A method, device, equipment and storage medium for classifying lung nodules

Abstract

The present application discloses a method, device, equipment and storage medium for pulmonary nodule feature classification, which relates to the field of data processing technology, including: determining the computed tomography report corresponding to the pulmonary nodule to be detected and the serum peptide spectrum data of the corresponding serum sample; using a preset deep learning model and a preset machine learning model to classify the computed tomography report, respectively, to obtain a first classification result and a second classification result, and using a preset serum peptide spectrum nodule classification model to classify the serum peptide spectrum data to obtain a third classification result; performing consistency judgment on the first classification result and the second classification result to obtain a judgment result, and performing an integrated regression operation on the judgment result and the third classification result to obtain a target benign classification result and a target malignant classification result. In this way, the present application combines the computed tomography report of the pulmonary nodule to be detected and the corresponding serum peptide spectrum data to detect the characteristics of the pulmonary nodule, thereby improving the accuracy of discrimination.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method, device, equipment and storage medium for pulmonary nodule feature classification. Background Art

[0002] In the process of feature analysis of CT (Computed Tomography) image reports of lung nodules to predict lung tumors, the commonly used technology is mainly for modeling and prediction of single or isolated lung nodules. The existing approach to feature analysis of CT image reports uses deep learning to screen text features based on a large number of medical vocabulary. The process is complicated and often cannot accurately determine the specific feature information of multiple nodules. However, feature information that is not within the rules cannot be accurately identified through text rule screening alone, and screening through machine learning models alone has problems such as low accuracy and inability to match the corresponding nodules.

[0003] It can be seen that how to improve the accuracy of distinguishing lung nodule features is a problem to be solved in this field. Summary of the invention

[0004] In view of this, the purpose of the present invention is to provide a method, device, equipment and storage medium for classifying lung nodules, which detects the characteristics of lung nodules by combining the computed tomography report of the lung nodules to be detected and the corresponding serum peptide spectrum data, thereby improving the accuracy of discrimination. The specific scheme is as follows:

[0005] In a first aspect, the present application provides a method for classifying features of pulmonary nodules, comprising:

[0006] Determining a computed tomography report corresponding to the pulmonary nodule to be detected and serum peptide spectrum data corresponding to the serum sample of the pulmonary nodule to be detected;

[0007] The computed tomography report is respectively classified into benign and malignant types for the lung nodules to be detected using a preset deep learning model and a preset machine learning model to obtain a first classification result and a second classification result, and the serum peptide spectrum data is classified and processed using a preset serum peptide spectrum nodule classification model to obtain a corresponding third classification result; wherein the preset deep learning model is a model directly constructed based on the medical keywords corresponding to the text in the computed tomography report of the lung nodule sample, and the preset machine learning model is a target machine learning model constructed by using the first initial machine learning model and / or the first text screening rule, and the second initial machine learning model and / or the second text screening rule to screen the computed tomography report of the lung nodule sample; the first initial machine learning model and / or the first text screening rule are used to screen the nodule location, and the second initial machine learning model and / or the second text screening rule are used to screen the nodule feature information corresponding to the nodule location;

[0008] A consistency judgment is performed on the first classification result and the second classification result to obtain a corresponding judgment result, and an integrated regression operation is performed on the judgment result and the third classification result to obtain a target benign classification result and a target malignant classification result corresponding to the pulmonary nodule to be detected.

[0009] Optionally, the computer tomography report corresponding to the pulmonary nodule to be detected and the serum peptide spectrum data corresponding to the serum sample of the pulmonary nodule to be detected include:

[0010] Using a preset word library, the text in the computer tomography report corresponding to the lung nodules to be detected is segmented to obtain corresponding segmentation results;

[0011] The peak intensity of the characteristic peak in the serum peptide spectrum data corresponding to the serum sample of the pulmonary nodule to be detected is extracted, and the extracted peak intensity is normalized to obtain the serum peptide spectrum characteristic peak corresponding to the serum sample of the pulmonary nodule to be detected.

[0012] Optionally, the using a preset deep learning model and a preset machine learning model to classify the computed tomography report into benign and malignant pulmonary nodules to be detected, respectively, to obtain a first classification result and a second classification result, includes:

[0013] Using a preset deep learning model, the word segmentation result corresponding to the computer tomography report is processed for benign or malignant classification of the pulmonary nodules to be detected to obtain a corresponding first classification result;

[0014] The computer tomography report corresponding to the pulmonary nodule to be detected is screened for nodule location and multi-nodule features by using the first initial machine learning model and / or the first text screening rule, and the second initial machine learning model and / or the second text screening rule, to obtain a nodule location screening result corresponding to the pulmonary nodule to be detected and corresponding multi-nodule feature information;

[0015] The nodule location screening results and the multi-nodule feature information are processed through a preset machine learning model to classify the benign or malignant pulmonary nodules to be detected, and a corresponding second classification result is obtained.

[0016] Optionally, the nodule location and multi-nodule features of the computed tomography report corresponding to the pulmonary nodule to be detected are screened by the first initial machine learning model and / or the first text screening rule, and the second initial machine learning model and / or the second text screening rule to obtain the nodule location screening result corresponding to the pulmonary nodule to be detected and the corresponding multi-nodule feature information, including:

[0017] Performing nodule location screening on the computed tomography report by using the first initial machine learning model and / or the first text screening rule to obtain a nodule location screening result characterizing the location of the pulmonary nodule to be detected;

[0018] Based on the second initial machine learning model and / or the second text screening rule, the nodule feature information between any two lung nodule positions in the nodule position screening results is screened to obtain multi-nodule feature information corresponding to the lung nodule to be detected.

[0019] Optionally, performing consistency determination on the first classification result and the second classification result to obtain a corresponding determination result includes:

[0020] Based on the binary classification principle, a consistency judgment is performed on the classification labels and corresponding probability values ​​representing the pulmonary nodules to be detected in the first classification result and the second classification result to obtain a judgment result representing the consistency between the first classification result and the second classification result.

