Tumor disease risk prediction and drug screening method and system based on artificial intelligence

Abstract

The present invention discloses a tumor disease risk prediction and drug screening method and system based on artificial intelligence, which relates to the field of tumor disease risk prediction technology, in order to solve the problem that in the prior art, targeted drug recommendations cannot be made according to the patient's tumor condition. The present invention realizes accurate stratification of patient tumor risks by confirming the degree of tumor occurrence risk, progression risk, and treatment response risk. Through accurate risk stratification and personalized drug recommendations, unnecessary waste of medical resources can be reduced, and limited medical resources can be more reasonably allocated and used. Personalized drug recommendations help patients obtain more suitable treatment plans, thereby improving the efficiency and effectiveness of treatment. Convolutional neural networks and recurrent neural networks are used to perform deep processing on image data and text data, respectively, and key features of their respective fields are effectively extracted.

Description

Technical Field

[0001] The present invention relates to the technical field of tumor disease risk prediction, and specifically to a tumor disease risk prediction and drug screening method and system based on artificial intelligence. Background Art

[0002] Tumor disease risk prediction refers to evaluating the possibility of future tumor development by analyzing individual internal factors and external environmental factors.

[0003] Patent application with publication number CN107201401A discloses a multi-factor model for predicting the risk of breast cancer and its establishment method, which is mainly constructed by combining multiple high-risk factors with germline gene mutations. It mainly consists of two parts: conventional high-risk factors and gene mutation factors. Data statistics are performed on statistically significant breast cancer risk factors screened out by univariate analysis, and then analyzed by Logistic regression. The multi-factor model for predicting the risk of breast cancer can perform early screening and early diagnosis for this group of people, and take active intervention measures. It is particularly suitable for predicting the high-risk group of familial breast cancer, and is of great significance to reducing the incidence of breast cancer in my country and improving the postoperative survival rate of this group of people. Although the above scheme solves the problem of risk prediction, the following problems still exist in actual operation:

[0004] 1. The patient’s medical data is not collected more accurately, resulting in inaccurate tumor data.

[0005] 2. The acquired tumor data were not further processed, resulting in poor training results when the tumor data was used for model training in the later stage.

[0006] 3. The acquired tumor data are not used to effectively predict tumor risk, resulting in the inability to make targeted drug recommendations based on the risk and type of the tumor. Summary of the invention

[0007] The purpose of the present invention is to provide an artificial intelligence-based tumor disease risk prediction and drug screening method and system, which realizes accurate stratification of patient tumor risk by confirming the degree of tumor occurrence risk, progression risk, and treatment response risk. Through accurate risk stratification and personalized drug recommendations, unnecessary waste of medical resources can be reduced, and limited medical resources can be more reasonably allocated and used. Personalized drug recommendations help patients obtain more suitable treatment plans, thereby improving the efficiency and effectiveness of treatment. Convolutional neural networks and recurrent neural networks are used to perform deep processing of image data and text data respectively, effectively extracting the key features of their respective fields, which can solve the problems in the prior art.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] The tumor disease risk prediction and drug screening method based on artificial intelligence includes the following steps:

[0010] S1: Acquisition of patient medical data: Acquiring the patient's medical data from the database and integrating the medical data with the patient's basic data into a data set. After the data set integration is completed, the patient's integrated data is obtained;

[0011] S2: Collection of relevant tumor information: Perform tumor data analysis on the integrated patient data, confirm the analyzed tumor-related data, and mark the confirmed tumor-related data as target tumor-related data;

[0012] S3: Tumor data preprocessing: preprocess the target tumor-related data, and obtain standard tumor-related data after the data preprocessing is completed;

[0013] S4: Tumor data risk prediction: Import standard tumor-related data into the risk prediction model for risk prediction, and mark the tumor information according to the risk threshold, and annotate the marked tumor information as predicted tumor data;

[0014] S5: Predicted tumor drug markers: Import the predicted tumor data into the drug prediction model for tumor drug analysis, and determine the drug recommendations for patients based on the analysis results.

[0015] Preferably, obtaining the patient's medical data from the database in S1 and integrating the medical data with the patient's basic data into a data set includes:

[0016] Obtaining the patient's medical data from a database, wherein the patient's medical data includes clinical data, genomic data, imaging data, and test data;

[0017] Basic patient data includes demographic data, physical characteristics data, and medical history data;

[0018] Integrate the patient's medical data and basic patient data into data sets;

[0019] Among them, the format of the patient's medical data and basic patient data is first unified;

[0020] Create a data dictionary in the database, and match the medical data and basic data of patients in a unified format in the data dictionary;

[0021] After data matching is completed, the medical data and basic data of the same patient are integrated;

[0022] After the integration is completed, the patient's integrated data is obtained.

[0023] Preferably, performing tumor data analysis on the integrated patient data in S2 and confirming the analyzed tumor-related data include:

[0024] Retrieve clinical data, genomic data, imaging data, and test data from integrated patient data;

[0025] Obtain the first tumor information from clinical data, genomic data, and imaging data respectively;

[0026] The first tumor information includes tumor data on the type, stage, location, size, growth rate, and whether the tumor has metastasized obtained from the clinical data;

[0027] Genomic data obtains genetic information of tumors;

[0028] Imaging data obtains structural information about the spatial location, shape, size, and boundaries of the tumor in the body;

[0029] The detection data obtains biomarker information of the tumor.

[0030] Preferably, performing tumor data analysis on the integrated patient data in S2 and confirming the analyzed tumor-related data further includes:

[0031] Confirm the first tumor information with omics data;

[0032] Among them, omics data include genomic data, transcriptomic data, and proteomic data;

[0033] The genomic data uses a high-throughput sequencing method to sequence the tumor tissue from the first tumor information, and the genomic DNA sequence information of the tumor cells is obtained after sequencing the tumor tissue;

[0034] The transcriptome data uses RNA sequencing to sequence RNA in the tumor tissue from the first tumor information, and quantitative analysis and differential expression analysis are performed during sequencing to finally obtain gene expression information;

[0035] The proteomic data uses mass spectrometry to separate and identify proteins in tumor tissues from the first tumor information, obtain information on the type, quantity and function of proteins, and at the same time, identify and authenticate tumor markers by comparing protein spectra in the database;

[0036] Based on the omics data confirmed in the first tumor information, the omics data is used as target tumor-related data of the patient.

