Rule learning-based tumor radiotherapy reaction prediction system and method

Abstract

The application discloses a tumor radiotherapy reaction prediction system and method based on rule learning, relates to the medical field, and solves the problem of poor prediction effect of the existing tumor radiotherapy reaction prediction system, and comprises the following modules: a disease analysis module: tumor marker examination is performed on a target tumor patient, corresponding marker screening intervals are set according to the examination results for tumor markers, patient disease cycle analysis data are obtained, a data matching module: a disease matching model is created according to the patient disease cycle analysis data, real-time disease consistent patients are matched for the target tumor patient through the disease matching model, target patient matching data are obtained, and a reaction prediction module: radiotherapy reaction prediction is performed on the target tumor patient according to the target patient matching data, and the application has high tumor radiotherapy reaction prediction efficiency and high accuracy of prediction results.

Description

TECHNICAL FIELD

[0001] The application belongs to the medical field, and relates to a rule learning technology, in particular to a tumor radiotherapy reaction prediction system and method based on rule learning. BACKGROUND

[0002] The existing tumor radiotherapy reaction prediction system has the following defects when predicting the radiotherapy reaction of a tumor patient:

[0003] 1. The existing tumor radiotherapy reaction prediction system cannot set multiple tumor marker examinations for a target tumor patient, cannot set a corresponding marker screening interval for each tumor marker examination according to the examination results of the target tumor patient within a medical index monitoring period, and cannot create a disease matching model based on the obtained marker screening interval as a model creation basis, so as to ensure the training effect of the disease matching model, and finally lead to the lack of accuracy of the tumor radiotherapy reaction prediction result.

[0004] 2. The existing tumor radiotherapy reaction prediction system cannot create a disease matching model according to patient disease cycle analysis data, cannot match a real-time disease consistent patient for the target tumor patient according to the disease matching model, cannot analyze the tumor radiotherapy reaction of the real-time disease consistent patient, and cannot predict the radiotherapy reaction of the target tumor patient according to the analysis result, so as to easily lead to low efficiency and lack of real-time of the tumor radiotherapy reaction prediction process.

[0005] Therefore, the tumor radiotherapy reaction prediction system and method based on rule learning are proposed. SUMMARY

[0006] In view of the defects of the prior art, the application aims to provide a tumor radiotherapy reaction prediction system and method based on rule learning, and aims to improve the accuracy and prediction efficiency of the tumor radiotherapy reaction prediction system.

[0007] In order to achieve the above-mentioned purpose, the application adopts the following technical scheme: a tumor radiotherapy reaction prediction system based on rule learning, comprising:

[0008] A disease analysis module: performing tumor marker examination on a target tumor patient, setting a corresponding marker screening interval for the tumor marker according to the examination result, and obtaining patient disease cycle analysis data;

[0009] A data matching module: creating a disease matching model according to the patient disease cycle analysis data, matching a real-time disease consistent patient for the target tumor patient through the disease matching model, and obtaining target patient matching data;

[0010] A reaction prediction module: predicting the radiotherapy reaction of the target tumor patient according to the target patient matching data.

[0011] Further, the patient condition cycle analysis data is obtained, specifically as follows:

[0012] The tumor patients in need of tumor radiotherapy reaction prediction are obtained, and a plurality of tumor patients are obtained, and a target tumor patient is randomly selected from the plurality of tumor patients;

[0013] In the process of analyzing the condition of the target tumor patient, the tumor markers examined by the target tumor patient are obtained, and a plurality of different types of tumor markers are obtained, and the plurality of different types of tumor markers are Z1 tumor marker to Za tumor marker;

[0014] In the process of analyzing the condition of the target tumor patient, the time point when the target tumor patient is definitely diagnosed as a tumor patient is set as a first cycle characteristic time point, the time point corresponding to the current time is set as a second cycle characteristic time point, the intermediate time point between the first cycle characteristic time point and the second cycle characteristic time point is set as a third cycle characteristic time point, and the period between the third cycle characteristic time point and the second cycle characteristic time point is set as a medical index monitoring period;

[0015] The Z1 tumor marker of the target tumor patient in the medical index monitoring period is monitored for index analysis, and the Z1 marker screening interval is obtained according to the analysis result.

[0016] Further, the Z1 marker screening interval is obtained, specifically as follows:

[0017] The Z1 tumor marker examination of the target tumor patient is obtained, and a plurality of Z1 tumor marker examinations are obtained, and the plurality of Z1 tumor marker examinations are renamed as X1 marker examination to Xb marker examination according to the order of examination;

[0018] The detection results of X1 marker examination to Xb marker examination are analyzed, and the first marker characteristic value and the second marker characteristic value are obtained according to the analysis result;

[0019] The product of the first marker characteristic value and the second marker characteristic value is calculated, and the third marker characteristic value is obtained;

[0020] The sum of the first marker characteristic value and the third marker characteristic value is calculated, and the upper limit of the Z1 marker screening interval is obtained. The difference between the first marker characteristic value and the third marker characteristic value is calculated, and the lower limit of the Z1 marker screening interval is obtained. The numerical interval composed of the upper limit of the Z1 marker screening interval and the lower limit of the Z1 marker screening interval is set as the Z1 marker screening interval.

