Prediction model training method and system based on early stage of pancreatic cancer, medium and equipment
By combining periodontitis indicators and machine learning technology, training and optimization of prediction models, the problem of low prediction accuracy in early stage pancreatic cancer is solved, and higher prediction accuracy and reliability are achieved, providing support for early diagnosis and treatment.
Patent Information
- Application Number
- CN202411342740.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-25
- Publication Date
- 2025-02-18
AI Technical Summary
The prior art is difficult to effectively predict the early stages of pancreatic cancer, which leads to patients being diagnosed in late stages, reducing the success rate of treatment.
Combining periodontitis index and machine learning technology, a variety of prediction models (such as random forest, support vector machine and logistic regression model) are trained, and the model is optimized through grid search and ten-fold cross-validation to improve the accuracy of early prediction of pancreatic cancer and reliability.
It improves the accuracy and reliability of early prediction of pancreatic cancer, provides strong support for the early diagnosis and treatment of pancreatic cancer, and reduces the treatment difficulties caused by advanced diagnosis.
Smart Images
Figure CN119480089A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of medical information processing, and in particular to a method, system, medium and equipment for training a prediction model based on early stage pancreatic cancer. Background Art
[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.
[0003] Pancreatic cancer is a rapidly developing malignant digestive tract tumor with extremely poor treatment and prognosis, with a 5-year survival rate of less than 10%. Because the human pancreas is located behind the peritoneum and the complex anatomical structure around it makes it particularly difficult to directly observe the state of the pancreas through conventional imaging methods. In addition, conventional biomarkers used to detect pancreatic cancer, such as serum cancer antigen 19-9 (CA19-9), have a high sensitivity only in advanced gastrointestinal adenocarcinoma, and have poor sensitivity for small tumors in the curable stage. Therefore, most pancreatic cancer patients are currently diagnosed in the advanced stage with cancer cell metastasis, which greatly reduces the success rate of patient treatment.
[0004] In recent years, the medical community has begun to pay attention to the potential connection between periodontitis and pancreatic cancer. Periodontitis is a chronic infectious disease caused by plaque microorganisms that destroy periodontal supporting tissues. It can trigger systemic inflammation and immune responses, and become a potential risk factor for many systemic diseases. Many studies have shown that patients with periodontitis have a higher risk of pancreatic cancer, which suggests that periodontitis may be an early warning sign of pancreatic cancer. However, research on the specific mechanism of action between periodontitis and pancreatic cancer is still in its early stages, and the potential connection between the two has not yet been clarified. Therefore, the early prediction method of pancreatic cancer based on periodontitis indicators has great research value and clinical application prospects. Summary of the invention
[0005] In order to solve the technical problems existing in the above-mentioned background technology, the present invention provides a method, system, medium and equipment for training a prediction model based on early pancreatic cancer. The present invention aims to combine periodontitis indicators and machine learning technology to improve the accuracy and reliability of early prediction of pancreatic cancer and provide strong support for the early diagnosis and treatment of pancreatic cancer.
[0006] In order to achieve the above object, the present invention adopts the following technical solution:
[0007] The first aspect of the present invention provides a prediction model training method based on early stage pancreatic cancer.
[0008] A prediction model training method based on early pancreatic cancer, comprising:
[0009] Obtain the patient's periodontitis indicators and pancreatic cancer test report results to construct a data set; divide the data set into a training set and a test set;
[0010] Based on the periodontitis indicators in the training set, relevant features were extracted;
[0011] Using relevant features as input and corresponding pancreatic cancer detection report results as output, multiple prediction models are trained;
[0012] Grid search and ten-fold cross validation are used to optimize multiple prediction models to obtain multiple trained prediction models;
[0013] The test set is used to make predictions for multiple trained prediction models, and the prediction model with the best output result is selected as the final prediction model.
[0014] Furthermore, the periodontitis indicators include: plaque index, presence or absence of furcation lesions, periodontal probing depth, and range of attachment loss.
[0015] Furthermore, the multiple prediction models include a random forest model, which includes multiple decision trees and a combiner. Each decision tree classifies the input of the decision tree based on its learned rules to obtain a classification result; the combiner integrates the classification results of all decision trees and takes the classification results with the largest number as the final output result.
