Method and System for Constructing a Prediction Model for the Effect of Tumor Chemotherapy Based on Artificial Intelligence

Abstract

The present invention provides a method and system for constructing a tumor chemotherapy effect prediction model based on artificial intelligence. Based on the data type, the feature extraction method is determined, and the target tumor manifestation data is subjected to feature extraction according to the feature extraction method to obtain tumor manifestation features, providing accurate training samples for the training of the model. Based on the data acquisition time, the target tumor manifestation features are divided into training features and verification features, and an initial tumor chemotherapy effect prediction model is trained based on the training features, providing a prediction model for tumor chemotherapy effect prediction. Finally, the initial tumor chemotherapy effect prediction model is verified and optimized based on the verification features to obtain a target tumor chemotherapy effect prediction model, ensuring the prediction accuracy of the model, ensuring the evaluation effect of tumor chemotherapy effect evaluation in terms of timeliness, objectivity, standardization, index integrity, etc., ultimately improving the treatment efficiency of tumor patients and ensuring the treatment effect.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and particularly to a method and system for constructing a tumor chemotherapy effect prediction model based on artificial intelligence. Background Art

[0002] The methods for evaluating the effect of tumor chemotherapy mainly include the following:

[0003] 1. Imaging evaluation: Evaluating the tumor treatment effect through imaging examinations such as CT and MRI, including changes in tumor size, shape, etc.

[0004] 2. Hematological evaluation: Evaluating the impact of chemotherapy on tumors through hematological examinations, such as plasma tumor markers, peripheral blood cell counts, etc.

[0005] 3. Histological evaluation: Evaluating the killing effect of chemotherapy on tumor cells through tumor tissue biopsy.

[0006] 4. Biochemical index evaluation: Comprehensively evaluating the chemotherapy effect by measuring the biochemical indexes of patients, such as serum tumor markers, inflammatory factors, etc.

[0007] 5. Cytological evaluation: Evaluating the damage of chemotherapy to the bone marrow and the impact on hematopoietic cells such as white blood cells and red blood cells through cytological examinations, such as bone marrow cell morphology and classification.

[0008] 6. Patient quality of life evaluation: Comprehensively evaluating the impact of chemotherapy on the patient's quality of life through the patient's quality of life evaluation questionnaire.

[0009] The above methods for evaluating the effect of tumor chemotherapy have the following defects:

[0010] 1. Some evaluation methods require waiting for a period of time. For example, imaging evaluation needs to wait for a period of time after treatment to observe the effect, resulting in a time delay in the evaluation results.

[0011] 2. Some evaluation methods may be subjective. For example, the quality of life evaluation questionnaire may be affected by the patient's subjective feelings and is not objective enough.

[0012] 3. Different evaluation methods may use different criteria and indicators, lacking a unified evaluation standard, resulting in poor comparability of evaluation results.

[0013] 4. The indicators of some evaluation methods may be affected by other factors. For example, plasma tumor markers may be affected by factors such as inflammation and other diseases, affecting the accuracy of the evaluation results.

[0014] 5. A single evaluation method may not be able to comprehensively reflect the overall effect of chemotherapy, and multiple evaluation methods need to be combined for evaluation.

[0015] Generally speaking, the existing methods for evaluating the effect of tumor chemotherapy have certain limitations and defects in terms of timeliness, objectivity, standardization, and index integrity. Summary of the Invention

[0016] The present invention provides a method and system for constructing a tumor chemotherapy effect prediction model based on artificial intelligence to solve the problems raised in the background technology.

[0017] A method for constructing a tumor chemotherapy effect prediction model based on artificial intelligence includes:

[0018] S1: Collect different types of tumor manifestation data from tumor patients, and based on the data type of the tumor manifestation data, match the corresponding data processing method to clean the tumor manifestation data to obtain the target tumor manifestation data;

[0019] S2: Based on the data type, determine the feature extraction method, extract features from the target tumor manifestation data according to the feature extraction method to obtain tumor manifestation features, and select features from the tumor manifestation features based on the model input features to obtain the target tumor manifestation features;

[0020] S3: Based on the data acquisition time, divide the target tumor manifestation features into training features and verification features, and train an initial tumor chemotherapy effect prediction model based on the training features;

[0021] S4: Verify and optimize the initial tumor chemotherapy effect prediction model based on the verification features to obtain the target tumor chemotherapy effect prediction model.

