CN110031624A - Tumor markers detection system based on multiple neural networks classifier, method, terminal, medium - Google Patents

Abstract

This application provides a tumor marker detection system, method, terminal, and medium based on a multi-neural network classifier, mainly using the data preprocessing method of principal component analysis and feature extraction, based on random forest, support vector machine, BP neural network, and limit The combined classifier model of the learning machine classifier model trains the data, and finally obtains a liver cancer tumor marker classifier whose accuracy, specificity and sensitivity meet the clinical diagnosis, which helps clinicians reduce the misdiagnosis rate in the initial diagnosis of liver cancer and improve liver cancer tumors. Marker detection accuracy.

Description

Tumor marker detection system, method, terminal, medium

technical field

The present application relates to the technical field of neural network classifiers, and in particular, to a tumor marker detection system, method, terminal and medium based on multiple neural network classifiers.

Background technique

Primary liver cancer is the fourth most common malignant tumor and the third leading cause of tumor death in my country, which seriously threatens the life and health of the Chinese people. At present, the detection methods for early diagnosis of liver cancer are mainly divided into two categories:

One is a detection method using serum tumor markers of liver cancer as a reference index. my country's latest "Primary Liver Cancer Diagnosis and Treatment Specifications 2017 Edition" pointed out that the liver cancer tumor marker alpha-fetoprotein AFP is a commonly used and important method for the diagnosis of liver cancer. suggest that there may be liver cancer. Although AFP has high sensitivity and specificity for the diagnosis of liver cancer, false positives can still occur in 40% of early stage liver cancer and 15%-20% of advanced liver cancer patients. The other is imaging and pathological means represented by ultrasound, computer X-ray, magnetic resonance imaging, digital subtraction angiography and liver pathological puncture.

However, the former detection method mainly relies on the actual work experience of doctors, and requires long-term practice to achieve an excellent decision-making level, which is greatly affected by the subjective experience of doctors and external interference factors; Most technologies have limited ability to diagnose liver cancer at an early stage, which not only has a certain impact on the patient's body, but also is expensive and not universal.

There are many types of serum tumor markers for liver cancer, but no single tumor marker can diagnose liver cancer clinically. The existence of each tumor marker has certain reference value in diagnosis, but it also has its own limitations. Clinically, doctors generally refer to the parameters of multiple tumor markers for joint detection and diagnosis. The factors of human experience are easy to cause errors in diagnosis. The use of machine learning methods can effectively improve the accuracy of combined diagnosis of multiple tumor markers. Computer-aided diagnosis algorithms based on machine learning have been continuously optimized and improved in recent years, which brings the possibility of constructing a multi-tumor marker combined diagnosis model. In terms of algorithms, some foreign scholars have compared the actual effects of 179 different classifiers on 121 different datasets. The research results show that the effects of different classifiers in different classification scenarios are different. For example, the random forest RF is the strongest on average, but it only won the first place in 9.9% of the data sets; the average level of support vector machine SVM It is close behind, taking first place on 10.7% of the datasets.

Therefore, there is an urgent need in the art for a technical solution that can effectively improve the accuracy of combined diagnosis of multiple tumor markers.

Application content

In view of the above-mentioned shortcomings of the prior art, the purpose of the present application is to provide a tumor marker detection system, method, terminal, and medium based on a multi-neural network classifier, which are used to solve the problems in the prior art.

In order to achieve the above object and other related objects, a first aspect of the present application provides a tumor marker detection system based on a multi-neural network classifier, which includes: a sample collection module for collecting sample data of tumor marker detection samples, The sample data includes test set data and training set data; a data preprocessing module is used to filter out a plurality of abnormal indicators that are highly correlated with tumors according to the sample data; a sample statistical analysis module is used to The sample data data is used to evaluate and analyze the detection results of a single abnormal index, and/or to evaluate and analyze the detection results of multiple joint abnormal indicators; the classifier model training module is used to perform a plurality of classifier models based on the training set data. Model training; a classifier model evaluation and testing module is used to test the trained classifier model based on the test set data, and compare the test results with the evaluation and analysis results of the sample statistical analysis module. to judge the validity of the tested classifier model; the classifier diagnosis method application module is used to detect the tumor marker data by using the classifier model judged to be valid, so as to output the diagnostic result information.

In some embodiments of the first aspect of the present application, the sample collection module collects clinical real sample data and classifies the collected sample data.

