A tumor gene detection report auxiliary analysis system

Abstract

This invention relates to the field of data screening technology, specifically to a tumor gene testing report auxiliary analysis system. The system comprises: a data acquisition module, which acquires medical record feature vectors and habit vectors from medical record data; a feature analysis module, which analyzes the differences between medical record feature vectors to obtain disease similarity feature values ​​and establishes classifications based on the consistency of habit vectors, calculating feature parameters of historical medical record data in each category; a habit influence analysis module, which analyzes the differences between the dimensions of habit vectors and the differences between the feature parameters of medical record data, constructing an influence curve for each dimension; and a medical record screening and auxiliary analysis module, which combines the differences between the habit vectors of the test medical record data and historical medical record data, the influence curves, and the disease similarity feature values ​​to screen suitable medical records. These suitable medical records can then more effectively assist in the tumor gene testing analysis of the test medical record data.

Description

Technical Field

[0001] This invention relates to the field of data screening technology, specifically to a tumor gene detection report auxiliary analysis system. Background Technology

[0002] Liver, bile duct, and pancreatic tumors are among the most common, clinically relevant, and challenging malignant tumors of the digestive system. Molecular subtyping for targeted therapy has become an effective treatment option. However, in some patients, abnormal results may occur during tumor tissue biopsy, such as completely negative blood tests. This could be due to a low tumor burden or ongoing effective treatment, preventing the detection of the minimum threshold. Therefore, big data analysis of medical records from patients with similar medical histories can provide valuable information to assist physicians in their analysis.

[0003] Existing technologies for big data mining of historical patient medical records typically compare the similarity of the medical record to be tested with the detection data in historical medical records to screen out matching medical records with auxiliary or reference value. However, due to the large volume of medical record data and the fact that patients' long-term lifestyle habits can also affect the manifestation of tumor diseases and provide additional auxiliary information, relying solely on the detection data in medical record data for medical big data mining will result in low accuracy and poor efficiency of the finally screened matching medical records, leading to a certain degree of error in the results of auxiliary analysis. Summary of the Invention

[0004] To address the impact of long-term lifestyle habits on tumor disease manifestations and to provide additional auxiliary information, it is crucial to understand that relying solely on test data from medical records for medical data mining can lead to errors in the final results. This invention aims to provide an auxiliary analysis system for tumor gene testing reports, and the specific technical solution adopted is as follows:

[0005] This invention proposes a tumor gene detection report auxiliary analysis system, the system comprising:

[0006] The data acquisition module is used to acquire medical record data, which includes at least medical record feature vectors and habit vectors, and is divided into historical medical record data and medical record data to be tested.

[0007] The feature analysis module is used to analyze the differences between medical record feature vectors and determine the disease similarity feature value between any two medical record data. In all historical medical record data, historical medical record data are classified based on the consistency between habit vectors, and feature parameters of each historical medical record data are determined in each category based on the disease similarity feature value.

[0008] The habit influence analysis module is used to analyze the differences between the dimensions of the habit vector and the differences between the feature parameters in historical medical record data, thereby determining the influence curve of each dimension;

[0009] The medical record screening and auxiliary analysis module is used to combine the differences between the habitual vectors of the medical record data to be tested and the historical medical record data, the influence curves corresponding to each dimension, and the disease similarity feature values ​​to determine the fusion similarity between the medical record data; based on the fusion similarity, suitable medical records are screened from the historical medical record data for auxiliary analysis of the tumor gene detection report of the medical record data to be tested.

[0010] Furthermore, the method for obtaining the disease similarity feature values ​​includes:

[0011] In all medical record data, the Euclidean distance between the feature vectors of any two medical records is calculated, and negative correlation mapping and normalization are performed to obtain the disease similarity feature values ​​between the two medical records.

[0012] Furthermore, the method for obtaining the feature parameters includes:

[0013] In all historical medical record data, historical medical record data with the same habit vector are grouped into one category to obtain all categories;

[0014] Within each category, the normalized mean of the disease similarity features among all historical medical records is used as the feature parameter for each historical medical record within that category.