[0022] Preferably, in S1, based on the data type of the tumor manifestation data, matching the corresponding data processing method to clean the tumor manifestation data to obtain the target tumor manifestation data includes:

[0023] Based on the data type of the tumor manifestation data, match the data processing method from the data processing method library, and obtain the processing range index and processing numerical index that need to be specifically determined in the processing method;

[0024] Obtain the target data features of the data type of the tumor manifestation data, and based on the target data features, set corresponding specific values for the processing range index and the processing numerical index;

[0025] Based on the data processing method and the specific values, set the data cleaning rules, and clean the tumor manifestation data according to the data cleaning rules to obtain the target tumor manifestation data.

[0026] Preferably, in S2, based on the data type, determining the feature extraction method includes:

[0027] Based on the data type, determine the recognition method for the data content, establish the data statistics rule for the target tumor manifestation data based on the recognition method, and also establish the data dimensionality reduction rule for the target tumor manifestation data based on the recognition method;

[0028] Based on the data statistics rule and the data dimensionality reduction rule, generate the initial feature extraction method for the target tumor manifestation data;

[0029] Based on the preset model to analyze the data format standard, determine the format requirements for the target tumor manifestation data, and optimize the initial feature extraction method in terms of the extracted data format based on the format requirements to obtain the final feature extraction method.

[0030] Preferably, in S2, perform feature extraction on the target tumor manifestation data according to the feature extraction method to obtain tumor manifestation features, including:

[0031] Based on the feature extraction method, determine the marking positions for the target tumor manifestation data, and segment the target tumor manifestation data according to the marking positions to obtain multiple data segments;

[0032] Perform data feature extraction on the multiple data segments according to the feature extraction method, and integrate the extracted features to obtain tumor manifestation features.

[0033] Preferably, in S2, based on the model input features, select features from the tumor manifestation features to obtain the target tumor manifestation features, including:

[0034] Based on the model input features, determine the standard input information volume, and based on the standard input information volume, determine the selection index;

[0035] Based on the selection index, determine the index feature values of the tumor manifestation features, and perform information volume mapping on the index feature values to obtain the actual information volume of the tumor manifestation features;

[0036] Obtain the first manifestation features with an appearance rate greater than the preset threshold from the tumor manifestation features, and regard other tumor manifestation features as the second manifestation features. Assign the first weight to the first manifestation features and the second weight to the second manifestation features;

[0037] Based on the actual information volume and the first weight, determine the first score value of the first manifestation features, and based on the actual information volume and the second weight, determine the second score value of the second manifestation features;

[0038] Select the tumor manifestation features with the first score value and the second score value greater than the preset score value as the first target manifestation features;

[0039] Obtain the performance features to be analyzed whose second score value obtained from the second performance feature is less than or equal to the preset score value but greater than the preset minimum score value, determine the correlation between the performance features to be analyzed and the first performance feature based on the actual information volume, and select the performance features to be analyzed with a correlation greater than the preset correlation as the second target performance features;

[0040] Take the first target performance feature and the second target performance feature as the target tumor performance features.

[0041] Preferably, in S3, based on the data acquisition time, divide the target tumor performance features into training features and verification features, including:

[0042] Take the target tumor performance features obtained before the preset time as training features;

[0043] Take the target tumor performance features obtained after the preset time as verification features.

[0044] Preferably, in S3, train an initial tumor chemotherapy effect prediction model based on the training features, including:

[0045] Based on the feature type, perform the first label annotation on the training features, based on the feature content, perform the second label annotation on the training features, and based on the tumor chemotherapy effect results corresponding to the training features, perform the third label annotation on the training features;

[0046] Based on the first label annotation, divide the training features into multiple feature training groups, based on the second label annotation and the first label annotation, input the feature training groups into an artificial neural network model for training, and obtain an initial tumor chemotherapy effect prediction model according to the training results.

[0047] Preferably, in S4, verify and optimize the initial tumor chemotherapy effect prediction model based on the verification features to obtain a target tumor chemotherapy effect prediction model, including:

[0048] Based on the feature type, divide the verification features into multiple first verification feature groups, and based on the tumor chemotherapy effect features, divide the verification features into multiple second verification feature groups;

[0049] Extract the parts with the same features from the first verification feature group and the second verification feature group to form a third verification feature group;

[0050] Based on the first verification feature group and the third verification feature group, the initial tumor chemotherapy effect prediction model is verified to obtain the first accuracy rate. Based on the second verification feature group and the third verification feature group, the initial tumor chemotherapy effect prediction model is verified to obtain the second accuracy rate. Based on the third verification feature group, the initial tumor chemotherapy effect prediction model is verified to obtain the third accuracy rate;