In some embodiments of the first aspect of the present application, the tumor markers include liver cancer tumor markers; wherein, the classification method of the sample data includes: dividing the sample data into liver cancer group sample data and non-liver cancer group sample data. , and/or divided into liver cancer group sample data, liver disease group sample data and healthy group sample data.

In some embodiments of the first aspect of the present application, the data preprocessing module is configured to perform feature selection based on a principal component analysis algorithm, so as to perform dimensionality reduction processing on a plurality of abnormal indicators initially screened according to the sample data, And obtain the abnormal index with high correlation with the tumor.

In some embodiments of the first aspect of the present application, the abnormal indicators with high correlation with tumors include abnormal indicators with high correlation with liver cancer tumors, which include alpha-fetoprotein, and also include: abnormal prothrombin, Carcinoembryonic antigen, carbohydrate antigen 199, carbohydrate antigen 242, carbohydrate antigen 211, carbohydrate antigen 125, sialic acid, or ferritin.

In some embodiments of the first aspect of the present application, the evaluation indicators for the evaluation and analysis of the test results include sensitivity, specificity, and accuracy; wherein, the sensitivity represents the proportion of samples that are actually positive, and the specificity represents the actual positive ratio. The proportion of negative samples judged to be negative, and the accuracy indicates the proportion of true positives and true negatives to the total number of subjects tested.

In some embodiments of the first aspect of the present application, the classifier model includes: a random forest model, a support vector machine model, a BP neural network model, and an extreme learning machine model.

In some embodiments of the first aspect of the present application, the sample statistical analysis module selects some or all of the abnormal indicators from the single abnormal indicators as the combined abnormal indicators according to the ROC curve.

In order to achieve the above object and other related objects, a second aspect of the present application provides a method for detecting tumor markers based on a multi-neural network classifier, which includes: collecting sample data of a tumor marker detection sample, wherein the sample data includes: Test set data and training set data; screen out a plurality of abnormal indicators with high correlation with the tumor according to the sample data; perform detection result evaluation and analysis on a single abnormal indicator based on the sample data data, and/or combine multiple abnormal indicators The abnormal index of the test results is evaluated and analyzed to generate corresponding evaluation and analysis results; model training is performed on multiple classifier models based on the training set data; based on the test set data, the trained classifier models are tested, and Compare the test results with the evaluation and analysis results to judge the validity of the tested classifier model; use the classifier model judged to be effective to detect the tumor marker data, and output the diagnostic result information accordingly .

In order to achieve the above object and other related objects, a third aspect of the present application provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the multi-neural network classifier-based algorithm is implemented. Tumor marker detection methods.

In order to achieve the above purpose and other related purposes, a fourth aspect of the present application provides a detection terminal, comprising: a processor and a memory; the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory , so that the detection terminal executes the multi-neural network classifier-based tumor marker detection method.

As mentioned above, the multi-neural network classifier-based tumor marker detection system, method, terminal, and medium of the present application have the following beneficial effects: The data preprocessing method of component analysis and feature extraction, based on random forest, support vector machine, BP neural network, and the combined classifier model of extreme learning machine classifier model, train the data, and finally get an accuracy, specificity, sensitivity The liver cancer tumor marker classifier that meets the clinical diagnosis can help clinicians reduce the misdiagnosis rate in the initial diagnosis of liver cancer and improve the detection accuracy of liver cancer tumor markers.

Description of drawings

FIG. 1 is a schematic diagram of a tumor marker detection system based on a multi-neural network classifier according to an embodiment of the present application.

FIG. 2 shows a schematic diagram of ROC curves drawn based on six liver cancer tumor markers in an example of the present application.

FIG. 3 is a schematic diagram of a method for detecting tumor markers based on a multi-neural network classifier according to an embodiment of the present application.

FIG. 4 is a schematic structural diagram of a detection terminal in an embodiment of the present application.

Detailed ways

The embodiments of the present application are described below through specific specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. The present application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that the following embodiments and features in the embodiments may be combined with each other under the condition of no conflict.

It should be noted that, in the following description, reference is made to the accompanying drawings, which describe several embodiments of the present application. It is to be understood that other embodiments may be utilized and mechanical, structural, electrical, as well as operational changes may be made without departing from the spirit and scope of the present application. The following detailed description should not be considered limiting, and the scope of embodiments of the present application is limited only by the claims of the issued patent. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application. Spatially related terms, such as "upper," "lower," "left," "right," "below," "below," "lower," "above," "upper," etc., may be used in the text for ease of description The relationship of one element or feature shown in the figures to another element or feature.

