Brain tumor multidimensional data analysis method and system

Abstract

The invention relates to the technical field of data mining, in particular to a brain tumor multi-dimensional data analysis method and system, and the method comprises the following steps: based on a brain tumor multi-dimensional data set, integrating image data, clinical data and gene data, distributing an initial weight of each type of data, and matching the criticality of each type of data in a target diagnosis task. According to the method, various types of data are ensured to be processed according to the actual contribution of the data in judgment through weight optimization, so that the brain tumor data set better meets clinical requirements, the sensitivity and specificity of brain tumor judgment are improved, mining of complementarity among different data sources is improved through refined feature fusion, and the accuracy of brain tumor judgment is improved. Pathological features which are difficult to find through a single data source become more obvious, clustering analysis and data quality evaluation technologies are adopted, data processing comprehensiveness is ensured, more reliable support is provided for judgment and decision, the whole analysis process is optimized, and data are increased in quantity and improved in quality remarkably.

Description

Technical Field

[0001] The present invention relates to the technical field of data mining, and in particular to a method and system for multidimensional data analysis of brain tumors. Background Art

[0002] Data mining technology aims to discover patterns, associations and knowledge from large amounts of data, and uses statistics, machine learning and artificial intelligence techniques to classify, predict, associate and cluster data. Data mining is widely used in business intelligence, financial analysis, bioinformatics, healthcare and other fields. Through data mining, valuable information can be extracted from complex data sets to help decision makers make data-based decisions.

[0003] Among them, the multidimensional data analysis method of brain tumors involves the use of data mining technology to analyze various types of data related to brain tumors, such as patients' clinical data, imaging data, and genetic data. The purpose is to improve the diagnostic accuracy of brain tumors, predict disease progression, and the probability of responding to treatment. By integrating and analyzing data from different sources, doctors can more comprehensively analyze the characteristics of brain tumors and provide patients with more accurate judgments.

[0004] Existing technologies lack sufficient adjustment mechanisms to adapt to the particularity and heterogeneity of data, and fail to fully adjust and optimize the processing weights for different types of medical data, resulting in the inability to effectively identify the data features that are most critical to diagnosis. In addition, existing technologies lack sufficient flexibility and depth in data fusion and feature extraction, and cannot accurately capture the complex relationships between disease markers, limiting their ability to provide medical support. Existing technologies also have imprecise data quality control and ignore key anomalies in the data set, resulting in inaccurate brain tumor data sets and affecting early detection of the disease. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a brain tumor multidimensional data analysis method and system.

[0006] In order to achieve the above object, the present invention adopts the following technical scheme: a multidimensional data analysis method for brain tumors, comprising the following steps: S1: Based on the brain tumor multidimensional data set, imaging data, clinical data and genetic data are integrated, and the initial weight of each data is assigned to match its criticality in the target diagnosis task. The fusion process is optimized according to the differences in the data to construct a data fusion expression set. S2: Based on the data fusion expression set, optimize the data structure, extract features of multiple data sources, and fuse the features of differentiated data sources to refine the mutual complementarity between data and obtain a feature fusion set; S3: Based on the feature fusion set, key diagnostic features are extracted and analyzed by density clustering, potential feature combinations associated with brain tumors are identified, and compared with known brain tumor judgment criteria to obtain feature matching analysis information; S4: Based on the feature matching analysis information, the data quality is evaluated using the isolation forest, abnormal data points in the data set are marked, a correction process is developed for the data points, and the modifications are applied to the data set through manual review to obtain data quality control results.

[0007] The present invention is improved in that the steps of allocating the initial weight of each type of data are specifically as follows: S111: Based on the brain tumor multidimensional dataset, the imaging data, clinical data and genetic data were integrated, and data cleaning and standardization were performed to generate the cleaned dataset; S112: Based on the cleaned data set, quantitatively evaluate the influence of multiple types of data in the diagnosis task and generate a data influence score; S113: Based on the data influence score, the formula is used:

[0008] Calculate the initial weight of each data type and generate the weight distribution result, where: Represents data type The initial weight of Represents data type The impact score, Represents the adjustment coefficient.

