Tumor radiotherapy effect evaluation method and system based on machine learning

Abstract

The invention discloses a tumor radiotherapy effect evaluation method and system based on machine learning. The method comprises the following steps: collecting clinical information of a patient; performing data preprocessing on the clinical information of the patient to obtain standard patient data; according to the standard patient data, a chi-square test model is adopted for screening, and patient data features related to prediction indexes are obtained; inputting the patient data features into a pre-constructed multi-task deep learning model, inputting screened features, and predicting the survival state of the patient after radiotherapy to obtain a tumor volume estimation result; wherein the tumor volume estimation result comprises tumor prediction results of a first period, a second period and a third period after treatment; wherein the multi-task deep learning model is a model obtained by training according to an original sample. According to the method, the labeling quantity of the medical image is increased, and the labeling efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of machine learning technology, and in particular to a method and system for evaluating tumor radiotherapy effects based on machine learning. Background Art

[0002] At present, the application of machine learning technology in medical image analysis is developing rapidly, especially in disease classification, segmentation, detection and image registration. Deep learning models can assist doctors in making more accurate and efficient diagnosis and treatment decisions by learning from a large amount of medical image data. With the advancement of technology, the accuracy of medical images has been continuously improved, providing rich data resources for machine learning algorithms and promoting the development of the field of medical image analysis.

[0003] In the existing technology, deep learning methods are used for medical image analysis to improve the accuracy and efficiency of diagnosis. These methods analyze patient characteristics and clinical data, build models to predict individual patient prognosis, and assist clinical decision-making and personalized treatment.

[0004] In summary, in the existing analysis methods, the number of medical image annotations is small, the cost is high, a lot of professional knowledge support is required, and the annotation process is time-consuming and laborious. Summary of the invention

[0005] The present invention provides a method and system for evaluating the effect of tumor radiotherapy based on machine learning, so as to solve the problems of high cost and low efficiency of medical image annotation in the prior art.

[0006] In a first aspect, in order to solve the above technical problems, the present invention provides a method for evaluating the effect of tumor radiotherapy based on machine learning, comprising: Collect clinical information of patients; Performing data preprocessing on the patient's clinical information to obtain standard patient data; Based on the standard patient data, a chi-square test model is used to screen and obtain patient data characteristics related to the prediction index; Inputting the patient data features into a pre-built multi-task deep learning model, inputting the screened features, predicting the patient's survival status after radiotherapy, and obtaining a tumor volume estimation result; wherein the tumor volume estimation result includes tumor prediction results for the first cycle, the second cycle, and the third cycle after treatment; Among them, the multi-task deep learning model is a model trained based on original samples.

[0007] As an optional implementation, the data preprocessing of the patient clinical information to obtain standard patient data includes: Cleaning the missing values ​​and abnormal values ​​of the patient's clinical information, standardizing the patient's age and BMI so that the patient's age and BMI are normally distributed, and obtaining cleaned patient data; Numerically encode the cleaned patient data using label encoding to obtain encoded data; Normalizing the target area GTVsum, the maximum diameter in the target area, and the target area volume in the encoded data to obtain standard patient data; The patient's clinical information includes age, gender, BMI, GTVsum, maximum diameter within the target area, target volume, stage, tumor type, and surrounding organs.

[0008] As an optional implementation, the method of numerically encoding the cleaned patient data using label coding to obtain encoded data includes: Gender male and female were coded as 1 and 0 respectively; pathological types adenocarcinoma, squamous cell carcinoma, and others were coded as 2, 3, and 4 respectively; smoking status yes and no were coded as 1 and 0 respectively; GTVsum esophagus, stomach, cardia, and left hilum were coded as 5, 6, 7, and 8 respectively.

[0009] As an optional implementation, the standard patient data is screened using a chi-square test model to obtain patient data features related to the prediction index, including: Screening is performed using the following chi-square test formula: ; in are the patient data features associated with the predictor, is the patient data, is the expected value of the patient data characteristic if the null hypothesis is true.

[0010] As an optional implementation, the multi-task deep learning model includes: Building a neural network model through a multi-task learning algorithm, inputting the original samples into the neural network model, and performing model training; When the number of training times of the neural network model is greater than a preset number of times, or when the loss function value of the neural network model is less than a preset loss threshold, the training is determined to be completed and a multi-task deep learning model is obtained.

[0011] As an optional implementation, the original sample is input into a neural network model to perform model training, including: A ten-fold cross validation is performed on the original samples, and the parameters with the highest classification accuracy are selected, and the parameters with the highest classification accuracy are input into the neural network model to perform model training; wherein the parameters with the highest classification accuracy include the number of hidden neurons, learning rate, number of iterations, Batch, Size, and activation function.