In this application, unless otherwise expressly specified and limited, terms such as "installation", "connection", "connection", "fixing", "fixing" and other terms should be understood in a broad sense, for example, it may be a fixed connection or a It is a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, or it can be the internal communication between two components. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific situations.

Also, as used herein, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context dictates otherwise. It should be further understood that the terms "comprising", "comprising" indicate the presence of a stated feature, operation, element, component, item, kind, and/or group, but do not exclude one or more other features, operations, elements, components, The existence, appearance or addition of items, categories, and/or groups. The terms "or" and "and/or" as used herein are to be construed to be inclusive or to mean any one or any combination. Thus, "A, B or C" or "A, B and/or C" means "any of the following: A; B; C; A and B; A and C; B and C; A, B and C" . Exceptions to this definition arise only when combinations of elements, functions, or operations are inherently mutually exclusive in some way.

There are many types of serum tumor markers for liver cancer, but no single tumor marker can diagnose liver cancer clinically. The existence of each tumor marker has certain reference value in diagnosis, but it also has its own limitations. Clinically, doctors generally refer to the parameters of multiple tumor markers for joint detection and diagnosis. The factors of human experience are easy to cause errors in diagnosis. The use of machine learning methods can effectively improve the accuracy of combined diagnosis of multiple tumor markers. Computer-aided diagnosis algorithms based on machine learning have been continuously optimized and improved in recent years, which brings the possibility of constructing a multi-tumor marker combined diagnosis model. In terms of algorithms, some foreign scholars have compared the actual effects of 179 different classifiers on 121 different datasets. The research results show that the effects of different classifiers in different classification scenarios are different. For example, the random forest RF is the strongest on average, but it only won the first place in 9.9% of the data sets; the average level of support vector machine SVM It is close behind, taking first place on 10.7% of the datasets.

In view of the above technical problems, the present application provides a tumor marker detection system, method, terminal, and medium based on a multi-neural network classifier to effectively solve these problems. The main idea of this application is to provide a multi-neural network classifier-based detection scheme for liver cancer tumor markers, data preprocessing methods using principal component analysis and feature extraction, based on random forests, support vector machines, BP neural networks, and extreme learning The combined classifier model of the machine classifier model trains the data, and finally obtains a liver cancer tumor marker classifier whose accuracy, specificity and sensitivity meet the clinical diagnosis, which helps clinicians reduce the misdiagnosis rate in the initial diagnosis of liver cancer and improve liver cancer tumor markers. object detection accuracy.

As shown in FIG. 1 , a schematic diagram of a tumor marker detection system based on a multi-neural network classifier in an embodiment of the present application is shown. The system includes a sample collection module 11 , a data preprocessing module 12 , a sample statistical analysis module 13 , a classifier model training module 14 , a classifier model evaluation and testing module 15 , and a classifier diagnosis method application module 16 .

The sample collection module 11 is used for collecting sample data of the tumor marker detection sample.

In one embodiment, in order to ensure the training effect of the system, the detection samples of liver cancer tumor markers preferably meet the requirements that the data source is real data and that the sample data is classified. Specifically, the detection samples of liver cancer tumor markers are derived from real cases, which can ensure the authenticity of the detection results of the detection system compared with the data amplified by machine learning methods. In addition, the detection samples of liver cancer tumor markers can be divided into two categories: a liver cancer group and a non-liver cancer group, or can be divided into three categories: a liver cancer group, a liver disease group, and a healthy group.

It is worth noting that the number of samples in each group should not be too low, and the number of samples between groups should be kept in the same order of magnitude as possible. Meanwhile, the samples are divided into training set and test set, and the ratio of training set and test set is preferably 7:3. Both the test set and the training set need to contain all classes of samples.

The data preprocessing module 12 is configured to screen out a plurality of abnormal indicators that are highly correlated with liver cancer tumors according to the sample data.

In one embodiment, the data preprocessing module removes attributes irrelevant to subjects before screening out abnormal indicators, thereby improving the accuracy of screening results. In this embodiment, the data preprocessing module performs multiple rounds of screening and finally determines 6 abnormal indicators that are highly correlated with liver cancer tumors.

Specifically, after eliminating irrelevant attribute data, more than 20 abnormal indicators are first screened. The data were filled with missing values by the multiple imputation method, and the mining association rule algorithm was used to screen out about 10 abnormal indicators with the strongest correlation with primary liver cancer from the previously screened more than 20 abnormal indicators. Finally, the principal component analysis algorithm is used for feature selection to reduce the dimension of the data.