[0009] The present invention is improved in that the step of obtaining the data fusion expression set is specifically as follows: S121: Identify the criticality of the data set in the target diagnostic task, combine the difference scale and type between the data, perform standardization and feature encoding, and obtain the key feature data set; S122: Using linear combination and nonlinear transformation to process the key feature data set, using the formula:

[0010] Get data fusion expression set ,in, No. The key feature input of class data, It is The weight coefficient of the class data, It is The weight matrix corresponding to the class data, No. The bias term corresponding to the class data, Indicates the number of data sources.

[0011] The present invention is improved in that the step of fusing the differentiated data source features is specifically as follows: S211: Based on the data fusion expression set, key features are identified from the imaging, clinical and genetic data, including imaging texture features, clinical biochemical indicators and gene mutation points, to obtain deep feature analysis results; S212: Based on the deep feature analysis results, data structure optimization is performed using the formula:

[0012] Normalize the difference scale features to the same interval to obtain the standardized feature results ,in, represents the original feature value extracted from the data source, and are the minimum and maximum values ​​in the feature set; S213: Based on the standardized feature results, multiple types of features are integrated, and multi-feature weights are automatically adjusted to obtain data source fusion processing results.

[0013] The present invention is improved in that the step of acquiring the feature fusion set is specifically as follows: S221: Based on the processed data source features, calculate the correlation coefficient of each data feature, perform complementarity evaluation, and generate a feature correlation matrix; S222: Based on the feature correlation matrix, the formula is:

[0014] Perform weighted fusion of features to obtain the feature fusion set ,in, Indicates The current value of a feature, It is The weight of the feature, is the total number of features.

[0015] The present invention is improved in that the step of identifying the potential feature combination is specifically as follows: S311: Based on the feature fusion set, data processing is performed, including standardization and noise reduction, to obtain processed feature data; S312: clustering the processed feature data using a density clustering algorithm, calculating the local density between data points, identifying and marking potential key areas, using the formula:

[0016] Calculate the The cluster contribution of data points ,in, Indicates the first data points, Representation and Adjacent data points, It is a data point Neighborhood of is the scale parameter, express and The Euclidean distance between S313: Based on the cluster contribution, a feature set with the best contribution is screened to obtain a potential feature combination associated with brain tumors.

[0017] The present invention is improved in that the step of acquiring the feature matching analysis information is specifically as follows: S321: Based on the potential feature combination, calculate the similarity score between the potential feature combination and the known brain tumor judgment criteria, score each feature combination, and obtain a preliminary scoring result; S322: Based on the preliminary scoring results, calculate the significance using the formula:

[0018] Get feature matching analysis information, where: represents the Z score, is the average feature matching score, is the expected matching score, is the sample standard deviation, is the sample size.

[0019] The present invention is improved in that the steps of obtaining the data quality control result are specifically as follows: S411: Based on the feature matching analysis information, using the isolation forest algorithm, identifying and marking abnormal data points in the data set to obtain abnormal data identification results; S412: Based on the abnormal data identification result, manual review and data point verification are performed, and the correction score is calculated using the formula:

[0020] The data correction results after manual review are obtained, among which, is the corrected score, Representative The weight of the data point, Indicates The binary value of a data point during manual review, is the total number of data points; S413: Based on the data correction result, data quality control is performed to re-evaluate the consistency and reliability of the brain tumor data set to obtain a data quality control result.