[0021] Optionally, performing an integrated regression operation on the determination result and the third classification result to obtain a target benign classification result and a target malignant classification result corresponding to the pulmonary nodule to be detected includes:

[0022] According to the determination result, a first classification label and a corresponding first probability value having consistent representations are selected from the first classification result and the second classification result, and a risk value of the first classification label and the corresponding first probability value is calculated by a preset risk value calculation formula to obtain a first risk value; the first risk value includes two groups of risk values ​​corresponding to the first classification result and the second classification result respectively;

[0023] An integrated regression operation is performed on the first risk value and the second risk value corresponding to the third classification result through a first preset integrated regression model to obtain a target benign classification result and a target malignant classification result corresponding to the pulmonary nodule to be detected.

[0024] Optionally, performing an integrated regression operation on the determination result and the third classification result to obtain a target benign classification result and a target malignant classification result corresponding to the pulmonary nodule to be detected includes:

[0025] According to the determination result, a second classification label and a corresponding second probability value representing inconsistency are selected from the first classification result and the second classification result, and a risk value of the second classification label and the corresponding second probability value is calculated by a preset risk value calculation formula to obtain a third risk value; the third risk value includes two groups of risk values ​​corresponding to the first classification result and the second classification result respectively;

[0026] An integrated regression operation is performed on the third risk value and the second risk value corresponding to the third classification result through a second preset integrated regression model to obtain a target benign classification result and a target malignant classification result corresponding to the pulmonary nodule to be detected.

[0027] In a second aspect, the present application provides a pulmonary nodule feature classification device, comprising:

[0028] A module for determining pulmonary nodules to be detected, used to determine a computer tomography report corresponding to the pulmonary nodules to be detected and serum peptide spectrum data corresponding to the serum sample of the pulmonary nodules to be detected;

[0029] A lung nodule classification module is used to classify the CT report into benign or malignant types for the pulmonary nodules to be detected using a preset deep learning model and a preset machine learning model, respectively, to obtain a first classification result and a second classification result, and to classify the serum peptide spectrum data using a preset serum peptide spectrum nodule classification model to obtain a corresponding third classification result; wherein the preset deep learning model is a model directly constructed based on the medical keywords corresponding to the text in the CT report of the pulmonary nodule sample, and the preset machine learning model is a target machine learning model constructed by using the first initial machine learning model and / or the first text screening rule, and the second initial machine learning model and / or the second text screening rule to screen the CT report of the pulmonary nodule sample; the first initial machine learning model and / or the first text screening rule are used to screen the nodule location, and the second initial machine learning model and / or the second text screening rule are used to screen the nodule feature information corresponding to the nodule location;

[0030] The classification result processing module is used to make consistency judgment on the first classification result and the second classification result to obtain a corresponding judgment result, and to perform an integrated regression operation on the judgment result and the third classification result to obtain a target benign classification result and a target malignant classification result corresponding to the lung nodule to be detected.

[0031] In a third aspect, the present application provides an electronic device, including:

[0032] Memory, used to store computer programs;

[0033] A processor is used to execute the computer program to implement the lung nodule feature classification method as described above.

[0034] In a fourth aspect, the present application provides a computer-readable storage medium for storing a computer program, which, when executed by a processor, implements the lung nodule feature classification method as described above.

[0035] It can be seen that the present application first determines the computed tomography report corresponding to the pulmonary nodule to be detected and the serum peptide spectrum data corresponding to the serum sample of the pulmonary nodule to be detected; then, the computed tomography report is classified into benign and malignant types for the pulmonary nodule to be detected using a preset deep learning model and a preset machine learning model, respectively, to obtain a first classification result and a second classification result, and the serum peptide spectrum data is classified using a preset serum peptide spectrum nodule classification model to obtain a corresponding third classification result; wherein, the preset deep learning model is a model directly constructed based on the medical keywords corresponding to the text in the computed tomography report of the pulmonary nodule sample, and the preset machine learning model is a model directly constructed based on the medical keywords corresponding to the text in the computed tomography report of the pulmonary nodule sample, and the preset machine learning model is a model directly constructed based on the medical keywords corresponding to the text in the preset deep learning model and the preset machine learning model is a model directly constructed based on the medical keywords corresponding to the text in the computed tomography report of the pulmonary nodule sample, and the preset machine learning model is a model directly constructed based on the medical keywords corresponding to the text in the preset deep learning model and the preset machine learning model is a model directly constructed based on the medical keywords corresponding to the text in the serum peptide spectrum nodule sample, ... Or the first text screening rule, and the second initial machine learning model and / or the second text screening rule construct a target machine learning model based on the information screened from the computed tomography report of the lung nodule sample; the first initial machine learning model and / or the first text screening rule are used to screen the nodule location, and the second initial machine learning model and / or the second text screening rule are used to screen the nodule feature information corresponding to the nodule location; then, the first classification result and the second classification result are judged for consistency to obtain the corresponding judgment result, and the judgment result and the third classification result are integrated and regressed to obtain the target benign classification result and the target malignant classification result corresponding to the lung nodule to be detected. In this way, the present application can detect the characteristics of lung nodules in combination with the computed tomography report of the lung nodules to be detected and the serum peptide spectrum data of the corresponding serum samples. Specifically, the nodule positions and corresponding nodule characteristics extracted from the computed tomography report are processed respectively by a preset deep learning model and a preset machine learning model, and the corresponding first classification results and second classification results can be obtained. In combination with the third classification result corresponding to the serum peptide spectrum data, the target benign classification results and target malignant classification results for the lung nodules to be detected can be obtained. In this way, the nodule characteristics can be accurately identified and the discrimination accuracy can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0037] Figure 1 A flow chart of a lung nodule feature classification method disclosed in this application;

[0038] Figure 2 A flowchart of a specific lung nodule feature classification method disclosed in this application;

[0039] Figure 3 This is a schematic diagram of the structure of a pulmonary nodule feature classification device disclosed in this application;

[0040] Figure 4 This is a structural diagram of an electronic device disclosed in this application. DETAILED DESCRIPTION

[0041] 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.

[0042] See also Figure 1 As shown, the embodiment of the present invention discloses a method for classifying lung nodules, including:

[0043] Step S11, determining the computer tomography report corresponding to the pulmonary nodule to be detected and the serum peptide spectrum data corresponding to the serum sample of the pulmonary nodule to be detected.