[0015] Furthermore, the method for obtaining the influence curve includes:

[0016] Choose one dimension from the habitual vector as the dimension to be tested, and use the remaining dimensions as the analysis dimensions;

[0017] Historical medical records with the same analysis dimension values ​​in the habitual vector are grouped into one class to obtain a cluster;

[0018] In each cluster, the absolute value of the difference between the values ​​of any two historical medical records under the dimension to be tested in the habitual vector is used as the dimension difference factor, and the absolute value of the difference between the feature parameters of any two historical medical records is used as the feature difference factor.

[0019] Under the dimension to be tested, curve fitting is performed on the dimensional difference factor and feature difference factor among the historical medical records in all clusters to obtain the influence curve corresponding to the dimension to be tested.

[0020] Furthermore, under the dimension to be measured, curve fitting is performed on the dimensional difference factor and feature difference factor among historical medical record data in all clusters to obtain the influence curve corresponding to the dimension to be measured, including:

[0021] Under the dimension to be tested, the dimensional difference factor is used as the horizontal axis and the feature difference factor is used as the vertical axis. Based on the least squares method, curve fitting is performed on all dimensional difference factors and feature difference factors among the medical record data to obtain the influence curve corresponding to the dimension to be tested.

[0022] Furthermore, the method for obtaining the fusion similarity includes:

[0023] By comparing the differences between the target medical record data and the habitual vectors of each historical medical record data, the target dimension can be determined.

[0024] Based on the numerical differences between the test medical record data and each historical medical record data in the target dimension, and combined with the influence curve corresponding to the target dimension, the influence value in the target dimension is obtained.

[0025] The value obtained by negatively mapping the impact values ​​of the medical record data to be tested and each historical medical record data under all target dimensions is used as the adjustment parameter;

[0026] The normalized value of the sum of the adjustment parameters and the disease similarity feature values ​​between the test medical record data and each historical medical record data is used as the fusion similarity between the test medical record data and each historical medical record data.

[0027] Furthermore, the method for obtaining the target dimension includes:

[0028] In the custom vectors of the medical record data to be tested and each historical medical record data, the values ​​under the same dimension are subtracted, and the dimension with a non-zero difference is taken as the target dimension.

[0029] Furthermore, the method for obtaining the influence value includes:

[0030] The absolute value of the numerical difference between the habitual vectors of the medical record data to be tested and each historical medical record data under each target dimension is used as the independent variable, and then mapped in the influence curve corresponding to each target dimension to obtain the influence value under each target dimension.

[0031] Furthermore, the step of screening suitable medical records from historical medical record data based on fusion similarity for auxiliary analysis of tumor gene testing reports of the medical record data to be tested includes:

[0032] All historical medical record data are sorted in descending order according to their fusion similarity to obtain a sorted sequence.

[0033] In the sorting sequence, a preset number of historical medical records are selected as the matching medical records for the medical records to be tested before filtering.

[0034] The system acquires the medical record data to be tested and the tumor gene testing report of the matching medical record data, compares the textual differences, and marks the different positions for auxiliary analysis.

[0035] Furthermore, the method for obtaining the medical record feature vector includes:

[0036] Obtain the medical record text from the medical record data, and perform bag-of-words model transformation on the text data in the medical record text to obtain text vectors;

[0037] Numerical data in medical record text is concatenated with text vectors to form the medical record feature vector.

[0038] The present invention has the following beneficial effects:

[0039] The data acquisition module obtains medical record feature vectors (used to characterize pathological indicators) and habit vectors (reflecting patients' behavioral habits) from medical record data. This dual data source cross-validation mechanism effectively avoids single-dimensional data bias, making subsequent analysis results more accurate. In the feature analysis module, a difference-driven method for calculating disease similarity feature values ​​is employed. By analyzing the differences between medical record feature vectors, potential association patterns between different medical records can be identified in a large amount of data. Furthermore, a classification strategy is established based on the consistency of habit vectors, allowing the calculation of feature parameters for historical medical record data in each category to automatically adapt to the distribution patterns of disease similarity feature values ​​for different patient groups. Given that differences in lifestyle habits may lead to changes in disease manifestations, in cases with large volumes of medical record data, the habit influence analysis module analyzes the differences between the dimensions of habit vectors and the differences between the feature parameters of medical record data. Influence curves corresponding to each dimension are constructed to reflect the impact of each dimension of the habit vector on patients' disease manifestations. Finally, in the medical record screening and auxiliary analysis module, the differences between the habitual vectors of the test medical record data and the historical medical record data are combined with the influence curve to adjust the disease similarity feature values ​​between the test medical record data and the historical medical record data, thereby determining a comprehensive index—fusion similarity. Fusion similarity can more comprehensively reflect the similarity between two medical record data. In turn, the matching medical records screened from a large number of medical record data by this index can more effectively and accurately assist doctors in analyzing the tumor gene testing report of the test medical record data. Attached Figure Description

[0040] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This is a system block diagram of a tumor gene detection report assistance system provided in one embodiment of the present invention;

[0042] Figure 2 This is a flowchart of a method for obtaining an influence curve according to an embodiment of the present invention;

[0043] Figure 3 This is a flowchart of a method for obtaining fusion similarity according to an embodiment of the present invention;

[0044] Figure 4 This is a schematic diagram of the system structure of a tumor gene detection report auxiliary analysis system provided in one embodiment of the present invention. Detailed Implementation

[0045] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a tumor gene detection report auxiliary analysis system proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0046] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0047] The following description, in conjunction with the accompanying drawings, details the specific scheme of the tumor gene detection report auxiliary analysis system provided by the present invention.

[0048] Please see Figure 1 The diagram illustrates a system block diagram of a tumor gene detection report auxiliary analysis system provided in an embodiment of the present invention. The system includes: a data acquisition module 101, a feature analysis module 102, a habit influence analysis module 103, and a medical record screening and auxiliary analysis module 104.

[0049] The data acquisition module 101 is used to acquire medical record data, which includes at least medical record feature vectors and habit vectors, and is divided into historical medical record data and medical record data to be tested.

[0050] In the field of digestive system tumor diagnosis and treatment, liver cancer, bile duct cancer, and pancreatic cancer have received considerable attention due to their high incidence, clinical complexity, and treatment difficulty. Genetic testing technology analyzes relevant genetic material in tissues or blood to detect related mutations, such as common ALK, EGFR, and KRAS mutations. This allows for targeted drug administration to address the specific mutations, improving treatment outcomes. However, because it is difficult to obtain tumor tissue for biopsy in some patients, blood from the surrounding tissue is often tested. This test result may be abnormal; for example, a completely negative blood test might indicate a low tumor burden or ongoing effective treatment, resulting in extremely low levels of the detection factor (usually ctDNA) in the released blood, failing to reach the detection limit. Therefore, medical big data mining is often used to analyze historical patient records to obtain data similar to those of the patient being tested, providing doctors with supplementary diagnostic information.

[0051] Generally, cancer is considered a chronic disease, and its development requires a long process. Existing literature studies have pointed out that long-term dietary culture and lifestyle habits can affect the detection parameters and risk of disease. Therefore, in this embodiment of the invention, the patient's habitual behavior and other characteristics are also considered as factors in the similarity of medical record data. Thus, in the data acquisition module, medical record data is acquired. The medical record data is divided into historical medical record data and medical record data to be tested. The acquisition of medical record data can be based on the database of the hospital management system, and the medical record data should include at least the medical record feature vector and the habit vector.

[0052] Preferably, in one embodiment of the present invention, the method for obtaining the medical record feature vector and the habit vector includes:

[0053] To obtain the medical record text from the patient's medical record data, since the medical record text usually includes the patient's basic information and relevant information about disease testing, it can be roughly divided into numerical data (blood pressure, ECG, etc.) and text data (endoscopic examination, CT scan, etc.). For text data, the bag-of-words model can be used to transform it to obtain text vectors; for numerical data, no processing is required. Then, the numerical data in the medical record text is concatenated with the text vector to obtain the medical record feature vector in the medical record data. For example, if a patient has a liver tumor, and their initial blood test report shows a white blood cell count of 7.92 and a lymphocyte count of 2.53, then the feature vector of the blood test is (7.92, 2.53). The description of the chest X-ray is "mild elevation of the right diaphragm." Based on the expert knowledge base, a keyword database is established for the X-rays of liver tumor patients, and each word is uniquely encoded. For example, "right diaphragm" is (1,0,0), and "elevation" is (0,0,1). Then the description is expressed as (1,0,1). Therefore, the medical record feature vector is (7.92, 2.53, 1,0,1).