[0021] A brain tumor multidimensional data analysis system, the system comprising: The data source adjustment module integrates imaging data, clinical data, and genetic data based on the brain tumor multidimensional data set, adjusts the weight of each data source to match its criticality in the diagnostic task, optimizes the processing differences between multiple data sources, and constructs a data fusion expression set; The feature optimization module extracts and processes the features of multiple data sources based on the data fusion expression set, performs feature optimization, explores the mutual complementarity between data, and merges the differentiated data source features to obtain a feature fusion set; The diagnostic feature clustering module applies density clustering analysis to key diagnostic features based on the feature fusion set, identifies potential feature combinations associated with brain tumors, and compares them with known brain tumor judgment criteria to obtain feature matching analysis information; The data anomaly assessment module uses the isolation forest algorithm to assess anomalies in the data set based on the feature matching analysis information, identifies data quality issues, and generates anomaly data markers; The data correction application module formulates a correction process based on the abnormal data marking, corrects the marked abnormal data points, and refines the integrity of the data set through manual review of the corrected data to obtain data quality control results.

[0022] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, weight optimization is used to ensure that various types of data are processed according to their actual contributions in the judgment, so that the brain tumor data set is more in line with clinical needs, and the sensitivity and specificity of brain tumor judgment are improved. The refined feature fusion improves the mining of complementarity between different data sources, making pathological features that are difficult to discover through a single data source more obvious. Cluster analysis and data quality assessment technology are used to ensure the comprehensiveness of data processing and provide more reliable support for judgment decisions. The optimization of the entire analysis process has not only increased the data in quantity, but also significantly improved its quality, which plays a key role in auxiliary judgment of diseases. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A flowchart of a multidimensional data analysis method for brain tumors proposed by the present invention; Figure 2 A flowchart for allocating the initial weight of each type of data in the present invention; Figure 3 This is a flow chart for obtaining the data fusion expression set in the present invention; Figure 4 It is a flow chart of the fusion of differentiated data source features in the present invention; Figure 5 This is a flow chart for obtaining a feature fusion set in the present invention; Figure 6 is a flow chart for identifying potential feature combinations in the present invention; Figure 7 This is a flowchart for obtaining feature matching analysis information in the present invention; Figure 8 The figure is a flow chart for obtaining data quality control results in the present invention. DETAILED DESCRIPTION

[0024] 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.

[0025] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating positions or positional relationships, are based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0026] Example See also Figure 1 The present invention provides a technical solution: a method for multidimensional data analysis of brain tumors, comprising the following steps: S1: Based on the brain tumor multidimensional data set, imaging data, clinical data and genetic data are integrated, and the initial weight of each data is assigned to match its criticality in the target diagnosis task. The fusion process is optimized according to the differences in the data to construct a data fusion expression set. S2: Based on the data fusion expression set, optimize the data expression, extract the features of multiple data sources, and fuse the features of differentiated data sources to refine the mutual complementarity between data and obtain the feature fusion set; S3: Based on the feature fusion set, key diagnostic features are extracted and analyzed through density clustering, potential feature combinations associated with brain tumors are identified, and compared with known brain tumor judgment criteria to obtain feature matching analysis information; S4: Based on the feature matching analysis information, the data quality is assessed using isolation forests, abnormal data points in the data set are marked, a correction process is developed for the data points, and the modifications are applied to the data set through manual review to obtain data quality control results.

[0027] The data fusion expression set includes a data interface in a unified format, a standardized processing flow, and a data integrity rating. The feature fusion set includes feature normalization indicators, feature correlation analysis results, and feature effectiveness evaluation results. The feature matching analysis information includes matching accuracy, potential diagnostic indicators, and differential feature analysis results. The data quality control results include anomaly point location, data integrity repair information, and audit adjustment records.