[0021] Further, the first marker characteristic value and the second marker characteristic value are obtained, specifically as follows:

[0022] Respectively acquire the marker examination result values corresponding to the X1 marker examination to the Xb marker examination, to obtain the X1 marker examination result value to the Xb marker examination result value;

[0023] The X1 marker examination result value to the Xb marker examination result value is calculated, and the first marker characteristic value is obtained;

[0024] The X1 result value change rate is obtained by calculating the X1 marker examination result value and the X2 marker examination result value;

[0025] The X1 result value change rate is calculated.

[0026] Further, the target patient matching data is acquired, specifically as follows:

[0027] Acquire the target patient matching data, and acquire the Z1 marker screening interval to the Za marker screening interval according to the target patient matching data;

[0028] The tumor type corresponding to the target tumor patient is acquired, and the target type tumor is obtained, the historical patients admitted by the diagnosis institution for the target type tumor are acquired, and a plurality of historical patients are obtained, and a disease matching model is created according to the historical patients;

[0029] The real-time admitted patients admitted by the diagnosis institution for the target type tumor are acquired, and a plurality of real-time admitted patients are obtained;

[0030] Each real-time admitted patient is matched with the target tumor patient for disease consistency using the patient disease matching model, a plurality of real-time disease consistent patients are obtained, and the plurality of real-time disease consistent patients obtained are defined as target patient matching data.

[0031] Further, the disease matching model is created, specifically as follows:

[0032] The plurality of historical admitted patients obtained are divided into disease matching consistent patients and disease matching inconsistent patients, and historical patient marking data is obtained;

[0033] The historical patient marking data is divided into a historical patient training set and a sample patient test set according to a preset training test ratio;

[0034] A rule learning model is created through an existing artificial intelligence platform, the rule learning model is classified and trained using the historical patient training set, and each historical admitted patient in the historical patient training set is trained once until the rule learning model is trained;

[0035] The rule learning model is tested for classification using a sample patient test set, and the classification accuracy is obtained. When the classification accuracy is greater than or equal to the target classification accuracy, the rule learning model training is completed, and the patient condition matching model is obtained. When the classification accuracy is less than the target classification accuracy, the rule learning model is continuously trained using the historical patient training set until the classification accuracy is greater than or equal to the target classification accuracy.

[0036] Further, the historical patient label data is obtained, specifically as follows:

[0037] An arbitrary sample patient is selected from the obtained plurality of historical patients, and a historical condition monitoring period is set for the sample patient;

[0038] The Z1 tumor marker examination performed by the target tumor patient within the historical condition monitoring period is obtained, obtaining a plurality of Z1 tumor marker historical examinations, and the examination result values corresponding to each Z1 tumor marker historical examination are obtained, and the average of the obtained plurality of examination result values is calculated to obtain a Z1 tumor marker analysis value;

[0039] The process of obtaining the Z1 tumor marker analysis value is repeated, and the examination result analysis is performed on the Z2 tumor marker to the Za tumor marker, and the Z2 tumor marker analysis value to the Za tumor marker analysis value is obtained according to the analysis result;

[0040] If the Z1 tumor marker analysis value to the Za tumor marker analysis value all correspond to the Z1 marker screening interval to the Za marker screening interval, the sample patient is divided into a condition matching consistent patient;

[0041] If any one of the Z1 tumor marker analysis value to the Za tumor marker analysis value does not correspond to the corresponding marker screening interval, the sample patient is divided into a condition matching inconsistent patient;

[0042] The process of dividing the sample patient into types is repeated, and each historical patient is divided to obtain the historical patient label data.

[0043] Further, the radiotherapy reaction prediction of the target tumor patient is as follows:

[0044] The target patient matching data is obtained, and a plurality of real-time condition consistent patients are obtained according to the target patient matching data;

[0045] The radiotherapy reactions of the plurality of real-time condition consistent patients after receiving radiotherapy are obtained, obtaining a plurality of different types of radiotherapy reactions, and an arbitrary sample radiotherapy reaction is selected from the obtained plurality of different types of radiotherapy reactions;

[0046] The sample radiotherapy reaction is analyzed by the number of occurrences, and the reaction occurrence proportion corresponding to the sample radiotherapy reaction is obtained according to the analysis result;

[0047] The process of obtaining the reaction occurrence proportion corresponding to the sample radiotherapy reaction is repeated, and the reaction occurrence proportion corresponding to each different type of radiotherapy reaction is obtained respectively, and a plurality of reaction occurrence proportions are obtained;

[0048] The obtained plurality of reaction occurrence proportions are compared in value size, the radiotherapy reaction corresponding to the maximum reaction occurrence proportion is set as the predicted radiotherapy reaction, and the predicted radiotherapy reaction is output.

[0049] Further, the reaction occurrence proportion corresponding to the sample radiotherapy reaction is obtained, and the specific process is as follows:

[0050] Among the plurality of real-time condition consistent patients, the number of patients who appear the sample radiotherapy reaction after receiving radiotherapy is counted to obtain a first characteristic patient quantity value, and the number of patients who do not appear the sample radiotherapy reaction after receiving radiotherapy is counted to obtain a second characteristic patient quantity value;

[0051] The first characteristic patient quantity value and the second characteristic patient quantity value are calculated to obtain the reaction occurrence proportion corresponding to the sample radiotherapy reaction;

[0052] The reaction occurrence proportion corresponding to the sample radiotherapy reaction is calculated.