[0051] Based on the difference between the first accuracy rate and the third accuracy rate, the feature type prediction difference is determined. Based on the difference between the second accuracy rate and the third accuracy rate, the tumor chemotherapy effect feature prediction difference is determined;

[0052] The learning network layer regarding the feature type in the initial tumor chemotherapy effect prediction model is obtained. Based on the feature type prediction difference, the weight parameters of the learning network layer are optimized to obtain the first optimization result;

[0053] Based on the feature type prediction difference, the weight parameters of all the learning network layers in the initial tumor chemotherapy effect prediction model are optimized to obtain the second optimization result;

[0054] The optimization parameters related to the first optimization result are obtained from the second optimization result. Based on the first optimization result, the optimization parameters are adjusted to obtain the target optimization parameters;

[0055] Based on the second optimization result and the target optimization parameters, a target tumor chemotherapy effect prediction model is generated.

[0056] Preferably, the step of obtaining the optimization parameters related to the first optimization result from the second optimization result, and adjusting the optimization parameters based on the first optimization result to obtain the target optimization parameters includes:

[0057] Determine whether the parameter difference between the optimization parameters and the optimization parameters in the first optimization result is within a preset range;

[0058] If so, use the optimization parameters as the target optimization parameters;

[0059] Otherwise, based on the parameter difference, match an optimization weight for the optimization parameters, and based on the optimization weight, obtain the target optimization parameters.

[0060] Preferably, it includes:

[0061] A data collection and processing module, configured to collect different types of tumor manifestation data from tumor patients, and based on the data type of the tumor manifestation data, match the corresponding data processing method to clean the tumor manifestation data to obtain the target tumor manifestation data;

[0062] A feature extraction module, configured to determine a feature extraction method based on the data type, extract features from the target tumor manifestation data according to the feature extraction method to obtain tumor manifestation features, and select features from the tumor manifestation features based on the model input features to obtain target tumor manifestation features;

[0063] A model training module, configured to divide the target tumor manifestation features into training features and verification features based on the data acquisition time, and train an initial tumor chemotherapy effect prediction model based on the training features;

[0064] A model optimization module, configured to verify and optimize the initial tumor chemotherapy effect prediction model based on the verification features to obtain a target tumor chemotherapy effect prediction model.

[0065] Compared with the prior art, the present invention has achieved the following beneficial effects:

[0066] By collecting different types of tumor manifestation data from tumor patients, and based on the data type of the tumor manifestation data, matching the corresponding data processing method to clean the tumor manifestation data to obtain the target tumor manifestation data, providing a comprehensive and accurate data basis for the training of the model. Based on the data type, determine the feature extraction method, extract features from the target tumor manifestation data according to the feature extraction method to obtain tumor manifestation features, and select features from the tumor manifestation features based on the model input features to obtain target tumor manifestation features, providing accurate training samples for the training of the model. Based on the data acquisition time, divide the target tumor manifestation features into training features and verification features, and train an initial tumor chemotherapy effect prediction model based on the training features, providing a prediction model for predicting the tumor chemotherapy effect. Finally, verify and optimize the initial tumor chemotherapy effect prediction model based on the verification features to obtain a target tumor chemotherapy effect prediction model, ensuring the prediction accuracy of the model, ensuring the evaluation effect of the tumor chemotherapy effect in terms of timeliness, objectivity, standardization, index integrity, etc., and ultimately improving the treatment efficiency of tumor patients and ensuring the treatment effect.

[0067] Other features and advantages of the present invention will be described in the following specification, and, in part, will be obvious from the specification, or will be understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained by the structures specifically pointed out in this application document.

[0068] The technical solutions of the present invention will be further described in detail below through the accompanying drawings and embodiments. Description of the Drawings

[0069] The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention. In the accompanying drawings:

[0070] Figure 1 This is a flowchart of a method for constructing an artificial intelligence-based tumor chemotherapy effect prediction model in an embodiment of the present invention;

[0071] Figure 2 This is a flowchart of obtaining target tumor manifestation data in an embodiment of the present invention;

[0072] Figure 3 This is a structural diagram of an artificial intelligence-based tumor chemotherapy effect prediction model construction system in an embodiment of the present invention. Detailed implementation manners

[0073] The following describes the preferred embodiments of the present invention with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only for explaining and illustrating the present invention, and are not used to limit the present invention.