[0037] Preferably, data preprocessing is performed on the target tumor related data in S3, including:

[0038] Classify target tumor-related data into text data and image data;

[0039] After the classification is completed, the text data and image data are preprocessed;

[0040] The data preprocessing of the text data is to remove noise and repeated data in the text data, then segment the text in the text data into words, and perform part-of-speech recognition on the segmented words, map the recognized words into a continuous vector space for semantic capture, and perform feature vocabulary training based on the captured semantics, wherein the feature vocabulary training is performed using a training model, and the trained feature vocabulary is standardized, and finally the preprocessed text data is obtained;

[0041] The data preprocessing of the image data is to convert the image data into grayscale images, perform geometric correction on the image after the grayscale image conversion is completed, adjust the parameters of the brightness, contrast, clarity and sharpness of the image after the geometric correction, use smoothing and median filtering methods to remove noise in the image after the parameter adjustment is completed, and then use the gradient operator method to detect the edge information in the image, extract the feature image of the image according to the edge information, normalize the feature image, and finally obtain the preprocessed image data;

[0042] The preprocessed image data and text data are uniformly labeled as standard tumor-related data.

[0043] Preferably, the data preprocessing of the image data includes:

[0044] Performing grayscale image conversion on the image data to generate a grayscale image corresponding to the image data;

[0045] Performing geometric correction on the grayscale image to obtain a grayscale image after geometric correction, and using the grayscale image after geometric correction as target image data;

[0046] Extracting the grayscale value of each pixel of the target image data as a first grayscale value;

[0047] Extracting the grayscale value of each pixel of the image data before geometric correction corresponding to the target image data as the second grayscale value;

[0048] Compare the first grayscale value of each pixel of the target image data with the second grayscale value of the pixel at the same position of the image data before geometric correction to obtain a grayscale difference;

[0049] The grayscale deviation coefficient of the target image data after geometric correction is obtained by using the grayscale difference; wherein the grayscale deviation coefficient is obtained by the following formula:

[0050] ;

[0051] Where S represents the grayscale deviation coefficient; n represents the number of pixels in the target image data; H 01i represents the first gray value of the i-th pixel in the target image data; H 02i represents the second grayscale value of the image data before geometric correction corresponding to the i-th pixel; H z01 Indicates the central gray value corresponding to the target image data; H z02 represents the central grayscale value corresponding to the image data before geometric correction; s represents the adjustment coefficient, and the adjustment coefficient is obtained by the following formula:

[0052] ;

[0053] Where s represents the adjustment coefficient; a represents the number of pixels whose grayscale value of the target image data is reduced relative to the image data before geometric correction; b represents the number of pixels whose grayscale value of the target image data is increased relative to the image data before geometric correction; H 01ai represents the first grayscale value corresponding to the pixel whose grayscale value decreases at the ith pixel; H 02ai H represents the second grayscale value corresponding to the pixel whose grayscale value is reduced by the i-th pixel; 01bi H represents the first grayscale value corresponding to the pixel point whose grayscale value is increased by the i-th pixel; 02bi represents the second grayscale value corresponding to the pixel point whose grayscale value is increased by the i-th pixel;

[0054] Comparing the grayscale deviation coefficient with a preset grayscale deviation coefficient threshold;

[0055] When the grayscale deviation coefficient exceeds a preset grayscale deviation coefficient threshold, the grayscale value of the target image data is adjusted.

[0056] Preferably, when the grayscale deviation coefficient exceeds a preset grayscale deviation coefficient threshold, grayscale value adjustment is performed on the target image data, including:

[0057] When the grayscale deviation coefficient exceeds a preset grayscale deviation coefficient threshold, the first grayscale value of each pixel of the target image data and the second grayscale value of the pixel of the image data corresponding to each pixel of the target image data are retrieved;

[0058] Using a grayscale difference between a first grayscale value of each pixel of the target image data and a second grayscale value of a pixel of the image data corresponding to each pixel of the target image data as a first grayscale difference;

[0059] Retrieving the first grayscale value of the pixel points adjacent to each pixel point in the target image data and the second grayscale value of the pixel points adjacent to each pixel point in the image data corresponding to each pixel point;

[0060] Obtaining a second grayscale difference value by using a first grayscale value of a pixel point adjacent to each pixel point in the target image data and a second grayscale value corresponding to a pixel point adjacent to each pixel point in the image data;

[0061] The target grayscale value of each pixel in the target image data is obtained by using the first grayscale difference and the second grayscale difference; wherein the target grayscale value is obtained by the following formula:

[0062] ;

[0063] Among them, H s Indicates the target gray value; H 01i represents the first gray value of the i-th pixel in the target image data; H 02i represents the second grayscale value of the image data before geometric correction corresponding to the i-th pixel; k represents the number of pixels adjacent to each pixel in the target image data; H 01j H represents the first gray value corresponding to the jth adjacent pixel; 02j represents the second grayscale value corresponding to the jth adjacent pixel; S represents the grayscale deviation coefficient; S y Indicates the preset grayscale deviation coefficient threshold; H 01 Represents the first grayscale value of the pixel in the target image data;

[0064] Performing grayscale adjustment on each pixel in the target image data according to the target grayscale value to obtain the target image data after grayscale adjustment;

[0065] The brightness and contrast are adjusted for the target image data after the grayscale adjustment to obtain the target image data after the brightness and contrast adjustment.

[0066] Preferably, adjusting the brightness and contrast of the target image data after the grayscale adjustment to obtain the target image data after the brightness and contrast adjustment includes:

[0067] Extracting the grayscale adjustment amplitude of each pixel of the target image data after grayscale adjustment;

[0068] The adjusted target brightness value is obtained by using the grayscale adjustment amplitude of each pixel point; wherein the target brightness value is obtained by the following formula:

[0069] ;

[0070] Among them, G s Indicates the target brightness value; G 0 represents the brightness value of the target image data before brightness adjustment; n represents the number of pixels of the target image data; H fiIndicates the grayscale adjustment amplitude corresponding to the i-th pixel. When H fi When H is less than 0, it means that the gray value of the pixel has been downgraded. fi When it is greater than 0, it means that the grayscale value of the pixel has been adjusted upward; 01i represents the first gray value of the i-th pixel in the target image data; H 02i represents the second grayscale value of the image data before geometric correction corresponding to the i-th pixel; g represents the sign adjustment coefficient. When g=-1, When , let g = 1;

[0071] The adjusted target contrast value is obtained by using the grayscale adjustment amplitude of each pixel point; wherein the target contrast value is obtained by the following formula:

[0072] ;

[0073] Among them, D s Indicates the target contrast value; D 0 Indicates the contrast value of the target image data before contrast adjustment; H fi Indicates the grayscale adjustment amplitude corresponding to the i-th pixel. When H fi When H is less than 0, it means that the gray value of the pixel has been downgraded. fi When it is greater than 0, it means that the grayscale value of the pixel has been adjusted upward; 01i represents the first gray value of the i-th pixel in the target image data; H 02i represents the second grayscale value of the image data before geometric correction corresponding to the i-th pixel;

[0074] The brightness and contrast of the target image data after grayscale adjustment are adjusted according to the target brightness value and the target contrast value to obtain the target image data after brightness and contrast adjustment.