[0053] The tumor radiotherapy reaction prediction method based on rule learning includes the following steps:

[0054] Step S1: The tumor marker of the target tumor patient is checked, the corresponding marker screening interval is set according to the checking result, and the patient condition cycle analysis data is obtained;

[0055] Step S2: A condition matching model is created according to the patient condition cycle analysis data, and the real-time condition consistent patients are matched for the target tumor patient through the condition matching model, and the target patient matching data is obtained;

[0056] Step S3: Radiotherapy reaction prediction is performed for the target tumor patient according to the target patient matching data.

[0057] As described above, due to the adoption of the above technical solutions, the present application has the following advantages:

[0058] 1、The present application is directed to a plurality of tumor marker tests for target tumor patients, and according to the test results of the target tumor patients in the medical index monitoring period, a corresponding marker screening interval is set for each tumor marker test, and the obtained marker screening interval is used as a model creation basis to create a disease matching model, thereby ensuring the training effect of the disease matching model and improving the accuracy of the tumor radiotherapy reaction prediction result.

[0059] 2、The present application creates a disease matching model according to the patient disease cycle analysis data, matches real-time disease consistent patients for the target tumor patients according to the disease matching model, analyzes the tumor radiotherapy reaction of the real-time disease consistent patients, and predicts the radiotherapy reaction of the target tumor patients according to the analysis result, thereby improving the tumor radiotherapy reaction prediction efficiency and the real-time of the prediction process. BRIEF DESCRIPTION OF DRAWINGS

[0060] In order to facilitate the understanding of those skilled in the art, the present application will be further described below in conjunction with the drawings.

[0061] Figure 1 The overall system block diagram of the present application is shown in the figure.

[0062] Figure 2 The implementation step diagram of the present application is shown in the figure. DETAILED DESCRIPTION

[0063] The technical solutions of the present application will be described below in conjunction with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0064] Embodiment one

[0065] Please refer to Figure 1 The disease matching model in the present application involves artificial intelligence technology. The present application provides a technical solution: a tumor radiotherapy reaction prediction system based on rule learning, which includes a disease analysis module, a data matching module, a reaction prediction module and a server. The disease analysis module, data matching module and reaction prediction module are connected with the server respectively, and the server controls the disease analysis module, data matching module and reaction prediction module respectively.

[0066] The disease analysis module performs tumor marker tests on target tumor patients, sets corresponding marker screening intervals for tumor markers according to the test results, and obtains patient disease cycle analysis data.

[0067] Specifically as follows:

[0068] Obtaining a plurality of tumor patients in need of tumor radiotherapy reaction prediction, and randomly selecting a target tumor patient from the obtained plurality of tumor patients;

[0069] It should be noted here that:

[0070] In this application, the target tumor types referred to here can be epithelial tissue tumors, mesenchymal tissue tumors, and neural cell tumors.

[0071] In the process of analyzing the condition of the target tumor patient, a plurality of different types of tumor markers examined by the target tumor patient are obtained, and the plurality of different types of tumor markers obtained are Z1 tumor marker to Za tumor marker;

[0072] It should be noted here that:

[0073] In this application, Z here refers to the symbol corresponding to the tumor marker, and a here refers to the number value corresponding to the tumor marker, and a is an integer greater than 0;

[0074] In this application, Z1 tumor marker referred to here can be alpha-fetoprotein, Z2 tumor marker referred to here can be carcinoembryonic antigen, and Z3 tumor marker referred to here can be carbohydrate antigen 125;

[0075] In the process of analyzing the condition of the target tumor patient, the time point when the target tumor patient is definitely diagnosed as a tumor patient is set as a first cycle characteristic time point, the time point corresponding to the current time is set as a second cycle characteristic time point, the intermediate time point between the first cycle characteristic time point and the second cycle characteristic time point is set as a third cycle characteristic time point, and the period between the third cycle characteristic time point and the second cycle characteristic time point is set as a medical index monitoring period.

[0076] The Z1 tumor marker of the target tumor patient in the medical index monitoring period is analyzed, and the Z1 marker screening interval is obtained according to the analysis result;

[0077] Specifically as follows:

[0078] The Z1 tumor marker examination of the target tumor patient is obtained, a plurality of Z1 tumor marker examinations are obtained, and the plurality of Z1 tumor marker examinations are renamed as X1 marker examination to Xb marker examination according to the order of examination;

[0079] It should be noted here that:

[0080] In the present application, X refers to the symbol of the Z1 tumor marker examination, and b refers to the number of Z1 tumor marker examinations.