[0054] To obtain habit vectors, quantification can be based on questionnaires. Each dimension / option represents a behavioral habit. For example, in a questionnaire, the options are based on frequency. If the options are: daily, 5-6 times per week, 3-4 times per week, 1-2 times per week, 1-3 times per month, less than once per month, these can be converted to 0, 0.2, 0.4, 0.6, 0.8, and 1, respectively. If the answer to habit 1 is daily and the answer to habit 2 is 1-3 times per month, then these can be converted to 0 and 0.8, respectively. The first dimension represents habit 1 and the second dimension represents habit 2, so the habit vector is (0, 0.8).

[0055] It should be noted that the dimensions of the habit vectors of all medical record data should be consistent, and zero-padding can be used when necessary; similarly, the dimensions of the feature vectors of all medical record data should also be consistent, and zero-padding can also be used; the bag-of-words model is a well-known technique, and the specific process will not be elaborated here.

[0056] In this embodiment of the invention, the collection and acquisition of medical record data are authorized by the relevant users, and the process does not violate relevant laws and regulations, nor does it violate public order and good morals.

[0057] The feature analysis module 102 is used to analyze the differences between medical record feature vectors, determine the disease similarity feature value between any two medical record data; in all historical medical record data, the historical medical record data is classified based on the consistency between habit vectors, and the feature parameters of each historical medical record data are determined based on the disease similarity feature value in each category.

[0058] According to existing biological consensus, genes are a type of hereditary material that typically determines the external expression of cellular regulation. In patients with hepatobiliary and pancreatic tumors, if the mutated genes are the same, the resulting disease characteristics may have a high degree of similarity. Therefore, patients with similar disease characteristics may also have similar mutation characteristics in their gene testing reports, thus similar disease characteristics may have certain informational reference value. Therefore, the differences between medical record feature vectors can be analyzed to determine the disease similarity feature values ​​between medical record data, reflecting the similarity between medical record data. Furthermore, given that tumors are generally considered chronic diseases with a long development process, long-term dietary culture and lifestyle habits can affect disease examination parameters and disease risk. Therefore, this embodiment of the invention considers the impact of patients' lifestyle habits on disease manifestations. Thus, in all historical medical record data, historical medical record data were classified based on the consistency between habit vectors, and the disease similarity feature values ​​between historical medical record data were fused in each category to determine the feature parameters of each historical medical record data, thus initially fusing habit features with medical record features.

[0059] Preferably, in one embodiment of the present invention, the method for obtaining disease similarity feature values ​​includes:

[0060] In all medical record data, the Euclidean distance between the feature vectors of any two medical records is calculated. A larger Euclidean distance indicates a lower similarity between the two feature vectors. Therefore, the Euclidean distance is negatively correlated and normalized to correct the logical relationship, thus obtaining the disease similarity feature value between the two medical records. A larger disease similarity feature value indicates a higher similarity between the two medical records, and thus a higher reference value for mutual auxiliary analysis. This negative correlation mapping and normalization can be performed using the formula... ,in, Let x represent an exponential function with the natural constant e as the base, and let x represent the independent variable.

[0061] At this point, we can obtain the disease similarity feature values ​​between any two medical records (including historical medical records and medical records to be tested).

[0062] Cancer patients' disease manifestations are usually the result of long-term accumulation, going through early, middle and late stages of development. Poor lifestyle and dietary habits can increase the risk of digestive system tumors. Therefore, historical medical records can be classified based on the consistency between habit vectors. When the number of historical patients is large, their lifestyle distribution can be considered to be approximately Gaussian. Thus, in each category, feature parameters of each historical medical record can be determined based on the disease similarity feature values ​​between historical medical records, which can reflect the overall characteristics of historical medical records in each category.