[0012] As an optional implementation, the patient data features are input into a pre-built multi-task deep learning model, the screened features are input, the survival status of the patient after radiotherapy is predicted, and the tumor volume estimation result is obtained, including: The tumor prediction results at the first cycle after treatment were obtained by the following formula:

[0013] The tumor prediction results at the second cycle after treatment were obtained by the following formula:

[0014] The tumor prediction results at the third cycle after treatment were obtained by the following formula:

[0015] in is the initial tumor volume; k is the decay constant, which depends on the effect of the treatment.

[0016] As an optional implementation, after the standard patient data is screened using a chi-square test model to obtain patient data features related to the prediction index, the method includes: The patient data features are eliminated using the Pearson correlation coefficient to obtain dimension-reduced data features; Predicting the patient's survival status after radiotherapy based on the dimension reduction data features; Among them, the Pearson correlation coefficient dimension reduction operation is performed by the following formula; ; Among them, r is the Pearson correlation coefficient, which is used to express the correlation between each feature and tumor recurrence and metastasis. is the target area regression rate corresponding to the i-th feature, is the i-th patient data feature, and They are respectively , The average value of .

[0017] In a second aspect, the present invention provides a tumor radiotherapy effect evaluation system based on machine learning, comprising: Data collection module, used to collect clinical information of patients; A data preprocessing module, used for preprocessing the patient's clinical information to obtain standard patient data; A feature selection module, for screening the standard patient data using a chi-square test model to obtain patient data features related to the prediction index; A multi-task learning module, which is used to input the patient data features into a pre-built multi-task deep learning model, input the screened features, predict the patient's survival status after radiotherapy, and obtain a tumor volume estimation result; Among them, the multi-task deep learning model is a model trained based on original samples.

[0018] In a third aspect, the present invention also provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements any one of the above-mentioned methods for evaluating the effect of tumor radiotherapy based on machine learning.

[0019] In a fourth aspect, the present invention also provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned methods for evaluating the effect of tumor radiotherapy based on machine learning.

[0020] Compared with the prior art, the present invention has the following beneficial effects: The present invention discloses a method for evaluating the effect of tumor radiotherapy based on machine learning, including obtaining patient clinical information; performing data preprocessing according to the patient clinical information to obtain standard patient data; using a chi-square test model to screen according to the standard patient data to obtain patient data features related to prediction indicators; inputting the patient data features into a pre-constructed multi-task deep learning model, inputting the screened features, predicting the patient's survival status after radiotherapy, and obtaining a tumor volume estimation result; wherein the tumor volume estimation result includes the tumor prediction results of the first cycle, the second cycle, and the third cycle after treatment. Compared with the prior art, the present invention discloses a method for evaluating the effect of tumor radiotherapy based on machine learning, which encodes the tumor volume and the maximum diameter numerical value, outputs the patient's survival probability through model training, and estimates the tumor volume through a physical formula to achieve tumor state estimation, provide patients with more comprehensive treatment information, facilitate the judgment of prognosis, and increase the number of annotations of medical images and improve annotation efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a flow chart of a method for evaluating the effect of tumor radiotherapy based on machine learning provided in the first embodiment of the present invention; Figure 2 is a schematic flow chart of the data preprocessing method of step 12 provided in the first embodiment of the present invention; Figure 3 It is a structural schematic diagram of a tumor radiotherapy effect evaluation system based on machine learning provided in the second embodiment of the present invention. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] At present, the application of machine learning technology in medical image analysis is developing rapidly, especially in disease classification, segmentation, detection and image registration. Deep learning models can assist doctors in making more accurate and efficient diagnosis and treatment decisions by learning from a large amount of medical image data. With the advancement of technology, the accuracy of medical images has been continuously improved, providing rich data resources for machine learning algorithms and promoting the development of the field of medical image analysis.

[0024] In the prior art, deep learning methods are used for medical image analysis to improve the accuracy and efficiency of diagnosis. These methods analyze patient characteristics and clinical data to build models to predict individual patient prognosis, assist clinical decision-making and personalized treatment. However, in the prior art analysis methods, the number of medical image annotations is small, the cost is high, a lot of professional knowledge support is required, and the annotation process is time-consuming and laborious.

[0025] In order to solve the above technical problems, refer to Figure 1 The first embodiment of the present invention provides a method for evaluating the effect of tumor radiotherapy based on machine learning, comprising the following steps: S11, collect patient clinical information; S12, performing data preprocessing on the patient's clinical information to obtain standard patient data; S13, screening using a chi-square test model based on the standard patient data to obtain patient data features related to the prediction index; S14, inputting the patient data features into a pre-built multi-task deep learning model, inputting the screened features, predicting the patient's survival status after radiotherapy, and obtaining a tumor volume estimation result; wherein the tumor volume estimation result includes tumor prediction results for the first cycle, the second cycle, and the third cycle after treatment; Among them, the multi-task deep learning model is a model trained based on original samples.