[0044] In an embodiment of the present application, it is first necessary to determine the computed tomography report corresponding to the lung nodule to be detected, there are multiple different nodules in the report, and the serum peptide spectrum data corresponding to the serum sample of the lung nodule to be detected; in a specific embodiment, the determination of the computed tomography report corresponding to the lung nodule to be detected and the serum peptide spectrum data corresponding to the serum sample of the lung nodule to be detected can include: using a preset vocabulary to perform word segmentation processing on the text in the computed tomography report corresponding to the lung nodule to be detected to obtain corresponding word segmentation results; extracting the peak intensity of the characteristic peak in the serum peptide spectrum data corresponding to the serum sample of the lung nodule to be detected, and normalizing the extracted peak intensity to obtain the serum peptide spectrum characteristic peak corresponding to the serum sample of the lung nodule to be detected. Specifically, the CT text can be preprocessed first to remove some worthless descriptions (descriptions of other parts outside the lungs, descriptions of no abnormalities such as not seen or none, hospital information, patient information, etc.), convert English punctuation into Chinese punctuation, convert numbers and their units, etc.; then the CT text can be segmented through a preset vocabulary to obtain the corresponding segmentation results; and, for the serum peptide spectrum data of the serum samples of the lung nodules to be detected, the serum peptide spectrum fingerprint of the serum polypeptide molecules can be detected by a mass spectrometer, and then the peak intensity of the characteristic peaks in the blood data can be extracted and normalized to obtain the final serum peptide spectrum data.

[0045] Step S12, respectively using a preset deep learning model and a preset machine learning model to classify the computed tomography report into benign or malignant types for the lung nodules to be detected, so as to obtain a first classification result and a second classification result, and using a preset serum peptide spectrum nodule classification model to classify the serum peptide spectrum data to obtain a corresponding third classification result; wherein the preset deep learning model is a model directly constructed based on the medical keywords corresponding to the text in the computed tomography report of the lung nodule sample, and the preset machine learning model is a target machine learning model constructed by using the first initial machine learning model and / or the first text screening rule, and the second initial machine learning model and / or the second text screening rule to screen the computed tomography report of the lung nodule sample; the first initial machine learning model and / or the first text screening rule are used to screen the nodule location, and the second initial machine learning model and / or the second text screening rule are used to screen the nodule feature information corresponding to the nodule location.

[0046] In the embodiment of the present application, the first classification result is obtained by classifying the benign and malignant lung nodules to be detected in the computed tomography report through a preset deep learning model; the corresponding second classification result is obtained by classifying the benign and malignant lung nodules to be detected in the computed tomography report through a preset machine learning model; wherein, the medical keywords obtained by processing the computed tomography report of the lung nodule sample using the set word segmentation rules can be directly trained to obtain a deep learning model; and for the machine learning model, it is necessary to screen the nodule location of the computed tomography report of the lung nodule sample, and it is necessary to screen the relevant nodule feature information according to the nodule location, and then train the machine learning model based on the multi-nodule feature information. For the serum peptide spectrum data, the serum peptide spectrum data can be classified and processed by the pre-trained serum peptide spectrum classification model to obtain the corresponding third classification result.

[0047] In a specific embodiment, the use of a preset deep learning model and a preset machine learning model to classify the CT report into benign or malignant for the pulmonary nodule to be detected, respectively, to obtain a first classification result and a second classification result, may include: using a preset deep learning model to classify the word segmentation result corresponding to the CT report into benign or malignant for the pulmonary nodule to be detected, to obtain a corresponding first classification result; using a first initial machine learning model and / or a first text screening rule, and a second initial machine learning model and / or a second text screening rule to screen the CT report corresponding to the pulmonary nodule to be detected for nodule location and multi-nodule features, to obtain a nodule location screening result and corresponding multi-nodule feature information corresponding to the pulmonary nodule to be detected; using a preset machine learning model to classify the nodule location screening result and the multi-nodule feature information into benign or malignant for the pulmonary nodule to be detected, to obtain a corresponding second classification result. It can be seen that the first classification result is obtained by directly processing the word segmentation result corresponding to the CT report of the pulmonary nodule to be detected by the preset deep learning model. In the process of processing the CT report of the pulmonary nodules to be detected by the preset machine learning model, the CT report of the pulmonary nodules to be detected can be processed by any one of the first initial machine learning model and the first text screening rule or a combination of both, and the second machine learning model and the second text screening rule or a combination of both to obtain the corresponding nodule location and multi-nodule feature information; specifically, the first initial machine learning model and the second initial machine learning model can be used to process the CT report of the pulmonary nodules to be detected to obtain the nodule location and multi-nodule feature information; or, the first initial machine learning model and the first text screening rule can be used in combination with the second initial machine learning model and the second text screening rule to process the CT report of the pulmonary nodules to be detected to obtain the nodule location and multi-nodule feature information. or, processing the computed tomography report of the pulmonary nodules to be detected through the first text screening rule and the second text screening rule to obtain the nodule location and multi-nodule feature information; or, processing the computed tomography report of the pulmonary nodules to be detected through the first initial machine learning model and the second text screening rule to obtain the nodule location and multi-nodule feature information; or, processing the computed tomography report of the pulmonary nodules to be detected through the first text screening rule and the second initial machine learning model to obtain the nodule location and multi-nodule feature information; finally, the nodule location screening results and multi-nodule feature information obtained in the above steps are processed by a preset machine learning model (i.e., the target machine learning model trained in advance) for benign and malignant classification of the pulmonary nodules to be detected, to obtain the corresponding second classification result.

