method and system for predicting and analyzing the benefits of chemotherapy for gastric cancer. Background Art

[0002] Gastric cancer is a common malignant tumor of the digestive system, and its treatment effect and prognosis are directly related to the patient's quality of life and life expectancy. Chemotherapy is one of the important means of treating gastric cancer, and its treatment plan selection and efficacy prediction have important guiding significance for clinical decision-making. With the continuous development of the concept of precision medicine, how to accurately evaluate and predict the chemotherapy benefits of gastric cancer patients and realize the formulation of individualized treatment plans has become one of the key topics of current research. However, the existing chemotherapy benefit prediction methods often rely too much on single-dimensional clinical data, such as only considering imaging features or basic biochemical indicators, while ignoring the dynamic change characteristics of patients during treatment and the correlation between multi-dimensional data. This one-sided evaluation method easily leads to insufficient accuracy of the prediction results, making it difficult to provide comprehensive and reliable decision support for clinicians. At the same time, due to the lack of dynamic monitoring and trend analysis of various indicators during the patient's treatment process, it is also difficult for the existing prediction methods to adjust the treatment plan in time, affecting the optimization and improvement of the treatment effect. Summary of the invention

[0003] The main purpose of the present invention is to provide a method and system for predicting the benefit of gastric cancer chemotherapy, which can improve the data basis integrity of benefit prediction and provide more reliable data support for subsequent analysis.

[0004] To achieve the above object, the present invention provides a method for predicting and analyzing the benefit of chemotherapy for gastric cancer, comprising: Obtain clinical chemotherapy index data, imaging data and basic index data of the program objectives, and perform standard integration processing to obtain a clinical integrated data set; Extracting image structure and basic targets from the clinical integrated data set to obtain image structured feature vectors and target feature data; Acquire knowledge graph data of chemotherapy regimens, and perform association analysis between the knowledge graph data and the target feature data to obtain a corresponding initial regimen group; Acquire the dynamic monitoring data of the patient, and perform trend prediction on the clinical integrated data set according to the dynamic monitoring data to obtain corresponding dynamic trend information; Inputting the image structured feature vector, the target feature data, the dynamic monitoring data and the dynamic trend information into a preset four-factor joint prediction model to perform effect trend prediction to obtain a corresponding benefit prediction result; The initial scheme group and the benefit forecast results are analyzed as a whole to obtain a corresponding overall benefit information report.

[0005] Furthermore, the clinical chemotherapy index data, imaging data and basic index data of the target of the regimen are obtained, and standard integration processing is performed to obtain a clinical integrated data set, including: Performing standard matrix conversion on the acquired clinical chemotherapy indicator data to obtain a corresponding clinical indicator data matrix; Performing registration processing on the imaging data according to the clinical indicator data matrix to obtain a registered image data set; Performing feature segmentation on the registered image data set to obtain an image feature sequence; Performing region matching processing on the basic indicator data according to the image feature sequence to obtain basic region data; Performing morphological analysis on the basic regional data to obtain a regional basic morphological vector; Performing gene molecular typing processing on the basic indicator data according to the regional basic morphological vector to obtain a gene feature set; Performing differential expression analysis on the gene feature set to obtain a gene expression profile; The clinical indicator data matrix, the image feature sequence, the regional basic morphological vector and the gene expression spectrum are integrated and normalized to obtain the clinical integrated data set.

[0006] Furthermore, the image structure and basic target extraction are performed on the clinical integrated data set to obtain image structured feature vectors and target feature data, including: Performing multi-scale image segmentation processing on the clinical integrated data set to obtain multiple initial target area image blocks; Performing spatial position correlation analysis on the initial target region image block to obtain region position features; Performing three-dimensional matrix quantization construction on the regional position features to obtain a three-dimensional spatial feature matrix; Performing texture extraction on the initial target area image block according to the three-dimensional spatial feature matrix to obtain a corresponding target texture vector; Performing hierarchical clustering on the clinical integrated data set according to the target texture vector to obtain basic classification information; Performing image structure conversion on the basic classification information to obtain the image structured feature vector; Performing target association on the target texture vector according to the image structured feature vector to obtain target association data; The target associated data is subjected to feature fusion according to the basic classification information to obtain the target feature data.

[0007] Furthermore, the step of acquiring knowledge graph data of chemotherapy regimens and performing association analysis between the knowledge graph data and the target feature data to obtain a corresponding initial regimen group includes: Perform multi-dimensional classification and labeling on the chemotherapy regimen to obtain basic data of the label regimen; Performing structural processing on the drug association graph according to the basic data of the labeling scheme to obtain an initial knowledge graph; Expanding node attributes of the initial knowledge graph to obtain the knowledge graph data; Perform similarity calculation on the target feature data according to the knowledge graph data to obtain feature matching data; Recursively traverse and analyze the feature matching degree data to obtain a set of candidate solutions; Perform clinical indicator attribute combination optimization according to the candidate solution set to obtain multiple groups of optimization solution sequences; Performing time series correlation analysis on the plurality of optimization scheme sequences to obtain scheme time series correlation data; The plurality of groups of optimization solution sequences are screened and scored according to the solution time series association data to obtain the initial solution group.

[0008] Furthermore, the acquiring of the patient's dynamic monitoring data and performing trend prediction on the clinical integrated data set according to the dynamic monitoring data to obtain corresponding dynamic trend information include: Extracting indicators from the dynamic monitoring data to obtain a corresponding dynamic monitoring indicator sequence; Performing time segment processing on the clinical integrated data set according to the dynamic monitoring indicator sequence to obtain segmented monitoring data; Performing multi-level recursive analysis on the segmented monitoring data to obtain corresponding monitoring representation information; Performing image dynamic curve comparison on the monitoring characterization information to obtain a load change curve; Performing trend fitting processing on the load change curve to obtain a corresponding fitting curve equation; Prognostic status prediction is performed according to the fitting curve equation to obtain a preliminary trend prediction result; The preliminary trend prediction result and the monitoring characterization information are trend-fused to obtain final dynamic trend information.

[0009] Furthermore, the image structured feature vector, the target feature data, the dynamic monitoring data and the dynamic trend information are input into a preset four-factor joint prediction model to perform effect trend prediction to obtain corresponding benefit prediction results, including: Inputting the image structured feature vector, the target feature data, the dynamic monitoring data and the dynamic trend information into the feature mapping layer of the four-factor joint prediction model, and performing feature fusion to obtain fused feature data; The fused feature data is input into the feature enhancement layer of the four-factor joint prediction model, a correlation matrix is ​​calculated through the feature enhancement layer, and a feature association graph is constructed, and graph structure feature extraction and feature weight allocation are performed on the feature association graph to obtain an enhanced feature vector; Inputting the enhanced feature vector into the prediction modeling layer of the four-factor joint prediction model, constructing a model for the enhanced feature vector through the prediction modeling layer, and performing prediction analysis to obtain a model prediction result; The model prediction results are input into the result integration layer of the four-factor joint prediction model, and the confidence calculation and result optimization of the model prediction results are performed through the result integration layer to obtain and output the benefit prediction results including the final score and confidence interval.