[0081] The marker examination result values corresponding to the X1 marker examination to the Xb marker examination are obtained respectively, to obtain the X1 marker examination result value to the Xb marker examination result value;

[0082] The X1 marker examination result value to the Xb marker examination result value is subjected to average calculation to obtain the first marker characteristic value;

[0083] The X1 marker examination result value and the X2 marker examination result value are subjected to calculation to obtain the X1 result value change rate;

[0084] The X1 result value change rate is calculated, and the specific formula is as follows:

[0085] ;

[0086] Wherein, Blx1 is the X1 result value change rate, Zx1 is the X1 marker examination result value, and Zx2 is the X2 marker examination result value;

[0087] The X1 result value change rate is repeatedly calculated to obtain the result value change rate between any two continuous marker examination result values except the X1 marker examination result value and the X2 marker examination result value, to obtain the X2 result value change rate to the Xb-1 result value change rate;

[0088] The X1 result value change rate to the Xb-1 result value change rate is subjected to numerical value size comparison, and the result value change rate with the largest numerical value is set as the second marker characteristic value;

[0089] The product of the first marker characteristic value and the second marker characteristic value is calculated to obtain the third marker characteristic value;

[0090] The sum of the first marker characteristic value and the third marker characteristic value is calculated to obtain the upper limit of the Z1 marker screening interval, and the difference between the first marker characteristic value and the third marker characteristic value is calculated to obtain the lower limit of the Z1 marker screening interval. The numerical value interval composed of the upper limit of the Z1 marker screening interval and the lower limit of the Z1 marker screening interval is set as the Z1 marker screening interval;

[0091] The acquisition process of the Z1 marker screening interval is repeated to obtain the monitoring index analysis of the Z2 tumor marker to the Za tumor marker, and the Z2 marker screening interval to the Za marker screening interval is obtained according to the analysis result;

[0092] The Z1 marker screening interval to the Za marker screening interval is defined as the patient condition cycle analysis data;

[0093] The data matching module creates a disease matching model according to the patient disease cycle analysis data, and matches real-time disease consistent patients for the target tumor patient according to the disease matching model to obtain target patient matching data;

[0094] Specifically as follows:

[0095] Target patient matching data is obtained, and Z1 marker screening interval to Za marker screening interval is obtained according to the target patient matching data;

[0096] The tumor type corresponding to the target tumor patient is obtained, and a target type tumor is obtained. The historical admitted patients admitted by the medical institution for the target type tumor are obtained, and a plurality of historical admitted patients are obtained, and a disease matching model is created according to the historical admitted patients;

[0097] It should be noted here that:

[0098] In this application, the historical admitted patients referred to here are specifically admitted patients who have completed treatment;

[0099] In this application, if the tumor type suffered by the target tumor patient is an epithelial tissue tumor, the plurality of historical admitted patients obtained here are also epithelial tissue tumors;

[0100] Specifically as follows:

[0101] The plurality of historical admitted patients obtained are divided into disease matching consistent patients and disease matching inconsistent patients to obtain historical patient marking data;

[0102] Specifically as follows:

[0103] An arbitrary sample admitted patient is selected from the plurality of historical admitted patients, and a historical disease monitoring period is set for the sample admitted patient;

[0104] The Z1 tumor marker examination performed by the target tumor patient within the historical disease monitoring period is obtained to obtain a plurality of Z1 tumor marker historical examinations. The examination result value corresponding to each Z1 tumor marker historical examination is obtained, and the plurality of examination result values are averaged to obtain a Z1 tumor marker analysis value;

[0105] The process of obtaining the Z1 tumor marker analysis value is repeated, and the examination result analysis is performed on Z2 tumor marker to Za tumor marker. The Z2 tumor marker analysis value to the Za tumor marker analysis value is obtained according to the analysis result;

[0106] If the Z1 tumor marker analysis value to the Za tumor marker analysis value all correspond to the Z1 marker screening interval to the Za marker screening interval, the sample admitted patient is divided into the disease matching consistent patient;

[0107] If any one of the Z1 tumor marker analysis value to the Za tumor marker analysis value is not in the corresponding marker screening interval, the sample admitted patient is divided into the disease matching inconsistent patient;

[0108] The process of dividing the type of the sample admitted patient is repeated, and each historical admitted patient is divided to obtain historical patient marking data;

[0109] Specifically as follows:

[0110] The historical patient marking data is divided into a historical patient training set and a sample patient test set according to a preset training test ratio;

[0111] A rule learning model is created through an existing artificial intelligence platform, the rule learning model is classified and trained using the historical patient training set, and each historical admitted patient in the historical patient training set trains the rule learning model once;

[0112] The rule learning model is classified and tested using the sample patient test set, and the classification accuracy is obtained, when the classification accuracy is greater than or equal to the target classification accuracy, the rule learning model training is completed, and the patient disease matching model is obtained, when the classification accuracy is less than the target classification accuracy, the rule learning model is continuously trained using the historical patient training set, until the classification accuracy is greater than or equal to the target classification accuracy.

[0113] A plurality of real-time admitted patients with tumors of the target type admitted by the admitting institution are obtained;

[0114] It should be noted here that:

[0115] In this application, the admitting institution referred to here is the institution where the target tumor patient is treated, and the real-time admitted patient referred to here is the patient who is being treated in the institution;

[0116] The plurality of real-time admitted patients referred to here are all patients who have undergone reflex treatment.