[0075] Preferably, in S4, the standard tumor-related data is imported into the risk prediction model for risk prediction, and tumor information is marked according to the risk threshold, including:

[0076] Extracting preprocessed image data and text data from standard tumor-related data;

[0077] After extraction, the image data and text data are fused;

[0078] Before data fusion, the image data is processed using a convolutional neural network, and the text data is processed using a recurrent neural network. After the data processing is completed, data fusion is performed;

[0079] The fused data is divided into a training set and a validation set. The random partitioning method is used to use 70% of the melted data as the training set and 30% of the data as the validation set.

[0080] Preferably, in S4, the standard tumor-related data is imported into the risk prediction model for risk prediction, and tumor information is marked according to the risk threshold, further comprising:

[0081] Retrieving a risk prediction model from a database, and importing a training set into the risk prediction model to train the risk prediction model;

[0082] After the risk prediction model is trained, the validation set is used to evaluate the indicators of the trained risk prediction model, and the prediction performance of the risk prediction model is judged according to the indicator evaluation results;

[0083] Among them, the evaluation indicators include the accuracy, recall, F1 index, and AUC value of the model;

[0084] When the prediction performance of the risk prediction model is judged to be not within the qualified range, the model parameters, feature dimensions and model structure of the risk prediction model are adjusted;

[0085] When the prediction performance of the risk prediction model is judged to be within the qualified range, the text data and image data in the trained risk prediction model are extracted;

[0086] Comparing the text data and image data in the trained risk prediction model with the standard medical data, wherein the standard medical data is obtained from the database;

[0087] Determine the prediction results of text data and image data in the risk prediction model based on the data comparison results;

[0088] Among them, the prediction results include the risk of tumor occurrence, progression, and treatment response;

[0089] And the prediction results are marked as predicted tumor data.

[0090] The tumor disease risk prediction and drug screening system based on artificial intelligence includes:

[0091] Predictive Tumor Outcomes Drug Analysis Unit for:

[0092] Confirm the risk level of tumor occurrence, progression, and treatment response based on predicted tumor data;

[0093] Among them, the risk levels are low risk, medium risk and high risk;

[0094] Importing predicted tumor data of each risk level and standard tumor-related data into a drug prediction model, wherein the drug prediction model is obtained from a database;

[0095] First, confirm the first compound data according to the risk level, wherein the first compound data is used to screen drugs of the same risk level according to the risk level of the predicted tumor data;

[0096] Then, the second compound data is screened out based on the medical records of the patient in the standard tumor-related data;

[0097] The first compound data and the second compound data are compared, and the same compound data are screened out according to the comparison result, and the same compound data are used as recommended drugs for the patient.

[0098] Compared with the prior art, the present invention has the following beneficial effects:

[0099] 1. The artificial intelligence-based tumor disease risk prediction and drug screening method and system provided by the present invention can integrate the patient's medical data and basic data, so that hospitals can manage medical resources more effectively. Through the confirmation of genomic data, transcriptomic data and proteomic data, the genetic background, gene expression and protein function of the tumor can be deeply understood. Advanced biotechnology means such as high-throughput sequencing, RNA sequencing and mass spectrometry are used. These technologies are highly sensitive and accurate, and can ensure the scientificity and reliability of the data.

[0100] 2. The artificial intelligence-based tumor disease risk prediction and drug screening method and system provided by the present invention utilize convolutional neural networks and recurrent neural networks to perform in-depth processing of image data and text data, respectively, effectively extracting key features of their respective fields, and comparing the trained risk prediction model with standard medical data, which can comprehensively evaluate the model's predictive ability in terms of tumor occurrence risk, progression risk, and treatment response risk. Through this risk prediction method based on big data and artificial intelligence, the molecular mechanisms of tumor occurrence, development, and treatment can be further revealed.

[0101] 3. The artificial intelligence-based tumor disease risk prediction and drug screening method and system provided by the present invention realizes the accurate stratification of patient tumor risks by confirming the degree of tumor occurrence risk, progression risk, and treatment response risk. Through accurate risk stratification and personalized drug recommendations, unnecessary waste of medical resources can be reduced, and limited medical resources can be more reasonably allocated and used. Personalized drug recommendations help patients obtain more suitable treatment plans, thereby improving the efficiency and effectiveness of treatment. BRIEF DESCRIPTION OF THE DRAWINGS

[0102] Figure 1A schematic diagram of the tumor disease risk prediction and drug screening steps of the present invention;

[0103] Figure 2 It is a schematic diagram of the process of tumor disease risk prediction and drug screening method of the present invention. DETAILED DESCRIPTION

[0104] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0105] In order to solve the problem that the existing technology does not collect the patient's medical data more accurately, resulting in inaccurate patient tumor data, please refer to Figure 1 and Figure 2 , this embodiment provides the following technical solutions:

[0106] The tumor disease risk prediction and drug screening method based on artificial intelligence includes the following steps:

[0107] S1: Acquisition of patient medical data: Acquiring the patient's medical data from the database and integrating the medical data with the patient's basic data into a data set. After the data set integration is completed, the patient's integrated data is obtained;

[0108] Among them, by integrating patients’ medical data and basic data, hospitals can manage medical resources more effectively;

[0109] S2: Collection of relevant tumor information: Perform tumor data analysis on the integrated patient data, confirm the analyzed tumor-related data, and mark the confirmed tumor-related data as target tumor-related data;

[0110] Among them, the integration of clinical data, genomic data, imaging data, and detection data helps to more accurately describe the characteristics and status of tumors;

[0111] S3: Tumor data preprocessing: preprocess the target tumor-related data, and obtain standard tumor-related data after the data preprocessing is completed;

[0112] Among them, the preprocessed data is more in line with the requirements of machine learning or deep learning models, can support more complex model training, and improve the performance and effect of the model;

[0113] S4: Tumor data risk prediction: Import standard tumor-related data into the risk prediction model for risk prediction, and mark the tumor information according to the risk threshold, and annotate the marked tumor information as predicted tumor data;

[0114] Among them, comprehensive prediction results help doctors to have a more comprehensive understanding of the patient's tumor condition and thus formulate a more reasonable treatment plan;

[0115] S5: Predicted tumor drug markers: Import the predicted tumor data into the drug prediction model for tumor drug analysis, and determine the drug recommendation for the patient based on the analysis results;

[0116] Among them, through precise risk stratification and personalized drug recommendations, unnecessary waste of medical resources can be reduced and limited medical resources can be allocated and used more reasonably.

[0117] In S1, the medical data of patients is obtained from the database, and the medical data is integrated with the basic data of patients into a data set, including:

[0118] Obtaining the patient's medical data from a database, wherein the patient's medical data includes clinical data, genomic data, imaging data, and test data;

[0119] Basic patient data includes demographic data, physical characteristics data, and medical history data;

[0120] Integrate the patient's medical data and basic patient data into data sets;

[0121] Among them, the format of the patient's medical data and basic patient data is first unified;

[0122] Create a data dictionary in the database, and match the medical data and basic data of patients in a unified format in the data dictionary;

[0123] After data matching is completed, the medical data and basic data of the same patient are integrated;

[0124] After the integration is completed, the patient's integrated data is obtained.