[0010] Furthermore, the enhanced feature vector is input into the prediction modeling layer of the four-factor joint prediction model, the enhanced feature vector is modeled by the prediction modeling layer, and prediction analysis is performed to obtain a model prediction result, including: Performing vector decomposition on the enhanced feature vector through the prediction modeling layer to obtain an image sub-vector, a basic sub-vector, a monitoring sub-vector and a trend sub-vector; Performing nonlinear transformation and feature mapping on the image sub-vector through the prediction modeling layer to obtain a corresponding image model, and performing image prediction in combination with the image sub-vector to obtain an image prediction result; The prediction modeling layer performs feature space mapping and function transformation on the basic sub-vector to obtain a corresponding basic model, and performs information prediction in combination with the basic sub-vector to obtain a basic prediction result; Performing time series feature extraction and sequence modeling on the monitoring sub-vector through the prediction modeling layer to obtain a corresponding monitoring model, and performing monitoring analysis in combination with the monitoring sub-vector to obtain a monitoring prediction result; The trend sub-vector is subjected to dynamic feature analysis and trend modeling by the prediction modeling layer to obtain a corresponding trend model, and trend analysis is performed in combination with the trend sub-vector to obtain a trend prediction result; The image prediction result, the basic prediction result, the monitoring prediction result and the trend prediction result are fused to obtain the model prediction result.

[0011] Furthermore, the initial scheme group and the benefit forecast result are analyzed as a whole to obtain a corresponding overall benefit information report, including: Performing group decomposition processing on each chemotherapy regimen in the initial regimen group to obtain regimen combination data; Performing hierarchical matching on the benefit prediction results according to the scheme combination data to obtain a scheme benefit corresponding matrix; Calculate the correlation of the scheme benefit correspondence matrix to obtain corresponding benefit correlation data; Sorting the initial solution group according to the benefit correlation data to obtain a target solution sequence; Performing target action analysis on the target solution sequence to obtain a target action relationship diagram; Performing a periodic evaluation on the target solution sequence according to the target action relationship diagram to obtain a solution periodic score; Dynamically integrating the scheme cycle score and the benefit correlation data to obtain a scheme comprehensive index; Screening the target solution sequence according to the solution comprehensive index to obtain a final target solution set; Performing a time series analysis on the final target solution set to obtain a solution time series curve; The benefit information report is obtained by integrating information according to the scheme timing curve and the scheme comprehensive index.

[0012] The present invention also provides a gastric cancer chemotherapy benefit prediction and analysis device, which is applied to any one of the gastric cancer chemotherapy benefit prediction and analysis methods described above, comprising: An acquisition module, which is used to obtain clinical chemotherapy index data, imaging data and basic index data of the program target, and perform standard integration processing to obtain a clinical integrated data set; An analysis module, the analysis module is used to extract image structure and basic targets from the clinical integrated data set to obtain image structured feature vectors and target feature data; An association module, the association module is used to obtain knowledge graph data of chemotherapy regimens, and perform association analysis between the knowledge graph data and the target feature data to obtain a corresponding initial regimen group; A processing module, the processing module is used to obtain the dynamic monitoring data of the patient, perform trend prediction on the clinical integrated data set according to the dynamic monitoring data, and obtain corresponding dynamic trend information; A control module, wherein the control module is used to input the image structured feature vector, the target feature data, the dynamic monitoring data and the dynamic trend information into a preset four-factor joint prediction model to perform effect trend prediction and obtain a corresponding benefit prediction result; An execution module is used to analyze the initial solution group and the benefit prediction result as a whole to obtain a corresponding overall benefit information report.

[0013] The present invention provides a method and system for predicting and analyzing the benefits of chemotherapy for gastric cancer, which has the following beneficial effects: By integrating the clinical chemotherapy indicators, imaging data and basic indicator data of patients in a standardized manner, the effective fusion of multi-dimensional data is achieved, the basic integrity of the data for benefit prediction is improved, and more reliable data support is provided for subsequent analysis. By extracting the features of image structure and basic targets and combining the association analysis of the knowledge graph of chemotherapy regimens, the individual characteristics of patients can be more accurately evaluated, thereby improving the scientificity and pertinence of the initial treatment regimen selection. By introducing dynamic monitoring data for trend prediction, real-time tracking of changes in various indicators during the patient's treatment process is achieved, which helps to timely discover potential problems and adjust the regimen, and improves the dynamic response ability of the treatment process. Through the application of the four-factor joint prediction model, the imaging features, target features, dynamic monitoring and trend information are comprehensively analyzed, overcoming the limitations of single-dimensional prediction and significantly improving the accuracy and reliability of benefit prediction. By analyzing the initial regimen group and the benefit prediction results as a whole, a systematic benefit information report is formed, which provides a comprehensive decision-making basis for clinicians, helps to achieve precise adjustment and optimization of the treatment regimen, and ultimately improves the treatment effect and prognosis quality of gastric cancer patients. At the same time, the application of this method also provides reliable technical support for the individualized treatment of patients with different types of gastric cancer, and has strong clinical practical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is a flow chart of a method for predicting and analyzing the benefits of chemotherapy for gastric cancer provided by the present invention; Figure 2 This is a structural diagram of a gastric cancer chemotherapy benefit prediction and analysis system provided by the present invention.

[0015] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0017] The present invention is further described below in conjunction with the accompanying drawings and specific implementation methods.

[0018] Reference Figure 1As shown, the present invention provides 1. A method for predicting and analyzing the benefits of chemotherapy for gastric cancer, characterized in that it comprises: Step S1: Acquire the clinical chemotherapy index data, imaging data and basic index data of the program target, and perform standard integration processing to obtain a clinical integrated data set; Step S2: extracting the image structure and basic targets of the clinical integrated data set to obtain image structured feature vectors and target feature data; Step S3: Obtain knowledge graph data of chemotherapy regimens, perform association analysis on the knowledge graph data and target feature data, and obtain a corresponding initial regimen group; Step S4: acquiring the patient's dynamic monitoring data, and performing trend prediction on the clinical integrated data set according to the dynamic monitoring data to obtain corresponding dynamic trend information; Step S5: inputting the image structured feature vector, target feature data, dynamic monitoring data and dynamic trend information into a preset four-factor joint prediction model to perform effect trend prediction and obtain corresponding benefit prediction results; Step S6: Analyze the initial solution group and the benefit forecast results as a whole to obtain a corresponding overall benefit information report.

[0019] Based on the steps described above, think about the process as follows: Step S1: Comprehensively collect and standardize patient data. Collect clinical chemotherapy index data of patients, including laboratory test results such as blood routine, liver function, and renal function, as well as previous chemotherapy records, adverse reaction records, etc. The second is to obtain imaging data, mainly including original image data of CT, MRI, ultrasound and other examinations and corresponding clinical diagnosis reports. Collect basic index data of patients, such as age, gender, body mass index, family history, past history and other basic information. After the data collection is completed, standardization is carried out, including data format unification, missing value processing, outlier identification and processing, etc. In the standardization process, a unified data dictionary and coding rules should be established to ensure that data from different sources can be effectively integrated. The time series characteristics are also considered during data integration, and the examination results at different time points are arranged according to a unified time axis to form a complete clinical data time series. The final clinical integrated data set should have good structure and analyzability, laying the foundation for subsequent feature extraction and analysis.