[0026] In step S11, clinical information of the patient needs to be collected.

[0027] The patient clinical information specifically includes the patient's age, gender, BMI, GTVsum, maximum diameter within the target area, target volume, stage, tumor type, information on surrounding organs and survival status after treatment. The data source is the electronic medical record system database of the hospital's radiotherapy department.

[0028] refer to Figure 2 In step S12, the patient clinical information is preprocessed to obtain standard patient data, including: S121, cleaning the missing values ​​and abnormal values ​​of the patient's clinical information, standardizing the patient's age and BMI so that the patient's age and BMI are normally distributed, and obtaining cleaned patient data; S122, using label coding to numerically encode the cleaned patient data to obtain encoded data; S123, normalizing the target area GTVsum, the maximum diameter in the target area, and the target area volume in the encoded data to obtain standard patient data; The patient's clinical information includes age, gender, BMI, GTVsum, maximum diameter within the target area, target volume, stage, tumor type, and surrounding organs.

[0029] In step S121, the missing values ​​and abnormal values ​​of the patient's clinical information are cleaned, and the patient's age and BMI are standardized so that the patient's age and BMI are normally distributed, thereby obtaining cleaned patient data.

[0030] In a specific embodiment, the patient clinical information is cleaned of missing values ​​and abnormal values, and the patient age and BMI are standardized so that the patient age and BMI are normally distributed to obtain the cleaned patient data, including: Forming the original data set: The clinical information of the patient, including age, gender, smoking history, tumor stage, tumor pathological type, tumor size and location, tumor stage and KPS, is stored in a CSV format file to form the original data set; Outlier cleaning: According to the data type distribution in the original data set, some outliers are pre-cleaned. Outliers exceeding 3 times the standard deviation are considered outliers. Maximum likelihood estimation is needed to find missing values ​​and obtain data after outlier cleaning. Feature cleaning: performing feature cleaning on the data after outlier cleaning: staging lung cancer patients, discretizing the staging features into original staging, and then taking the logarithm to obtain the data after feature cleaning; Missing value filling: The missing values ​​of the data after the feature cleaning are counted. If the missing value ratio does not exceed 30%, no missing value filling is performed. If the missing value ratio is greater than 30%, the missing value is filled. The missing value filling method is: calculate the number of patients of the type corresponding to the missing value, divide the data type into several segments, each segment represents a type of data, and for the missing numerical type, use the data type as a label, perform feature encoding, fill the missing value, and obtain the missing value filled data, which is the cleaned patient data.

[0031] In step S122, label coding is used to numerically encode the cleaned patient data to obtain encoded data.

[0032] In a specific embodiment, the step of numerically encoding the cleaned patient data using label coding to obtain encoded data includes: Gender male and female were coded as 1 and 0 respectively; pathological types adenocarcinoma, squamous cell carcinoma, and others were coded as 2, 3, and 4 respectively; smoking status yes and no were coded as 1 and 0 respectively; GTVsum esophagus, stomach, cardia, and left hilum were coded as 5, 6, 7, and 8 respectively.

[0033] It should be noted that the encoding method is a combination of label encoding and numerical encoding for the encoding of the patient's clinical information. Categorical variables such as gender, pathological classification, smoking status, GTVsum, etc. are encoded using label encoding to convert different categories into numerical values ​​for easy processing by the machine learning model. Numerical variables such as age, BMI, target area GTVsum, maximum diameter within the target area, and target area volume are standardized and normalized to eliminate the influence of different dimensions and make model training more stable and efficient.

[0034] Among them, the advantages of this encoding method include: Simplify model processing: Convert categorical variables into numerical types to facilitate mathematical operations and learning of the model.

[0035] Improve model generalization ability: The data distribution after standardization and normalization is more even, which helps to improve the generalization ability and prediction accuracy of the model.

[0036] Speed ​​up training: Normalized data can reduce the number of iterations during model training, thereby speeding up training.

[0037] Improve model performance: The processed data helps to avoid model bias caused by dimensional differences in dependent variables and improve model performance.

[0038] Facilitates feature selection: Encoded data is easier to perform feature selection and dimensionality reduction, which helps to screen out the features that have the greatest impact on the prediction results.

[0039] In step S123, the target area GTVsum, the maximum diameter within the target area, and the target area volume in the encoded data are normalized to obtain standard patient data.