[0074] Embodiment 1:

[0075] An embodiment of the present invention provides a method for constructing an artificial intelligence-based tumor chemotherapy effect prediction model, as Figure 1 shown, including:

[0076] S1: Collect different types of tumor manifestation data from tumor patients, and based on the data types of the tumor manifestation data, match the corresponding data processing methods to clean the tumor manifestation data to obtain target tumor manifestation data;

[0077] S2: Based on the data types, determine the feature extraction method, extract features from the target tumor manifestation data according to the feature extraction method to obtain tumor manifestation features, and select features from the tumor manifestation features based on the model input features to obtain target tumor manifestation features;

[0078] S3: Based on the data acquisition time, divide the target tumor manifestation features into training features and verification features, and train an initial tumor chemotherapy effect prediction model based on the training features;

[0079] S4: Verify and optimize the initial tumor chemotherapy effect prediction model based on the verification features to obtain a target tumor chemotherapy effect prediction model.

[0080] In this embodiment, different types of tumor manifestation data from tumor patients include different types of tumor manifestation data from tumor patients.

[0081] The beneficial effects of the above design solution are as follows: By collecting different types of tumor manifestation data from tumor patients, and based on the data types of the tumor manifestation data, matching corresponding data processing methods to clean the tumor manifestation data, target tumor manifestation data is obtained, providing a comprehensive and accurate data basis for the training of the model. Based on the data types, a feature extraction method is determined, and the target tumor manifestation data is subjected to feature extraction according to the feature extraction method to obtain tumor manifestation features. Based on the model input features, feature selection is performed from the tumor manifestation features to obtain target tumor manifestation features, providing accurate training samples for the training of the model. Based on the data acquisition time, the target tumor manifestation features are divided into training features and verification features, and an initial tumor chemotherapy effect prediction model is trained based on the training features, providing a prediction model for tumor chemotherapy effect prediction. Finally, the initial tumor chemotherapy effect prediction model is verified and optimized based on the verification features to obtain a target tumor chemotherapy effect prediction model, ensuring the prediction accuracy of the model, ensuring the evaluation effect of tumor chemotherapy effect evaluation in terms of timeliness, objectivity, standardization, index integrity, etc., and ultimately improving the treatment efficiency of tumor patients and ensuring the treatment effect.

[0082] Example 2:

[0083] Based on Example 1, an embodiment of the present invention provides a method for constructing a tumor chemotherapy effect prediction model based on artificial intelligence, as Figure 2 shown. In S1, based on the data types of the tumor manifestation data, matching corresponding data processing methods to clean the tumor manifestation data to obtain target tumor manifestation data includes:

[0084] Based on the data types of the tumor manifestation data, matching data processing methods from the data processing method library, and obtaining the processing range indicators and processing numerical indicators that need to be specifically determined in the processing method;

[0085] Obtaining the target data features of the data types of the tumor manifestation data, and setting corresponding specific values for the processing range indicators and processing numerical indicators based on the target data features;

[0086] Setting data cleaning rules based on the data processing method and specific values, and cleaning the tumor manifestation data according to the data cleaning rules to obtain target tumor manifestation data.

[0087] The beneficial effects of the above design solution are as follows: Providing appropriate data cleaning rules, improving the quality of the obtained target tumor manifestation data, and providing a data basis for model evaluation.

[0088] Example 3:

[0089] Based on Embodiment 1, an embodiment of the present invention provides a method for constructing a tumor chemotherapy effect prediction model based on artificial intelligence. In S2, based on the data type, the feature extraction method is determined, including:

[0090] Based on the data type, determine the recognition method for the data content, and based on the recognition method, establish the data statistics rule for the target tumor manifestation data, and also establish the data dimensionality reduction rule for the target tumor manifestation data;

[0091] Based on the data statistics rule and the data dimensionality reduction rule, generate the initial feature extraction method for the target tumor manifestation data;

[0092] Based on the preset model analysis data format standard, determine the format requirements for the target tumor manifestation data, and based on the format requirements, optimize the initial feature extraction method in terms of the extracted data format to obtain the final feature extraction method.

[0093] The beneficial effect of the above design is that by determining the feature extraction method in combination with the data type in terms of the data content and data format, the extraction efficiency and accuracy of the obtained feature extraction method are guaranteed.

[0094] Embodiment 4:

[0095] Based on Embodiment 1, an embodiment of the present invention provides a method for constructing a tumor chemotherapy effect prediction model based on artificial intelligence, characterized in that in S2, the target tumor manifestation data is subjected to feature extraction according to the feature extraction method to obtain tumor manifestation features, including:

[0096] Based on the feature extraction method, determine the marking positions of the target tumor manifestation data, and segment the target tumor manifestation data according to the marking positions to obtain multiple data segments;

[0097] Perform data feature extraction on the multiple data segments according to the feature extraction method, and integrate the extracted features to obtain tumor manifestation features.