[0016] Furthermore, the classification process of each decision tree includes: predicting according to the presence or absence of root bifurcation disease characteristics, and outputting classification results of excluding pancreatic cancer, confirming pancreatic cancer and the first suspected pancreatic cancer; for the result of the first suspected pancreatic cancer, predicting according to the plaque index characteristics, and outputting classification results of excluding pancreatic cancer, confirming pancreatic cancer and the second suspected pancreatic cancer; for the result of the second suspected pancreatic cancer, predicting according to the periodontal probing depth characteristics, and outputting classification results of excluding pancreatic cancer, confirming pancreatic cancer and the third suspected pancreatic cancer; for the result of the third suspected pancreatic cancer, predicting according to the attachment loss range characteristics, and outputting classification results of excluding pancreatic cancer, confirming pancreatic cancer and the fourth suspected pancreatic cancer; summarizing and outputting the classification results of excluding pancreatic cancer, confirming pancreatic cancer and suspected pancreatic cancer.
[0017] Furthermore, the multiple prediction models include a support vector machine model, which places periodontitis indicators into a feature space; defines two hyperplanes based on the periodontitis indicators; obtains expressions of the two hyperplanes based on the training set to divide the feature space into three parts, namely, excluding pancreatic cancer, confirming pancreatic cancer, and suspected pancreatic cancer; and inputs relevant features of the periodontitis indicators in the test set into the feature space to obtain classification results.
[0018] Furthermore, the multiple prediction models include a logistic regression model, which linearly fits the relevant features of the periodontitis indicators in the training set with the corresponding pancreatic cancer detection report results to obtain a multivariate linear regression equation; maps the multivariate linear regression equation to [0, 1] through a sigmoid activation function, and obtains a classification threshold D based on the pancreatic cancer detection report results; inputs the relevant features of the periodontitis indicators in the prediction set into the multivariate linear regression equation to obtain an equation result; maps the equation result to [0, 1] through a sigmoid activation function to obtain a mapping result; compares the mapping result with the classification threshold to obtain a classification result.
[0019] Furthermore, the grid search and ten-fold cross validation are used to optimize multiple prediction models, and the method includes: defining a hyperparameter network for each prediction model; traversing all possible hyperparameter combinations on the hyperparameter network, and using ten-fold cross validation to evaluate the performance of each hyperparameter combination, and selecting the hyperparameter combination with the best performance as the optimal hyperparameter of the prediction model.
[0020] The second aspect of the present invention provides a prediction model training system based on early stage pancreatic cancer.
[0021] A prediction model training system based on early pancreatic cancer, comprising:
[0022] A data acquisition module is configured to: obtain the patient's periodontitis index and pancreatic cancer test report results to construct a data set; divide the data set into a training set and a test set;
[0023] A feature extraction module is configured to: extract relevant features based on the periodontitis indicators in the training set;
[0024] A training module is configured to: take relevant features as input and corresponding pancreatic cancer detection report results as output, and train multiple prediction models;
[0025] An optimization module is configured to: optimize multiple prediction models by using grid search and ten-fold cross validation to obtain multiple trained prediction models;
[0026] The test module is configured to: use the test set to predict the multiple trained prediction models, and select the prediction model with the best output result as the final prediction model.
[0027] A third aspect of the present invention provides a computer-readable storage medium.
[0028] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps in the method for training a prediction model based on early stage pancreatic cancer as described in the first aspect above.
[0029] A fourth aspect of the present invention provides a computer device.
[0030] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the method for training a prediction model based on early stage pancreatic cancer as described in the first aspect above are implemented.
[0031] Compared with the prior art, the present invention has the following beneficial effects:
[0032] The present invention proposes a prediction model training method, system, medium and equipment based on early pancreatic cancer, combines periodontitis indicators and machine learning technology to train a prediction model based on periodontitis indicators to achieve early prediction of pancreatic cancer; the prediction model is built into the detection system, and the user inputs the corresponding periodontitis indicators in the operation interface to call the prediction model and complete the prediction. The problem of low accuracy and poor reliability of early prediction of pancreatic cancer in the prior art is solved, the accuracy of early prediction of pancreatic cancer is improved, and strong support is provided for the early diagnosis and treatment of pancreatic cancer. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.