[0020] Step S2: Structured processing of imaging data. By using computer vision and image processing technology, quantitative features such as tumor size, shape, density, texture and other features are extracted from medical images such as CT and MRI. This process usually uses radiomics methods, including three main links: image segmentation, feature extraction and feature selection. To accurately locate the tumor area, image segmentation can use deep learning methods such as U-Net for automatic segmentation. Feature extraction calculates features of multiple dimensions such as shape features, statistical features, texture features, etc. to form a high-dimensional feature vector. At the same time, target features are also extracted from basic indicator data, including key information such as clinical pathological characteristics, genetic test results, and immune indicators of patients. After dimensionality reduction and feature selection, these features are screened out with feature subsets that are strongly correlated with chemotherapy efficacy. The final feature vector should be able to comprehensively characterize the patient's disease status and prognosis-related factors.

[0021] Step S3: Collect and organize professional knowledge related to chemotherapy for gastric cancer, including indications, contraindications, medication specifications, efficacy evaluation criteria, etc. of various chemotherapy regimens. This information needs to be organized according to the structure of the knowledge graph to establish semantic associations between entities. The knowledge graph should contain multiple types of nodes such as drug entities, regimen entities, indication entities, and various relationships between them. After the construction is completed, the patient's target feature data needs to be mapped to the knowledge graph, and the most suitable candidate regimen set for the patient is found through semantic reasoning and similarity calculation. This process needs to take into account the patient's specific conditions, such as age, physical condition, gene mutation type, etc., and match them with the rules in the knowledge graph to finally form a screened initial regimen group. This regimen group should include multiple optional chemotherapy regimens, with the applicable conditions and expected effects of each regimen.

[0022] Step S4: Establish a complete dynamic monitoring system, including regular blood tests, imaging examinations, symptom assessments and other multi-dimensional monitoring indicators. Monitoring data should include treatment response indicators (such as changes in tumor size, tumor marker levels), toxicity indicators (such as blood cell counts, liver and kidney function), and patient quality of life indicators. Based on data collection, time series analysis methods such as autoregressive models and exponential smoothing methods are needed to predict the changing trends of monitoring indicators. This prediction process needs to combine the baseline data collected in step S1 and the feature data extracted in step S2 to comprehensively evaluate the patient's response trend to the current treatment plan.

[0023] Step S5: Integrate all the key information from the previous steps to build a comprehensive prediction model. The four key factors include: the image structured feature vector obtained from step S2, the target feature data, and the dynamic monitoring data and dynamic trend information in step S4. The prediction model can use machine learning methods, such as random forests, deep neural networks and other algorithms, to establish a multi-input and multi-output prediction system. Historical case data needs to be used for verification and optimization during model training to ensure the accuracy and stability of the prediction. The prediction content should include short-term efficacy prediction (such as recent tumor response) and long-term prognosis prediction (such as progression-free survival, overall survival), as well as the possible risk of adverse reactions. This prediction process needs to consider the interaction between various factors to obtain more accurate prediction results.

[0024] Step S6: Overall benefit analysis report This last step is a summary and evaluation of the entire prediction and analysis process. The initial regimen group is systematically compared and analyzed with the benefit prediction results. The analysis should include the expected benefits and risk assessment of each candidate regimen, taking into account multiple dimensions such as efficacy, side effects, and quality of life. The form of the report should not only meet the professional needs of doctors, but also be easy for patients to understand. The report should include: quantitative prediction indicators (such as predicted survival value, probability of adverse reactions, etc.), qualitative comprehensive evaluation (such as analysis of regimen advantages and limitations), and personalized suggestions (such as dose adjustment suggestions, adjuvant treatment measures, etc.). The report should also include uncertainty analysis of the prediction results to help doctors fully consider various possibilities when making treatment decisions. The overall benefit information report should provide an intuitive and reliable reference for clinical decision-making and promote the implementation of precision treatment.

[0025] The present invention provides a method for predicting and analyzing the benefit of chemotherapy for gastric cancer. By standardizing and integrating the clinical chemotherapy indexes, imaging data and basic index data of the patient, the effective integration of multi-dimensional data is achieved, the basic data integrity of the benefit prediction is improved, and more reliable data support is provided for subsequent analysis. By extracting the features of the image structure and the basic target and combining the association analysis of the knowledge map of the chemotherapy regimen, the individual characteristics of the patient can be more accurately evaluated, thereby improving the scientificity and pertinence of the initial treatment regimen selection. By introducing dynamic monitoring data for trend prediction, real-time tracking of the changes in various indicators during the patient's treatment process is achieved, which helps to timely discover potential problems and adjust the regimen, and improves the dynamic response ability of the treatment process. Through the application of the four-factor joint prediction model, the image features, target features, dynamic monitoring and trend information are comprehensively analyzed, the limitations of single-dimensional prediction are overcome, and the accuracy and reliability of the benefit prediction are significantly improved. By analyzing the initial regimen group and the benefit prediction results as a whole, a systematic benefit information report is formed, which provides a comprehensive decision-making basis for clinicians, helps to achieve precise adjustment and optimization of the treatment regimen, and ultimately improves the treatment effect and prognosis quality of gastric cancer patients. At the same time, the application of this method also provides reliable technical support for the individualized treatment of patients with different types of gastric cancer and has strong clinical practical value.

[0026] In one embodiment, clinical chemotherapy index data, imaging data, and basic index data of the regimen target are obtained, and standard integration processing is performed to obtain a clinical integrated data set, including: The acquired clinical chemotherapy index data were converted to a standard matrix to obtain the corresponding clinical index data matrix. During the standard matrix conversion process, the chemotherapy index data of the patient were organized and arranged according to the time series and index type to construct a multidimensional data matrix. The matrix contains key clinical indicators such as chemotherapy cycles, doses, adverse reactions, etc., and the dimensional differences are eliminated through standardization.

[0027] The imaging data is registered according to the clinical index data matrix to obtain the registered image data set. The registration process uses a combination of rigid transformation and non-rigid transformation to establish the spatial correspondence between image data at different time points. Clinical index data is introduced as a constraint in the registration process to improve the registration accuracy.

[0028] The registered image data set is segmented to obtain an image feature sequence. The feature segmentation uses a multi-scale analysis method to extract the morphological, texture and functional features of the lesion area.

[0029] The basic indicator data is processed by regional matching according to the image feature sequence to obtain the basic regional data. The regional matching process establishes the spatial correspondence between the image features and the basic indicators, and achieves accurate matching through feature similarity measurement. The matching results reflect the correlation between the biological characteristics of different regions and clinical manifestations.