[0040] In a specific embodiment, the target area GTVsum, the maximum diameter in the target area, and the target area volume in the encoded data are normalized to obtain standard patient data, including: Numerical feature processing: normalizing the data features of the missing value filling, wherein age and chemotherapy course are treated as having a mean of 0 and a unit difference quotient of 1, and performing log (x+1) normalization processing on the numerical features in the data features of the missing value filling to obtain data of numerical feature processing; Discretization processing: In the data processed by the numerical feature, gender and pathological classification are encoded as 0 / 1; the functional status score standard (KPS) is discretized, and the score is normalized into three levels to obtain standard patient data.

[0041] In step S13, the standard patient data is screened using a chi-square test model to obtain patient data features related to the prediction index, including: Screening is performed using the following chi-square test formula: ; in are the patient data features associated with the predictor, is the patient data, is the expected value of the patient data characteristic if the null hypothesis is true.

[0042] It should be noted that the use of the chi-square test model for screening is a key step of the present invention, and its purpose is to screen out the features most relevant to predicting the treatment effect from the numerous clinical data features. The specific operation process is to first construct a contingency table, list the frequencies of different values ​​of each feature and the treatment results, and then calculate the chi-square statistic, which reflects the degree of deviation between the observed frequency and the expected frequency under the assumption that the feature is unrelated to the treatment result. By comparing the statistic with the critical value of the chi-square distribution, it can be determined whether the feature is independent of the treatment result, that is, whether it is related. Selecting those features that are significantly related to the treatment results as model input can effectively improve the accuracy and efficiency of model prediction.

[0043] Using the chi-square test model for feature screening to eliminate features that are not closely related to treatment outcomes can reduce the model's dimensionality, thereby reducing the computing resources required for model training and prediction. At the same time, screening out the most relevant features can also improve the model's prediction accuracy because the model only focuses on those factors that have the greatest impact on the prediction results.

[0044] The Pearson correlation coefficient was used to calculate the correlation between each feature and tumor recurrence and metastasis. The three features with the strongest correlation were eliminated, and the feature dimension was reduced to 26 to obtain the reduced dimension data. The calculation formula is as follows: ; Among them, r is the Pearson correlation coefficient, which is used to express the correlation between each feature and tumor recurrence and metastasis. is the corresponding feature of tumor recurrence and metastasis of the ith tumor, is the i-th patient data feature, and They are respectively , The average value of The correlation between each feature and the target area regression rate is calculated based on the dimensionality reduction data to obtain the correlation data between each feature and the target area regression rate. The calculation formula is as follows: ; Among them, r is the Pearson correlation coefficient, which is used to express the correlation between each feature and the target area regression rate. is the target area regression rate corresponding to the i-th feature, is the i-th patient data feature, and They are respectively , The average value of According to the correlation data between each feature and the target area regression rate, the features were eliminated in the dimensionality reduction data, and the features with a correlation with the target area regression rate less than 0.3 and a correlation with tumor recurrence greater than 0.9 were removed to obtain the model features after screening; The model characteristics after screening include: gender, stage, pathological classification, previous history, KPS, hemoglobin content before and after chemotherapy, WBC count, ANC count, NK cell percentage, T lymphocyte percentage, absolute value of T lymphocytes, T / CD3, Treg / CD8, T / Treg, C-reactive protein content, liver enzymes GPT, PT, Alb, and LDH.

[0045] In step S14, the patient data features are input into a pre-built multi-task deep learning model, and the screened features are input to predict the patient's survival status after radiotherapy to obtain a tumor volume estimation result; wherein the tumor volume estimation result includes tumor prediction results for the first cycle, second cycle, and third cycle after treatment.

[0046] In one implementation, the process of constructing the multi-task deep learning model includes: The multi-task deep learning model constructs a neural network model through a multi-task learning algorithm, and inputs the original samples into the neural network model to perform model training; When the number of training times of the neural network model is greater than a preset number of times, or when the loss function value of the neural network model is less than a preset loss threshold, the training is determined to be completed and a multi-task deep learning model is obtained.

[0047] The original sample is input into the neural network model to perform model training, including: A ten-fold cross validation is performed on the original samples, and the parameters with the highest classification accuracy are selected, and the parameters with the highest classification accuracy are input into the neural network model to perform model training; wherein the parameters with the highest classification accuracy include the number of hidden neurons, learning rate, number of iterations, Batch, Size, and activation function.

[0048] It should be noted that 10-fold cross validation is a commonly used statistical method to evaluate and compare the predictive performance of machine learning models. During the model training process, using 10-fold cross validation can improve the reliability and accuracy of model evaluation. This method divides the dataset into ten parts. In each iteration, one of the parts is used as the test set and the remaining nine parts are used as the training set. This process is repeated ten times, and a different part is selected as the test set each time, ensuring that each data point is used for one test; In general, 10-fold cross validation is a powerful tool that improves the accuracy and reliability of model evaluation by evaluating the model on multiple training and test sets. This approach helps to select the best model parameters and improve the generalization and robustness of the model, thereby achieving better performance in practical applications.