[0034] Figure 1 It is a flow chart of the prediction model training method based on early stage pancreatic cancer disclosed in the present invention;
[0035] Figure 2 It is a prediction flow chart using random forest as a prediction model disclosed in the present invention;
[0036] Figure 3 It is a structural diagram of a decision tree disclosed in the present invention;
[0037] Figure 4 It is a prediction flow chart of the present invention using a support vector machine as a prediction model;
[0038] Figure 5 It is a structural diagram of an application device of a pancreatic cancer prediction system based on periodontitis indicators disclosed in the present invention;
[0039] Figure 6 It is a structural diagram of a pancreatic cancer prediction system based on periodontitis indicators disclosed in the present invention;
[0040] Among them, 401, tablet computer, 402, smart phone, 403, PC, 404, server, 405, database, 501, patient information input module, 502, prediction result output module, 503, prediction module, 504, prediction model retraining module, 505, data storage module. DETAILED DESCRIPTION
[0041] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.
[0042] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.
[0043] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.
[0044] Embodiment 1
[0045] This embodiment provides a prediction model training method based on early pancreatic cancer, including:
[0046] Obtain the patient's periodontitis indicators and pancreatic cancer test report results to construct a data set; divide the data set into a training set and a test set;
[0047] Based on the periodontitis indicators in the training set, relevant features were extracted;
[0048] Using relevant features as input and corresponding pancreatic cancer detection report results as output, multiple prediction models are trained;
[0049] Grid search and ten-fold cross validation are used to optimize multiple prediction models to obtain multiple trained prediction models;
[0050] The test set is used to make predictions for multiple trained prediction models, and the prediction model with the best output result is selected as the final prediction model.
[0051] The present invention aims to combine periodontitis indicators and machine learning technology to improve the accuracy and reliability of early prediction of pancreatic cancer and provide strong support for the early diagnosis and treatment of pancreatic cancer.
[0052] The present embodiment is described in detail below with reference to the accompanying drawings. Figure 1 As shown:
[0053] Step 1: Collect data sets of periodontitis patients: including periodontitis indicators (plaque index, soft plaque index, gingival bleeding index, presence of furcation lesions, periodontal probing depth, and range of attachment loss, etc.) and patients' pancreatic cancer test report results, including confirmed pancreatic cancer, suspected pancreatic cancer, and excluded pancreatic cancer.
[0054] Step 2: Data preprocessing: including data gap filling and data cleaning. Data gap filling uses KNN interpolation to fill in the missing data. Data cleaning is to remove variables or special discrete values that are not related to the prediction to obtain a data set of 1,000 patients.
[0055] Step 3: Dataset division: The dataset is randomly divided into 10 groups, 8 groups are training sets and 2 groups are test sets.
[0056] Step 4: Feature extraction: Normalize the dataset and extract features.
[0057] Step 5: Model training: Since different machine learning models have different working principles and assumptions, different models may show different performance on different data sets and tasks. Therefore, it is necessary to select multiple models to find the most suitable prediction model. In addition, by comparing the performance of different models on the same data set, the effectiveness of the model can be verified. Therefore, three machine learning models, random forest, support vector machine and logistic regression, are selected for training. Figure 2 As shown in Figure 1, random forest is an ensemble learning method that integrates multiple estimators (decision trees) internally. Each decision tree will classify the sample based on the rules it has learned. Finally, the classification results of all decision trees are integrated through the combiner to obtain the final cancer prediction result.
[0058] The classification structure of the decision tree is as follows Figure 3 As shown. According to the prediction of the presence or absence of root bifurcation lesions, the output is the results of excluding pancreatic cancer, confirming pancreatic cancer and the first suspected pancreatic cancer; for the result of the first suspected pancreatic cancer, the prediction is made according to the plaque index, and the output is the results of excluding pancreatic cancer, confirming pancreatic cancer and the second suspected pancreatic cancer; for the result of the second suspected pancreatic cancer, the prediction is made according to the periodontal probing depth index, and the output is the results of excluding pancreatic cancer, confirming pancreatic cancer and the third suspected pancreatic cancer; for the result of the third suspected pancreatic cancer, the prediction is made according to the attachment loss range index, and the output is the results of excluding pancreatic cancer, confirming pancreatic cancer and the fourth suspected pancreatic cancer.