[0098] The beneficial effect of the above design is that by performing data feature extraction on the multiple data segments according to the feature extraction method and integrating the extracted features to obtain tumor manifestation features, the integrity and accuracy of the obtained tumor manifestation features are guaranteed.

[0099] Embodiment 5:

[0100] Based on Embodiment 1, an embodiment of the present invention provides a method for constructing a tumor chemotherapy effect prediction model based on artificial intelligence. In S2, based on the model input features, feature selection is performed from the tumor manifestation features to obtain the target tumor manifestation features, including:

[0101] Based on the model input features, determine the standard input information amount, and based on the standard input information amount, determine the selection index;

[0102] Based on the selection index, determine the index eigenvalue of the tumor manifestation feature, and perform information amount mapping on the index eigenvalue to obtain the actual information amount of the tumor manifestation feature;

[0103] Obtain the first manifestation features with an occurrence rate greater than a preset threshold from the tumor manifestation features, and use other tumor manifestation features as the second manifestation features. Assign a first weight to the first manifestation features and a second weight to the second manifestation features;

[0104] Based on the actual information amount and the first weight, determine the first score value of the first manifestation features, and based on the actual information amount and the second weight, determine the second score value of the second manifestation features;

[0105] Select the tumor manifestation features with the first score value and the second score value greater than a preset score value as the first target manifestation features;

[0106] Obtain the to-be-analyzed manifestation features with the second score value less than or equal to the preset score value but greater than the preset lowest score value from the second manifestation features, and based on the actual information amount, determine the correlation degree between the to-be-analyzed manifestation features and the first manifestation features. Select the to-be-analyzed manifestation features with a correlation degree greater than the preset correlation degree as the second target manifestation features;

[0107] Use the first target manifestation features and the second target manifestation features as the target tumor manifestation features.

[0108] In this embodiment, the selection indexes include imaging indexes, hematological indexes, biochemical indexes, histological indexes, etc.

[0109] The beneficial effects of the above design scheme are as follows: Based on the model input features, combined with the occurrence rate of the features, feature selection is performed from the tumor manifestation features, and the features with a low occurrence rate are selected considering the correlation, and finally used as the target tumor manifestation features, ensuring the efficiency and accuracy of the obtained target tumor manifestation features.

[0110] Embodiment 6:

[0111] Based on Embodiment 1, the embodiment of the present invention provides a method for constructing a tumor chemotherapy effect prediction model based on artificial intelligence. In S3, based on the data acquisition time, divide the target tumor manifestation features into training features and verification features, including:

[0112] Use the target tumor manifestation features obtained before the preset time of data acquisition as training features;

[0113] Use the target tumor manifestation features obtained after the preset time of data acquisition as verification features.

[0114] The beneficial effects of the above design are as follows: providing appropriate data for the training and verification of the model to ensure the effects of model training and verification.

[0115] Example 7:

[0116] Based on Example 1, an embodiment of the present invention provides a method for constructing a tumor chemotherapy effect prediction model based on artificial intelligence. In S3, an initial tumor chemotherapy effect prediction model is trained based on the training features, including:

[0117] Performing a first label annotation on the training features based on the feature type, performing a second label annotation on the training features based on the feature content, and performing a third label annotation on the training features based on the tumor chemotherapy effect results corresponding to the training features;

[0118] Based on the first label annotation, dividing the training features into multiple feature training groups, based on the second label annotation and the first label annotation, inputting the feature training groups into an artificial neural network model for training, and obtaining an initial tumor chemotherapy effect prediction model according to the training results.

[0119] The beneficial effects of the above design are as follows: By performing annotations on the training features from the feature type, training features, and tumor chemotherapy effect results and then training, the training effect of the model is ensured, and the model accuracy of the obtained initial tumor chemotherapy effect prediction model is ensured.