[0028] See also Figure 2 , the steps for assigning the initial weight of each data are as follows: S111: Based on the brain tumor multidimensional dataset, the imaging data, clinical data and genetic data were integrated, and data cleaning and standardization were performed to generate the cleaned dataset; Integrate imaging data, clinical data and genetic data to ensure data integrity and compatibility, including the unification of various data formats and the processing of missing values ​​in the data set. For example, imaging data usually exists in DICOM format, while clinical data may be stored in tabular form, and genetic data may appear in the form of sequences or expression spectra. Standardize the data to ensure the accuracy of subsequent analysis, and then apply data cleaning technology to remove irrelevant information and outliers, such as removing non-tissue structure interference in images and filtering out obvious erroneous inputs in clinical data. Genetic data cleaning needs to remove low-quality sequences according to bioinformatics standards to ensure data quality and provide a basis for further data analysis.

[0029] S112: Based on the cleaned data set, quantitatively evaluate the influence of multiple types of data in the diagnosis task and generate data influence scores; In the cleaned comprehensive data set, principal component analysis (PCA) and random forest algorithms were used for feature selection and importance scoring. PCA identified the principal components that best represent the data variability through dimensionality reduction technology, while random forest constructed multiple decision trees to evaluate the average information gain of each feature in random sampling. This not only enhances the model's ability to interpret the data, but also finds the most influential features in the data while reducing the risk of overfitting. After scoring the influence of each data type, a data influence score table was obtained that objectively reflects the impact of each type of data on diagnosis.

[0030] S113: Based on data impact scoring, the formula is:

[0031] Calculate the initial weight of each data type and generate the weight distribution result, where: Represents data type The initial weight of Represents data type The impact score, Represents the adjustment coefficient, which is used to adjust the distribution of weights according to the requirements of the diagnostic task. is the exponential function, used to calculate the exponential power of the base of the natural logarithm; There are three data types, with impact scores: , and ,Pick , first compute the exponential term of each term:

[0032]

[0033]

[0034] Then calculate the sum of the exponential terms:

[0035] Calculate the weight for each data type:

[0036]

[0037]

[0038] This result indicates that the third data type contributes the most to the diagnosis task and should therefore be given a higher weight in data fusion.

[0039] See also Figure 3 , the specific steps for obtaining the data fusion expression set are: S121: Identify the criticality of the data set in the target diagnostic task, combine the difference scale and type between the data, perform standardization and feature encoding, and obtain the key feature data set; In the feature selection stage, data monitoring is used to identify which features have a decisive impact on the diagnostic task. This involves the ability to extract information from large amounts of data. Each data type, such as imaging, clinical, and genetic, has different characteristics. Imaging data focuses on changes in structure and morphology, clinical data focuses on abnormalities in physiological parameters, and genetic data provides the molecular basis for disease development. Through integrated learning and correlation analysis, features with high complementarity and diagnostic relevance are selected to ensure that the selected features can cover all aspects of the disease. The features are then standardized to ensure comparability and compatibility in subsequent steps. The standardization process includes adjusting the scale of the features so that features from different data sources can be compared under the same metric to generate a key feature data set in a unified format.

[0040] S122: Use linear combination and nonlinear transformation to process the key feature data set using the formula:

[0041] Get data fusion expression set ,in, No. The key feature input of class data, It is The weight coefficient of the class data, It is The weight matrix corresponding to the class data, No. The bias term corresponding to the class data, Indicates the number of data sources; There are three data types, are 0.3, 0.5 and 0.2 respectively, is [1, 2, 3], is [0.5, 1, 1.5] and For [0.1, 0.1, 0.1], calculate the weighted input for each feature:

[0042]

[0043]

[0044] Sum to get the fusion feature:

[0045] The results show that the features fused in this way can better reflect the comprehensive impact of different data sources on the diagnosis task and improve the accuracy of judgment.

[0046] See also Figure 4 ,The specific steps of fusing the features of differentiated data sources are: S211: Based on the data fusion expression set, key features are identified from imaging, clinical and genetic data, including image texture features, clinical biochemical indicators and gene mutation points, to obtain deep feature analysis results; Feature extraction is performed to identify key diagnostic features of each data source. For example, texture and shape features of images are extracted from imaging data, key physiological indicators such as heart rate and blood pressure are measured from clinical data, and specific gene expression patterns are analyzed from genetic data. Features are systematically converted into numerical representations for further statistical analysis and machine learning processing to ensure that the information of each data source is fully utilized and in-depth feature analysis results are obtained.