[0030] Morphological analysis is performed on the basic regional data to obtain the basic regional morphological vector. Morphological analysis uses quantitative description methods to extract characteristic parameters such as the size, shape, and boundary of the region. The morphological vector contains feature descriptions of multiple dimensions and comprehensively describes the geometric and topological characteristics of the region.

[0031] According to the regional basic morphological vector, the basic indicator data is processed by gene molecular typing to obtain a gene feature set. The molecular typing process combines morphological features with gene expression data to establish a phenotype-genotype mapping relationship. The typing results reflect the molecular biological characteristics of the tumor and its correlation with prognosis.

[0032] Differential expression analysis of gene feature sets was performed to obtain gene expression profiles. Differential analysis uses statistical methods to identify key genes related to chemotherapy response. Expression profiles contain information such as gene expression levels and regulatory relationships, revealing the molecular mechanisms of chemotherapy sensitivity.

[0033] The clinical indicator data matrix, imaging feature sequence, regional basic morphological vector and gene expression spectrum are integrated and normalized to obtain a clinical integrated data set. The normalization process uses a multimodal data fusion method to eliminate the heterogeneity between different data sources and construct a unified feature representation. The integrated data set contains multiple levels of biological information, providing comprehensive data support for subsequent predictive analysis.

[0034] This embodiment realizes the unified expression and analysis of multi-source heterogeneous data by standardizing and integrating the clinical chemotherapy index data, imaging data and basic index data of gastric cancer patients. Standard matrix conversion and registration processing ensure the spatial correspondence of data at different time points and improve the accuracy of data analysis. The feature segmentation and region matching process establishes the association between image features and basic indicators, revealing the biological characteristics of the lesion area. The combination of morphological analysis and gene molecular typing realizes multi-level feature extraction from phenotype to genotype. Differential expression analysis identifies key genes related to chemotherapy response and provides molecular-level evidence support for predictive analysis. The multimodal data fusion method eliminates the heterogeneity between different data sources, constructs a unified feature representation system, provides a comprehensive data basis for the prediction of the chemotherapy benefit of gastric cancer, and improves the reliability and accuracy of the prediction results.

[0035] In one embodiment, image structure and basic target extraction are performed on the clinical integrated data set to obtain image structure feature vectors and target feature data, including: The clinical integrated data set is subjected to multi-scale image segmentation processing to obtain multiple initial target area image blocks. The image multi-scale segmentation processing adopts an adaptive threshold segmentation algorithm to perform layered segmentation on the medical image data by setting different scale parameters, and obtain the corresponding target area image blocks at each scale. In this process, the minimum scale threshold and the maximum scale threshold are set to ensure the rationality and integrity of the segmentation results.

[0036] The spatial position association analysis is performed on the image blocks of the initial target area to obtain the regional position features. The spatial position association analysis is based on the relative position relationship of the image blocks, establishes a spatial topological structure model, calculates the distance matrix and position weight coefficient between each image block, and forms a complete regional spatial feature description.

[0037] The regional position features are quantitatively constructed by three-dimensional matrix to obtain a three-dimensional spatial feature matrix. In the process of three-dimensional matrix quantitative construction, the regional position features are mapped to the three-dimensional coordinate space, and the feature data is discretized by voxel processing method to construct a feature description matrix with spatial hierarchy.

[0038] The texture of the initial target area image block is extracted according to the three-dimensional spatial feature matrix to obtain the corresponding target texture vector. The texture extraction process uses a multi-dimensional texture analysis method, combined with feature description operators such as gray-level co-occurrence matrix and local binary pattern, to extract deep features of the target area and form a high-dimensional texture feature vector.

[0039] According to the target texture vector, the clinical integrated data set is hierarchically clustered to obtain basic classification information. Hierarchical clustering uses a hierarchical clustering algorithm to calculate the Euclidean distance between texture feature vectors, construct a feature similarity matrix, and perform feature classification based on the similarity threshold.

[0040] The basic classification information is converted into an image structure to obtain an image structured feature vector. The image structure conversion process establishes a feature mapping model to convert discrete classification information into continuous feature representations, and forms a standardized structural feature vector through feature dimensionality reduction and normalization.

[0041] According to the image structured feature vector, the target texture vector is associated with the target to obtain the target association data. In the target association process, a feature matching model is established, and the cosine similarity calculation method is used to perform similarity matching on the feature vectors. The matching threshold is set to filter the valid association items and generate the target association matrix.

[0042] The target associated data is feature fused based on the basic classification information to obtain the target feature data. The feature fusion process adopts a weighted fusion strategy, assigns weight coefficients according to the importance of different features, integrates the associated data through the feature combination optimization algorithm, and finally generates a comprehensive target feature representation.

[0043] This embodiment extracts the image structure and basic targets of the clinical integrated data set, and realizes the accurate hierarchical processing and feature extraction of medical image data based on multi-scale segmentation and spatial position association analysis. A complete spatial feature description system is established by using three-dimensional matrix quantization construction and texture extraction technology, which improves the accuracy and completeness of feature expression. Combining hierarchical clustering and image structure conversion methods, a standardized feature vector representation is constructed to enhance the interpretability of the data. Through target association and feature fusion strategies, the effective integration of multidimensional features is achieved, and the comprehensive performance of feature expression is improved. While ensuring the accuracy of feature extraction, this method significantly improves the computational efficiency, providing reliable data support and analysis basis for the prediction of the benefits of chemotherapy for gastric cancer.

[0044] In one embodiment, knowledge graph data of chemotherapy regimens is obtained, and the knowledge graph data is associated with target feature data to obtain a corresponding initial regimen group, including: The multi-dimensional classification and labeling of chemotherapy regimens includes information on drug type, route of administration, dosage range, and use cycle. By establishing a drug attribute label library for each chemotherapy drug, labeling of single-drug and combination drug regimens is performed to form standardized labeling regimen basic data. The labeling regimen basic data covers clinical application characteristics such as drug mechanism of action, adverse reactions, and contraindications.

[0045] The structured processing of the drug association graph converts the basic data of the label scheme into a graph structure representation. In the graph structure, nodes represent specific drugs or schemes, and edges represent the interaction relationship between drugs, including synergy, antagonism, and incompatibility. By establishing a hierarchical node relationship, an initial knowledge graph reflecting the combination rules of chemotherapy schemes is constructed. This knowledge graph contains information such as the strength of interaction between drugs and clinical use experience.

[0046] The node attribute expansion phase supplements the attributes of the nodes in the initial knowledge graph. The expanded attributes include clinical indicators such as drug efficacy data, adverse reaction incidence, and patient tolerance scores. By integrating multi-source clinical data, the attribute characteristics of the nodes are enriched to form a complete knowledge graph data. This knowledge graph data fully reflects the clinical application characteristics of chemotherapy regimens.

[0047] In the similarity calculation phase, the target feature data of the patient is matched and analyzed with the knowledge graph data. The target features include clinical features such as tumor stage, pathological type, and previous treatment history. A multidimensional similarity calculation model is used to comprehensively evaluate the matching degree between features and generate feature matching data. This matching data reflects the compatibility of the scheme with the patient's characteristics.