[0120] Example 8:

[0121] Based on Example 1, an embodiment of the present invention provides a method for constructing a tumor chemotherapy effect prediction model based on artificial intelligence. In S4, the initial tumor chemotherapy effect prediction model is verified and optimized based on the verification features to obtain a target tumor chemotherapy effect prediction model, including:

[0122] Dividing the verification features into multiple first verification feature groups based on the feature type, and dividing the verification features into multiple second verification feature groups based on the tumor chemotherapy effect features;

[0123] Extracting the parts with the same features from the first verification feature group and the second verification feature group to form a third verification feature group;

[0124] Verifying the initial tumor chemotherapy effect prediction model based on the first verification feature group and the third verification feature group to obtain a first accuracy rate, verifying the initial tumor chemotherapy effect prediction model based on the second verification feature group and the third verification feature group to obtain a second accuracy rate, and verifying the initial tumor chemotherapy effect prediction model based on the third verification feature group to obtain a third accuracy rate;

[0125] Determine the feature type prediction difference based on the difference between the first accuracy rate and the third accuracy rate, and determine the tumor chemotherapy effect feature prediction difference based on the difference between the second accuracy rate and the third accuracy rate;

[0126] Obtain the learning network layer regarding the feature type in the initial tumor chemotherapy effect prediction model, and optimize the weight parameters of the learning network layer based on the feature type prediction difference to obtain a first optimization result;

[0127] Optimize the weight parameters of all the learning network layers in the initial tumor chemotherapy effect prediction model based on the feature type prediction difference to obtain a second optimization result;

[0128] Obtain the optimization parameters related to the first optimization result from the second optimization result, and adjust the optimization parameters based on the first optimization result to obtain the target optimization parameters;

[0129] Generate a target tumor chemotherapy effect prediction model based on the second optimization result and the target optimization parameters.

[0130] The beneficial effects of the above design are as follows: Based on the feature type, determine the learning network layers that need to be optimized locally in the initial tumor chemotherapy effect prediction model, and based on the tumor chemotherapy effect features, determine the learning network layers that need to be optimized globally in the initial tumor chemotherapy effect prediction model, and combine and analyze the two, and finally jointly optimize to obtain the target tumor chemotherapy effect prediction model, ensuring the accuracy of the obtained target tumor chemotherapy effect prediction model, ensuring the evaluation effects of tumor chemotherapy effect evaluation in terms of timeliness, objectivity, standardization, index integrity, etc., ultimately improving the treatment efficiency of tumor patients and ensuring the treatment effect.

[0131] Example 9:

[0132] Based on Example 8, the embodiment of the present invention provides a method for constructing a tumor chemotherapy effect prediction model based on artificial intelligence. The step of obtaining the optimization parameters related to the first optimization result from the second optimization result and adjusting the optimization parameters based on the first optimization result to obtain the target optimization parameters includes:

[0133] Judge whether the parameter difference between the optimization parameters and the optimization parameters in the first optimization result is within a preset range;

[0134] If so, use the optimization parameters as the target optimization parameters;

[0135] Otherwise, based on the parameter difference, match an optimization weight for the optimization parameters, and obtain the target optimization parameters based on the optimization weight.

[0136] The beneficial effects of the above design are as follows: By comprehensively analyzing the two optimization results and using the local first optimization result to adjust part of the overall second optimization result, the superiority of the obtained target optimization parameters is ensured, providing a basis for model optimization.

[0137] Embodiment 10:

[0138] Based on Embodiment 1, an embodiment of the present invention provides a construction system for a tumor chemotherapy effect prediction model based on artificial intelligence, as Figure 3 shown, including:

[0139] A data collection and processing module, configured to collect different types of tumor manifestation data from tumor patients, and based on the data types of the tumor manifestation data, match corresponding data processing methods to clean the tumor manifestation data to obtain target tumor manifestation data;

[0140] A feature extraction module, configured to determine a feature extraction method based on the data type, extract features from the target tumor manifestation data according to the feature extraction method to obtain tumor manifestation features, and select features from the tumor manifestation features based on the model input features to obtain target tumor manifestation features;

[0141] A model training module, configured to divide the target tumor manifestation features into training features and verification features based on the data acquisition time, and train an initial tumor chemotherapy effect prediction model based on the training features;

[0142] A model optimization module, configured to verify and optimize the initial tumor chemotherapy effect prediction model based on the verification features to obtain a target tumor chemotherapy effect prediction model.

[0143] In this embodiment, different types of tumor manifestation data from tumor patients include different types of tumor manifestation data from tumor patients.