In this example, the final screening of liver cancer tumor markers are: alpha-fetoprotein (AFP), other auxiliary reference tumor markers are abnormal prothrombin (PIVKA-II), carcinoembryonic antigen (CEA)), sugar Class antigen 199 (CA199), carbohydrate antigen 242 (CA242), carbohydrate antigen 211 (CA211), carbohydrate antigen 125, sialic acid, ferritin, etc.

It should be noted that the multiple imputation method described in this embodiment refers to the process of replacing each missing value with a vector containing multiple imputed values, that is, replacing each missing value with a series of possible values. The algorithm for mining association rules in this embodiment is, for example, the Apriori algorithm, the core idea of which is to mine frequent itemsets through two stages of candidate set generation and plot downward closure detection.

The sample statistical analysis module 13 is configured to perform detection result evaluation and analysis on a single abnormality index based on the sample data data, and/or perform detection result evaluation and analysis on a plurality of combined abnormality indicators.

In one embodiment, descriptive analysis is performed on the selected tumor marker sample data respectively using statistical software spss12.0. Descriptive analysis refers to an analysis method that analyzes the collected data and obtains various quantitative characteristics that reflect objective phenomena. It includes data central tendency analysis, data dispersion degree analysis, data frequency distribution analysis, and so on.

Specifically, single tumor marker assays and combined assays were analyzed separately. A nonparametric test was used to judge whether the difference between different groups was statistically significant, with α=0.05 as the test standard. The diagnostic sensitivity, specificity, accuracy and ROC curve were used as evaluation indicators. The calculation formulas are shown in the following formulas 1) to 3):

Sensitivity TPR=TP/(TP+FN)×100%; Formula 1)

Specificity TNR=TN/(TN+FP)×100%; Formula 2)

Accuracy=(TP+TN)/(TP+TN+FP+FN)×100%; Equation 3)

Among them, sensitivity refers to the proportion of samples that are actually positive, specificity refers to the proportion of samples that are actually negative, and accuracy refers to the proportion of true positives and true negatives in the number of subjects. In the formula, TP means true positive, TN means true negative, FP means false positive, and FN means false negative.

In one embodiment, the sample statistical analysis module selects some or all of the abnormality indexes from the single abnormality indexes as the combined abnormality indexes according to the ROC curve. The ROC curve refers to the receiver operating characteristic curve, which is a comprehensive index reflecting the continuous variables of sensitivity and specificity. It uses the composition method to reveal the relationship between sensitivity and specificity. Multiple different critical values are used to calculate a series of sensitivity and specificity, and then draw a curve with sensitivity as the ordinate and (1-specificity) as the abscissa. The larger the area under the curve, the higher the diagnostic accuracy. . On the ROC curve, the point closest to the upper left of the graph is the critical value with high sensitivity and specificity.

It is worth noting that although the existence of each tumor marker has certain reference value in diagnosis, it also has its own limitations. At present, no single tumor marker can diagnose liver cancer in clinical practice. The present application proposes the combined detection of multiple tumor markers such as AFP, abnormal prothrombin, carcinoembryonic antigen, carbohydrate antigen 199, carbohydrate antigen 242, carbohydrate antigen 211, etc., which improves the accuracy of liver cancer diagnosis and provides a computer-aided diagnostic method. It can help clinicians reduce missed diagnosis and misdiagnosis, and has extremely high clinical value for the detection of early liver cancer.

The classifier model training module 14 is configured to perform model training on a plurality of classifier models based on the training set data.

In one embodiment, based on the data preprocessed by the data preprocessing module, model training is performed on classifier models such as random forest model, support vector machine model, BP neural network model, and extreme learning machine model. Among them, the random forest model is resampled based on the Bootstrap method; the implementation function model of the support vector machine algorithm is C-SVC; the BP neural network uses the genetic algorithm to improve the network, and the input layer of the neural network is 5-10 neurons. The hidden layer is 15-20 neurons and the output layer is 1 neuron.

In one embodiment, different expected output values are set according to different types of sample data. Taking the sample data of the liver cancer group, the sample data of the liver disease group and the sample data of the healthy group as examples, the expected output value of the healthy group is 0.1, the expected output value of the liver disease group is 0.5, and the expected output value of the liver cancer group is 0.9. Get the trained model.