[0059] During the training process, each decision tree will randomly extract the same number of periodontitis indicators and pancreatic cancer test report results from the data set, and continuously classify the prediction results of pancreatic cancer based on the obtained periodontitis indicator data until the decision tree depth or classification accuracy reaches the specified requirements. After multiple trainings, multiple decision tree classification structures are obtained. When a classification prediction task is required, the patient's periodontitis indicators are input, and all decision trees will classify and predict the patient's pancreatic cancer test results according to their own classification structure, and upload the classification prediction results to the combiner. The combiner will collect the classification prediction results of all decision numbers and output the classification results with the largest number as the final result.
[0060] Support vector machine (SVM) achieves classification prediction by using hyperplanes that separate samples of different categories in feature space. Figure 4 As shown in Figure 1, the collected periodontitis indicators are placed in the feature space (since there are multiple periodontitis indicators, the feature space should be multidimensional. Here, for the convenience of expression, two periodontitis indicators are used to simplify the feature space into a two-dimensional plane for representation). During the training process, two hyperplanes w are defined according to the pancreatic cancer diagnosis results. T i .x+b i =0, i=1,2, where x represents the periodontitis index feature vector, and w is the hyperplane normal vector w=(w1,w1,...,w n ), b is the displacement. For a given periodontitis dataset T = {x1, x2, ..., x n}, the geometric interval of the hyperplane with respect to the sample points of the data set can be expressed as According to the principle of maximizing geometric intervals, the Lagrange multiplier method is used to calculate w and b, and the expressions of two hyperplanes are obtained. The two hyperplanes will divide the feature space into three parts, corresponding to the three results of the pancreatic cancer detection report. In the prediction stage, the patient's periodontitis index is input into the above feature space, and the model will determine which part of the feature space the input periodontitis index belongs to, and then output the classification prediction result.
[0061] Logistic regression can be divided into two parts: linear regression and sigmoid function activation. In the training phase, the periodontitis index and pancreatic cancer test report results are linearly fitted to obtain a multivariate linear regression equation f(x) = w T x+b, where f(x) is the pancreatic cancer test report result, x is the characteristic vector composed of periodontitis indicators, and w TThe weight parameters of the features are learned through maximum likelihood estimation or gradient descent algorithm. Secondly, the obtained multivariate linear regression equation result f(x) is mapped to [0, 1] through the sigmoid activation function, and a classification threshold D is obtained according to the pancreatic cancer detection report result. In the prediction stage, the patient's periodontitis index is brought into the multivariate linear regression equation to obtain the equation result f T (x), f is activated by the sigmoid function. T (x) is mapped to [0, 1], and the mapping result is compared with the classification threshold D obtained in the training stage, so that the classification prediction of pancreatic cancer can be realized through periodontitis indicators.
[0062] Step 6: Model optimization: Optimize the model using grid search and ten-fold cross validation. Grid search is an exhaustive search method used to determine the optimal hyperparameter combination of the model. First, a hyperparameter grid is predefined for each of the three prediction models (random forest: number of decision trees (n_estimators) and maximum depth (max_depth); support vector machine: kernel function (kernel) and penalty coefficient (C); logistic regression: regularization reinforcement strength (C) and solver (solver)). Second, all possible hyperparameter combinations are traversed on different hyperparameter grids, and the performance of each combination is evaluated using cross validation. Finally, the combination with the best performance is selected as the optimal hyperparameter of the model. The cross validation adopts the ten-fold validation method, that is, the training data set is divided into 10 non-overlapping subsets, and one of the subsets is selected as the validation set each time, and the remaining 9 subsets are used as the training set to train the model. This process is repeated 10 times, and each subset is used as the validation set once. Finally, the average of the 10 validation results is calculated as the performance indicator of the model under a certain hyperparameter combination.
[0063] Step 7: Get prediction results: Select the model with the best output results as the final prediction model and output the prediction results.