[0048] Recursive traversal analysis conducts in-depth mining of feature matching data. Based on the preset matching threshold, the scheme combinations with higher matching degree are screened out. Through the recursive search algorithm, the feasibility of different scheme combinations is explored to form a candidate scheme set. The candidate scheme set contains multiple potential treatment scheme combinations.

[0049] Clinical indicator attribute combination optimization optimizes the candidate solution set. Based on multi-dimensional clinical indicators such as efficacy, safety, and economy, the solution combination is comprehensively evaluated. A multi-objective optimization algorithm is used to generate a sequence of solutions with different optimization characteristics. The optimized solution sequence reflects the optimal solution combination under different clinical goals.

[0050] Time series correlation analysis evaluates the time dimension characteristics of the optimization scheme sequence. By analyzing the time series dependency between schemes, the continuity and transition of the treatment scheme are evaluated. Based on the time series analysis model, the associated data reflecting the time series characteristics of the scheme is generated. The associated data reflects the dynamic adjustment rules of the treatment scheme.

[0051] In the screening and scoring phase, multiple groups of optimization scheme sequences are screened. Based on the scheme time series correlation data, the overall coordination of the scheme sequence is evaluated. The final initial scheme group is determined through a comprehensive scoring model.

[0052] This embodiment constructs a complete knowledge graph data system by multi-dimensional classification and annotation and graph processing of chemotherapy regimens, and realizes standardized management and intelligent matching of chemotherapy regimens. Based on multi-dimensional similarity calculation and recursive traversal analysis, candidate regimens that highly match patient characteristics are accurately identified, thereby improving the accuracy of regimen screening. Combined with clinical indicator attribute combination optimization and time series correlation analysis, an optimized regimen sequence with dynamic adjustment characteristics is generated to ensure the continuity and reliability of the treatment regimen. The regimen is finally screened through a comprehensive scoring model, which improves the efficiency of formulating individualized treatment plans while ensuring clinical standardization. This method establishes a scientific and complete chemotherapy regimen screening system, overcomes the limitations of traditional empirical regimen selection, provides clinicians with more objective and reliable decision support, and effectively improves the therapeutic effect of gastric cancer chemotherapy and the accuracy of prognosis evaluation.

[0053] In one embodiment, the dynamic monitoring data of the patient is obtained, and trend prediction is performed on the clinical integrated data set according to the dynamic monitoring data to obtain corresponding dynamic trend information, including: The dynamic monitoring data is subjected to index extraction to obtain the corresponding dynamic monitoring index sequence. The dynamic monitoring data includes clinical data such as the patient's blood index, vital signs, and imaging examination results. During the index extraction process, specific extraction rules are set for different types of monitoring data to extract key feature parameters and form a standardized index sequence. This index sequence reflects the dynamic change trend of the patient during the treatment process.

[0054] The clinical integrated data set is processed by time segmentation according to the dynamic monitoring indicator sequence to obtain segmented monitoring data. Time segmentation processing divides the indicator sequence based on the preset time window and sampling interval. The clinical integrated data set contains complete information such as the patient's baseline characteristics, treatment plan, follow-up records, etc. Segmentation processing ensures the continuity and timeliness of data analysis.

[0055] Multi-level recursive analysis is performed on the segmented monitoring data to obtain the corresponding monitoring representation information. Recursive analysis uses a hierarchical data processing model to extract data features at different time scales. The monitoring representation information includes multi-dimensional information such as the patient's response characteristics to chemotherapy, toxic and side effects, and prognosis-related factors.

[0056] The dynamic image curve is compared with the monitoring characterization information to obtain the load change curve. During the dynamic curve comparison process, the patient's imaging examination results are correlated with the monitoring characterization information. The load change curve reflects the change pattern of tumor load with treatment time.

[0057] The load change curve is subjected to trend fitting to obtain the corresponding fitting curve equation. The trend fitting uses a nonlinear regression model to establish a fitting equation that mathematically describes the change in tumor load. The fitting process takes data noise and measurement errors into account, and the fitting accuracy is improved through an optimization algorithm.

[0058] The prognosis status is predicted based on the fitting curve equation to obtain preliminary trend prediction results. The prognosis status prediction is based on the changing characteristics of the fitting curve and combined with the clinical prognosis evaluation criteria to quantitatively evaluate the treatment effect of the patient. The prediction results include key indicators such as disease progression risk and survival expectation.

[0059] The preliminary trend prediction results and monitoring characterization information are trend-fused to obtain the final dynamic trend information. Trend fusion adopts a multi-source data integration method, comprehensively considering the reliability of the prediction results and the timeliness of the monitoring information. Dynamic trend information provides an objective basis for clinical decision-making and guides the timely adjustment of chemotherapy regimens.

[0060] This embodiment achieves comprehensive tracking and evaluation of various clinical indicators during chemotherapy by establishing a multi-level dynamic monitoring and analysis mechanism. The time segmentation processing method is used to process the clinical integrated data set, which ensures the timeliness and continuity of data analysis and improves the accuracy of the prediction results. Based on the recursive analysis model, the segmented monitoring data is deeply mined, and the patient's response characteristics to chemotherapy are effectively extracted, providing a reliable basis for the evaluation of treatment effects. The correlation analysis of imaging data and clinical indicators is realized through dynamic curve comparison technology, which accurately reflects the law of tumor load changes. The nonlinear regression model is used for trend fitting, and an accurate mathematical description is established to overcome the influence of data noise in traditional evaluation methods. Finally, trend fusion is achieved through multi-source data integration method, which not only ensures the reliability of the prediction results, but also provides timely clinical decision support, providing a scientific basis for the dynamic adjustment of chemotherapy regimens.

[0061] In one embodiment, the image structured feature vector, target feature data, dynamic monitoring data and dynamic trend information are input into a preset four-factor joint prediction model to perform effect trend prediction, and the corresponding benefit prediction results are obtained, including: The image structured feature vector, target feature data, dynamic monitoring data and dynamic trend information are input into the feature mapping layer of the four-factor joint prediction model, and feature fusion is performed to obtain fused feature data. The feature mapping layer reduces the dimension and extracts features of the four types of input feature data through a multi-layer neural network structure to extract key information from each type of feature data. The feature fusion process uses an attention mechanism to adaptively weight different types of features to achieve effective feature fusion. The fused feature data contains the core information and interrelationships of various features.

[0062] The fused feature data is input into the feature enhancement layer of the four-factor joint prediction model. The feature enhancement layer calculates the correlation matrix and constructs a feature association graph. The feature association graph is used to extract graph structure features and assign feature weights to obtain an enhanced feature vector. The feature enhancement layer uses a graph neural network structure to calculate the correlation between features through the correlation matrix and construct a feature association graph to depict the topological relationship between features. The graph structure feature extraction process uses graph convolution operations to extract high-order feature information contained in the feature association graph. The feature weight assignment is based on the feature importance score, and the extracted features are weighted to generate an enhanced feature vector.