[0047] S212: Based on the results of deep feature analysis, data expression optimization is performed using the formula:

[0048] Normalize the difference scale features to the same interval to obtain the standardized feature results ,in, represents the original feature value extracted from the data source, and are the minimum and maximum values ​​in the feature set; For a particular set of features, the minimum , the maximum value , combined with an original eigenvalue , then the normalized eigenvalues ​​are calculated as follows:

[0049] This result shows that the original eigenvalue of 20 is converted to 0.2 after normalization, and the conversion process maintains the relative differences of the data and the integrity of the original information, making it more suitable for use in different statistical and machine learning algorithms.

[0050] S213: Based on the standardized feature results, multiple types of features are integrated, and the weights of multiple features are automatically adjusted to obtain the data source fusion processing results; After completing the standardization of features, the data is applied to the machine learning model for synthesis and optimization. The use of algorithms such as principal component analysis (PCA) not only improves the efficiency of data processing, but also optimizes the combination of features, so that the information obtained from different data sources can retain the maximum diagnostic value while reducing redundancy. In this way, a highly optimized data source fusion processing result that reflects multiple biomarkers can be created.

[0051] See also Figure 5 , the specific steps for obtaining the feature fusion set are: S221: Based on the processed data source features, calculate the correlation coefficient of each data feature, perform complementarity evaluation, and generate a feature correlation matrix; For features from multiple data sources, correlation analysis is performed. By calculating the correlation coefficient of the features, not only can the direct correlation between single features be identified, but also the common impact of different feature combinations on the target variable can be explored. This involves extracting key diagnostic features from various data sources such as imaging, clinical, and genetic data, and performing systematic statistical evaluation of the features, such as calculating the Pearson or Spearman correlation coefficient to determine linear or nonlinear relationships. The analysis helps to select the feature combination that contributes the most to diagnosis, provides a basis for subsequent feature fusion, and generates a feature correlation matrix.

[0052] S222: Based on the feature correlation matrix, the formula is:

[0053] Perform weighted fusion of features to obtain the feature fusion set ,in, Indicates The current value of a feature, It is The weight of the feature, is the total number of features; Consider three eigenvalues = 0.8 (contribution of image features), = 0.5 (contribution of clinical characteristics), = 0.3 (contribution of gene features), and the weights are =0.5, =0.3, =0.2, calculated according to the formula:

[0054]

[0055] The results show that after comprehensively considering the contribution of the characteristics of each data source, a comprehensive characteristic value of 0.61 was obtained, reflecting the comprehensive contribution of each data source to the diagnosis result.

[0056] See also Figure 6 , the identification steps of potential feature combinations are as follows: S311: Based on the feature fusion set, data processing is performed, including standardization and noise reduction, to obtain processed feature data; Data preprocessing, including standardization and noise reduction, is performed by calculating the mean and standard deviation of each data point and applying the Z-score formula to standardize each eigenvalue. At the same time, the sliding window method is used to smooth the data to reduce the impact of random noise, ensure data quality, provide a basis for cluster analysis, and generate a processed feature data set.

[0057] S312: Use density clustering algorithm to cluster the processed feature data, calculate the local density between data points, identify and mark potential key areas, and use the formula:

[0058] Calculate the The cluster contribution of data points , used to evaluate the influence of each point in its cluster, where Indicates the first data points, Representation and Adjacent data points, It is a data point Neighborhood of is a scale parameter used to control the point pairs in the neighborhood The degree of influence, express and The Euclidean distance between There is a data point , which is in the neighborhood There are three data points in and , the distances are 0.5, 1.0 and 1.5 respectively, , then the contribution of each neighborhood point is:

[0059]

[0060]

[0061] The calculation results are:

[0062]

[0063] The calculation results show that it has a high local density in its neighborhood and is a potential clustering core point, which is helpful to screen out key diagnostic features related to brain tumors from a large amount of data.