[0117] Each real-time admitted patient is matched with the target tumor patient for disease consistency using the patient disease matching model, a plurality of real-time disease consistent patients are obtained, and the plurality of real-time disease consistent patients obtained are defined as target patient matching data;

[0118] The reaction prediction module performs radiotherapy reaction prediction for the target tumor patient according to the target patient matching data.

[0119] Specifically as follows:

[0120] Obtaining target patient matching data, and obtaining a plurality of real-time condition consistent patients according to the target patient matching data;

[0121] Obtaining a plurality of different types of radiotherapy reactions of the plurality of real-time condition consistent patients after receiving radiotherapy, and randomly selecting a sample radiotherapy reaction from the plurality of different types of radiotherapy reactions;

[0122] It should be noted here that:

[0123] The radiotherapy reaction referred to here can be nausea, diarrhea, and abdominal pain.

[0124] In the plurality of real-time condition consistent patients, the number of patients who have the sample radiotherapy reaction after receiving radiotherapy is counted to obtain a first characteristic patient number value, and the number of patients who do not have the sample radiotherapy reaction after receiving radiotherapy is counted to obtain a second characteristic patient number value;

[0125] The first characteristic patient number value and the second characteristic patient number value are calculated to obtain a reaction occurrence proportion corresponding to the sample radiotherapy reaction;

[0126] The reaction occurrence proportion corresponding to the sample radiotherapy reaction is calculated, specifically as follows:

[0127]

[0128] Wherein, Yfb is the reaction occurrence proportion corresponding to the sample radiotherapy reaction, Hst1 is the first characteristic patient number value, and Hst2 is the second characteristic patient number value.

[0129] The process of obtaining the reaction occurrence proportion corresponding to the sample radiotherapy reaction is repeated to obtain a reaction occurrence proportion corresponding to each type of radiotherapy reaction, and a plurality of reaction occurrence proportions are obtained.

[0130] The plurality of reaction occurrence proportions are compared in value size, and the radiotherapy reaction corresponding to the maximum reaction occurrence proportion is set as the predicted radiotherapy reaction, and the predicted radiotherapy reaction is output.

[0131] Embodiment two

[0132] Based on the same invention, another concept is proposed, that is, a tumor radiotherapy reaction prediction method based on rule learning, including the following steps:

[0133] Please refer to Figure 2 ​, step S1: tumor marker examination is performed on the target tumor patient, and a corresponding marker screening interval is set for the tumor marker according to the examination result, to obtain patient condition cycle analysis data;

[0134] The step S1 further includes the following steps:

[0135] A tumor patient in need of tumor radiotherapy reaction prediction is obtained, to obtain a plurality of tumor patients, and a target tumor patient is arbitrarily selected from the plurality of tumor patients;

[0136] It should be noted here that:

[0137] In this application, the target tumor types referred to here can be epithelial tissue tumors, mesenchymal tissue tumors, and neural cell tumors.

[0138] In the process of analyzing the condition of the target tumor patient, the tumor markers examined by the target tumor patient are obtained, to obtain a plurality of different types of tumor markers, and the plurality of different types of tumor markers are Z1 tumor markers to Za tumor markers;

[0139] It should be noted here that:

[0140] In this application, Z here refers to the symbol corresponding to the tumor marker, and a here refers to the number value corresponding to the tumor marker, and a is an integer greater than 0;

[0141] In this application, Z1 tumor marker referred to here can be alpha-fetoprotein, Z2 tumor marker referred to here can be carcinoembryonic antigen, and Z3 tumor marker referred to here can be carbohydrate antigen 125;

[0142] In the process of analyzing the condition of the target tumor patient, the time point when the target tumor patient is definitely diagnosed as a tumor patient is set as a first cycle characteristic time point, the time point corresponding to the current time is set as a second cycle characteristic time point, the intermediate time point between the first cycle characteristic time point and the second cycle characteristic time point is set as a third cycle characteristic time point, and the period between the third cycle characteristic time point and the second cycle characteristic time point is set as a medical index monitoring period;

[0143] The Z1 tumor marker of the target tumor patient in the medical index monitoring period is analyzed, and the Z1 marker screening interval is obtained according to the analysis result;

[0144] Specifically as follows:

[0145] Obtaining the Z1 tumor marker examination of the target tumor patient, obtaining multiple Z1 tumor marker examinations, and renaming the multiple Z1 tumor marker examinations as X1 marker examination to Xb marker examination according to the order of the examinations;

[0146] It should be noted here that:

[0147] In this application, X here refers to the symbol of the Z1 tumor marker examination, and b here refers to the number of examinations corresponding to the Z1 tumor marker examination.