It is worth noting that due to the different effects of different classifiers in different classification scenarios, for example, random forest RF is the strongest on average, but it only won the first place in 9.9% of the data sets; support vector machine SVM's The average is a close second, taking first place on 10.7% of the datasets. Therefore, the present application provides a method for assisted diagnosis of liver cancer tumor markers based on a multi-neural network classifier, which is superior in accuracy to the assisted diagnosis of liver cancer tumor markers based on a single neural network model.

The classifier model evaluation and testing module 15 is used to test the trained classifier model based on the test set data, and compare the test results with the evaluation and analysis results of the sample statistical analysis module, so as to judge the quality of the classifier model. The validity of the tested classifier model.

Specifically, for each trained classifier model, it needs to be verified on the test set data. When the accuracy index is greater than the evaluation analysis result output by the sample statistical analysis module, the classifier model is considered to be valid, otherwise Adjust the parameters and retrain a new classifier model through the classifier model evaluation and testing module 15 .

The classifier diagnosis method application module 16 is used for detecting the tumor marker data by using the classifier model that is judged to be valid, and outputting diagnosis result information accordingly. For example, a patient or a doctor can use the clinically detected liver cancer tumor marker data as the input data of the classifier model provided in this application, and the classifier model automatically gives the diagnosis result information for reference.

It should be understood that the division of each module of the above system is only a division of logical functions, and may be fully or partially integrated into a physical entity in actual implementation, or may be physically separated. And these modules can all be implemented in the form of software calling through processing elements; they can also all be implemented in hardware; some modules can also be implemented in the form of calling software through processing elements, and some modules can be implemented in hardware. For example, the classifier model training module can be a separately established processing element, or can be integrated into a certain chip of the above-mentioned system. In addition, it can also be stored in the memory of the above-mentioned system in the form of program code. A processing element invokes and executes the functions of the above classifier model training module. The implementation of other modules is similar. In addition, all or part of these modules can be integrated together, and can also be implemented independently. The processing element described here may be an integrated circuit with signal processing capability. In the implementation process, each step of the above-mentioned method or each of the above-mentioned modules can be completed by an integrated logic circuit of hardware in the processor element or an instruction in the form of software.

For example, the above modules may be one or more integrated circuits configured to implement the above methods, such as: one or more specific integrated circuits (Application Specific Integrated Circuit, ASIC for short), or one or more microprocessors ( digital signal processor, referred to as DSP), or, one or more Field Programmable Gate Array (Field Programmable Gate Array, referred to as FPGA) and the like. For another example, when one of the above modules is implemented in the form of processing element scheduling program code, the processing element may be a general-purpose processor, such as a central processing unit (Central Processing Unit, CPU for short) or other processors that can call program codes. For another example, these modules can be integrated together and implemented in the form of a system-on-a-chip (SOC for short).

In order to facilitate the understanding of those skilled in the art, a number of cases in a hepatobiliary surgery hospital from March 2017 to October 2018 are taken as the research objects.

The sample collection module collects 139 primary liver cancer cases and 100 non-primary liver cancer cases (including 20 benign tumor cases and 80 other liver disease cases), among which 120 cases are selected as the training set (including primary liver cancer cases). The remaining 119 cases were used as the test set (including 69 primary liver cancer cases and 50 non-primary liver cancer cases).

The data preprocessing module first removes attributes irrelevant to patients, and selects more than 20 abnormal indicators. The missing values were filled in a small part of the data by the multiple imputation method, and the Apriori correlation algorithm was used to obtain about 10 abnormal indicators with the strongest correlation with primary liver cancer. Finally, the data were analyzed by the feature selection and principal component analysis algorithm. Dimensional reduction treatment, the final screening of liver cancer tumor markers are alpha-fetoprotein (AFP), carcinoembryonic antigen (CEA), abnormal prothrombin (PIVKA-II), carbohydrate antigen 199 (CA199), carbohydrate antigen 242 ( CA242), and carbohydrate antigen 211 (CA211).

The sample statistical analysis module uses the statistical software spss12.0 to descriptively analyze the selected tumor marker sample data. The single tumor marker detection and combined detection were analyzed separately. A nonparametric test was used to judge whether the difference between different groups was statistically significant, with α=0.05 as the test standard. The diagnostic sensitivity, specificity, accuracy and ROC curve were used as evaluation indicators.

Taking the BP neural network as an example for analysis, as shown in FIG. 2 , the ROC curves drawn by the six liver cancer tumor markers as the research objects in this example are shown. Among them, the diagonal line 21 is the reference line, the curve 22 represents the carbohydrate antigen 211, the curve 23 represents the cancer blank antigen, the curve 24 represents the carbohydrate antigen 242, the curve 25 represents the carbohydrate antigen 199, the curve 26 represents the alpha-fetoprotein, and the curve 27 Indicates abnormal prothrombin.