[0144] The beneficial effects of the above design are as follows: By collecting different types of tumor manifestation data from tumor patients, and based on the data types of the tumor manifestation data, matching the corresponding data processing methods to clean the tumor manifestation data, obtaining the target tumor manifestation data, providing a comprehensive and accurate data basis for the training of the model. Based on the data types, determining the feature extraction method, extracting features from the target tumor manifestation data according to the feature extraction method to obtain tumor manifestation features, and based on the model input features, selecting features from the tumor manifestation features to obtain the target tumor manifestation features, providing accurate training samples for the training of the model. Based on the data acquisition time, dividing the target tumor manifestation features into training features and validation features, and training an initial tumor chemotherapy effect prediction model based on the training features, providing a prediction model for tumor chemotherapy effect prediction. Finally, validating and optimizing the initial tumor chemotherapy effect prediction model based on the validation features to obtain the target tumor chemotherapy effect prediction model, ensuring the prediction accuracy of the model, ensuring the evaluation effects of tumor chemotherapy effect evaluation in terms of timeliness, objectivity, standardization, index integrity, etc., ultimately improving the treatment efficiency of tumor patients and ensuring the treatment effect.

[0145] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of this application document and its equivalent technologies, the present invention also intends to include these changes and modifications.

Claims

1. A method for constructing a tumor chemotherapy effect prediction model based on artificial intelligence, characterized in that: include: S1: Collect different types of tumor manifestation data from tumor patients, and clean the tumor manifestation data by matching corresponding data processing methods based on the data type of the tumor manifestation data to obtain target tumor manifestation data; S2: Based on the data type, determine the feature extraction method, extract features from the target tumor performance data according to the feature extraction method to obtain tumor performance features, and select features from the tumor performance features based on the model input features to obtain the target tumor performance features; S3: Based on the data acquisition time, the target tumor performance characteristics are divided into training characteristics and verification characteristics, and an initial tumor chemotherapy effect prediction model is obtained by training based on the training characteristics; S4: Based on the validation features, the initial tumor chemotherapy effect prediction model is verified and optimized to obtain the target tumor chemotherapy effect prediction model, which specifically includes: Based on the feature type, the verification feature is divided into a plurality of first verification feature groups, and based on the tumor chemotherapy effect feature, the verification feature is divided into a plurality of second verification feature groups; Extracting parts with the same features from the first verification feature group and the second verification feature group to form a third verification feature group; Based on the first verification feature group and the third verification feature group, the initial tumor chemotherapy effect prediction model is verified to obtain a first accuracy rate; based on the second verification feature group and the third verification feature group, the initial tumor chemotherapy effect prediction model is verified to obtain a second accuracy rate; based on the third verification feature group, the initial tumor chemotherapy effect prediction model is verified to obtain a third accuracy rate; Based on the difference between the first accuracy rate and the third accuracy rate, the prediction difference of the feature type is determined; based on the difference between the second accuracy rate and the third accuracy rate, the prediction difference of the tumor chemotherapy effect feature is determined; Obtaining a learning network layer for feature types in an initial tumor chemotherapy effect prediction model, and optimizing weight parameters of the learning network layer based on the feature type prediction difference to obtain a first optimization result; Based on the predicted differences in the tumor chemotherapy effect characteristics, the weight parameters of all learning network layers in the initial tumor chemotherapy effect prediction model are optimized to obtain a second optimization result; Obtaining optimization parameters related to the first optimization result from the second optimization result, and adjusting the optimization parameters based on the first optimization result to obtain target optimization parameters; Based on the second optimization result and the target optimization parameters, a target tumor chemotherapy effect prediction model is generated.

2. The method for constructing a tumor chemotherapy effect prediction model based on artificial intelligence according to claim 1, characterized in that: In S1, based on the data type of the tumor manifestation data, the tumor manifestation data is cleaned by matching the corresponding data processing method to obtain the target tumor manifestation data, including: Based on the data type of the tumor manifestation data, matching the data processing method from the data processing method library, and obtaining the processing range index and processing value index that need to be specifically determined in the processing method; Obtaining target data features of the data type of the tumor performance data, and setting corresponding specific values ​​for the processing range indicator and the processing value indicator based on the target data features; Data cleaning rules are set based on the data processing method and specific values, and the tumor performance data is cleaned according to the data cleaning rules to obtain target tumor performance data.

3. The method for constructing a tumor chemotherapy effect prediction model based on artificial intelligence according to claim 1, characterized in that: In S2, based on the data type, a feature extraction method is determined, including: Determine a method for identifying data content based on the data type, establish a data statistical rule for the target tumor performance data based on the identification method, and establish a data dimension reduction rule for the target tumor performance data based on the identification method; Based on data statistical rules and data dimension reduction rules, an initial feature extraction method for target tumor performance data is generated; The data format standard is analyzed based on the preset model to determine the format requirements for the target tumor performance data. Based on the format requirements, the initial feature extraction method is optimized in the extracted data format to obtain the final feature extraction method.