[0063] Preferably, in one embodiment of the present invention, the method for obtaining the feature parameters includes:

[0064] In all historical medical records, those with the same habit vector are grouped into one category to obtain all categories. At this point, the historical medical records in each category have consistent habitual behavior characteristics, and there is also a disease similarity feature value between every two historical medical records.

[0065] Then, within each category, the mean of the disease similarity feature values ​​among all historical medical records is normalized and used as the feature parameter for each historical medical record within that category. The larger the feature parameter, the higher the overall symptom similarity of the historical medical records under that habitual behavior feature category. Normalization is a technique well-known to those skilled in the art, and the choice of normalization function can be linear normalization or standard normalization, etc. The specific normalization method is not limited here.

[0066] The habit influence analysis module 103 is used to analyze the differences between the dimensions of the habit vector and the differences between the feature parameters in historical medical record data, thereby determining the influence curve of each dimension.

[0067] Given that different habits can have varying impacts on disease presentation, the concept of fixed variables can be used to perform pattern learning on the similarity of single-dimensional changes in the habit vector, thereby obtaining the impact of single-dimensional habit changes on disease similarity. In this embodiment of the invention, the differences between dimensions of the habit vector in historical medical record data are analyzed, and combined with the differences between feature parameters of historical medical record data, the influence curve of each dimension in the habit vector is determined.

[0068] Preferably, in one embodiment of the present invention, the method for obtaining the influence curve includes:

[0069] Please see Figure 2 The diagram illustrates a method flowchart for obtaining an influence curve according to an embodiment of the present invention. The method includes the following steps:

[0070] Step S301: Select one dimension from the habitual vector as the dimension to be tested, and the remaining dimensions as the analysis dimensions. Then, perform cluster analysis on all historical medical records based on the values ​​under the analysis dimensions to obtain clusters.

[0071] In the habit vector, one dimension is randomly selected as the dimension to be tested, and the remaining dimensions are used as analysis dimensions. Historical medical records with the same values ​​in the analysis dimensions of the habit vector are grouped into one category, resulting in a cluster. At this point, all historical medical records in the cluster have the same value under the analysis dimension. Therefore, in subsequent processes, we can mainly focus on analyzing the impact of the dimension to be tested on the symptoms. For example, if the habit vector has four dimensions, a1-a4, and a1 is used as the dimension to be tested, then in all historical medical records, the values ​​of the habit vector under a2-a4 are compared, and historical medical records with the same a2, a3, and a4 are grouped into one category.

[0072] Step S302: In each cluster, analyze the numerical differences of historical medical record data under the measured dimension of the habitual vector to obtain the dimensional difference factor between any two historical medical record data.

[0073] In each cluster, the absolute value of the difference between the values ​​of any two historical medical records under the tested dimension of the habit vector is used as the dimension difference factor. The larger the dimension difference factor, the greater the difference between the habit vectors of the two historical medical records under the tested dimension, and the greater the influence of the tested dimension may be.

[0074] Thus, in each cluster, every two historical medical records have a dimensionality difference factor under the dimension to be tested.

[0075] Step S303: In each cluster, analyze the differences between the feature parameters of historical medical records to obtain the feature difference factor between any two historical medical records.

[0076] In each cluster, the absolute value of the difference between the feature parameters of any two historical medical records is used as the feature difference factor. The larger the feature difference factor, the greater the difference between the feature parameters of the two historical medical records, and the lower the degree of similarity of the symptoms.

[0077] Thus, in each cluster, there is a characteristic difference factor between every two historical medical records.

[0078] Step S304: Under the dimension to be tested, perform curve fitting on the dimensional difference factor and feature difference factor among the historical medical record data in all clusters to obtain the influence curve corresponding to the dimension to be tested.

[0079] After the aforementioned steps are completed, at this point, under the dimension to be tested, any two historical medical records in each cluster will have a dimension difference factor and a feature difference factor, which can form a combination in the form of a two-dimensional coordinate.