[0125] Specifically, by unifying the formats of medical data and basic data, data islands caused by inconsistent data formats are eliminated, ensuring data consistency and comparability. This helps reduce errors and deviations in subsequent data analysis, combining medical data (including clinical, genomics, imaging, and test data) with basic data (demographics, physical characteristics, and medical history) to form a complete patient data view. This comprehensive data set can more comprehensively reflect the patient's health status and condition, providing doctors with richer information support. By integrating patients' medical data and basic data, hospitals can more effectively manage medical resources, such as reasonably allocating beds and optimizing diagnosis and treatment processes. This helps improve the hospital's operational efficiency and service quality.

[0126] Perform tumor data analysis on the integrated patient data in S2 and confirm the analyzed tumor-related data, including:

[0127] Retrieve clinical data, genomic data, imaging data, and test data from integrated patient data;

[0128] Obtain the first tumor information from clinical data, genomic data, and imaging data respectively;

[0129] The first tumor information includes tumor data on the type, stage, location, size, growth rate, and whether the tumor has metastasized obtained from the clinical data;

[0130] Genomic data obtains genetic information of tumors;

[0131] Imaging data obtains structural information about the spatial location, shape, size, and boundaries of the tumor in the body;

[0132] The detection data obtains biomarker information of the tumor.

[0133] Confirm the first tumor information with omics data;

[0134] Among them, omics data include genomic data, transcriptomic data, and proteomic data;

[0135] The genomic data uses a high-throughput sequencing method to sequence the tumor tissue from the first tumor information, and the genomic DNA sequence information of the tumor cells is obtained after sequencing the tumor tissue;

[0136] The transcriptome data uses RNA sequencing to sequence RNA in the tumor tissue from the first tumor information, and quantitative analysis and differential expression analysis are performed during sequencing to finally obtain gene expression information;

[0137] The proteomic data uses mass spectrometry to separate and identify proteins in tumor tissues from the first tumor information, obtain information on the type, quantity and function of proteins, and at the same time, identify and authenticate tumor markers by comparing protein spectra in the database;

[0138] Based on the omics data confirmed in the first tumor information, the omics data is used as target tumor-related data of the patient.

[0139] Specifically, by integrating clinical data, genomic data, imaging data and detection data, it covers multi-dimensional information of tumors. This comprehensive data integration helps to more accurately describe the characteristics and status of tumors. Through the confirmation of genomic data, transcriptomic data and proteomic data, we can deeply understand the genetic background, gene expression and protein function of tumors, and provide strong support for the precise diagnosis and treatment of tumors. Genome sequencing can reveal the genomic DNA sequence information of tumor cells, which helps to discover tumor-related gene mutations and variations. Transcriptomic sequencing can quantitatively analyze gene expression and discover differentially expressed genes, thereby revealing the molecular mechanism of tumor occurrence and development. Proteomic analysis can identify the types, quantities and functions of proteins in tumor tissues, further verifying and supplementing the results of genomics and transcriptomics. Advanced biotechnology methods such as high-throughput sequencing, RNA sequencing and mass spectrometry are used. These technologies are highly sensitive and accurate, and can ensure the scientificity and reliability of data. By comparing protein spectra in the database, tumor markers can be identified and identified, providing an important basis for early diagnosis, prognosis evaluation and treatment monitoring of tumors.

[0140] In order to solve the problem in the prior art that the acquired tumor data is not further processed, resulting in poor training effect when the tumor data is used for model training in the later stage, please refer to Figure 1 and Figure 2 , this embodiment provides the following technical solutions:

[0141] Preprocess the target tumor-related data in S3, including:

[0142] Classify target tumor-related data into text data and image data;

[0143] After the classification is completed, the text data and image data are preprocessed;

[0144] The data preprocessing of the text data is to remove noise and repeated data in the text data, then segment the text in the text data into words, and perform part-of-speech recognition on the segmented words, map the recognized words into a continuous vector space for semantic capture, and perform feature vocabulary training based on the captured semantics, wherein the feature vocabulary training is performed using a training model, and the trained feature vocabulary is standardized, and finally the preprocessed text data is obtained;

[0145] The data preprocessing of the image data is to convert the image data into grayscale images, perform geometric correction on the image after the grayscale image conversion is completed, adjust the parameters of the brightness, contrast, clarity and sharpness of the image after the geometric correction, use smoothing and median filtering methods to remove noise in the image after the parameter adjustment is completed, and then use the gradient operator method to detect the edge information in the image, extract the feature image of the image according to the edge information, normalize the feature image, and finally obtain the preprocessed image data;

[0146] The preprocessed image data and text data are uniformly labeled as standard tumor-related data.

[0147] Specifically, the data is clearly divided into text data and image data, which helps to adopt the most appropriate preprocessing method for different types of data and improve the accuracy and efficiency of subsequent analysis. In text data processing, removing noise and duplicate data can reduce data redundancy, improve data quality, and make model training more efficient and accurate. Steps such as text segmentation, part-of-speech recognition, semantic capture, and feature vocabulary training can deeply mine useful information in the text and convert text data into a format that is easier for machines to understand and process, providing strong support for subsequent model training. Standardized feature vocabulary ensures consistency and comparability of data and helps to improve the generalization ability of the model. Steps such as grayscale image conversion, geometric correction, and parameter adjustment can significantly improve the accuracy of the model. Improve image quality and reduce interference factors in images, laying a solid foundation for subsequent feature extraction. Steps such as noise removal, edge information detection and feature image extraction can accurately capture key information in images and improve the utilization efficiency of image data. Normalization processing ensures the comparability and consistency of image data in different dimensions, which helps to improve the stability and accuracy of the model. The preprocessed image data and text data are uniformly labeled as standard tumor-related data, which helps to achieve data standardization and normalization, and facilitates data sharing and comparative analysis between different studies or projects. The preprocessed data is more in line with the requirements of machine learning or deep learning models, and can support more complex model training, improving the performance and effect of the model.