[0064] S313: Based on the clustering contribution, the feature set with the best contribution is screened to obtain the potential feature combination associated with brain tumors; According to the results of density clustering, a set of data points with high contribution was selected to represent the key features in the high-density area. The clustering pattern and association rules of the feature points were further analyzed. By identifying the common characteristics of the high-contribution points, the potential feature combinations related to brain tumors were determined, which improved the accuracy of diagnosis and generated potential feature combinations associated with brain tumors.

[0065] See also Figure 7 , the specific steps for obtaining feature matching analysis information are: S321: Based on the potential feature combination, calculate the similarity score with the known brain tumor judgment criteria, score each feature combination, and obtain a preliminary scoring result; Based on the potential feature combination, the feature dimensions related to the brain tumor judgment criteria are determined, including image features and biomarkers, etc. The numerical parameters of each feature dimension are extracted from the existing standard literature, and the weight of each feature in the judgment process is calculated using the weighted average method. The specific process is to extract the texture features and edge information of the image data, and eliminate the deviation between data from different sources through the standardization algorithm. For biomarkers, the correlation analysis model is used to evaluate their association with tumor type and growth stage. After defining the weight distribution, the similarity score with the standard model is calculated for each feature combination. This process uses the Euclidean distance formula to calculate the difference between the feature vectors of the standard model and each feature combination. The difference is squared and then accumulated to obtain the distance value, which is then normalized. Finally, the similarity score of each feature combination is obtained, the feature combination that best meets the standard is screened, and the results are summarized. S322: Based on the preliminary scoring results, calculate the significance using the formula:

[0066] Determine whether there is a statistically significant match between the feature combination and the brain tumor criteria, and obtain feature matching analysis information, wherein: represents the Z score, is the average feature matching score, is the expected matching score, is the sample standard deviation, is the sample size; During the feature matching process, an average feature similarity score is obtained , expected matching score , standard deviation , sample size , the calculation process is as follows:

[0067] The results showed that the obtained feature matching scores were statistically significant, indicating that the matching results were significantly correlated with the judgment criteria for brain tumors, supporting the value of the features in diagnosis.

[0068] See also Figure 8 , the specific steps for obtaining data quality control results are: S411: Based on the feature matching analysis information, the isolation forest algorithm is used to identify and mark abnormal data points in the data set to obtain abnormal data identification results; Based on the feature matching analysis information, the feature set used for analysis is obtained, and the basic statistical information of the data set is prepared. The isolation forest is used to process the features. By analyzing the feature distribution of data points one by one, the outlier distance between the data point and the normal data is calculated, and the outlier distance is further converted into a scoring standard. Based on the scoring, the recognition threshold of the outlier point is set, and the score of each data point is compared with the threshold to confirm which data points are abnormal. Finally, all data points that meet the anomaly judgment criteria are marked to obtain the abnormal data recognition result.

[0069] S412: Based on the abnormal data identification results, manual review and data point verification are performed, and the correction score is calculated using the formula:

[0070] The data correction results after manual review are obtained, among which, is the corrected score, Representative The weight of the data point, Indicates The binary value of a data point in the manual review process, 1 represents the review passed (that is, the data point is normal), 0 represents the review failed (that is, the data point is abnormal), is the total number of data points; If the number of data points , where the weight of the data point , and the manual review results , indicating whether the data point has passed the review (1 for passed, 0 for failed), the calculation process is as follows:

[0071]

[0072] The results show that the overall data set review pass rate is 0.6, or 60%, which means that most key data points meet the quality requirements, helping to ensure the overall quality control of the data set.