Comparing the area under each ROC curve, it can be seen that the ROC curves of abnormal prothrombin and alpha-fetoprotein, that is, the area under curve 27 and curve 26 are the largest, so they have the greatest diagnostic value for liver cancer; while carbohydrate antigen 242 and carbohydrate The ROC curve of antigen 211, that is, curve 24 and curve 22 are close to the diagonal line 21, and have little diagnostic value for combined detection of liver cancer. Therefore, combined detection can discard these two indicators as training input. Compared with the test results of each single index in the joint detection group, the difference was statistically significant (P<0.05). Table 1 below is the evaluation index table of the samples.

Table 1 Sample Statistical Analysis Table

The classifier model training module conducts model training on classifier models such as random forest model, support vector machine model, BP neural network model, and extreme learning machine model based on the data preprocessed by the data preprocessing module. Among them, the random forest model is resampled based on the Bootstrap method; the implementation function model of the support vector machine algorithm is C-SVC; the BP neural network uses the genetic algorithm to improve the network, and the input layer of the neural network is 5-10 neurons. The hidden layer is 15-20 neurons and the output layer is 1 neuron.

Different expected output values are set according to different types of sample data. Taking the sample data of the liver cancer group, the sample data of the liver disease group and the sample data of the healthy group as examples, the expected output value of the healthy group is 0.1, the expected output value of the liver disease group is 0.5, and the expected output value of the liver cancer group is 0.9. Get the trained model. The trained model needs to select the tumor marker with the highest diagnostic value as the joint detection index according to the ROC curve. In this embodiment, alpha-fetoprotein and abnormal prothrombin are selected as the input index of the joint detection.

The classifier model evaluation and testing module For each trained classifier model, it needs to be verified on the test set data. When the accuracy index is greater than the evaluation analysis result output by the sample statistical analysis module, the classifier is considered to be The model is valid, otherwise adjust the parameters, and retrain a new classifier model through the classifier model evaluation and testing module. The final classifier model is constructed based on the four classifiers output by the classifier model training module using the idea of decision tree.

In this embodiment, the performance of the four classifiers on the test set data is shown in Table 2. Comparing Table 2 with Table 1 above, since the accuracy in Table 2 is higher than that in Table 1, the classifier model is considered to be effective.

classifier model specificity Sensitivity Accuracy BP neural network 98.60% 76.314% 89.92% Support Vector Machine SVM 98% 95.65% 96.64% Random Forest RF 100% 97.10 98.32% Extreme Learning Machine ELM 90% 91.30% 90.76%

The classifier diagnosis method application module uses the classifier model judged to be valid to detect the tumor marker data, and outputs the diagnosis result information accordingly. For example, a patient or a doctor can use the clinically detected liver cancer tumor marker data as the input data of the classifier model provided in this application, and the classifier model automatically gives the diagnosis result information for reference.

As shown in FIG. 3 , a schematic flowchart of a method for detecting tumor markers based on a multi-neural network classifier according to an embodiment of the present application is shown.

In some embodiments, the method may be applied to a controller, such as an ARM controller, an FPGA controller, a SoC controller, a DSP controller, or an MCU controller, among others. In some embodiments, the method may also be applied to include a memory, a memory controller, one or more processing units (CPUs), peripheral interfaces, RF circuits, audio circuits, speakers, microphones, input/output (I/ O) Computers with components such as subsystems, display screens, other output or control devices, and external ports; such computers include, but are not limited to, such as desktop computers, notebook computers, tablet computers, smart phones, smart TVs, personal digital assistants (Personal Digital Assistants) Digital Assistant, referred to as PDA) and other personal computers. In other embodiments, the method can also be applied to a server, and the server can be arranged on one or more entity servers according to various factors such as function and load, and can also be composed of a distributed or centralized server cluster.

In this embodiment, the method includes step S31, step S32, step S33, step S34, step S35, and step S36.

In step S31, sample data of the tumor marker detection sample is collected, wherein the sample data includes test set data and training set data.

In step S32, a plurality of abnormal indicators with high correlation with the tumor are screened out according to the sample data.

In step S33, based on the sample data data, a detection result evaluation and analysis is performed on a single abnormality index, and/or a detection result evaluation and analysis is performed on a plurality of combined abnormality indicators, so as to generate a corresponding evaluation and analysis result.