[0148] Specifically, the data preprocessing of image data includes:

[0149] Performing grayscale image conversion on the image data to generate a grayscale image corresponding to the image data;

[0150] Performing geometric correction on the grayscale image to obtain a grayscale image after geometric correction, and using the grayscale image after geometric correction as target image data;

[0151] Extracting the grayscale value of each pixel of the target image data as a first grayscale value;

[0152] Extracting the grayscale value of each pixel of the image data before geometric correction corresponding to the target image data as the second grayscale value;

[0153] Compare the first grayscale value of each pixel of the target image data with the second grayscale value of the pixel at the same position of the image data before geometric correction to obtain a grayscale difference;

[0154] The grayscale deviation coefficient of the target image data after geometric correction is obtained by using the grayscale difference; wherein the grayscale deviation coefficient is obtained by the following formula:

[0155] ;

[0156] Where S represents the grayscale deviation coefficient; n represents the number of pixels in the target image data; H 01i represents the first gray value of the i-th pixel in the target image data; H 02i represents the second grayscale value of the image data before geometric correction corresponding to the i-th pixel; H z01 Indicates the central gray value corresponding to the target image data; H z02 represents the central grayscale value corresponding to the image data before geometric correction; s represents the adjustment coefficient, and the adjustment coefficient is obtained by the following formula:

[0157] ;

[0158] Where s represents the adjustment coefficient; a represents the number of pixels whose grayscale value of the target image data is reduced relative to the image data before geometric correction; b represents the number of pixels whose grayscale value of the target image data is increased relative to the image data before geometric correction; H 01ai represents the first grayscale value corresponding to the pixel whose grayscale value decreases at the ith pixel; H 02ai H represents the second grayscale value corresponding to the pixel whose grayscale value is reduced by the i-th pixel; 01bi H represents the first grayscale value corresponding to the pixel point whose grayscale value is increased by the i-th pixel; 02bi represents the second grayscale value corresponding to the pixel point whose grayscale value is increased by the i-th pixel;

[0159] Comparing the grayscale deviation coefficient with a preset grayscale deviation coefficient threshold;

[0160] When the grayscale deviation coefficient exceeds a preset grayscale deviation coefficient threshold, the grayscale value of the target image data is adjusted.

[0161] The technical effect of the above technical solution is: through grayscale image conversion, the complexity of image processing is simplified while retaining the main information of the image. Geometric correction can correct the geometric deformation of the image during shooting or transmission, making the image more accurate and clear.

[0162] By comparing the grayscale value of each pixel in the image data before and after geometric correction, the grayscale difference can be accurately calculated, and then the grayscale deviation coefficient can be obtained. The grayscale deviation coefficient reflects the degree of change of the grayscale value of the image during the geometric correction process and is an important indicator for evaluating image quality and processing effect.

[0163] According to the comparison result between the grayscale deviation coefficient and the preset threshold, the grayscale value of the target image data can be adjusted to further optimize the image processing effect. When the grayscale deviation coefficient exceeds the preset threshold, it means that the geometric correction has a great impact on the grayscale value of the image, and adjustment is required to restore or improve the image quality. By introducing the adjustment coefficient, the calculation method of the grayscale deviation coefficient can be flexibly adjusted according to the specific situation and processing requirements of the image, so that the image processing can be more adapted to different application scenarios and image characteristics.

[0164] In summary, this technical solution realizes the preprocessing and optimization of image data through steps such as grayscale image conversion, geometric correction, grayscale value comparison and adjustment, improves image quality and processing effect, and enhances the flexibility and adaptability of image processing.

[0165] Specifically, when the grayscale deviation coefficient exceeds a preset grayscale deviation coefficient threshold, grayscale value adjustment is performed on the target image data, including:

[0166] When the grayscale deviation coefficient exceeds a preset grayscale deviation coefficient threshold, the first grayscale value of each pixel of the target image data and the second grayscale value of the pixel of the image data corresponding to each pixel of the target image data are retrieved;

[0167] Using a grayscale difference between a first grayscale value of each pixel of the target image data and a second grayscale value of a pixel of the image data corresponding to each pixel of the target image data as a first grayscale difference;

[0168] Retrieving the first grayscale value of the pixel points adjacent to each pixel point in the target image data and the second grayscale value of the pixel points adjacent to each pixel point in the image data corresponding to each pixel point;

[0169] Obtaining a second grayscale difference value by using a first grayscale value of a pixel point adjacent to each pixel point in the target image data and a second grayscale value corresponding to a pixel point adjacent to each pixel point in the image data;

[0170] The target grayscale value of each pixel in the target image data is obtained by using the first grayscale difference and the second grayscale difference; wherein the target grayscale value is obtained by the following formula:

[0171] ;

[0172] Among them, H s Indicates the target gray value; H 01i represents the first gray value of the i-th pixel in the target image data; H 02i represents the second grayscale value of the image data before geometric correction corresponding to the i-th pixel; k represents the number of pixels adjacent to each pixel in the target image data; H 01j H represents the first gray value corresponding to the jth adjacent pixel; 02j represents the second grayscale value corresponding to the jth adjacent pixel; S represents the grayscale deviation coefficient; S y Indicates the preset grayscale deviation coefficient threshold; H 01 Represents the first grayscale value of the pixel in the target image data;

[0173] Performing grayscale adjustment on each pixel in the target image data according to the target grayscale value to obtain the target image data after grayscale adjustment;

[0174] The brightness and contrast are adjusted for the target image data after the grayscale adjustment to obtain the target image data after the brightness and contrast adjustment.

[0175] The technical effect of the above technical solution is: when the grayscale deviation coefficient exceeds the preset threshold, the solution accurately adjusts the grayscale value of the target image data by calculating the grayscale difference (first grayscale difference and second grayscale difference) of each pixel and its adjacent pixels in detail. This adjustment method based on local grayscale changes can process images more carefully and reduce grayscale distortion caused by geometric correction. By considering the grayscale value of each pixel and its adjacent pixels, the solution not only considers the grayscale change of a single pixel when calculating the target grayscale value, but also takes into account the continuity between pixels. This helps to maintain a smooth transition of the image and reduce the abruptness caused by grayscale adjustment.

[0176] The grayscale deviation coefficient (S) and the preset threshold (Sy) are introduced into the formula to make the grayscale adjustment process adaptive. When the grayscale deviation of the image data is large, the adjustment intensity will increase accordingly to better restore or improve the image quality. On the basis of grayscale adjustment, the solution further adjusts the brightness and contrast of the target image data. This helps to improve the visual effect of the image and make the image more vivid and clear. By optimizing the brightness and contrast, the important information in the image can be better highlighted while reducing background interference. Combining multiple steps such as grayscale adjustment, brightness and contrast optimization, this technical solution can significantly improve the overall quality of the target image data. Whether in terms of the uniformity of grayscale distribution, image continuity or visual effect, the processed image is closer to the real scene, which is conducive to subsequent image analysis and processing.

[0177] In summary, this technical solution achieves high-quality preprocessing of target image data by accurately calculating grayscale differences, considering pixel continuity, introducing an adaptive adjustment mechanism, and optimizing brightness and contrast, laying a solid foundation for subsequent image processing and analysis.