[0080] Then, under the dimension to be tested, the dimensional difference factor between historical medical records is used as the x-axis and the feature difference factor is used as the y-axis. Based on the least squares method, curve fitting is performed on all the dimensional difference factors and feature difference factors between historical medical records to obtain the influence curve corresponding to the dimension to be tested. This influence curve can be regarded as the influence curve of the change of the dimension to be tested in the habit vector learned from historical medical records on the similarity of disease manifestations.

[0081] Thus, the influence curve corresponding to each dimension in the habit vector can be obtained through the above method.

[0082] The medical record screening and auxiliary analysis module 104 is used to combine the differences between the habitual vectors of the medical record data to be tested and the historical medical record data, the influence curves corresponding to each dimension, and the disease similarity feature values ​​to determine the fusion similarity between the medical record data; based on the fusion similarity, suitable medical records are screened from the historical medical record data for auxiliary analysis of the tumor gene detection report of the medical record data to be tested.

[0083] Based on the aforementioned steps, the influence curve of each dimension in the habitual vector on the symptom representation can be obtained. The horizontal axis of the influence curve represents the dimensional difference factor, and the vertical axis represents the feature difference factor. Therefore, in this module, the differences between the habitual vectors of the test medical record data and historical medical record data can be compared. This allows the similarity between the test medical record data and historical medical record data to be determined in the influence curve. Furthermore, this similarity is combined with the disease similarity feature values ​​between them to obtain the fusion similarity. The fusion similarity can more comprehensively represent the similarity of symptom presentation between the test medical record data and each historical medical record. Finally, based on this indicator, suitable medical records can be more accurately selected from numerous historical medical record data for auxiliary analysis of tumor gene testing reports of the test medical record data.

[0084] Preferably, in one embodiment of the present invention, the method for obtaining fusion similarity includes:

[0085] Please see Figure 3 The diagram illustrates a flowchart of a method for obtaining fused similarity in one embodiment of the present invention. The method includes the following steps:

[0086] Step S401: Compare the differences between the target medical record data and the habitual vectors of each historical medical record data to determine the target dimension.

[0087] In the custom vectors of the medical record data to be tested and each historical medical record data, the values ​​under the same dimension are compared, and the dimension with differences, that is, the dimension with a difference of non-zero, is taken as the target dimension.

[0088] Step S402: Based on the numerical differences between the medical record data to be tested and each historical medical record data under the target dimension, and combined with the influence curve corresponding to the target dimension, obtain the influence value under the target dimension.

[0089] Under each target dimension, the absolute value of the difference between the habitual vector values ​​of the test medical record data and each historical medical record data is used as the independent variable. The independent variable here is the dimension difference factor between the test medical record data and each historical medical record data under the target dimension. This is then mapped onto the influence curve corresponding to each target dimension to obtain the influence value under each target dimension. The influence value here is the feature difference factor between the test medical record data and each historical medical record data under the target dimension. The larger the value, the greater the influence of the target dimension, which means that the symptom similarity between the test medical record data and the historical medical record data should be reduced.

[0090] Step S403: Combine the influence values ​​of the test medical record data and each historical medical record data under all target dimensions to obtain the fusion similarity between the test medical record data and each historical medical record data.

[0091] Based on the foregoing analysis, it is known that when the influence value is larger, the similarity of symptoms between the test medical record data and the historical medical record data should be reduced. Therefore, the influence values ​​of the test medical record data and each historical medical record data under all target dimensions are negatively correlated to correct the logical relationship and obtain the adjustment parameter. The larger the adjustment parameter is, the higher the similarity.

[0092] Finally, the sum of the adjustment parameters between the test medical record data and each historical medical record data, along with the disease similarity feature value, is normalized and used as the fusion similarity between the test medical record data and each historical medical record data. It is easy to understand that the higher the fusion similarity, the higher the similarity between the test medical record data and the historical medical record data after considering the influence of habitual behaviors. Normalization is a technique well-known to those skilled in the art; the normalization function can be linear normalization or standard normalization, etc., and the specific normalization method is not limited here.