[0064] Step 8: Model evaluation: Based on the test results in step (7), the model effect is measured using three evaluation indicators: accuracy, precision, and recall. The model with the best comprehensive performance is selected as the final prediction model to predict early pancreatic cancer as a basis for doctors' auxiliary diagnosis. As shown in Table 1, the comprehensive evaluation performance of random forest is the best. The hyperparameter settings of random forest are shown in Table 2. The initial number of decision trees is 300 and the maximum depth is 30. After grid search optimization, the number of decision trees is 800 and the maximum depth is 25.
[0065] Table 1 Model evaluation
[0066] Model Accuracy Accuracy Recall Random Forest 0.94 0.97 0.95 Support Vector Machine 0.94 0.94 0.87 Logistic Regression 0.94 0.96 0.92
[0067] Table 2 Hyperparameter settings of random forest
[0068] Hyperparameters Hyperparameter Values N_estimators 800 Max_depth 25 Min_samples_split None Min_samples_leaf None Max_leaf_nodes None
[0069] Embodiment 2
[0070] This embodiment provides a prediction model training system based on early pancreatic cancer, including:
[0071] A data acquisition module is configured to: obtain the patient's periodontitis index and pancreatic cancer test report results to construct a data set; divide the data set into a training set and a test set;
[0072] A feature extraction module is configured to: extract relevant features based on the periodontitis indicators in the training set;
[0073] A training module is configured to: take relevant features as input and corresponding pancreatic cancer detection report results as output, and train multiple prediction models;
[0074] An optimization module is configured to: optimize multiple prediction models by using grid search and ten-fold cross validation to obtain multiple trained prediction models;
[0075] The test module is configured to: use the test set to predict the multiple trained prediction models, and select the prediction model with the best output result as the final prediction model.
[0076] Figure 5 The application device architecture diagram of the pancreatic cancer prediction system based on periodontitis indicators is shown. Figure 5 As shown, the system application device architecture includes a tablet computer 401, a smart phone 402, a PC 403, a server 404, and a database 405. Users can use the tablet computer 401, the smart phone 402, and the PC 403 to interact with the server 404 through the network, and the server 404 regularly stores data in the database 405 to achieve long-term data storage.
[0077] Figure 6 The pancreatic cancer prediction system based on periodontitis indicators disclosed in the present invention is as follows: Figure 6As shown, it includes: a patient information input module 501, a prediction result output module 502, a prediction module 503, a prediction model retraining module 504 and a data storage module 505. In the specific use process, the patient information input module 501 can realize the interaction between the user and the prediction model, and the user fills in the patient's periodontitis index through the patient information input module 501. After clicking the data submission, the patient's periodontitis index data will be integrated and transferred to the prediction module 503. The prediction module 503 has a built-in random forest prediction model. After the prediction module 503 receives the integrated data, the prediction model is activated to start prediction. The prediction result is displayed through the prediction result output 502, and the patient's periodontitis index data and the prediction result will be stored in the data storage module 505, which is convenient for subsequent call and modification. In addition, the disclosed pancreatic cancer prediction system based on periodontitis indicators supports the retraining of the prediction model, calls the prediction model retraining module 504, and realizes the retraining of the prediction model by entering new periodontitis indicators and pancreatic cancer diagnosis results, combined with the initial training data, to further improve the performance of the prediction model.
[0078] Embodiment 3
[0079] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in the prediction model training method based on early stage pancreatic cancer as described in the first embodiment above are implemented.
[0080] Embodiment 4
[0081] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the early prediction model training method for pancreatic cancer described in the first embodiment are implemented.
[0082] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.
[0083] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0084] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0085] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0086] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.
[0087] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. A prediction model training method based on early pancreatic cancer, characterized in that: include: Obtain patients' periodontitis indicators and pancreatic cancer test report results to construct a data set; Divide the dataset into training and testing sets; Based on the periodontitis indicators in the training set, relevant features were extracted; Using relevant features as input and corresponding pancreatic cancer detection report results as output, multiple prediction models are trained; Grid search and ten-fold cross validation are used to optimize multiple prediction models to obtain multiple trained prediction models; The test set is used to make predictions for multiple trained prediction models, and the prediction model with the best output result is selected as the final prediction model.