[0073] S413: Based on the data correction results, perform data quality control, re-evaluate the consistency and reliability of the brain tumor data set, and obtain data quality control results; Based on the data correction results, data points marked as abnormal are collected, and the characteristic information of the abnormal data points is sorted out. Through manual review, combined with the known patterns and normal data standards in the brain tumor dataset, it is gradually evaluated whether the abnormal data is a data error or a measurement error. The existing correction process is used to adjust the characteristic values ​​of the abnormal data points, and the attribute values ​​are recalculated and corrected to keep them consistent with the normal data. Finally, through further global data quality assessment, the matching of the corrected data points with the overall brain tumor dataset is confirmed to obtain the data quality control results.

[0074] Brain tumor multidimensional data analysis system, the system includes: The data source adjustment module integrates imaging data, clinical data, and genetic data based on the brain tumor multidimensional data set, adjusts the weight of each data source to match its criticality in the diagnostic task, optimizes the processing differences between multiple data sources, and constructs a data fusion expression set; The feature optimization module extracts and processes the features of multiple data sources based on the data fusion expression set, performs feature optimization, explores the mutual complementarity between data, and merges the differentiated data source features to obtain the feature fusion set; The diagnostic feature clustering module uses density clustering to analyze key diagnostic features based on feature fusion sets, identifies potential feature combinations associated with brain tumors, and compares them with known brain tumor judgment criteria to obtain feature matching analysis information; The data anomaly assessment module uses the isolation forest algorithm to evaluate anomalies in the data set based on feature matching analysis information, identify data quality issues, and generate abnormal data markers; The data correction application module develops a correction process based on abnormal data marking, corrects the marked abnormal data points, and refines the integrity of the data set through manual review of the corrected data to obtain data quality control results.

[0075] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A method for multidimensional data analysis of brain tumors, characterized in that: The following steps are involved: Based on the brain tumor multidimensional data set, the imaging data, clinical data and genetic data are integrated, the initial weight of each data is assigned to match its criticality in the target diagnosis task, and the fusion process is optimized according to the differences in the data to construct a data fusion expression set; Based on the data fusion expression set, data structure optimization is performed, features of multiple data sources are extracted, and features of differentiated data sources are fused and processed to refine the mutual complementarity between data and obtain a feature fusion set; Based on the feature fusion set, key diagnostic features are extracted and analyzed by density clustering, potential feature combinations associated with brain tumors are identified, and compared with known brain tumor judgment criteria to obtain feature matching analysis information; Based on the feature matching analysis information, the data quality is assessed using isolation forests, abnormal data points in the data set are marked, a correction process is developed for the data points, and the modifications are applied to the data set through manual review to obtain data quality control results.

2. The brain tumor multidimensional data analysis method according to claim 1, characterized in that: The steps for allocating the initial weight of each type of data are specifically as follows: Based on the brain tumor multidimensional dataset, imaging data, clinical data and genetic data are integrated, and data cleaning and standardization are performed to generate a cleaned dataset; Based on the cleaned data set, quantitatively evaluate the influence of multiple types of data in the diagnosis task and generate a data influence score; Based on the data impact score, the formula is used: Calculate the initial weights for each data type, where Represents data type The initial weight of Represents data type The impact score, Represents the adjustment coefficient.

3. The brain tumor multidimensional data analysis method according to claim 1, characterized in that: The steps of obtaining the data fusion expression set are specifically as follows: Identify the criticality of the data set in the target diagnostic task, combine the difference scale and type between the data, perform standardization and feature encoding, and obtain the key feature data set; The key feature data set is processed by linear combination and nonlinear transformation, using the formula: Get data fusion expression set ,in, No. The key feature input of class data, It is The weight coefficient of the class data, It is The weight matrix corresponding to the class data, No. The bias term corresponding to the class data, Indicates the number of data sources.