[0049] It should be noted that the prediction process of the tumor volume estimation result includes: Performing a ten-fold cross validation on the original samples, selecting the parameters with the highest classification accuracy, inputting the parameters with the highest classification accuracy into the neural network model, using the time points of recurrence, metastasis, first cycle, second cycle, and third cycle of survival as supervision quantities, and performing model training in combination with the screened model features; when the number of training times of the neural network model is greater than a preset number of times, or when the loss function value of the neural network model is less than a preset loss threshold, determining that the training is completed, and obtaining a multi-task deep learning model; Use the test set for prediction: predict the patient's survival after treatment based on the model, and use MAE, accuracy, AUC, recall, and F1 value to evaluate the model; Estimation of tumor volume and maximum diameter: Calculate the tumor volume and maximum diameter indicators related to patient recurrence; Estimation of MRI indicators: calculation of MRI indicators related to patient recurrence, including tumor perfusion, tumor fibrosis, necrosis, intratumoral bleeding, and tumor edema; The first cycle, the second cycle and the third cycle refer to the third month, the sixth month and the twelfth month respectively.

[0050] In one implementation, the patient data features are input into a pre-built multi-task deep learning model, the screened features are input, the survival status of the patient after radiotherapy is predicted, and the tumor volume estimation result is obtained, including: The tumor prediction results at the first cycle after treatment were obtained by the following formula: (1) The tumor prediction results at the second cycle after treatment were obtained by the following formula: (2) The tumor prediction results at the third cycle after treatment were obtained by the following formula: (3) in is the initial tumor volume; k is the decay constant, which depends on the effect of the treatment.

[0051] The formulas in the above prediction results are used to estimate the tumor volume at different time periods after the patient receives radiotherapy. These formulas are based on an exponential decay model, in which the tumor volume decreases with time, and the decay rate is determined by the decay constant k, which reflects the effect of the treatment.

[0052] It should be noted that Initial tumor volume, which is the size of the tumor before treatment begins. This value can usually be obtained before treatment using medical imaging techniques (such as CT, MRI, or PET scans), and automated image analysis tools can be used to improve the accuracy and consistency of the measurement.

[0053] It should be noted that k: decay constant, represents the rate at which tumor volume decreases over time. This parameter depends on many factors, including the dose of radiotherapy, the type of tumor, the patient's overall health, etc. In practice, k is usually determined by analyzing historical data, which contains information on tumor volume changes in other patients who received similar treatments. In addition, machine learning models can be used to analyze this data to estimate the value of k.

[0054] The following describes the working process of the present invention using a common scenario as an example. Taking lung cancer patient data as an example, the specific implementation method of the present invention is as follows: Figure 2 , a method for evaluating the effect of tumor radiotherapy based on machine learning, comprising the following steps: Clinical data of lung cancer radiotherapy patients were collected from the electronic medical record system database of the hospital's radiotherapy department, including age, gender, BMI, GTVsum, maximum diameter within the target area, target volume, stage, tumor type, information on surrounding organs, and survival after treatment; The clinical data of the lung cancer radiotherapy patients are pre-cleaned according to the data type distribution in the original data set, and some outliers are pre-cleaned. Outliers exceeding 3 times the standard deviation are considered outliers, and the maximum likelihood estimation is required to find missing values ​​to obtain data after outlier cleaning; the data of lung cancer radiotherapy patients are uniformly processed to obtain the characteristics of lung cancer radiotherapy patients, including: age, gender, smoking history, tumor stage, tumor pathological type, tumor size and location, tumor stage, KPS, chemotherapy course, and chemotherapy cycle; Performing feature cleaning on the clinical data of the lung cancer radiotherapy patients, discretizing the staging data into original stages (stages I-III) according to the lung cancer stages, and then taking the logarithm to obtain feature-cleaned data; The data cleaned by the features are normalized and digitized, and the numerical features including age, chemotherapy cycle, chemotherapy course, etc. are normalized, wherein age and chemotherapy course are treated as having a mean of 0 and a unit difference quotient of 1, and the numerical features such as chemotherapy cycle, maximum tumor diameter, target area size, etc. are normalized by log (x+1) to obtain data processed by numerical features; The data processed by the numerical features are discretized using a feature screening algorithm, including: gender, pathological classification, and functional status scoring criteria; The Pearson correlation coefficient was used to calculate the correlation between each feature and tumor recurrence and metastasis, and the three features with the strongest correlation were eliminated, and the feature dimension was reduced to 26 to obtain the reduced dimension data; The correlation between each feature and the target area regression rate is calculated based on the dimension reduction data to obtain the correlation data between each feature and the target area regression rate; According to the correlation data between each feature and the target area regression rate, the features were eliminated in the dimensionality reduction data, and the features with a correlation of less than 0.3 with the target area regression rate and a correlation of more than 0.9 with tumor recurrence were removed to obtain the model features after screening, including: gender, stage, pathological classification, previous history, KPS, hemoglobin content before and after chemotherapy, WBC count, ANC count, NK cell percentage, T lymphocyte percentage, T lymphocyte absolute value, T / CD3, Treg / CD8, T / Treg, C-reactive protein content, liver enzymes GPT, PT, Alb, LDH; Conduct model training: select some of the filtered model features as the test set, input the test set, conduct model training, and use the multi-task learning algorithm to learn tumor recurrence and metastasis; Use the remaining data sets to predict the results: Input the model training set to obtain the patient's survival predictions three months, six months, and twelve months after treatment; Predict tumor volume and maximum diameter: Use the recurrence rate as the regression output value to train a neural network with a CNN structure and selected tumor features as input. Use the network to learn patient data and estimate tumor volume and maximum diameter. Predict MRI indicators related to tumor prognosis: MRI images are used, and a trained neural network model is used, with the recurrence rate as the regression output value. The network structure is CNN+RNN. The network model inputs MRI images and uses a convolutional neural network to estimate tumor perfusion, fibrosis, necrosis, hemorrhage, edema and other indicators; The model was scored using MAE, accuracy, AUC, recall, and F1 values ​​to predict tumor recurrence. The recurrence of lung cancer after treatment was used as the supervision quantity, and the regressor was used for learning. The logistic regressor was used for training to predict the prognosis of the patient in the first, second, and third cycles. The prediction results were stored through the multi-task learning algorithm and input into the test data set to obtain the prediction results. The tumor-related volume index and maximum diameter index related to recurrence were obtained through formulas (2) and (3). The MRI indexes related to patient recurrence, including perfusion, fibrosis, necrosis, hemorrhage, and edema, were estimated through convolutional neural networks. The model was scored using MAE, accuracy, AUC, recall, and F1 values.