[0048] In a specific embodiment, the process of constructing a serum peptide spectrum nodule classification model first determines the serum samples of patients with benign lung nodules and malignant lung nodules from representative clinical institutions in different regions and at different levels; then incubate the serum samples to enrich the serum polypeptide molecules; and detect the serum polypeptide fingerprints of patients with benign lung nodules and malignant lung nodules by a time-of-flight mass spectrometer. Further, it is necessary to preprocess the serum polypeptide fingerprints: remove the baseline, molecular weight calibration, noise reduction, spectrum normalization, peak screening, peak clustering; and then screen the characteristic peaks of the serum peptide spectrum: take the collection center as the unit, count the frequency of mass spectrum peaks of each center, calculate the average and standard deviation of the peak frequency of each center: X±σ, remove the mass spectrum peaks with a peak frequency lower than X-σ, and obtain a list of mass spectrum peaks with stable peaks in each collection center, so as to determine the number and position of mass spectrum peaks in each center. The mass spectrometry peaks of each center were counted and analyzed, and the mass spectrometry peaks that appeared in each center were screened out as the characteristic peaks of serum peptide spectrum; then the peak intensity of the characteristic peaks of serum peptide spectrum was extracted, and the peaks were normalized to obtain the serum peptide spectrum data. The serum peptide spectrum data of benign lung nodules and malignant lung nodules (including the characteristic peak information of serum peptide spectrum) were divided into 2 / 3 as the model training set and 1 / 3 as the model test set according to the proportion of the collection center and the benign and malignant nodules; the training set was used to train the SVM (Support Vector Machine) model, and the feature combination and parameters were optimized by genetic algorithm. The obtained features and parameters were then used to train the SVM model, and the performance of the final trained model was evaluated with the test set. If the evaluation is passed, it means that the features, algorithms and parameters meet the requirements, and the final serum peptide spectrum nodule classification model is obtained.

[0049] Step S13, performing consistency judgment on the first classification result and the second classification result to obtain a corresponding judgment result, and performing an integrated regression operation on the judgment result and the third classification result to obtain a target benign classification result and a target malignant classification result corresponding to the lung nodule to be detected.

[0050] In the embodiment of the present application, since both the first classification result and the second classification result are obtained based on the computed tomography report, the consistency of the two results can be determined, and the corresponding determination result and the third classification result obtained from the serum peptide spectrum data can be integrated and regressed, so that the target benign classification result and the target malignant classification result corresponding to the pulmonary nodule to be detected can be obtained. In a specific embodiment, the consistency determination of the first classification result and the second classification result to obtain the corresponding determination result may include: based on the binary classification principle, the consistency determination of the classification labels and corresponding probability values ​​characterizing the pulmonary nodules to be detected in the first classification result and the second classification result is performed to obtain a determination result characterizing the consistency between the first classification result and the second classification result. Specifically, the first classification result and the second classification result include classification label results and corresponding probability values. The consistency of each classification label and the corresponding probability value can be judged by the binary classification principle to determine whether the classification labels in the two classification results are consistent, as well as the difference in the probability values ​​corresponding to each classification label. If the classification labels are consistent and the probabilities are both greater than 0.5 or both less than 0.5, they are judged to be consistent. If the classification labels are consistent, but the difference between the probability values ​​is less than a preset threshold range (such as 0.51 and 0.49), they can also be judged to be consistent. Finally, a judgment result characterizing the consistency between the first classification result and the second classification result can be obtained.

[0051] Furthermore, the integrated regression operation on the judgment result and the third classification result to obtain a target benign classification result and a target malignant classification result corresponding to the pulmonary nodule to be detected may include: screening out a first classification label and a corresponding first probability value with consistent representation from the first classification result and the second classification result according to the judgment result, and calculating the risk value of the first classification label and the corresponding first probability value through a preset risk value calculation formula to obtain a first risk value; the first risk value includes two groups of risk values ​​corresponding to the first classification result and the second classification result, respectively; and performing an integrated regression operation on the first risk value and the second risk value corresponding to the third classification result through a first preset integrated regression model to obtain a target benign classification result and a target malignant classification result corresponding to the pulmonary nodule to be detected. Specifically, when an integrated regression operation is performed on the judgment result and the third classification result, firstly, based on the judgment result, a first classification label and a corresponding first probability value with consistent representation are screened out from the first classification result and the second classification result, and then the product of the first classification label and the first probability value is calculated and summed to obtain a first risk value for the first classification label and the first probability value; and the relevant classification labels and corresponding probability values ​​in the third classification result also correspond to a second risk value; then, an integrated regression operation is performed on the first risk value and the second risk value through a first preset integrated regression model to obtain a final target benign classification result and a target malignant classification result; wherein, the integrated regression operation refers to redistributing weights through an nls regression analysis method (Non-linear least squares).

[0052] Correspondingly, the integrated regression operation of the determination result and the third classification result to obtain the target benign classification result and the target malignant classification result corresponding to the pulmonary nodule to be detected may include: selecting the second classification label and the corresponding second probability value representing inconsistency from the first classification result and the second classification result according to the determination result, and calculating the risk value of the second classification label and the corresponding second probability value through a preset risk value calculation formula to obtain a third risk value; the third risk value includes two groups of risk values ​​corresponding to the first classification result and the second classification result respectively; the third risk value and the second risk value corresponding to the third classification result are integrated and regressed through a second preset integrated regression model to obtain the target benign classification result and the target malignant classification result corresponding to the pulmonary nodule to be detected. Specifically, for the second classification label and the corresponding second probability value representing inconsistency in the first classification result and the second classification result, the corresponding third risk value can be calculated; and the third classification result also has a fourth risk value corresponding to the relevant label and probability value, and then the third risk value and the fourth risk value are integrated and regressed through a second preset integrated regression model to obtain the final target benign classification result and the target malignant classification result corresponding to the pulmonary nodule to be detected.

[0053] It can be seen that the present application can detect the characteristics of lung nodules in combination with the computed tomography report of the lung nodules to be detected and the serum peptide spectrum data of the corresponding serum samples. Specifically, the nodule positions and corresponding nodule characteristics extracted from the computed tomography report are processed by a preset deep learning model and a preset machine learning model respectively, and the corresponding first classification results and second classification results can be obtained. In this way, the words characterizing the nodule characteristics in the computed tomography report can be accurately identified to obtain accurate nodule feature information; then, the target benign classification results and target malignant classification results for the lung nodules to be detected can be obtained in combination with the third classification results corresponding to the serum peptide spectrum data, so that the nodule characteristics can be accurately identified and the discrimination accuracy can be improved.