In step S34, model training is performed on a plurality of classifier models based on the training set data.

In step S35, the trained classifier model is tested based on the test set data, and the test results are compared with the evaluation and analysis results, so as to judge the validity of the tested classifier model.

In step S36, the tumor marker data is detected by using the classifier model that is judged to be valid, so as to output diagnosis result information.

It should be noted that the implementation of the tumor marker detection method based on the multi-neural network classifier in this embodiment is similar to the implementation of the tumor marker detection system based on the multi-neural network classifier above, so it will not be repeated.

In addition, those skilled in the art can understand that all or part of the steps of implementing the above method embodiments may be completed by hardware related to computer programs. The aforementioned computer program may be stored in a computer-readable storage medium. When the program is executed, the steps including the above method embodiments are executed; and the foregoing storage medium includes: ROM, RAM, magnetic disk or optical disk and other media that can store program codes.

As shown in FIG. 4 , a schematic structural diagram of a detection terminal in another embodiment of the present application is shown. The detection terminal provided in this example includes: a processor 41, a memory 42, a transceiver 43, a communication interface 44, and a system bus 45; the memory 42 and the communication interface 44 are connected to the processor 41 and the transceiver 43 through the system bus 45 and complete each other The memory 42 is used to store the computer program, the communication interface 44 and the transceiver 43 are used to communicate with other devices, and the processor 41 is used to run the computer program, so that the detection terminal executes each step of the above detection method.

The system bus mentioned above may be a Peripheral Component Interconnect (PCI for short) bus or an Extended Industry Standard Architecture (Extended Industry Standard Architecture, EISA for short) bus or the like. The system bus can be divided into address bus, data bus, control bus and so on. For ease of presentation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus. The communication interface is used to realize the communication between the database access device and other devices (eg client, read-write library and read-only library). The memory may include random access memory (Random Access Memory, RAM for short), and may also include non-volatile memory (non-volatile memory), such as at least one disk memory.

The above-mentioned processor may be a general-purpose processor, including a central processing unit (Central Processing Unit, referred to as CPU), a network processor (Network Processor, referred to as NP), etc.; may also be a digital signal processor (Digital Signal Processing, referred to as DSP) , Application Specific Integrated Circuit (ASIC for short), Field-Programmable Gate Array (FPGA for short) or other programmable logic devices, discrete gate or transistor logic devices, and discrete hardware components.

To sum up, the present application provides a tumor marker detection system, method, terminal, and medium based on a multi-neural network classifier. The extracted data preprocessing method is based on random forest, support vector machine, BP neural network, and the combined classifier model of the extreme learning machine classifier model to train the data, and finally obtain an accuracy, specificity, and sensitivity that meet the requirements of clinical diagnosis. The liver cancer tumor marker classifier helps clinicians reduce the misdiagnosis rate in the initial diagnosis of liver cancer and improve the detection accuracy of liver cancer tumor markers. Therefore, the present application effectively overcomes various shortcomings in the prior art and has high industrial application value.

The above-mentioned embodiments merely illustrate the principles and effects of the present application, but are not intended to limit the present application. Anyone skilled in the art can make modifications or changes to the above embodiments without departing from the spirit and scope of the present application. Therefore, all equivalent modifications or changes made by those with ordinary knowledge in the technical field without departing from the spirit and technical idea disclosed in this application should still be covered by the claims of this application.

Claims (18)