[0063] The enhanced feature vector is input into the prediction modeling layer of the four-factor joint prediction model. The enhanced feature vector is modeled and analyzed through the prediction modeling layer to obtain the model prediction result. The prediction modeling layer adopts a deep learning model structure to perform nonlinear transformation and feature combination on the enhanced feature vector to build a prediction model. In the prediction analysis process, the model is trained and verified with historical data, and the prediction accuracy of the model is improved by optimizing the loss function. The prediction probability distribution is output as the model prediction result.

[0064] The model prediction results are input into the result integration layer of the four-factor joint prediction model. The confidence of the model prediction results is calculated and the results are optimized through the result integration layer, and the benefit prediction results including the final score and confidence interval are obtained and output. The result integration layer calculates the confidence of the prediction results based on the Bayesian inference method and optimizes the prediction results using an integrated learning strategy. The final score is a weighted average of the prediction results of multiple models, and the confidence interval is calculated based on the predicted probability distribution, providing a reliable reference for clinical decision-making.

[0065] This embodiment uses a four-factor joint prediction model to predict the effectiveness of chemotherapy for gastric cancer, making the prediction process more accurate and reliable. The model fuses multi-dimensional feature data through the feature mapping layer, fully utilizing the key information in imaging features, target features and dynamic data, and improving the comprehensiveness of feature expression. The feature enhancement layer uses a graph neural network structure to construct a feature association graph, effectively capturing the complex correlation between features and enhancing the model's ability to understand the inherent laws of the data. The predictive modeling layer performs model training and verification based on a deep learning structure, significantly improving prediction accuracy. The result integration layer uses Bayesian inference and integrated learning methods to not only provide reliable prediction scores, but also gives corresponding confidence intervals, providing important decision-making references for clinicians to formulate personalized treatment plans, effectively improving the accuracy and reliability of chemotherapy plans.

[0066] In one embodiment, the enhanced feature vector is input into the prediction modeling layer of the four-factor joint prediction model, and the enhanced feature vector is modeled and predicted and analyzed by the prediction modeling layer to obtain the model prediction results, including: The model is constructed and predicted by inputting the enhanced feature vector into the prediction modeling layer of the four-factor joint prediction model. The prediction modeling layer performs vector decomposition operation on the input enhanced feature vector to generate four independent sub-vectors: image sub-vector, basic sub-vector, monitoring sub-vector and trend sub-vector. Each sub-vector carries the patient's image data features, basic information features, clinical monitoring features and disease development trend features.

[0067] For the image sub-vector, the prediction modeling layer uses a deep convolutional neural network for nonlinear transformation to map the image features to a high-dimensional feature space. The local and global features of the image are extracted through multi-layer convolution and pooling operations to establish an image model. The model combines the feature information in the image sub-vector, analyzes the patient's imaging manifestations, and outputs the image prediction results. The image prediction results include the predicted values ​​of key indicators such as tumor size and degree of infiltration.

[0068] For the basic sub-vector, the prediction modeling layer uses the support vector machine algorithm to map the feature space. The optimal classification hyperplane is constructed in the high-dimensional feature space, and the basic model is established through kernel function transformation. The model combines the age, gender, family history and other information in the basic sub-vector to predict the basic condition of the patient and output the basic prediction result. The basic prediction result reflects the overall health status of the patient and the prognosis risk assessment.

[0069] For the monitoring sub-vector, the prediction modeling layer uses a long short-term memory network to extract temporal features. The long-term dependencies of the data are captured through the gating mechanism to establish a monitoring model. The model combines the various biochemical indicators and physical sign data in the monitoring sub-vector, analyzes the dynamic changes of the patient, and outputs the monitoring prediction results. The monitoring prediction results include the change trends of various indicators and abnormal warning information.

[0070] For the trend sub-vector, the prediction modeling layer uses the time series analysis method to perform dynamic feature analysis. The trend model is established through the autoregressive moving average model. The model combines the disease progression data in the trend sub-vector, analyzes the law of disease development, and outputs the trend prediction results. The trend prediction results reflect the direction and speed of disease development.

[0071] After obtaining the four prediction results, the prediction modeling layer integrates the results through a weighted fusion algorithm. According to the reliability and importance of each prediction result, a weight coefficient is assigned, and the final model prediction result is obtained using a linear weighting method.

[0072] This embodiment achieves a comprehensive prediction and analysis of the chemotherapy benefits for gastric cancer by adopting enhanced feature vector decomposition and a four-factor joint prediction model. This method uses deep convolutional neural network processing on image sub-vectors to effectively extract patient image feature information and improve the accuracy of prediction results. The basic sub-vectors are mapped to feature space by the support vector machine algorithm, which enhances the ability to analyze the basic condition of patients. The long short-term memory network is used to process the monitoring sub-vectors, which realizes the dynamic tracking of the patient's clinical indicators and improves the timeliness of the prediction. The trend sub-vectors are processed using time series analysis methods to accurately grasp the laws of disease development.

[0073] In one embodiment, the initial solution group and the benefit forecast results are analyzed as a whole to obtain a corresponding overall benefit information report, including: When analyzing the initial regimen group and benefit prediction results as a whole, each chemotherapy regimen in the initial regimen group is decomposed into specific drug combinations, administration time, administration cycle and other basic data items to form a standardized regimen combination data structure. This decomposition adopts the hierarchical analysis method to systematically decompose each chemotherapy regimen according to the dimensions of treatment goals, administration methods, drug types, etc.

[0074] When performing stratified matching of benefit prediction results, a multi-dimensional mapping relationship is established based on the scheme combination data, and a corresponding relationship is established between the prediction results and the specific scheme elements. In the stratified matching process, three levels are set: drug action mechanism matching, drug administration sequence matching, and treatment cycle matching. The scheme benefit correspondence matrix is ​​generated by calculating the matching degree of each level. The matching rules include the similarity threshold of the drug action mechanism not less than 0.8, the overlap of the drug administration sequence not less than 0.7, and the matching degree of the treatment cycle not less than 0.75.

[0075] In the correlation calculation stage, the grey correlation analysis method is applied to the scheme benefit corresponding matrix to calculate the correlation degree between each scheme element and the benefit index. The correlation calculation adopts the minimum-maximum normalization processing, and the correlation threshold is set to 0.6. The correlation items below the threshold will be filtered. The initial scheme group is sorted based on the benefit correlation data. When generating the target scheme sequence, the weighted sorting algorithm is used, and the weight coefficient is determined according to the importance of each benefit index.

[0076] In the target effect analysis phase, a target effect relationship diagram is constructed and graph theory is used to analyze the interaction between the schemes. In the relationship diagram, nodes represent specific schemes, edges represent the interaction relationship between schemes, and edge weights reflect the intensity of the interaction. In the periodic evaluation process, an evaluation index system is set up, including dimensions such as the duration of efficacy, the degree of adverse reactions, and patient tolerance, and the score is based on a percentage system.