[0093] At this point, the fusion similarity between the medical record data to be tested and each historical medical record data can be obtained. Based on this indicator, suitable medical records can be selected for subsequent operations.

[0094] Preferably, in one embodiment of the present invention, based on fusion similarity, suitable medical records are screened from historical medical record data for auxiliary analysis of tumor gene detection reports of the medical record data to be tested, including:

[0095] All historical medical record data are sorted in descending order according to fusion similarity to obtain a sorted sequence; in the sorted sequence, a preset number of historical medical record data are selected as the matching medical records for the medical record data to be tested.

[0096] The system acquires the medical record data to be tested and the tumor gene testing report of the matching medical record data, compares the textual differences, and marks the different positions to help doctors conduct auxiliary analysis.

[0097] It should be noted that in this embodiment of the present invention, the preset quantity is set to 2, and the specific quantity can be adjusted according to the implementation scenario, and is not limited here.

[0098] In summary, this embodiment of the invention acquires medical record feature vectors (used to characterize pathological indicators) and habit vectors (reflecting patient behavioral habits) from medical record data through a data acquisition module. The cross-validation mechanism using dual data sources effectively avoids single-dimensional data bias. In the feature analysis module, a difference-driven method for calculating disease similarity feature values ​​is employed. By analyzing the differences between medical record feature vectors, potential association patterns between different medical records can be identified. Furthermore, a classification strategy is established based on the consistency of habit vectors, allowing the calculation of feature parameters for historical medical record data in each category to automatically adapt to the distribution patterns of disease similarity feature values ​​for different patient groups. Given that differences in lifestyle habits may lead to changes in disease manifestations, the habit influence analysis module analyzes the differences between the dimensions of the habit vectors and the differences between the feature parameters of the medical record data, constructing an influence curve for each dimension to reflect the impact of each dimension of the habit vector on the patient's disease manifestations. Finally, in the medical record screening and auxiliary analysis module, the differences between the habitual vectors of the test medical record data and the historical medical record data are combined with the influence curve to adjust the disease similarity feature values ​​between the test medical record data and the historical medical record data, thereby determining a comprehensive index—fusion similarity. Fusion similarity can more comprehensively reflect the similarity between two medical record data. In this way, the matching medical records screened by this index can more effectively and accurately assist doctors in analyzing the tumor gene testing report of the test medical record data.

[0099] It should be noted that the system provided in the above embodiments is only an example of the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above.

[0100] Please see Figure 4 This diagram illustrates a system architecture of a tumor gene detection report auxiliary analysis system according to an embodiment of the present invention. The system includes a processor 400, a memory 401, a bus 402, and a communication interface 403. The processor 400, communication interface 403, and memory 401 are connected via the bus 402. The memory 401 may include a high-speed random access memory, and the bus 402 may be an ISA bus, PCI bus, or EISA bus, etc. The processor 400 may be an integrated circuit chip with signal processing capabilities. The memory 401 stores at least one instruction, at least one program, code set, or instruction set. When the processor loads and executes the at least one instruction, at least one program, code set, or instruction set, it implements the steps of each module in the tumor gene detection report auxiliary analysis system.