[0148] Obtaining the marker examination result values corresponding to the X1 marker examination to the Xb marker examination, obtaining the X1 marker examination result value to the Xb marker examination result value;

[0149] Calculating the average of the X1 marker examination result value to the Xb marker examination result value, obtaining the first marker characteristic value;

[0150] Calculating the X1 result value change rate by the X1 marker examination result value and the X2 marker examination result value;

[0151] Calculating the X1 result value change rate, and the specific formula is as follows:

[0152] ;

[0153] Wherein, Blx1 is the X1 result value change rate, Zx1 is the X1 marker examination result value, and Zx2 is the X2 marker examination result value;

[0154] Repeating the calculation of the X1 result value change rate, obtaining the result value change rate between any two continuous marker examination result values except the X1 marker examination result value and the X2 marker examination result value, obtaining the X2 result value change rate to the Xb-1 result value change rate;

[0155] Comparing the numerical values of the X1 result value change rate to the Xb-1 result value change rate, setting the result value change rate with the largest numerical value as the second marker characteristic value;

[0156] Calculating the product of the first marker characteristic value and the second marker characteristic value, obtaining the third marker characteristic value;

[0157] Calculating the sum of the first marker characteristic value and the third marker characteristic value, obtaining the upper limit of the Z1 marker screening interval, calculating the difference between the first marker characteristic value and the third marker characteristic value, obtaining the lower limit of the Z1 marker screening interval, and setting the numerical interval composed of the upper limit of the Z1 marker screening interval and the lower limit of the Z1 marker screening interval as the Z1 marker screening interval;

[0158] Repeating the acquisition process of the Z1 marker screening interval, respectively acquiring Z2 tumor markers to Za tumor markers for monitoring index analysis, and acquiring Z2 marker screening interval to Za marker screening interval according to the analysis result;

[0159] Defining the Z1 marker screening interval to the Za marker screening interval as the patient condition cycle analysis data;

[0160] It should be noted here that:

[0161] The above step S1 sets multiple tumor marker checks for the target tumor patient, sets the corresponding marker screening interval for each tumor marker check according to the check results of the target tumor patient in the medical index monitoring period, and creates a condition matching model based on the acquired marker screening interval as the model creation basis. By constructing the condition matching model and matching the real-time condition consistent patients, a dynamic individualized reference group can be formed. This process is based on patient condition cycle analysis data, ensuring that the reference group and the target patient have high similarity in terms of tumor characteristics, disease development, etc. Thus, accurate data basis is provided for subsequent analysis. By analyzing the radiotherapy response of these patients, actual treatment data can be accumulated, and the potential correlation between radiotherapy response and disease characteristics can be revealed, providing a basis for predicting the response of the target patient. Ultimately, based on the analysis results, the radiotherapy response of the target patient is predicted, which can adjust the treatment plan in advance, select the most suitable radiotherapy dose and method, and achieve personalized treatment. This not only improves the accuracy of treatment, but also optimizes the allocation of medical resources and improves the overall medical efficiency, providing more effective radiotherapy support for patients.

[0162] Step S2: creating a condition matching model according to the patient condition cycle analysis data, and matching real-time condition consistent patients for the target tumor patient through the condition matching model to obtain target patient matching data;

[0163] The step S2 further includes the following steps:

[0164] Obtaining target patient matching data, and respectively acquiring Z1 marker screening interval to Za marker screening interval according to the target patient matching data;

[0165] Obtaining the tumor type corresponding to the target tumor patient to obtain the target type tumor, obtaining the historical admitted patients admitted by the admitting institution who have the target type tumor to obtain a plurality of historical admitted patients, and creating a condition matching model according to the historical admitted patients;

[0166] It should be noted here that:

[0167] In this application, the historical admitted patients referred to here specifically refer to admitted patients who have completed treatment;

[0168] In the present application, if the tumor type of the target tumor patient is an epithelial tumor, the plurality of historical patients obtained here are also epithelial tumors;

[0169] Specifically as follows:

[0170] The plurality of historical patients obtained are divided into patients with consistent disease conditions and patients with inconsistent disease conditions to obtain historical patient marking data;

[0171] Specifically as follows:

[0172] An arbitrary sample patient is selected from the plurality of historical patients, and a historical disease condition monitoring period is set for the sample patient;

[0173] The Z1 tumor marker examination performed by the target tumor patient within the historical disease condition monitoring period is obtained to obtain a plurality of Z1 tumor marker historical examinations, the examination result value corresponding to each Z1 tumor marker historical examination is obtained, and the average of the plurality of examination result values is calculated to obtain a Z1 tumor marker analysis value;

[0174] The process of obtaining the Z1 tumor marker analysis value is repeated, and the examination result analysis is performed on the Z2 tumor marker to the Za tumor marker, and the Z2 tumor marker analysis value to the Za tumor marker analysis value is obtained according to the analysis result;

[0175] If the Z1 tumor marker analysis value to the Za tumor marker analysis value all correspond to the Z1 marker screening interval to the Za marker screening interval, the sample patient is divided into a patient with consistent disease conditions;

[0176] If any one of the Z1 tumor marker analysis value to the Za tumor marker analysis value is not in the corresponding marker screening interval, the sample patient is divided into a patient with inconsistent disease conditions;

[0177] The process of dividing the sample patient into types is repeated, and each historical patient is divided to obtain historical patient marking data;

[0178] Specifically as follows:

[0179] The historical patient marking data is divided into a historical patient training set and a sample patient test set according to a preset training test ratio;

[0180] A rule learning model is created through an existing artificial intelligence platform, the rule learning model is classified and trained using the historical patient training set, and each historical patient in the historical patient training set is trained once until the rule learning model is trained;

[0181] The rule learning model is classified and tested using the sample patient test set, and the classification accuracy is obtained. When the classification accuracy is greater than or equal to the target classification accuracy, the rule learning model training is completed, and the patient condition matching model is obtained. When the classification accuracy is less than the target classification accuracy, the rule learning model is continuously trained using the historical patient training set until the classification accuracy is greater than or equal to the target classification accuracy.