1. a tumor marker detection system based on a multi-neural network classifier, is characterized in that, comprising: a sample collection module for collecting sample data of tumor marker detection samples, wherein the sample data includes test set data and training set data; a data preprocessing module, configured to screen out a plurality of abnormal indicators that are highly correlated with the tumor according to the sample data; A sample statistical analysis module, configured to perform detection result evaluation and analysis on a single abnormal index based on the sample data data, and/or perform detection result evaluation and analysis on a plurality of combined abnormal indicators; A classifier model training module for performing model training on multiple classifier models based on the training set data; The classifier model evaluation and testing module is used to test the trained classifier model based on the test set data, and compare the test results with the evaluation and analysis results of the sample statistical analysis module, so as to judge the the validity of the tested classifier model; The classifier diagnosis method application module is used for detecting the tumor marker data by using the classifier model judged to be effective, so as to output the diagnosis result information. 2 . The system according to claim 1 , wherein the sample collection module collects clinical real sample data and classifies the collected sample data. 3 . 3. The system according to claim 2, wherein the tumor markers comprise liver cancer tumor markers; wherein, the classification method of the sample data comprises: dividing the sample data into liver cancer group sample data and non-liver cancer group Sample data, and/or divided into liver cancer group sample data, liver disease group sample data and healthy group sample data. 4 . The system according to claim 1 , wherein the data preprocessing module is used to perform feature selection based on a principal component analysis algorithm, so as to perform dimension reduction on a plurality of abnormal indicators initially screened out according to the sample data. 5 . treatment, and obtain the abnormal index with high correlation with the tumor. 5. The system according to claim 4, wherein the abnormal index with high correlation with tumor comprises abnormal index with high correlation with liver cancer tumor, which includes alpha-fetoprotein, and also includes: abnormal thrombin Pro, carcinoembryonic antigen, carbohydrate antigen 199, carbohydrate antigen 242, carbohydrate antigen 211, carbohydrate antigen 125, sialic acid, or ferritin. 6. system according to claim 1, is characterized in that, the evaluation index of test result evaluation analysis comprises sensitivity, specificity, and accuracy; Wherein, sensitivity represents the ratio that is judged to be positive in the sample that is actually positive, specificity Represents the proportion of samples that are actually negative, and accuracy represents the ratio of true positives and true negatives to the total number of subjects tested. 7. The system according to claim 1, wherein the classifier model comprises: a random forest model, a support vector machine model, a BP neural network model, and an extreme learning machine model. 8 . The system according to claim 1 , comprising: the sample statistical analysis module selects some or all of the abnormal indicators from the single abnormal indicators as the joint abnormal indicators according to the ROC curve. 9 . 9. A method for detecting tumor markers based on a multi-neural network classifier, comprising: collecting sample data of tumor marker detection samples, wherein the sample data includes test set data and training set data; Screening out a plurality of abnormal indicators that are highly correlated with the tumor according to the sample data; Based on the sample data data, the detection result evaluation analysis is performed on a single abnormal index, and/or the detection result evaluation analysis is performed on a plurality of combined abnormal indicators, so as to generate a corresponding evaluation analysis result; Model training is performed on multiple classifier models based on the training set data; Test the trained classifier model based on the test set data, and compare the test results with the evaluation analysis results, so as to judge the validity of the tested classifier model; The tumor marker data is detected by using the classifier model that is judged to be effective, so as to output the diagnosis result information. 10 . The method according to claim 9 , wherein the method comprises: collecting clinical real sample data and classifying the collected sample data. 11 . 11. The method according to claim 10, wherein the tumor markers comprise liver cancer tumor markers; wherein, the classification method of the sample data comprises: dividing the sample data into liver cancer group sample data and non-liver cancer group Sample data, and/or divided into liver cancer group sample data, liver disease group sample data and healthy group sample data. 12. The method according to claim 9, wherein the method comprises: performing feature selection based on a principal component analysis algorithm, so as to perform dimensionality reduction processing on a plurality of abnormal indicators initially screened according to the sample data, and The abnormal index with high correlation with the tumor is obtained. 13. The method according to claim 12, wherein the abnormal index with high correlation with tumor comprises an abnormal index with high correlation with liver cancer tumor, which comprises alpha-fetoprotein, and further comprises: abnormal thrombin Pro, carcinoembryonic antigen, carbohydrate antigen 199, carbohydrate antigen 242, carbohydrate antigen 211, carbohydrate antigen 125, sialic acid, or ferritin. 14. method according to claim 9, is characterized in that, the evaluation index of test result evaluation analysis comprises sensitivity, specificity, and accuracy; Wherein, sensitivity represents the ratio that is judged to be positive in actual positive samples, specificity Represents the proportion of samples that are actually negative, and accuracy represents the ratio of true positives and true negatives to the total number of subjects tested. 15. The method according to claim 9, wherein the classifier model comprises: a random forest model, a support vector machine model, a BP neural network model, and an extreme learning machine model. 16 . The method according to claim 9 , wherein the method comprises: selecting part or all of the abnormality indexes from the single abnormality indexes as the joint abnormality indexes according to the ROC curve. 17 . 17. A computer-readable storage medium on which a computer program is stored, wherein when the computer program is executed by a processor, the multi-neural network classifier-based tumor according to any one of claims 9 to 16 is implemented. Marker detection methods. 18. A detection terminal, comprising: a processor and a memory; the memory is used to store computer programs; The processor is configured to execute the computer program stored in the memory, so that the terminal executes the method for detecting tumor markers based on a multi-neural network classifier according to any one of claims 9 to 16.