[0178] Specifically, adjusting the brightness and contrast of the target image data after the grayscale adjustment to obtain the target image data after the brightness and contrast adjustment includes:

[0179] Extracting the grayscale adjustment amplitude of each pixel of the target image data after grayscale adjustment;

[0180] The adjusted target brightness value is obtained by using the grayscale adjustment amplitude of each pixel point; wherein the target brightness value is obtained by the following formula:

[0181] ;

[0182] Among them, G s Indicates the target brightness value; G 0 represents the brightness value of the target image data before brightness adjustment; n represents the number of pixels of the target image data; H fi Indicates the grayscale adjustment amplitude corresponding to the i-th pixel. When H fi When H is less than 0, it means that the gray value of the pixel has been downgraded. fi When it is greater than 0, it means that the grayscale value of the pixel has been adjusted upward; 01i represents the first gray value of the i-th pixel in the target image data; H 02i represents the second grayscale value of the image data before geometric correction corresponding to the i-th pixel; g represents the sign adjustment coefficient. When g=-1, When , let g = 1;

[0183] The adjusted target contrast value is obtained by using the grayscale adjustment amplitude of each pixel point; wherein the target contrast value is obtained by the following formula:

[0184] ;

[0185] Among them, D s Indicates the target contrast value; D 0 Indicates the contrast value of the target image data before contrast adjustment; H fi Indicates the grayscale adjustment amplitude corresponding to the i-th pixel. When H fi When H is less than 0, it means that the gray value of the pixel has been downgraded. fi When it is greater than 0, it means that the grayscale value of the pixel has been adjusted upward; 01i represents the first gray value of the i-th pixel in the target image data; H 02i represents the second grayscale value of the image data before geometric correction corresponding to the i-th pixel;

[0186] The brightness and contrast of the target image data after grayscale adjustment are adjusted according to the target brightness value and the target contrast value to obtain the target image data after brightness and contrast adjustment.

[0187] The technical effect of the above technical solution is: the solution realizes precise adjustment of image brightness and contrast by calculating the grayscale adjustment amplitude of each pixel in detail and obtaining the adjusted target brightness value and target contrast value accordingly. This adjustment method based on the grayscale change of pixel points can process the image more carefully, so that the brightness and contrast of the image are more in line with visual requirements. On the basis of grayscale adjustment, the overall quality of the image can be further improved by optimizing brightness and contrast. Brightness adjustment can make the image brighter or dimmer, while contrast adjustment can enhance the difference between different areas in the image, making the image more vivid and clear.

[0188] The introduction of the sign adjustment coefficient (g) in the formula makes the brightness adjustment process adaptive. According to the increase or decrease of the grayscale value of the pixel, the direction of the brightness adjustment will also change accordingly to better restore or improve the image quality. By considering the grayscale adjustment amplitude of each pixel, the scheme can maintain the image detail information when adjusting the brightness and contrast. This helps to avoid losing important image features such as edges and textures during the adjustment process. Combined with the optimization of brightness and contrast, this technical solution can significantly improve the visual effect of the image. The adjusted image is more in line with the visual habits of the human eye, which is conducive to subsequent image observation, analysis and processing.

[0189] In summary, this technical solution achieves further optimization of the target image data after grayscale adjustment by accurately calculating the grayscale adjustment amplitude, adaptively adjusting the brightness, maintaining image details, and improving visual effects, providing higher quality image data for subsequent image processing and applications.

[0190] In order to solve the problem in the prior art that the acquired tumor data cannot be used to effectively predict tumor risk, resulting in the inability to make targeted drug recommendations based on tumor risk and type, please refer to Figure 1 and Figure 2 , this embodiment provides the following technical solutions:

[0191] In S4, standard tumor-related data are imported into the risk prediction model for risk prediction, and tumor information is marked according to the risk threshold, including:

[0192] Extracting preprocessed image data and text data from standard tumor-related data;

[0193] After extraction, the image data and text data are fused;

[0194] Before data fusion, the image data is processed using a convolutional neural network, and the text data is processed using a recurrent neural network. After the data processing is completed, data fusion is performed;

[0195] The fused data is divided into a training set and a validation set. The random partitioning method is used to use 70% of the melted data as the training set and 30% of the data as the validation set.

[0196] Retrieving a risk prediction model from a database, and importing a training set into the risk prediction model to train the risk prediction model;

[0197] After the risk prediction model is trained, the validation set is used to evaluate the indicators of the trained risk prediction model, and the prediction performance of the risk prediction model is judged according to the indicator evaluation results;

[0198] Among them, the evaluation indicators include the accuracy, recall, F1 index, and AUC value of the model;

[0199] When the prediction performance of the risk prediction model is judged to be not within the qualified range, the model parameters, feature dimensions and model structure of the risk prediction model are adjusted;

[0200] When the prediction performance of the risk prediction model is judged to be within the qualified range, the text data and image data in the trained risk prediction model are extracted;

[0201] Comparing the text data and image data in the trained risk prediction model with the standard medical data, wherein the standard medical data is obtained from the database;

[0202] Determine the prediction results of text data and image data in the risk prediction model based on the data comparison results;

[0203] Among them, the prediction results include the risk of tumor occurrence, progression, and treatment response;

[0204] And the prediction results are marked as predicted tumor data.

[0205] Specifically, convolutional neural networks and recurrent neural networks were used to perform deep processing on image data and text data respectively, effectively extracting the key features of their respective fields. Subsequently, the two different but complementary information sources were integrated through data fusion technology, which significantly enhanced the quality of the model's input data and provided a basis for more accurate risk prediction. The fused data was divided into a training set and a validation set using a random partitioning method to ensure the independence and representativeness of the training set and the validation set. The risk prediction model was trained with the training set, and then the model was rigorously evaluated with the validation set using indicators including accuracy, recall, F1 index, and AUC value. These indicators fully reflect the model's predictive performance. When the model's predictive performance does not meet the requirements, the solution provides flexible adjustment strategies, including adjusting model parameters, feature dimensions, and model structure, to optimize the model's predictive ability. This iterative optimization process helps to continuously improve the accuracy and robustness of the model. By comparing the trained risk prediction model with standard medical data, the model's predictive ability in terms of tumor occurrence risk, progression risk, and treatment response risk can be comprehensively evaluated. This comprehensive prediction result helps doctors to have a more comprehensive understanding of the patient's tumor condition, so as to formulate a more reasonable treatment plan. Through this risk prediction method based on big data and artificial intelligence, the molecular mechanism of tumor occurrence, development and treatment can be further revealed, providing new ideas and methods for basic research and clinical application of tumors. At the same time, this prediction method can also help promote the development of personalized tumor treatment and improve treatment effects and patient survival rates.

[0206] The tumor disease risk prediction and drug screening system based on artificial intelligence includes:

[0207] Predictive Tumor Outcomes Drug Analysis Unit for:

[0208] Confirm the risk level of tumor occurrence, progression, and treatment response based on predicted tumor data;

[0209] Among them, the risk levels are low risk, medium risk and high risk;

[0210] Importing predicted tumor data of each risk level and standard tumor-related data into a drug prediction model, wherein the drug prediction model is obtained from a database;

[0211] First, confirm the first compound data according to the risk level, wherein the first compound data is used to screen drugs of the same risk level according to the risk level of the predicted tumor data;

[0212] Then, the second compound data is screened out based on the medical records of the patient in the standard tumor-related data;

[0213] The first compound data and the second compound data are compared, and the same compound data are screened out according to the comparison result, and the same compound data are used as recommended drugs for the patient.