[0101] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0102] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0103] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A tumor gene detection report auxiliary analysis system, characterized in that, The system includes: The data acquisition module is used to acquire medical record data, which includes at least medical record feature vectors and habit vectors, and is divided into historical medical record data and medical record data to be tested. The feature analysis module is used to analyze the differences between medical record feature vectors and determine the disease similarity feature value between any two medical record data. In all historical medical record data, historical medical record data are classified based on the consistency between habit vectors, and feature parameters of each historical medical record data are determined in each category based on the disease similarity feature value. The habit influence analysis module is used to analyze the differences between dimensions and feature parameters of historical medical record data in the habit vector, thereby determining the influence curve for each dimension. The method for obtaining the influence curve includes: selecting one dimension in the habit vector as the dimension to be tested, and the remaining dimensions as the analysis dimensions; clustering historical medical record data with the same analysis dimension values ​​in the habit vector into a single category to obtain clusters; within each cluster, using the absolute value of the difference between any two historical medical record data values ​​in the dimension to be tested as the dimension difference factor, and using the absolute value of the difference between the feature parameters of any two historical medical record data as the feature difference factor; and performing curve fitting on the dimension difference factor and feature difference factor among all clusters of historical medical record data in the dimension to be tested to obtain the influence curve corresponding to the dimension to be tested. The medical record screening and auxiliary analysis module is used to combine the differences between the habitual vectors of the medical record data to be tested and the historical medical record data, the influence curves corresponding to each dimension, and the disease similarity feature values ​​to determine the fusion similarity between the medical record data. Based on the fusion similarity, suitable medical records are screened from the historical medical record data for auxiliary analysis of the tumor gene detection report of the medical record data to be tested. The method for obtaining the fusion similarity includes: comparing the differences between the habitual vectors of the medical record data to be tested and each historical medical record data to determine the target dimension; based on the numerical differences between the medical record data to be tested and each historical medical record data under the target dimension, and combined with the influence curves corresponding to the target dimension, obtaining the influence value under the target dimension; negatively mapping the influence values ​​of the medical record data to be tested and each historical medical record data under all target dimensions, and using the result as an adjustment parameter; normalizing the sum of the adjustment parameter and the disease similarity feature values ​​between the medical record data to be tested and each historical medical record data, and using the result as the fusion similarity between the medical record data to be tested and each historical medical record data.

2. The tumor gene detection report auxiliary analysis system according to claim 1, characterized in that, The method for obtaining the disease similarity feature values ​​includes: In all medical record data, the Euclidean distance between the feature vectors of any two medical records is calculated, and negative correlation mapping and normalization are performed to obtain the disease similarity feature values ​​between the two medical records.

3. The tumor gene detection report auxiliary analysis system according to claim 1, characterized in that, The method for obtaining the feature parameters includes: In all historical medical record data, historical medical record data with the same habit vector are grouped into one category to obtain all categories; Within each category, the normalized mean of the disease similarity features among all historical medical records is used as the feature parameter for each historical medical record within that category.

4. The tumor gene detection report auxiliary analysis system according to claim 1, characterized in that, Under the dimension to be measured, curve fitting is performed on the dimensional difference factor and feature difference factor among historical medical record data in all clusters to obtain the influence curve corresponding to the dimension to be measured, including: Under the dimension to be tested, the dimensional difference factor is used as the horizontal axis and the feature difference factor is used as the vertical axis. Based on the least squares method, curve fitting is performed on all dimensional difference factors and feature difference factors among the medical record data to obtain the influence curve corresponding to the dimension to be tested.

5. The tumor gene detection report auxiliary analysis system according to claim 1, characterized in that, The methods for obtaining the target dimension include: In the custom vectors of the medical record data to be tested and each historical medical record data, the values ​​under the same dimension are subtracted, and the dimension with a non-zero difference is taken as the target dimension.

6. The tumor gene detection report auxiliary analysis system according to claim 1, characterized in that, The method for obtaining the influence value includes: The absolute value of the numerical difference between the habitual vectors of the medical record data to be tested and each historical medical record data under each target dimension is used as the independent variable, and then mapped in the influence curve corresponding to each target dimension to obtain the influence value under each target dimension.

7. The tumor gene detection report auxiliary analysis system according to claim 1, characterized in that, The process of selecting suitable medical records from historical medical record data based on fusion similarity for auxiliary analysis of tumor gene testing reports of the medical record data to be tested includes: All historical medical record data are sorted in descending order according to their fusion similarity to obtain a sorted sequence. In the sorting sequence, a preset number of historical medical records are selected as the matching medical records for the medical records to be tested before filtering. The system acquires the medical record data to be tested and the tumor gene testing report of the matching medical record data, compares the textual differences, and marks the different positions for auxiliary analysis.

8. The tumor gene detection report auxiliary analysis system according to claim 1, characterized in that, The method for obtaining the medical record feature vector includes: Obtain the medical record text from the medical record data, and perform bag-of-words model transformation on the text data in the medical record text to obtain text vectors; Numerical data in medical record text is concatenated with text vectors to form the medical record feature vector.

Images