[0054] See also Figure 2 As shown, the embodiment of the present invention discloses a method for classifying lung nodules, including:

[0055] Step S21, determining the computer tomography report corresponding to the pulmonary nodule to be detected and the serum peptide spectrum data corresponding to the serum sample of the pulmonary nodule to be detected.

[0056] Step S22: Use a preset deep learning model to perform benign or malignant classification processing on the word segmentation results corresponding to the computed tomography report for the pulmonary nodules to be detected, and obtain a corresponding first classification result.

[0057] Step S23: perform nodule location screening on the computed tomography report using the first initial machine learning model and / or the first text screening rule to obtain a nodule location screening result that characterizes the location of the lung nodule to be detected.

[0058] In an embodiment of the present application, the word segmentation results corresponding to the computed tomography report of the lung nodules to be detected are processed by a first initial machine learning model corresponding to a preset machine learning model (an algorithm model such as a support vector machine, a random forest, etc. constructed based on words describing the location of lung nodules in multiple CT reports and a medical knowledge base) and / or a first text screening rule (a regularization method for sorting out words describing the location of lung nodules based on multiple CT reports and a medical knowledge base), and the words representing the location of the nodules in the computed tomography report can be determined to obtain a nodule location screening result; and for complex sentences, the sentence structure can be analyzed based on a regularization method of keywords such as and, or, and commas in the complex sentences (for example, a parallel sentence may contain multiple nodule locations, and an inverted sentence may indicate that the relevant description is located in the front part of the sentence and the nodule location is located in the back part), thereby effectively distinguishing the nodule location.

[0059] Step S24: based on the second initial machine learning model and / or the second text screening rule, the nodule feature information between any two lung nodule positions in the nodule position screening results is screened to obtain the multi-nodule feature information corresponding to the lung nodule to be detected.

[0060] Further, after obtaining the nodule location screening results of the computed tomography report of the pulmonary nodules to be detected, the words containing several pulmonary nodule information between any two pulmonary nodule locations in the nodule location screening results can be processed by the second initial machine learning model (multi-classification model constructed by support vector machine, random forest, etc.) and / or the second text screening rule (based on multiple CT reports and the regularization method of arranging the pulmonary nodule feature information description words in the medical knowledge base, such as nodule nature, nodule number, nodule size, etc.) corresponding to the preset machine learning model, and the words between the two nodule locations are extracted. These words contain relevant nodule information, and finally multiple nodule feature information can be obtained; when there is only one nodule location in the nodule location screening results, several words before and after the nodule location are extracted respectively. In addition, for complex sentences, the sentence structure can be analyzed based on the regularization method of keywords such as and, or, and comma in complex sentences (for example, a parallel sentence may contain multiple nodule information) to obtain multiple nodule feature information.

[0061] In a specific embodiment, in the process of constructing the final target machine learning model, the pre-prepared nodule features can be numerically converted according to specified standards to obtain a data matrix. For example, the risk level of ground glass properties in nodule properties is often higher than that of solid properties, and the numerical value can reflect the degree of risk in a clinical sense. The data matrix is ​​divided into a training set and a test set to construct a target machine learning classification model, that is, the preset machine learning model in the above embodiment.

[0062] Step S25: Using a preset machine learning model, the nodule location screening result and the multi-nodule feature information are processed for benign or malignant classification of the pulmonary nodules to be detected to obtain a corresponding second classification result.

[0063] After obtaining the nodule location screening results and related multi-nodule feature information of the computed tomography report of the lung nodules to be detected through the above steps, this information can be input into a preset machine learning model, and the machine learning model will perform benign or malignant classification processing on the nodule location screening results and multi-nodule feature information for the lung nodules to be detected, and finally obtain a second classification result corresponding to the lung nodules to be detected.

[0064] Step S26: classify the serum peptide spectrum data using a preset serum peptide spectrum nodule classification model to obtain a corresponding third classification result.

[0065] Step S27, performing consistency judgment on the first classification result and the second classification result to obtain a corresponding judgment result, and performing an integrated regression operation on the judgment result and the third classification result to obtain a target benign classification result and a target malignant classification result corresponding to the lung nodule to be detected.

[0066] For more specific processing procedures of the above steps S21, S22, S26 and S27, reference may be made to the corresponding contents disclosed in the aforementioned embodiments, which will not be described in detail here.

[0067] In a specific embodiment, 160 serum peptide spectrum samples containing lung CT texts can be selected, including 80 lung cancer samples and 80 non-cancer samples, and then the samples are divided into training sets and test sets by group, with the training set accounting for 80% and the test set accounting for 20%. Furthermore, the sample data needs to be preprocessed first, and the CT text data is processed in two ways (deep learning model and machine learning model). Specifically, in the process of processing based on the deep learning model, the commonly used word segmentation vocabulary is first sorted out and summarized. The jieba library (a Python Chinese word segmentation third-party library) can be used to segment the training set and test set CT texts, and then converted into hash values ​​to obtain serialized vectors, and then processed to obtain tensor matrices of equal length, that is, the word segmentation results A1 of the preprocessing result data training set and the word segmentation results A2 of the test set. Furthermore, a deep learning model is constructed using the training set result A1, which mainly includes a convolution layer, a pooling layer, a nonlinear transformation layer based on relu (Rectified Linear Unit) and a linear layer. The classification cross entropy loss function is used, and the test set data A2 is predicted to obtain the test set result. The product of the label and the probability is calculated and converted into a risk value, and finally the first classification result is obtained.