[0214] Specifically, by confirming the risk of tumor occurrence, progression risk, and treatment response risk (low risk, medium risk, and high risk), accurate stratification of patient tumor risk is achieved. This stratification not only helps doctors understand the patient's disease status more clearly, but also provides a more accurate basis for subsequent drug recommendations. Drugs of the same risk level (first compound data) are screened according to the risk level, and personalized drug recommendations are made in combination with the patient's medical records (second compound data). This method fully considers the patient's specific situation and disease characteristics, improves the pertinence and effectiveness of drug recommendations, and screens out the same compound data as recommended drugs by comparing the first compound data and the second compound data. This process is based on scientific data comparison and screening methods, avoids the risks that may be caused by blindly recommending drugs, ensures the rationality and safety of recommended drugs, and reduces unnecessary waste of medical resources through accurate risk stratification and personalized drug recommendations, so that limited medical resources can be more reasonably allocated and used. Personalized drug recommendations help patients obtain more suitable treatment plans, thereby improving the efficiency and effectiveness of treatment. At the same time, accurate risk stratification also helps doctors identify high-risk patients earlier, take more active intervention measures, and delay the progression of the disease.

[0215] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0216] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

Claims

1. A tumor disease risk prediction and drug screening method based on artificial intelligence, characterized in that: The steps include: S1: Acquisition of patient medical data: Acquiring the patient's medical data from the database and integrating the medical data with the patient's basic data into a data set. After the data set integration is completed, the patient's integrated data is obtained; S2: Collection of relevant tumor information: Perform tumor data analysis on the integrated patient data, confirm the analyzed tumor-related data, and mark the confirmed tumor-related data as target tumor-related data; S3: Tumor data preprocessing: preprocess the target tumor-related data, and obtain standard tumor-related data after the data preprocessing is completed; S4: Tumor data risk prediction: Import standard tumor-related data into the risk prediction model for risk prediction, and mark the tumor information according to the risk threshold, and annotate the marked tumor information as predicted tumor data; S5: Predicted tumor drug markers: Import the predicted tumor data into the drug prediction model for tumor drug analysis, and determine the drug recommendation for the patient based on the analysis results; Preprocess the target tumor-related data in S3, including: Classify target tumor-related data into text data and image data; After the classification is completed, the text data and image data are preprocessed; The data preprocessing of the text data is to remove noise and repeated data in the text data, then segment the text in the text data into words, and perform part-of-speech recognition on the segmented words, map the recognized words into a continuous vector space for semantic capture, and perform feature vocabulary training based on the captured semantics, wherein the feature vocabulary training is performed using a training model, and the trained feature vocabulary is standardized, and finally the preprocessed text data is obtained; The data preprocessing of the image data is to convert the image data into grayscale images, perform geometric correction on the image after the grayscale image conversion is completed, adjust the parameters of the brightness, contrast, clarity and sharpness of the image after the geometric correction, use smoothing and median filtering methods to remove noise in the image after the parameter adjustment is completed, and then use the gradient operator method to detect the edge information in the image, extract the feature image of the image according to the edge information, normalize the feature image, and finally obtain the preprocessed image data; The pre-processed image data and text data are uniformly labeled as standard tumor-related data; In S4, standard tumor-related data are imported into the risk prediction model for risk prediction, and tumor information is marked according to the risk threshold, including: Extracting pre-processed image data and text data from standard tumor-related data; After extraction, the image data and text data are fused; Before data fusion, the image data is processed using a convolutional neural network, and the text data is processed using a recurrent neural network. After the data processing is completed, data fusion is performed; Divide the fused data into a training set and a validation set; Retrieving a risk prediction model from a database, and importing a training set into the risk prediction model to train the risk prediction model; After the risk prediction model is trained, the validation set is used to evaluate the indicators of the trained risk prediction model, and the prediction performance of the risk prediction model is judged according to the indicator evaluation results; When the prediction performance of the risk prediction model is judged to be not within the qualified range, the model parameters, feature dimensions and model structure of the risk prediction model are adjusted; When the prediction performance of the risk prediction model is judged to be within the qualified range, the text data and image data in the trained risk prediction model are extracted; Comparing the text data and image data in the trained risk prediction model with the standard medical data, wherein the standard medical data is obtained from the database; Determine the prediction results of text data and image data in the risk prediction model based on the data comparison results; Among them, the prediction results include the risk of tumor occurrence, progression, and treatment response; And the prediction results are marked as predicted tumor data; For the predictive tumor drug markers in s5, it includes: confirming the risk of tumor occurrence, progression, and treatment response based on the predicted tumor data; Among them, the risk levels are low risk, medium risk and high risk; Importing predicted tumor data of each risk level and standard tumor-related data into a drug prediction model, wherein the drug prediction model is obtained from a database; First, confirm the first compound data according to the risk level, wherein the first compound data is used to screen drugs of the same risk level according to the risk level of the predicted tumor data; Then, the second compound data is screened out based on the medical records of the patient in the standard tumor-related data; The first compound data and the second compound data are compared, and the same compound data are screened out according to the comparison result, and the same compound data are used as recommended drugs for the patient.

2. The method for predicting tumor disease risk and screening drugs based on artificial intelligence according to claim 1, characterized in that: In S1, the medical data of patients is obtained from the database, and the medical data is integrated with the basic data of patients into a data set, including: Obtaining the patient's medical data from a database, wherein the patient's medical data includes clinical data, genomic data, imaging data, and test data; Basic patient data includes demographic data, physical characteristics data, and medical history data; Integrate the patient's medical data and basic patient data into data sets; Among them, the format of the patient's medical data and basic patient data is first unified; Create a data dictionary in the database, and match the medical data and basic data of patients in a unified format in the data dictionary; After data matching is completed, the medical data and basic data of the same patient are integrated; After the integration is completed, the patient's integrated data is obtained.

3. The method for predicting tumor disease risk and screening drugs based on artificial intelligence according to claim 2, characterized in that: Perform tumor data analysis on the integrated patient data in S2 and confirm the analyzed tumor-related data, including: Retrieve clinical data, genomic data, imaging data, and test data from integrated patient data; Obtain the first tumor information from clinical data, genomic data, and imaging data respectively; The first tumor information includes tumor data on the type, stage, location, size, growth rate, and whether the tumor has metastasized obtained from the clinical data; Genomic data obtains genetic information of tumors; Imaging data obtains structural information about the spatial location, shape, size, and boundaries of the tumor in the body; The detection data obtains biomarker information of the tumor.