2. The method for training a prediction model for early pancreatic cancer according to claim 1, characterized in that: The periodontitis indicators include: plaque index, presence or absence of furcation lesions, periodontal probing depth, and attachment loss range.
3. The method for training a prediction model for early pancreatic cancer according to claim 2, characterized in that: The multiple prediction models include a random forest model, which includes multiple decision trees and a combiner. Each decision tree classifies the input of the decision tree based on its learned rules to obtain a classification result. The combiner integrates the classification results of all decision trees and takes the classification results with the largest number as the final output result.
4. The method for training a prediction model for early pancreatic cancer according to claim 3, characterized in that: The classification process of each decision tree includes: predicting according to the presence or absence of root bifurcation disease characteristics, and outputting classification results of excluding pancreatic cancer, confirming pancreatic cancer and the first suspected pancreatic cancer; for the result of the first suspected pancreatic cancer, predicting according to the plaque index characteristics, and outputting classification results of excluding pancreatic cancer, confirming pancreatic cancer and the second suspected pancreatic cancer; for the result of the second suspected pancreatic cancer, predicting according to the periodontal probing depth characteristics, and outputting classification results of excluding pancreatic cancer, confirming pancreatic cancer and the third suspected pancreatic cancer; for the result of the third suspected pancreatic cancer, predicting according to the attachment loss range characteristics, and outputting classification results of excluding pancreatic cancer, confirming pancreatic cancer and the fourth suspected pancreatic cancer; summarizing and outputting the classification results of excluding pancreatic cancer, confirming pancreatic cancer and suspected pancreatic cancer.
5. The method for training a prediction model for early pancreatic cancer according to claim 2, characterized in that: The multiple prediction models include a support vector machine model that places periodontitis indicators into a feature space; According to the periodontitis index, two hyperplanes are defined; according to the training set, expressions of the two hyperplanes are obtained to divide the feature space into three parts, namely, excluding pancreatic cancer, confirming pancreatic cancer and suspected pancreatic cancer; the relevant features of the periodontitis index in the test set are input into the feature space to obtain the classification result.
6. The method for training a prediction model for early pancreatic cancer according to claim 2, characterized in that: The multiple prediction models include a logistic regression model, which linearly fits the relevant features of the periodontitis indicators in the training set with the corresponding pancreatic cancer detection report results to obtain a multiple linear regression equation; the multiple linear regression equation is mapped to [0, 1] through a sigmoid activation function, and a classification threshold D is obtained according to the pancreatic cancer detection report results; Inputting relevant features of the periodontitis indicators in the prediction set into the multivariate linear regression equation to obtain the equation result; The equation result is mapped to [0, 1] through the sigmoid activation function to obtain a mapping result; the mapping result is compared with the classification threshold to obtain a classification result.
7. The method for training a prediction model for early pancreatic cancer according to claim 1, characterized in that: The method of using grid search and ten-fold cross validation to optimize multiple prediction models includes: defining a hyperparameter network for each prediction model; traversing all possible hyperparameter combinations on the hyperparameter network, and using ten-fold cross validation to evaluate the performance of each hyperparameter combination, and selecting the hyperparameter combination with the best performance as the optimal hyperparameter of the prediction model.
8. A prediction model training system based on early pancreatic cancer, characterized in that: include: A data acquisition module is configured to: acquire the patient's periodontitis index and pancreatic cancer detection report results to construct a data set; Divide the dataset into training and testing sets; A feature extraction module is configured to: extract relevant features based on the periodontitis indicators in the training set; A training module is configured to: take relevant features as input and corresponding pancreatic cancer detection report results as output, and train multiple prediction models; An optimization module is configured to: optimize multiple prediction models by using grid search and ten-fold cross validation to obtain multiple trained prediction models; The test module is configured to: use the test set to predict the multiple trained prediction models, and select the prediction model with the best output result as the final prediction model.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in the prediction model training method based on early stage pancreatic cancer as described in any one of claims 1 to 7 are implemented.
10. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the prediction model training method based on early stage pancreatic cancer are implemented as described in any one of claims 1-7.