[0068] Correspondingly, in the process of processing based on the machine learning model, the training set data can be manually labeled to obtain 11 dimensional feature description words including "number of nodules", "nodule size", "nodule location", "nodule nature", "lobulation", "burr", "pleural indentation", "cavitation", "whether the edge of the nodule is clear", "calcification", and "vascular bundle sign", and the words are converted into a numerical matrix through the conversion rules as the result data B1 of the training set (i.e., the nodule location and multi-nodule characteristics of the lung nodule sample in the above embodiment); the test set can be processed by constructing multiple rule functions (the initial machine learning model and / or text screening rules of the above embodiment), which can specifically include eliminating some confusing words and sentences that are not related to the lung part, distinguishing and splitting sentences containing multiple parts and multiple nodules, extracting the above 11 dimensional feature description words for the relevant nodules in each sentence, obtaining the relevant description of each nodule of each sample, and converting the words into a numerical matrix through the conversion rules, i.e., the result data B2 of the test set (the nodule location and multi-nodule characteristics of the lung nodules to be detected in the above embodiment). Furthermore, the result data B1 based on the training set is selected, and the random forest model (target machine learning model) is determined by cross-validation to screen the best parameters, and the test set is predicted by the random forest model to obtain the corresponding test set prediction results; the test set prediction results contain the predicted nodes of multiple nodules in a sample; and the product of the label and the probability is calculated and summed to convert into a risk value, so as to screen out high-risk nodules, and the result corresponding to the maximum risk value is used as the final result of the model, that is, the second classification result.

[0069] Furthermore, a support vector machine model (the preset serum peptide spectrum nodule classification model of the above embodiment) is pre-constructed based on the serum peptide spectrum data of patients with benign lung nodules and malignant lung nodules from representative clinical institutions in different regions and at different levels, and the serum peptide spectrum data corresponding to the training set and the test set are predicted respectively, wherein the test set corresponds to the third classification result. Afterwards, the consistency between the first classification result and the second classification result is judged, and the data with high consistency and the corresponding serum peptide spectrum result are fitted to function 1, and the data with low consistency and the corresponding serum peptide spectrum result are fitted to function 2, and regression processing is performed respectively. Since the result after regression is a continuous value, the optimal threshold value can be determined by the ROC curve (Receiver Operating Characteristic, receiver operating characteristic curve), and it is classified as non-cancer if it is below the threshold value, and it is classified as lung cancer if it is above the threshold value. The following Tables 1 and 2 are the prediction results for the training set and the test set, respectively.

[0070] Table 1

[0071]

[0072] Table 2

[0073]

[0074] It can be seen that in the process of processing the computed tomography report of the present application, you can choose to directly use commonly used word segmentation vocabulary for processing; or choose a pre-set machine learning model or text filtering rule, or a combination of a machine learning model and a text filtering rule to filter nodule feature information; and perform hierarchical screening, first determine the nodule location, and then filter the nodule feature information based on the nodule location; in this way, information that accurately characterizes the nodule can be obtained from the computed tomography report; by processing the computed tomography report in these two ways to obtain the first classification result and the second classification result, and combining the third classification result corresponding to the serum peptide spectrum data, you can get the target benign classification result and the target malignant classification result for the lung nodule to be detected, so that you can accurately identify the nodule characteristics and improve the accuracy of the discrimination.

[0075] See also Figure 3 As shown, the embodiment of the present invention discloses a pulmonary nodule feature classification device, comprising:

[0076] The to-be-detected pulmonary nodule determination module 11 is used to determine the computed tomography report corresponding to the to-be-detected pulmonary nodule and the serum peptide spectrum data corresponding to the serum sample of the to-be-detected pulmonary nodule;

[0077] The pulmonary nodule classification module 12 is used to classify the CT report into benign or malignant types for the pulmonary nodules to be detected using a preset deep learning model and a preset machine learning model, respectively, to obtain a first classification result and a second classification result, and to classify the serum peptide spectrum data using a preset serum peptide spectrum nodule classification model to obtain a corresponding third classification result; wherein the preset deep learning model is a model directly constructed based on the medical keywords corresponding to the text in the CT report of the pulmonary nodule sample, and the pre-machine learning model is a target machine learning model constructed by using the first initial machine learning model and / or the first text screening rule, and the second initial machine learning model and / or the second text screening rule to screen the CT report of the pulmonary nodule sample; the first initial machine learning model and / or the first text screening rule are used to screen the nodule location, and the second initial machine learning model and / or the second text screening rule are used to screen the nodule feature information corresponding to the nodule location;

[0078] The classification result processing module 13 is used to perform consistency judgment on the first classification result and the second classification result to obtain a corresponding judgment result, and perform an integrated regression operation on the judgment result and the third classification result to obtain a target benign classification result and a target malignant classification result corresponding to the lung nodule to be detected.

[0079] It can be seen that the present application can detect the characteristics of lung nodules in combination with the computed tomography report of the lung nodules to be detected and the serum peptide spectrum data of the corresponding serum samples. Specifically, the nodule positions and corresponding nodule characteristics extracted from the computed tomography report are processed by a preset deep learning model and a preset machine learning model respectively, and the corresponding first classification results and second classification results can be obtained. In combination with the third classification result corresponding to the serum peptide spectrum data, the target benign classification results and target malignant classification results for the lung nodules to be detected can be obtained. In this way, the nodule characteristics can be accurately identified and the discrimination accuracy can be improved.

[0080] In a specific embodiment, the to-be-detected pulmonary nodule determination module 11 may include:

[0081] A word segmentation processing unit is used to perform word segmentation processing on the text in the computer tomography report corresponding to the pulmonary nodule to be detected using a preset word library to obtain a corresponding word segmentation result;

[0082] The peak extraction unit is used to extract the peak intensity of the characteristic peak in the serum peptide spectrum data corresponding to the serum sample of the pulmonary nodule to be detected, and normalize the extracted peak intensity to obtain the serum peptide spectrum characteristic peak corresponding to the serum sample of the pulmonary nodule to be detected.

[0083] In a specific embodiment, the pulmonary nodule classification module 12 may include:

[0084] A first classification submodule is used to classify the word segmentation result corresponding to the computer tomography report into benign or malignant pulmonary nodules to be detected by using a preset deep learning model to obtain a corresponding first classification result;

[0085] An information screening submodule is used to screen the computed tomography report corresponding to the pulmonary nodule to be detected for nodule location and multi-nodule features through a first initial machine learning model and / or a first text screening rule, and a second initial machine learning model and / or a second text screening rule, to obtain a nodule location screening result corresponding to the pulmonary nodule to be detected and corresponding multi-nodule feature information;

[0086] The second classification unit is used to classify the nodule position screening results and the multi-nodule feature information into benign or malignant types for the lung nodules to be detected through a preset machine learning model to obtain a corresponding second classification result.