4. The method for predicting tumor disease risk and screening drugs based on artificial intelligence according to claim 3, characterized in that: In S2, the patient integrated data is analyzed for tumor data, and the analyzed tumor-related data is confirmed, including: Confirm the first tumor information with omics data; Among them, omics data include genomic data, transcriptomic data, and proteomic data; The genomic data uses a high-throughput sequencing method to sequence the tumor tissue from the first tumor information, and the genomic DNA sequence information of the tumor cells is obtained after sequencing the tumor tissue; The transcriptome data uses RNA sequencing to sequence RNA in the tumor tissue from the first tumor information, and quantitative analysis and differential expression analysis are performed during sequencing to finally obtain gene expression information; The proteomic data uses mass spectrometry to separate and identify proteins in tumor tissues from the first tumor information, obtain information on the type, quantity and function of proteins, and at the same time, identify and authenticate tumor markers by comparing protein spectra in the database; Based on the omics data confirmed in the first tumor information, the omics data is used as target tumor-related data of the patient.

5. The method for tumor disease risk prediction and drug screening based on artificial intelligence according to claim 1, characterized in that: Data preprocessing of image data, including: Performing grayscale image conversion on the image data to generate a grayscale image corresponding to the image data; Performing geometric correction on the grayscale image to obtain a grayscale image after geometric correction, and using the grayscale image after geometric correction as target image data; Extracting the grayscale value of each pixel of the target image data as a first grayscale value; Extracting the grayscale value of each pixel of the image data before geometric correction corresponding to the target image data as the second grayscale value; Compare the first grayscale value of each pixel of the target image data with the second grayscale value of the pixel at the same position of the image data before geometric correction to obtain a grayscale difference; The grayscale deviation coefficient of the target image data after geometric correction is obtained by using the grayscale difference; wherein the grayscale deviation coefficient is obtained by the following formula: ； Where S represents the grayscale deviation coefficient; n represents the number of pixels in the target image data; H 01i represents the first gray value of the i-th pixel in the target image data; H 02i represents the second grayscale value of the image data before geometric correction corresponding to the i-th pixel; H z01 Indicates the central gray value corresponding to the target image data; H z02 represents the central grayscale value corresponding to the image data before geometric correction; s represents the adjustment coefficient, and the adjustment coefficient is obtained by the following formula: ； Where s represents the adjustment coefficient; a represents the number of pixels whose grayscale value of the target image data is reduced relative to the image data before geometric correction; b represents the number of pixels whose grayscale value of the target image data is increased relative to the image data before geometric correction; H 01ai represents the first grayscale value corresponding to the pixel whose grayscale value decreases at the ith pixel; H 02ai H represents the second grayscale value corresponding to the pixel whose grayscale value is reduced by the i-th pixel; 01bi H represents the first grayscale value corresponding to the pixel point whose grayscale value is increased by the i-th pixel; 02bi represents the second grayscale value corresponding to the pixel point whose grayscale value is increased by the i-th pixel; Comparing the grayscale deviation coefficient with a preset grayscale deviation coefficient threshold; When the grayscale deviation coefficient exceeds a preset grayscale deviation coefficient threshold, the grayscale value of the target image data is adjusted.

6. The method for predicting tumor disease risk and screening drugs based on artificial intelligence according to claim 5, characterized in that: When the grayscale deviation coefficient exceeds a preset grayscale deviation coefficient threshold, grayscale value adjustment is performed on the target image data, including: When the grayscale deviation coefficient exceeds a preset grayscale deviation coefficient threshold, the first grayscale value of each pixel of the target image data and the second grayscale value of the pixel of the image data corresponding to each pixel of the target image data are retrieved; Using a grayscale difference between a first grayscale value of each pixel of the target image data and a second grayscale value of a pixel of the image data corresponding to each pixel of the target image data as a first grayscale difference; Retrieving the first grayscale value of the pixel points adjacent to each pixel point in the target image data and the second grayscale value of the pixel points adjacent to each pixel point in the image data corresponding to each pixel point; Obtaining a second grayscale difference value by using a first grayscale value of a pixel point adjacent to each pixel point in the target image data and a second grayscale value corresponding to a pixel point adjacent to each pixel point in the image data; The target grayscale value of each pixel in the target image data is obtained by using the first grayscale difference and the second grayscale difference; wherein the target grayscale value is obtained by the following formula: ； Among them, H s Indicates the target gray value; H 01i represents the first gray value of the i-th pixel in the target image data; H 02i represents the second grayscale value of the image data before geometric correction corresponding to the i-th pixel; k represents the number of pixels adjacent to each pixel in the target image data; H 01j H represents the first gray value corresponding to the jth adjacent pixel; 02j represents the second grayscale value corresponding to the jth adjacent pixel; S represents the grayscale deviation coefficient; S y Indicates the preset grayscale deviation coefficient threshold; H 01 Represents the first grayscale value of the pixel in the target image data; Performing grayscale adjustment on each pixel in the target image data according to the target grayscale value to obtain the target image data after grayscale adjustment; The brightness and contrast are adjusted for the target image data after the grayscale adjustment to obtain the target image data after the brightness and contrast adjustment.

7. The method for predicting tumor disease risk and screening drugs based on artificial intelligence according to claim 6, characterized in that: The brightness and contrast of the target image data after the grayscale adjustment are adjusted to obtain the target image data after the brightness and contrast adjustment, including: Extracting the grayscale adjustment amplitude of each pixel of the target image data after grayscale adjustment; The adjusted target brightness value is obtained by using the grayscale adjustment amplitude of each pixel point; wherein the target brightness value is obtained by the following formula: ； Among them, G s represents the target brightness value; G0 represents the brightness value of the target image data before brightness adjustment; n represents the number of pixels of the target image data; H fi Indicates the grayscale adjustment amplitude corresponding to the i-th pixel. When H fi When H is less than 0, it means that the gray value of the pixel has been downgraded. fi When it is greater than 0, it means that the grayscale value of the pixel has been adjusted upward; 01i represents the first gray value of the i-th pixel in the target image data; H 02i represents the second grayscale value of the image data before geometric correction corresponding to the i-th pixel; g represents the sign adjustment coefficient. When g=-1, When , let g = 1; The adjusted target contrast value is obtained by using the grayscale adjustment amplitude of each pixel point; wherein the target contrast value is obtained by the following formula: ； Among them, D s represents the target contrast value; D0 represents the contrast value of the target image data before contrast adjustment; H fi Indicates the grayscale adjustment amplitude corresponding to the i-th pixel. When H fi When H is less than 0, it means that the gray value of the pixel has been downgraded. fi When it is greater than 0, it means that the grayscale value of the pixel has been adjusted upward; 01i represents the first gray value of the i-th pixel in the target image data; H 02i represents the second grayscale value of the image data before geometric correction corresponding to the i-th pixel; The brightness and contrast of the target image data after grayscale adjustment are adjusted according to the target brightness value and the target contrast value to obtain the target image data after brightness and contrast adjustment.

8. The method for predicting tumor disease risk and screening drugs based on artificial intelligence according to claim 1, characterized in that: The random partitioning method is used to select 70% of the fused data as the training set and 30% of the data as the validation set. The evaluation indicators include model accuracy, recall rate, F1 index, and AUC value.

Images