4. The method for constructing a tumor chemotherapy effect prediction model based on artificial intelligence according to claim 1, characterized in that: In S2, feature extraction is performed on the target tumor manifestation data according to a feature extraction method to obtain tumor manifestation features, including: Determine the marking position of the target tumor performance data based on the feature extraction method, and segment the target tumor performance data according to the marking position to obtain multiple data segments; Data features are extracted from multiple data segments according to the feature extraction method, and the extracted features are integrated to obtain tumor manifestation features.

5. The method for constructing a tumor chemotherapy effect prediction model based on artificial intelligence according to claim 1, characterized in that: In S2, based on the model input features, feature selection is performed from the tumor manifestation features to obtain the target tumor manifestation features, including: Determine a standard input information amount based on the model input characteristics, and determine a selection index based on the standard input information amount; Based on the selected indicators, the indicator characteristic values ​​of the tumor manifestation characteristics are determined, and the indicator characteristic values ​​are mapped for information volume to obtain the actual information volume of the tumor manifestation characteristics; Acquire a first manifestation feature whose occurrence rate is greater than a preset threshold from the tumor manifestation features, use other tumor manifestation features as second manifestation features, assign a first weight to the first manifestation feature, and assign a second weight to the second manifestation feature; Determine a first score value of the first performance feature based on the actual amount of information and the first weight, and determine a second score value of the second performance feature based on the actual amount of information and the second weight; Selecting a tumor expression feature whose first score value and second score value are greater than a preset score value as a first target expression feature; Obtaining a performance feature to be analyzed whose second score value is less than or equal to a preset score value but greater than a preset minimum score value from the second performance feature, and determining the correlation between the performance feature to be analyzed and the first performance feature based on the actual amount of information, and selecting the performance feature to be analyzed whose correlation is greater than the preset correlation as the second target performance feature; The first target expression feature and the second target expression feature are used as target tumor expression features.

6. The method for constructing a tumor chemotherapy effect prediction model based on artificial intelligence according to claim 1, characterized in that: In S3, based on the data acquisition time, the target tumor performance characteristics are divided into training characteristics and verification characteristics, including: Using the target tumor expression features acquired before a preset time as training features; The target tumor expression features obtained after the preset data acquisition time are used as verification features.

7. The method for constructing a tumor chemotherapy effect prediction model based on artificial intelligence according to claim 1, characterized in that: In S3, obtaining an initial tumor chemotherapy effect prediction model based on the training features includes: Based on the feature type, the training feature is labeled with a first label, based on the feature content, the training feature is labeled with a second label, and based on the tumor chemotherapy effect result corresponding to the training feature, the training feature is labeled with a third label; Based on the first label annotation, the training features are divided into multiple feature training groups. Based on the second label annotation and the first label annotation, the feature training groups are input into the artificial neural network model for training. According to the training results, an initial tumor chemotherapy effect prediction model is obtained.

8. The method for constructing a tumor chemotherapy effect prediction model based on artificial intelligence according to claim 1, characterized in that: The step of acquiring optimization parameters related to the first optimization result from the second optimization result, and adjusting the optimization parameters based on the first optimization result to obtain target optimization parameters includes: Determine whether a parameter difference between the optimization parameter and the optimization parameter in the first optimization result is within a preset range; If so, the optimized parameter is used as the target optimized parameter; Otherwise, based on the parameter difference, the optimization weights are matched to the optimization parameters, and based on the optimization weights, the target optimization parameters are obtained.

9. The system for constructing a method for constructing a tumor chemotherapy effect prediction model based on artificial intelligence according to claim 1, characterized in that: include: A data collection and processing module is used to collect different types of tumor manifestation data from tumor patients, and based on the data type of the tumor manifestation data, match the corresponding data processing method to clean the tumor manifestation data to obtain target tumor manifestation data; A feature extraction module is used to determine a feature extraction method based on the data type, perform feature extraction on the target tumor performance data according to the feature extraction method to obtain tumor performance characteristics, and perform feature selection from the tumor performance characteristics based on the model input characteristics to obtain the target tumor performance characteristics; A model training module, used to divide the target tumor performance characteristics into training characteristics and verification characteristics based on the data acquisition time, and train an initial tumor chemotherapy effect prediction model based on the training characteristics; The model optimization module is used to verify and optimize the initial tumor chemotherapy effect prediction model based on the verification features to obtain the target tumor chemotherapy effect prediction model.

Images