[0182] The real-time admitted patients admitted by the admitting institution for the target type of tumor are obtained, and a plurality of real-time admitted patients are obtained.

[0183] It should be noted here that:

[0184] In this application, the admitting institution referred to here is the institution where the target tumor patient is treated, and the real-time admitted patient referred to here is the patient who is being treated in the admitting institution.

[0185] The plurality of real-time admitted patients referred to here are patients who have undergone reflex treatment.

[0186] Each real-time admitted patient is matched with the target tumor patient for condition consistency using the patient condition matching model, a plurality of real-time condition consistent patients are obtained, and the obtained plurality of real-time condition consistent patients are defined as target patient matching data.

[0187] Step S2 sets a plurality of tumor marker examinations for the target tumor patient, sets a corresponding marker screening interval for each tumor marker examination according to the examination results of the target tumor patient within the medical index monitoring period, can provide a multi-dimensional dynamic monitoring framework for condition assessment, and can form an individualized condition feature map by integrating the change trend of different markers, avoiding the limitations of a single index.

[0188] Step S3: Radiotherapy reaction prediction is performed for the target tumor patient according to the target patient matching data.

[0189] The step S3 further includes the following steps:

[0190] The target patient matching data is obtained, and a plurality of real-time condition consistent patients are obtained according to the target patient matching data.

[0191] The radiotherapy reactions of the plurality of real-time condition consistent patients after receiving radiotherapy are obtained, a plurality of different types of radiotherapy reactions are obtained, and a sample radiotherapy reaction is arbitrarily selected from the obtained plurality of different types of radiotherapy reactions.

[0192] It should be noted here that:

[0193] The radiotherapy reaction referred to here can be nausea, diarrhea, and abdominal pain.

[0194] In the plurality of real-time condition consistent patients, the number of patients who have sample radiotherapy reactions after receiving radiotherapy is counted to obtain a first characteristic patient number value, and the number of patients who do not have sample radiotherapy reactions after receiving radiotherapy is counted to obtain a second characteristic patient number value;

[0195] The first characteristic patient number value and the second characteristic patient number value are calculated to obtain a reaction occurrence proportion corresponding to the sample radiotherapy reaction;

[0196] The reaction occurrence proportion corresponding to the sample radiotherapy reaction is calculated, specifically as follows:

[0197] ;

[0198] Wherein, Yfb is the reaction occurrence proportion corresponding to the sample radiotherapy reaction, Hst1 is the first characteristic patient number value, and Hst2 is the second characteristic patient number value;

[0199] The process of obtaining the reaction occurrence proportion corresponding to the sample radiotherapy reaction is repeated to obtain a plurality of reaction occurrence proportions corresponding to different types of radiotherapy reactions;

[0200] The plurality of reaction occurrence proportions are compared in value size, and the radiotherapy reaction corresponding to the maximum reaction occurrence proportion is set as a predicted radiotherapy reaction, and the predicted radiotherapy reaction is output.

[0201] The preferred embodiments disclosed above are only used to help explain the present application. The preferred embodiments do not describe all the details, nor limit the present application to the specific embodiments. Obviously, according to the content of the present application, many modifications and changes can be made. The present application is selected and described in detail to better explain the principles and practical applications of the present application, so that those skilled in the art can well understand and utilize the present application. The present application is limited by the claims and their entire scope and equivalents.