[0077] The calculation of the scheme comprehensive index adopts the dynamic weighted fusion method to integrate the scheme cycle score with the benefit correlation data. The time decay factor is set in the fusion process to reflect the importance of recent data. When screening the final target scheme set, the comprehensive index threshold is set to 0.75, and schemes above the threshold are retained. In the time series analysis stage, the time series analysis method is applied to the final target scheme set to construct the scheme time series curve to reflect the changing trend of the scheme benefit over time.

[0078] In the process of generating the benefit information report, the scheme time series curve and scheme comprehensive index data are integrated to form a comprehensive report containing scheme benefit evaluation, risk warnings, optimization suggestions, etc. The report content structure includes four parts: scheme overview, benefit analysis, risk assessment, and improvement suggestions. Each part is based on specific data support to ensure the objectivity and reliability of the report.

[0079] This embodiment uses a multi-level matching mechanism and a grey correlation analysis method to establish an accurate correspondence between scheme elements and benefit indicators, overcoming the problems of fuzzy benefit evaluation and difficult-to-quantify correlation in traditional analysis methods. The target action relationship diagram is introduced to analyze the mutual influence between schemes, and the dynamic weighted fusion method is combined to calculate the comprehensive index, thereby achieving a comprehensive evaluation of the benefits of chemotherapy schemes. The scheme time series curve is constructed through time series analysis to accurately reflect the trend of benefit changes and provide a reliable basis for scheme optimization. The final generated benefit information report covers analysis results in multiple dimensions, which not only ensures the objectivity of the evaluation results, but also provides targeted improvement suggestions, significantly improving the efficiency of formulating gastric cancer chemotherapy schemes and the treatment effect.

[0080] Reference Figure 2 As shown, the present invention also provides a gastric cancer chemotherapy benefit prediction and analysis device, which is applied to any one of the gastric cancer chemotherapy benefit prediction and analysis methods mentioned above, comprising: The acquisition module is used to obtain the clinical chemotherapy index data, imaging data and basic index data of the program target, and perform standard integration processing to obtain a clinical integrated data set; An analysis module is used to extract the image structure and basic targets of the clinical integrated data set to obtain the image structure feature vector and target feature data; An association module is used to obtain knowledge graph data of chemotherapy regimens, and to perform association analysis between the knowledge graph data and target feature data to obtain a corresponding initial regimen group; A processing module, the processing module is used to obtain the patient's dynamic monitoring data, and perform trend prediction on the clinical integrated data set according to the dynamic monitoring data to obtain corresponding dynamic trend information; A control module, which is used to input the image structured feature vector, target feature data, dynamic monitoring data and dynamic trend information into a preset four-factor joint prediction model to perform effect trend prediction and obtain corresponding benefit prediction results; The execution module is used to analyze the initial solution group and the benefit prediction results as a whole to obtain the corresponding overall benefit information report.

[0081] The present invention provides a gastric cancer chemotherapy benefit prediction and analysis system, which realizes the effective integration of multi-dimensional data by standard integration processing of the patient's clinical chemotherapy indicators, imaging data and basic indicator data, improves the basic data integrity of benefit prediction, and provides more reliable data support for subsequent analysis. By extracting the features of image structure and basic targets and combining the association analysis of chemotherapy scheme knowledge map, the individual characteristics of patients can be more accurately evaluated, thereby improving the scientificity and pertinence of the initial treatment scheme selection. By introducing dynamic monitoring data for trend prediction, real-time tracking of changes in various indicators during the patient's treatment process is achieved, which helps to timely discover potential problems and adjust the scheme, and improves the dynamic response ability of the treatment process. Through the application of the four-factor joint prediction model, the image features, target features, dynamic monitoring and trend information are comprehensively analyzed, which overcomes the limitations of single-dimensional prediction and significantly improves the accuracy and reliability of benefit prediction. By analyzing the initial scheme group and the benefit prediction results as a whole, a systematic benefit information report is formed, which provides a comprehensive decision-making basis for clinicians, helps to achieve precise adjustment and optimization of treatment schemes, and ultimately improves the treatment effect and prognosis quality of gastric cancer patients. At the same time, the application of this method also provides reliable technical support for the individualized treatment of patients with different types of gastric cancer and has strong clinical practical value.

[0082] It should be noted that technicians in the relevant technical field can clearly understand that for the convenience and conciseness of description, the specific working process of the system and each module described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0083] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for predicting and analyzing the effect of chemotherapy for gastric cancer, characterized in that: include: Obtain clinical chemotherapy index data, imaging data and basic index data of the program objectives, and perform standard integration processing to obtain a clinical integrated data set; Extracting image structure and basic targets from the clinical integrated data set to obtain image structured feature vectors and target feature data; Acquire knowledge graph data of chemotherapy regimens, and perform association analysis between the knowledge graph data and the target feature data to obtain a corresponding initial regimen group; Acquire the dynamic monitoring data of the patient, and perform trend prediction on the clinical integrated data set according to the dynamic monitoring data to obtain corresponding dynamic trend information; Inputting the image structured feature vector, the target feature data, the dynamic monitoring data and the dynamic trend information into a preset four-factor joint prediction model to perform effect trend prediction to obtain a corresponding benefit prediction result; The initial scheme group and the benefit forecast results are analyzed as a whole to obtain a corresponding overall benefit information report.

2. The method for predicting and analyzing the effect of chemotherapy for gastric cancer according to claim 1, characterized in that: The clinical chemotherapy index data, imaging data and basic index data of the target of the scheme are obtained, and standard integration processing is performed to obtain a clinical integrated data set, including: Performing standard matrix conversion on the acquired clinical chemotherapy indicator data to obtain a corresponding clinical indicator data matrix; Performing registration processing on the imaging data according to the clinical indicator data matrix to obtain a registered image data set; Performing feature segmentation on the registered image data set to obtain an image feature sequence; Performing region matching processing on the basic indicator data according to the image feature sequence to obtain basic region data; Performing morphological analysis on the basic regional data to obtain a regional basic morphological vector; Performing gene molecular typing processing on the basic indicator data according to the regional basic morphological vector to obtain a gene feature set; Performing differential expression analysis on the gene feature set to obtain a gene expression profile; The clinical indicator data matrix, the image feature sequence, the regional basic morphological vector and the gene expression spectrum are integrated and normalized to obtain the clinical integrated data set.

3. The method for predicting and analyzing the effect of chemotherapy for gastric cancer according to claim 1, characterized in that: The step of extracting the image structure and basic targets of the clinical integrated data set to obtain image structured feature vectors and target feature data includes: Performing multi-scale image segmentation processing on the clinical integrated data set to obtain multiple initial target area image blocks; Performing spatial position correlation analysis on the initial target region image block to obtain region position features; Performing three-dimensional matrix quantization construction on the regional position features to obtain a three-dimensional spatial feature matrix; Performing texture extraction on the initial target area image block according to the three-dimensional spatial feature matrix to obtain a corresponding target texture vector; Performing hierarchical clustering on the clinical integrated data set according to the target texture vector to obtain basic classification information; Performing image structure conversion on the basic classification information to obtain the image structured feature vector; Performing target association on the target texture vector according to the image structured feature vector to obtain target association data; The target associated data is subjected to feature fusion according to the basic classification information to obtain the target feature data.