[0055] In summary, the present invention is based on a method of machine learning, which performs data preprocessing, feature selection and deep learning algorithms on patients to achieve accurate assessment of patient prognosis. Through the model, the recurrence of patients 3 months, 6 months and 12 months after radiotherapy is predicted, and the survival of patients after treatment is estimated. According to the data characteristics of the patients, accurate and comprehensive prediction results are given, providing an objective evaluation method for the treatment of radiotherapy patients, which can assist medical staff in making judgments and improve the work efficiency of medical staff. At the same time, combined with physical formulas, a secondary evaluation of the patient is performed to make up for the shortcomings of the machine learning model. The method of the present invention can achieve accurate prognosis assessment, provide prediction results for patients, and facilitate patients to seek medical treatment.

[0056] Although the present invention has been disclosed as above in the form of preferred embodiments, the embodiments are not intended to limit the present invention. Any person familiar with this technology can make possible changes and modifications to the technical solutions of the present invention by using the methods and technical contents disclosed above without departing from the spirit and scope of the present invention. Therefore, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solutions of the present invention are within the protection scope of the technical solutions of the present invention.

[0057] Reference Figure 3 The second embodiment of the present invention provides a tumor radiotherapy effect evaluation system based on machine learning, comprising: Data collection module, used to collect clinical information of patients; A data preprocessing module, used for preprocessing the patient's clinical information to obtain standard patient data; A feature selection module, for screening the standard patient data using a chi-square test model to obtain patient data features related to the prediction index; A multi-task learning module, which is used to input the patient data features into a pre-built multi-task deep learning model, input the screened features, predict the patient's survival status after radiotherapy, and obtain a tumor volume estimation result; Among them, the multi-task deep learning model is a model trained based on original samples.

[0058] In an optional implementation, the data preprocessing module includes: A data cleaning unit is used to clean the missing values ​​and abnormal values ​​of the patient's clinical information, standardize the patient's age and BMI so that the patient's age and BMI are normally distributed, and obtain cleaned patient data; A numerical encoding unit, used for numerically encoding the cleaned patient data using label encoding to obtain encoded data; A normalization unit, used for normalizing the target area GTVsum, the maximum diameter in the target area, and the target area volume in the encoded data to obtain standard patient data; The patient's clinical information includes age, gender, BMI, GTVsum, maximum diameter within the target area, target volume, stage, tumor type, and surrounding organs.

[0059] In an optional implementation, the using label coding to numerically encode the cleaned patient data to obtain the encoded data includes: Gender male and female were coded as 1 and 0 respectively; pathological types adenocarcinoma, squamous cell carcinoma, and others were coded as 2, 3, and 4 respectively; smoking status yes and no were coded as 1 and 0 respectively; GTVsum esophagus, stomach, cardia, and left hilum were coded as 5, 6, 7, and 8 respectively.