Claims

1. A tumor radiotherapy response prediction system based on rule learning, characterized in that, include: Disease Analysis Module: Perform tumor marker tests on target tumor patients, set corresponding marker screening intervals based on the test results, and obtain patient disease cycle analysis data; Data matching module: Creates a disease matching model based on the analysis data of the patient's disease cycle, and matches the target tumor patient with patients whose real-time disease conditions are consistent with the model to obtain the target patient matching data; Response prediction module: Predicts radiotherapy response for target tumor patients based on target patient matching data; Specifically, the acquisition of patient disease cycle analysis data is as follows: We acquire multiple tumor patients who require prediction of tumor radiotherapy response, and then randomly select one target tumor patient from among the multiple acquired tumor patients. During the analysis of the disease status of patients with target tumors, Z1 tumor markers to Za tumor markers and medical indicator monitoring cycles are set for patients with target tumors. Z1 tumor markers were analyzed for target tumor patients in the medical indicator monitoring cycle, and the Z1 marker screening interval was obtained based on the analysis results. Specifically, the Z1 marker screening interval is obtained as follows: Z1 tumor marker tests were obtained from patients with the target tumor, resulting in multiple Z1 tumor marker tests. Based on the order of the tests, the multiple Z1 tumor marker tests were renamed from X1 marker tests to Xb marker tests. The detection results of X1 marker to Xb marker are analyzed, and the characteristic values ​​of the first marker and the second marker are obtained based on the analysis results. Calculate the product of the first marker feature value and the second marker feature value to obtain the third marker feature value; Calculate the sum of the first marker feature value and the third marker feature value to obtain the upper limit of the Z1 marker screening interval. Calculate the difference between the first marker feature value and the third marker feature value to obtain the lower limit of the Z1 marker screening interval. Set the numerical interval formed by the upper limit of the Z1 marker screening interval and the lower limit of the Z1 marker screening interval as the Z1 marker screening interval. Specifically, the feature values ​​of the first and second markers are obtained as follows: Obtain the marker inspection result values ​​corresponding to the X1 marker inspection to the Xb marker inspection respectively, and obtain the X1 marker inspection result value to the Xb marker inspection result value; The average value of the first marker is calculated by averaging the inspection results of markers X1 to Xb. The rate of change of the X1 marker test result value is obtained by calculating the X1 marker test result value and the X2 marker test result value; Similarly, the rate of change of the X2 result value to the rate of change of the Xb-1 ​​result value are obtained; Compare the numerical values ​​of the change rates from X1 to Xb-1, and set the change rate of the result value with the largest value as the feature value of the second marker. The prediction of radiotherapy response in patients with target tumors is as follows: Obtain target patient matching data, and based on the target patient matching data, obtain multiple patients with consistent real-time conditions; The radiotherapy reactions of multiple patients with consistent real-time conditions after receiving radiotherapy were obtained, resulting in a variety of different types of radiotherapy reactions. A sample radiotherapy reaction was randomly selected from the various types of radiotherapy reactions obtained. Analyze the frequency of radiotherapy reactions in the samples, and obtain the percentage of each reaction based on the analysis results; Obtain the percentage of occurrence of each different type of radiotherapy reaction to obtain multiple percentages of reactions. The percentages of the obtained multiple reactions are compared numerically, and the radiotherapy reaction corresponding to the largest percentage of the reaction is set as the predicted radiotherapy reaction. The predicted radiotherapy reaction is then output.

2. The tumor radiotherapy response prediction system based on rule learning according to claim 1, characterized in that, The target patient matching data was obtained as follows: Obtain target patient matching data, and obtain Z1 marker screening interval to Za marker screening interval based on the target patient matching data; The tumor type corresponding to the target tumor patient is obtained, the target tumor type is obtained, the historical patients with the target tumor type treated by the treatment institution are obtained, multiple historical patients are obtained, and a disease matching model is created based on the historical patients. The system retrieves real-time patient data of patients with the target type of tumor who are being treated at the medical institution, resulting in multiple real-time patient data. The patient condition matching model is used to match each real-time patient with the target tumor patient to obtain multiple real-time patients with consistent conditions. These multiple real-time patients with consistent conditions are defined as the target patient matching data.

3. The tumor radiotherapy response prediction system based on rule learning according to claim 2, characterized in that, The disease matching model is created as follows: The acquired historical patients were divided into patients with consistent disease conditions and patients with inconsistent disease conditions, thus obtaining historical patient labeling data. Historical patient labeled data is divided into a historical patient training set and a sample patient test set according to a preset training-test ratio; Create a rule learning model using an existing artificial intelligence platform, and use a historical patient training set to classify and train the rule learning model until each historical patient in the historical patient training set has trained the rule learning model once. The rule learning model is tested using a sample patient test set, and the classification accuracy is obtained. When the classification accuracy is greater than or equal to the target classification accuracy, the rule learning model is trained and a patient condition matching model is obtained. When the classification accuracy is less than the target classification accuracy, the rule learning model is trained again using the historical patient training set until the classification accuracy is greater than or equal to the target classification accuracy.

4. The tumor radiotherapy response prediction system based on rule learning according to claim 3, characterized in that, The historical patient-labeled data was obtained as follows: Randomly select one sample patient from the multiple historical patients obtained, and set a historical disease monitoring period for the sample patient. The Z1 tumor marker tests performed on the target tumor patients during the historical disease monitoring period were obtained, resulting in multiple historical Z1 tumor marker tests. The test result values ​​corresponding to each historical Z1 tumor marker test were obtained, and the average of the multiple test result values ​​was calculated to obtain the Z1 tumor marker analysis value. The results of the tests on tumor markers Z2 to Za were analyzed separately, and the analytical values ​​of tumor markers Z2 to Za were obtained based on the analysis results. If the Z1 tumor marker analysis value to the Za tumor marker analysis value all fall within the Z1 marker screening interval to the Za marker screening interval, then the patients in the sample will be classified as patients with consistent disease conditions. If any tumor marker analysis value from Z1 to Za is not within the corresponding marker screening range, then the patients in the sample will be classified as patients with inconsistent disease matching. Each historical patient was segmented to obtain historical patient labeling data.

5. The tumor radiotherapy response prediction system based on rule learning according to claim 1, characterized in that, The percentage of reactions corresponding to radiotherapy responses in the samples was obtained, as follows: Among multiple patients with consistent real-time conditions, the number of patients who showed a sample radiotherapy response after receiving radiotherapy was counted to obtain the number of patients with the first characteristic, and the number of patients who did not show a sample radiotherapy response after receiving radiotherapy was counted to obtain the number of patients with the second characteristic. The percentage of reactions corresponding to radiotherapy responses in the sample was obtained by calculating the number of patients with the first characteristic and the number of patients with the second characteristic. The percentage of reactions corresponding to radiotherapy responses in the samples was calculated.

Images