[0087] In another specific embodiment, the information screening submodule may include:

[0088] a nodule position screening unit, configured to perform nodule position screening on the computed tomography report by using the first initial machine learning model and / or the first text screening rule, so as to obtain a nodule position screening result characterizing the position of the pulmonary nodule for the pulmonary nodule to be detected;

[0089] A nodule feature screening unit is used to screen the nodule feature information between any two lung nodule positions in the nodule position screening results based on the second initial machine learning model and / or the second text screening rule to obtain multi-nodule feature information corresponding to the lung nodule to be detected.

[0090] In a specific embodiment, the classification result processing module may include:

[0091] A consistency judgment unit is used to perform consistency judgment on the classification labels and corresponding probability values ​​representing the pulmonary nodules to be detected in the first classification result and the second classification result based on the binary classification principle, so as to obtain a judgment result representing the consistency between the first classification result and the second classification result.

[0092] In a specific embodiment, the classification result processing module may include:

[0093] a first risk value calculation unit, configured to select first classification labels and corresponding first probability values ​​having consistent representations from the first classification results and the second classification results according to the determination result, and calculate the risk values ​​of the first classification labels and the corresponding first probability values ​​by using a preset risk value calculation formula to obtain a first risk value; the first risk value includes two groups of risk values ​​corresponding to the first classification results and the second classification results, respectively;

[0094] The first integrated regression unit is used to perform an integrated regression operation on the first risk value and the second risk value corresponding to the third classification result through a first preset integrated regression model to obtain a target benign classification result and a target malignant classification result corresponding to the pulmonary nodule to be detected.

[0095] In another specific embodiment, the classification result processing module may include:

[0096] a second risk value calculation unit, configured to select second classification labels and corresponding second probability values ​​representing inconsistencies from the first classification results and the second classification results according to the determination result, and calculate the risk values ​​of the second classification labels and the corresponding second probability values ​​by a preset risk value calculation formula to obtain a third risk value; the third risk value includes two groups of risk values ​​corresponding to the first classification results and the second classification results, respectively;

[0097] The second integrated regression unit is used to perform an integrated regression operation on the third risk value and the second risk value corresponding to the third classification result through a second preset integrated regression model to obtain a target benign classification result and a target malignant classification result corresponding to the pulmonary nodule to be detected.

[0098] Furthermore, the present application also discloses an electronic device. Figure 4 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content in the diagram cannot be considered as any limitation on the scope of use of the present application.

[0099] Figure 4 A schematic diagram of the structure of an electronic device 20 provided in an embodiment of the present application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the method for classifying pulmonary nodules in any of the aforementioned embodiments. In addition, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0100] In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present application, and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs and is not specifically limited here.

[0101] In addition, the memory 22, as a carrier for storing resources, can be a read-only memory, a random access memory, a disk or an optical disk, etc. The resources stored thereon can include an operating system 221, a computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0102] The operating system 221 is used to manage and control the hardware devices and computer program 222 on the electronic device 20, which can be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program that can be used to complete the pulmonary nodule feature classification method performed by the electronic device 20 disclosed in any of the aforementioned embodiments, the computer program 222 can further include a computer program that can be used to complete other specific tasks.

[0103] Furthermore, the present application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the aforementioned pulmonary nodule feature classification method is implemented. The specific steps of the method can be referred to the corresponding contents disclosed in the aforementioned embodiments, and will not be repeated here.

[0104] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0105] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0106] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0107] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0108] The technical solution provided by the present application is introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for general technicians in this field, according to the idea of ​​the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for classifying lung nodules, characterized in that: include: Determining a computed tomography report corresponding to the pulmonary nodule to be detected and serum peptide spectrum data corresponding to the serum sample of the pulmonary nodule to be detected; The computed tomography report is respectively classified into benign and malignant types for the lung nodules to be detected using a preset deep learning model and a preset machine learning model to obtain a first classification result and a second classification result, and the serum peptide spectrum data is classified and processed using a preset serum peptide spectrum nodule classification model to obtain a corresponding third classification result; wherein the preset deep learning model is a model directly constructed based on the medical keywords corresponding to the text in the computed tomography report of the lung nodule sample, and the preset machine learning model is a target machine learning model constructed by using the first initial machine learning model and / or the first text screening rule, and the second initial machine learning model and / or the second text screening rule to screen the computed tomography report of the lung nodule sample; the first initial machine learning model and / or the first text screening rule are used to screen the nodule location, and the second initial machine learning model and / or the second text screening rule are used to screen the nodule feature information corresponding to the nodule location; Performing consistency judgment on the first classification result and the second classification result to obtain a corresponding judgment result, and performing an integrated regression operation on the judgment result and the third classification result to obtain a target benign classification result and a target malignant classification result corresponding to the pulmonary nodule to be detected; The computer tomography report corresponding to the pulmonary nodule to be detected and the serum peptide spectrum data corresponding to the serum sample of the pulmonary nodule to be detected include: Using a preset word library, the text in the computer tomography report corresponding to the lung nodules to be detected is segmented to obtain corresponding segmentation results; Extracting the peak intensity of the characteristic peak in the serum peptide spectrum data corresponding to the serum sample of the pulmonary nodule to be detected, and normalizing the extracted peak intensity to obtain the serum peptide spectrum characteristic peak corresponding to the serum sample of the pulmonary nodule to be detected; The method of using a preset deep learning model and a preset machine learning model to classify the CT report into benign and malignant pulmonary nodules to be detected, respectively, to obtain a first classification result and a second classification result, includes: Using a preset deep learning model, the word segmentation result corresponding to the computer tomography report is processed for benign or malignant classification of the pulmonary nodules to be detected to obtain a corresponding first classification result; The computer tomography report corresponding to the pulmonary nodule to be detected is screened for nodule location and multi-nodule features by using the first initial machine learning model and / or the first text screening rule, and the second initial machine learning model and / or the second text screening rule, to obtain a nodule location screening result corresponding to the pulmonary nodule to be detected and corresponding multi-nodule feature information; The nodule location screening result and the multi-nodule feature information are processed by a preset machine learning model for benign and malignant classification of the pulmonary nodules to be detected, to obtain a corresponding second classification result; The nodule location and multi-nodule features of the computed tomography report corresponding to the pulmonary nodule to be detected are screened by the first initial machine learning model and / or the first text screening rule, and the second initial machine learning model and / or the second text screening rule to obtain the nodule location screening result and the corresponding multi-nodule feature information corresponding to the pulmonary nodule to be detected, including: Performing nodule location screening on the computed tomography report by using the first initial machine learning model and / or the first text screening rule to obtain a nodule location screening result characterizing the location of the pulmonary nodule to be detected; Based on the second initial machine learning model and / or the second text screening rule, the nodule feature information between any two lung nodule positions in the nodule position screening results is screened to obtain multi-nodule feature information corresponding to the lung nodule to be detected.