[0060] In an optional embodiment, the standard patient data is screened using a chi-square test model to obtain patient data features related to the prediction index, including: Screening is performed using the following chi-square test formula: ; in are the patient data features associated with the predictor, is the patient data, is the expected value of the patient data characteristic if the null hypothesis is true.

[0061] In an optional implementation, the multi-task deep learning model includes: Building a neural network model through a multi-task learning algorithm, inputting the original samples into the neural network model, and performing model training; When the number of training times of the neural network model is greater than the preset number, or when the neural network model When the loss function value is less than the preset loss threshold, the training is considered completed and a multi-task deep learning model is obtained.

[0062] In an optional implementation, the original sample is input into a neural network model for model training, including A ten-fold cross validation is performed on the original samples, and the parameters with the highest classification accuracy are selected, and the parameters with the highest classification accuracy are input into the neural network model to perform model training; wherein the parameters with the highest classification accuracy include the number of hidden neurons, learning rate, number of iterations, Batch, Size, and activation function.

[0063] In an optional embodiment, the patient data features are input into a pre-built multi-task deep learning model, the screened features are input, the survival status of the patient after radiotherapy is predicted, and the tumor volume estimation result is obtained, including: The tumor prediction results at the first cycle after treatment were obtained by the following formula: (1) The tumor prediction results at the second cycle after treatment were obtained by the following formula: (2) The tumor prediction results at the third cycle after treatment were obtained by the following formula: (3) in is the initial tumor volume; k is the decay constant, which depends on the effect of the treatment.

[0064] In an optional embodiment, after the standard patient data is screened using a chi-square test model to obtain patient data features related to the prediction index, the method includes: The patient data features are eliminated using the Pearson correlation coefficient to obtain dimension-reduced data features; Predicting the patient's survival status after radiotherapy based on the dimension reduction data features; Among them, the Pearson correlation coefficient dimension reduction operation is performed by the following formula; ; Among them, r is the Pearson correlation coefficient, which is used to express the correlation between each feature and tumor recurrence and metastasis. is the target area regression rate corresponding to the i-th feature, is the i-th patient data feature, and They are respectively , The average value of .

[0065] In summary, the present invention discloses that the present invention provides a tumor radiotherapy effect evaluation system based on machine learning, including: a data acquisition module, used to collect clinical information of patients; a data preprocessing module, used to perform data preprocessing on the clinical information of patients to obtain standard patient data; a feature selection module, used to screen according to the standard patient data using a chi-square test model to obtain patient data features related to prediction indicators; a multi-task learning module, used to input the patient data features into a pre-built multi-task deep learning model, input the screened features, predict the survival status of the patient after radiotherapy, and obtain a tumor volume estimation result; wherein the multi-task deep learning model is a model trained according to the original sample.

[0066] Compared with the prior art, the present invention discloses a tumor radiotherapy effect evaluation system based on machine learning. By encoding the tumor volume and maximum diameter numerical values, outputting the patient survival probability through model training, and estimating the tumor volume through physical formulas, the tumor state estimation is achieved, providing patients with more comprehensive treatment information, facilitating the judgment of prognosis, and increasing the number of medical image annotations and improving annotation efficiency.

[0067] It should be noted that the machine learning-based tumor radiotherapy effect evaluation system provided in an embodiment of the present invention is used to execute all the process steps of the machine learning-based tumor radiotherapy effect evaluation method in the above-mentioned embodiment. The working principles and beneficial effects of the two correspond one to one, and therefore will not be repeated here.

[0068] The embodiment of the present invention further provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a tumor radiotherapy effect evaluation program based on machine learning. When the processor executes the computer program, the steps in each of the above-mentioned embodiments of the tumor radiotherapy effect evaluation method based on machine learning are implemented, such as Figure 1 Alternatively, the processor implements the functions of each module / unit in the above-mentioned device embodiments when executing the computer program.

[0069] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the electronic device.

[0070] The electronic device may be a computing device such as a desktop computer, a notebook, a PDA, and a smart tablet. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art will appreciate that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. The electronic device may include more or fewer components than the above components, or may combine certain components, or different components. For example, the electronic device may also include input and output devices, network access devices, buses, etc.

[0071] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the electronic device, and uses various interfaces and lines to connect various parts of the entire electronic device.

[0072] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the electronic device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0073] Wherein, if the module / unit integrated in the electronic device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0074] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0075] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for evaluating the effect of tumor radiotherapy based on machine learning, characterized in that: include: Collect patient clinical information; Performing data preprocessing on the patient's clinical information to obtain standard patient data; Based on the standard patient data, a chi-square test model is used to screen and obtain patient data characteristics related to the prediction index; Inputting the patient data features into a pre-built multi-task deep learning model, inputting the screened features, predicting the patient's survival status after radiotherapy, and obtaining a tumor volume estimation result; wherein the tumor volume estimation result includes tumor prediction results for the first cycle, the second cycle, and the third cycle after treatment; Among them, the multi-task deep learning model is a model trained based on original samples.