4. The method for predicting and analyzing the effect of chemotherapy for gastric cancer according to claim 1, characterized in that: The step of acquiring knowledge graph data of chemotherapy regimens and performing association analysis between the knowledge graph data and the target feature data to obtain a corresponding initial regimen group includes: Perform multi-dimensional classification and labeling on the chemotherapy regimen to obtain basic data of the label regimen; Performing structural processing on the drug association graph according to the basic data of the labeling scheme to obtain an initial knowledge graph; Expanding node attributes of the initial knowledge graph to obtain the knowledge graph data; Perform similarity calculation on the target feature data according to the knowledge graph data to obtain feature matching data; Recursively traverse and analyze the feature matching degree data to obtain a set of candidate solutions; Perform clinical indicator attribute combination optimization according to the candidate solution set to obtain multiple groups of optimization solution sequences; Performing time series correlation analysis on the plurality of optimization scheme sequences to obtain scheme time series correlation data; The plurality of groups of optimization solution sequences are screened and scored according to the solution time series association data to obtain the initial solution group.

5. The method for predicting and analyzing the effect of chemotherapy for gastric cancer according to claim 1, characterized in that: The acquiring the dynamic monitoring data of the patient, and performing trend prediction on the clinical integrated data set according to the dynamic monitoring data to obtain corresponding dynamic trend information, includes: Extracting indicators from the dynamic monitoring data to obtain a corresponding dynamic monitoring indicator sequence; Performing time segment processing on the clinical integrated data set according to the dynamic monitoring indicator sequence to obtain segmented monitoring data; Performing multi-level recursive analysis on the segmented monitoring data to obtain corresponding monitoring representation information; Performing image dynamic curve comparison on the monitoring characterization information to obtain a load change curve; Performing trend fitting processing on the load change curve to obtain a corresponding fitting curve equation; Prognostic status prediction is performed according to the fitting curve equation to obtain a preliminary trend prediction result; The preliminary trend prediction result and the monitoring characterization information are trend-fused to obtain final dynamic trend information.

6. The method for predicting and analyzing the effect of chemotherapy for gastric cancer according to claim 1, characterized in that: The input of the image structured feature vector, the target feature data, the dynamic monitoring data and the dynamic trend information into a preset four-factor joint prediction model to perform effect trend prediction to obtain corresponding benefit prediction results, including: Inputting the image structured feature vector, the target feature data, the dynamic monitoring data and the dynamic trend information into the feature mapping layer of the four-factor joint prediction model, and performing feature fusion to obtain fused feature data; The fused feature data is input into the feature enhancement layer of the four-factor joint prediction model, a correlation matrix is ​​calculated through the feature enhancement layer, and a feature association graph is constructed, and graph structure feature extraction and feature weight allocation are performed on the feature association graph to obtain an enhanced feature vector; Inputting the enhanced feature vector into the prediction modeling layer of the four-factor joint prediction model, constructing a model for the enhanced feature vector through the prediction modeling layer, and performing prediction analysis to obtain a model prediction result; The model prediction results are input into the result integration layer of the four-factor joint prediction model, and the confidence calculation and result optimization of the model prediction results are performed through the result integration layer to obtain and output the benefit prediction results including the final score and confidence interval.

7. The method for predicting and analyzing the effect of chemotherapy for gastric cancer according to claim 6, characterized in that: The step of inputting the enhanced feature vector into the prediction modeling layer of the four-factor joint prediction model, constructing a model for the enhanced feature vector through the prediction modeling layer, and performing prediction analysis to obtain a model prediction result includes: Performing vector decomposition on the enhanced feature vector through the prediction modeling layer to obtain an image sub-vector, a basic sub-vector, a monitoring sub-vector and a trend sub-vector; Performing nonlinear transformation and feature mapping on the image sub-vector through the prediction modeling layer to obtain a corresponding image model, and performing image prediction in combination with the image sub-vector to obtain an image prediction result; The prediction modeling layer performs feature space mapping and function transformation on the basic sub-vector to obtain a corresponding basic model, and performs information prediction in combination with the basic sub-vector to obtain a basic prediction result; Performing time series feature extraction and sequence modeling on the monitoring sub-vector through the prediction modeling layer to obtain a corresponding monitoring model, and performing monitoring analysis in combination with the monitoring sub-vector to obtain a monitoring prediction result; The trend sub-vector is subjected to dynamic feature analysis and trend modeling by the prediction modeling layer to obtain a corresponding trend model, and trend analysis is performed in combination with the trend sub-vector to obtain a trend prediction result; The image prediction result, the basic prediction result, the monitoring prediction result and the trend prediction result are fused to obtain the model prediction result.

8. The method for predicting and analyzing the effect of chemotherapy for gastric cancer according to claim 1, characterized in that: The overall analysis of the initial scheme group and the benefit forecast results to obtain a corresponding overall benefit information report includes: Performing group decomposition processing on each chemotherapy regimen in the initial regimen group to obtain regimen combination data; Performing hierarchical matching on the benefit prediction results according to the scheme combination data to obtain a scheme benefit corresponding matrix; Calculate the correlation of the scheme benefit correspondence matrix to obtain corresponding benefit correlation data; Sorting the initial solution group according to the benefit correlation data to obtain a target solution sequence; Performing target action analysis on the target solution sequence to obtain a target action relationship diagram; Performing a periodic evaluation on the target solution sequence according to the target action relationship diagram to obtain a solution periodic score; Dynamically integrating the scheme cycle score and the benefit correlation data to obtain a scheme comprehensive index; Screening the target solution sequence according to the solution comprehensive index to obtain a final target solution set; Performing a time series analysis on the final target solution set to obtain a solution time series curve; The benefit information report is obtained by integrating information according to the scheme timing curve and the scheme comprehensive index.

9. A device for predicting and analyzing the benefits of chemotherapy for gastric cancer, characterized in that: The method for predicting and analyzing the effect of chemotherapy for gastric cancer as described in any one of claims 1 to 8 above comprises: An acquisition module, which is used to obtain clinical chemotherapy index data, imaging data and basic index data of the program target, and perform standard integration processing to obtain a clinical integrated data set; An analysis module, the analysis module is used to extract image structure and basic targets from the clinical integrated data set to obtain image structured feature vectors and target feature data; An association module, the association module is used to obtain knowledge graph data of chemotherapy regimens, and perform association analysis between the knowledge graph data and the target feature data to obtain a corresponding initial regimen group; A processing module, the processing module is used to obtain the dynamic monitoring data of the patient, perform trend prediction on the clinical integrated data set according to the dynamic monitoring data, and obtain corresponding dynamic trend information; A control module, wherein the control module is used to input the image structured feature vector, the target feature data, the dynamic monitoring data and the dynamic trend information into a preset four-factor joint prediction model to perform effect trend prediction and obtain a corresponding benefit prediction result; An execution module is used to analyze the initial solution group and the benefit prediction result as a whole to obtain a corresponding overall benefit information report.

Images