4. The brain tumor multidimensional data analysis method according to claim 1, characterized in that: The steps of fusing the differentiated data source features are specifically as follows: Based on the data fusion expression set, key features are identified from the imaging, clinical and genetic data, including imaging texture features, clinical biochemical indicators and gene mutation points, to obtain deep feature analysis results; Based on the deep feature analysis results, data structure optimization is performed using the formula: Normalize the difference scale features to the same interval to obtain the standardized feature results ,in, represents the original feature value extracted from the data source, and are the minimum and maximum values ​​in the feature set; Based on the standardized feature results, multiple types of features are integrated, and the weights of multiple features are automatically adjusted to obtain data source fusion processing results.

5. The brain tumor multidimensional data analysis method according to claim 1, characterized in that: The steps for obtaining the feature fusion set are specifically as follows: Based on the processed data source features, the correlation coefficient of each data feature is calculated, complementarity evaluation is performed, and a feature correlation matrix is ​​generated; Based on the feature correlation matrix, the formula is adopted: Perform weighted fusion of features to obtain the feature fusion set ,in, Indicates The current value of a feature, It is The weight of the feature, is the total number of features.

6. The brain tumor multidimensional data analysis method according to claim 1, characterized in that: The steps of identifying the potential feature combination are specifically as follows: Based on the feature fusion set, data processing is performed, including standardization and noise reduction, to obtain processed feature data; The processed feature data are clustered using a density clustering algorithm to calculate the local density between data points, identify and mark potential key areas, and use the formula: Calculate the The cluster contribution of data points ,in, Indicates the first data points, Representation and Adjacent data points, It is a data point Neighborhood of is the scale parameter, express and The Euclidean distance between Based on the cluster contribution, a feature set with the best contribution is screened to obtain a potential feature combination associated with brain tumors.

7. The brain tumor multidimensional data analysis method according to claim 1, characterized in that: The steps for obtaining the feature matching analysis information are specifically as follows: Based on the potential feature combination, calculating the similarity score with the known brain tumor judgment criteria, scoring each feature combination, and obtaining a preliminary scoring result; Based on the preliminary scoring results, the significance is calculated using the formula: Get feature matching analysis information, where: represents the Z score, is the average feature matching score, is the expected matching score, is the sample standard deviation, is the sample size.

8. The brain tumor multidimensional data analysis method according to claim 1, characterized in that: The steps for obtaining the data quality control results are specifically as follows: Based on the feature matching analysis information, an isolation forest algorithm is used to identify and mark abnormal data points in the data set to obtain an abnormal data identification result; Based on the abnormal data identification results, manual review and data point verification are performed, and the correction score is calculated using the formula: The data correction results after manual review are obtained, among which, is the corrected score, Representative The weight of the data point, Indicates The binary value of a data point during manual review, is the total number of data points; Based on the data correction results, data quality control is performed to re-evaluate the consistency and reliability of the brain tumor data set to obtain data quality control results.

9. Brain tumor multidimensional data analysis system, characterized in that: The method for multidimensional data analysis of brain tumors according to any one of claims 1 to 8 is implemented, wherein the system comprises: The data source adjustment module integrates imaging data, clinical data, and genetic data based on the brain tumor multidimensional data set, adjusts the weight of each data source to match its criticality in the diagnostic task, optimizes the processing differences between multiple data sources, and constructs a data fusion expression set; The feature optimization module extracts and processes the features of multiple data sources based on the data fusion expression set, performs feature optimization, explores the mutual complementarity between data, and merges the differentiated data source features to obtain a feature fusion set; The diagnostic feature clustering module applies density clustering analysis to key diagnostic features based on the feature fusion set, identifies potential feature combinations associated with brain tumors, and compares them with known brain tumor judgment criteria to obtain feature matching analysis information; The data anomaly assessment module uses the isolation forest algorithm to assess anomalies in the data set based on the feature matching analysis information, identifies data quality issues, and generates anomaly data markers; The data correction application module formulates a correction process based on the abnormal data marking, corrects the marked abnormal data points, and refines the integrity of the data set through manual review of the corrected data to obtain data quality control results.

Images