2. The method for evaluating the effect of tumor radiotherapy based on machine learning according to claim 1, characterized in that: The step of preprocessing the patient's clinical information to obtain standard patient data includes: Cleaning the missing values ​​and abnormal values ​​of the patient's clinical information, standardizing the patient's age and BMI so that the patient's age and BMI are normally distributed, and obtaining cleaned patient data; Numerically encode the cleaned patient data using label encoding to obtain encoded data; Normalizing the target area GTVsum, the maximum diameter in the target area, and the target area volume in the encoded data to obtain standard patient data; The patient's clinical information includes age, gender, BMI, GTVsum, maximum diameter within the target area, target volume, stage, tumor type, and surrounding organs.

3. The method for evaluating the effect of tumor radiotherapy based on machine learning according to claim 2, characterized in that: The step of using label coding to numerically encode the cleaned patient data to obtain encoded data includes: Gender male and female were coded as 1 and 0 respectively; pathological types adenocarcinoma, squamous cell carcinoma, and others were coded as 2, 3, and 4 respectively; smoking status yes and no were coded as 1 and 0 respectively; GTVsum esophagus, stomach, cardia, and left hilum were coded as 5, 6, 7, and 8 respectively.

4. The method for evaluating the effect of tumor radiotherapy based on machine learning according to claim 1, characterized in that: The standard patient data is screened using a chi-square test model to obtain patient data features related to the prediction index, including: Screening is performed using the following chi-square test formula: ； in are the patient data features associated with the predictor, is the patient data, is the expected value of the patient data characteristic if the null hypothesis is true.

5. The method for evaluating tumor radiotherapy effect based on machine learning according to claim 1, characterized in that: The multi-task deep learning model includes: Building a neural network model through a multi-task learning algorithm, inputting the original samples into the neural network model, and performing model training; When the number of training times of the neural network model is greater than a preset number of times, or when the loss function value of the neural network model is less than a preset loss threshold, the training is determined to be completed and a multi-task deep learning model is obtained.

6. The method for evaluating the effect of tumor radiotherapy based on machine learning according to claim 5, characterized in that: The original samples are input into the neural network model to perform model training, including: A ten-fold cross validation is performed on the original samples, and the parameters with the highest classification accuracy are selected, and the parameters with the highest classification accuracy are input into the neural network model to perform model training; wherein the parameters with the highest classification accuracy include the number of hidden neurons, learning rate, number of iterations, Batch, Size, and activation function.

7. The method for evaluating the effect of tumor radiotherapy based on machine learning according to claim 1, characterized in that: The patient data features are input into a pre-built multi-task deep learning model, the screened features are input, the survival status of the patient after radiotherapy is predicted, and the tumor volume estimation result is obtained, including: The tumor prediction results at the first cycle after treatment were obtained by the following formula: （1） The tumor prediction results at the second cycle after treatment were obtained by the following formula: （2） The tumor prediction results at the third cycle after treatment were obtained by the following formula: （3） in is the initial tumor volume; k is the decay constant, which depends on the effect of the treatment.

8. The method for evaluating tumor radiotherapy effect based on machine learning according to claim 1, characterized in that: After the standard patient data is screened using a chi-square test model to obtain patient data features related to the prediction index, the method includes: The patient data features are eliminated using the Pearson correlation coefficient to obtain dimension-reduced data features; Predicting the patient's survival status after radiotherapy based on the dimension reduction data features; Among them, the Pearson correlation coefficient dimension reduction operation is performed by the following formula; ； Among them, r is the Pearson correlation coefficient, which is used to express the correlation between each feature and tumor recurrence and metastasis. is the target area regression rate corresponding to the i-th feature, is the i-th patient data feature, and They are respectively , The average value of .

9. A tumor radiotherapy effect evaluation system based on machine learning, characterized in that: include: Data collection module, used to collect clinical information of patients; A data preprocessing module, used for preprocessing the patient's clinical information to obtain standard patient data; A feature selection module, for screening the standard patient data using a chi-square test model to obtain patient data features related to the prediction index; A multi-task learning module, used to input the patient data features into a pre-built multi-task deep learning model, input the screened features, predict the patient's survival status after radiotherapy, and obtain a tumor volume estimation result; wherein the tumor volume estimation result includes tumor prediction results for the first cycle, the second cycle, and the third cycle after treatment; Among them, the multi-task deep learning model is a model trained based on original samples.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the tumor radiotherapy method based on machine learning described in any one of claims 1 to 8 are implemented.

Images