2. The method for pulmonary nodule feature classification according to claim 1, characterized in that: The performing consistency determination on the first classification result and the second classification result to obtain a corresponding determination result includes: Based on the binary classification principle, a consistency judgment is performed on the classification labels and corresponding probability values ​​representing the pulmonary nodules to be detected in the first classification result and the second classification result to obtain a judgment result representing the consistency between the first classification result and the second classification result.

3. The method for pulmonary nodule feature classification according to claim 2, characterized in that: The integrated regression operation is performed on the determination result and the third classification result to obtain a target benign classification result and a target malignant classification result corresponding to the pulmonary nodule to be detected, including: According to the determination result, a first classification label and a corresponding first probability value having consistent representations are selected from the first classification result and the second classification result, and a risk value of the first classification label and the corresponding first probability value is calculated by a preset risk value calculation formula to obtain a first risk value; the first risk value includes two groups of risk values ​​corresponding to the first classification result and the second classification result respectively; An integrated regression operation is performed on the first risk value and the second risk value corresponding to the third classification result through a first preset integrated regression model to obtain a target benign classification result and a target malignant classification result corresponding to the pulmonary nodule to be detected.

4. The method for pulmonary nodule feature classification according to claim 2, characterized in that: The integrated regression operation is performed on the determination result and the third classification result to obtain a target benign classification result and a target malignant classification result corresponding to the pulmonary nodule to be detected, including: According to the determination result, a second classification label and a corresponding second probability value representing inconsistency are selected from the first classification result and the second classification result, and a risk value of the second classification label and the corresponding second probability value is calculated by a preset risk value calculation formula to obtain a third risk value; the third risk value includes two groups of risk values ​​corresponding to the first classification result and the second classification result respectively; An integrated regression operation is performed on the third risk value and the second risk value corresponding to the third classification result through a second preset integrated regression model to obtain a target benign classification result and a target malignant classification result corresponding to the pulmonary nodule to be detected.

5. A pulmonary nodule feature classification device, characterized in that: include: A module for determining pulmonary nodules to be detected, used to determine a computer tomography report corresponding to the pulmonary nodules to be detected and serum peptide spectrum data corresponding to the serum sample of the pulmonary nodules to be detected; A lung nodule classification module is used to classify the CT report into benign or malignant types for the pulmonary nodules to be detected using a preset deep learning model and a preset machine learning model, respectively, to obtain a first classification result and a second classification result, and to classify the serum peptide spectrum data using a preset serum peptide spectrum nodule classification model to obtain a corresponding third classification result; wherein the preset deep learning model is a model directly constructed based on the medical keywords corresponding to the text in the CT report of the pulmonary nodule sample, and the preset machine learning model is a target machine learning model constructed by using the first initial machine learning model and / or the first text screening rule, and the second initial machine learning model and / or the second text screening rule to screen the CT report of the pulmonary nodule sample; the first initial machine learning model and / or the first text screening rule are used to screen the nodule location, and the second initial machine learning model and / or the second text screening rule are used to screen the nodule feature information corresponding to the nodule location; A classification result processing module, used for performing consistency determination on the first classification result and the second classification result to obtain a corresponding determination result, and performing an integrated regression operation on the determination result and the third classification result to obtain a target benign classification result and a target malignant classification result corresponding to the pulmonary nodule to be detected; Wherein, the module for determining the pulmonary nodules to be detected includes: A word segmentation processing unit is used to perform word segmentation processing on the text in the computer tomography report corresponding to the pulmonary nodule to be detected using a preset word library to obtain a corresponding word segmentation result; A peak extraction unit, used to extract the peak intensity of the characteristic peak in the serum peptide spectrum data corresponding to the serum sample of the pulmonary nodule to be detected, and normalize the extracted peak intensity to obtain the serum peptide spectrum characteristic peak corresponding to the serum sample of the pulmonary nodule to be detected; Wherein, the pulmonary nodule classification module includes: A first classification submodule is used to classify the word segmentation result corresponding to the computer tomography report into benign or malignant pulmonary nodules to be detected by using a preset deep learning model to obtain a corresponding first classification result; An information screening submodule is used to screen the computed tomography report corresponding to the pulmonary nodule to be detected for nodule location and multi-nodule features through a first initial machine learning model and / or a first text screening rule, and a second initial machine learning model and / or a second text screening rule, to obtain a nodule location screening result corresponding to the pulmonary nodule to be detected and corresponding multi-nodule feature information; A second classification unit is used to classify the nodule position screening result and the multi-nodule feature information into benign or malignant types for the pulmonary nodules to be detected by using a preset machine learning model to obtain a corresponding second classification result; Wherein, the information screening submodule includes: a nodule position screening unit, configured to perform nodule position screening on the computed tomography report by using the first initial machine learning model and / or the first text screening rule, so as to obtain a nodule position screening result characterizing the position of the pulmonary nodule for the pulmonary nodule to be detected; A nodule feature screening unit is used to screen the nodule feature information between any two lung nodule positions in the nodule position screening results based on the second initial machine learning model and / or the second text screening rule to obtain multi-nodule feature information corresponding to the lung nodule to be detected.

6. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the method for classifying pulmonary nodules according to any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that: Used to store a computer program, which, when executed by a processor, implements the method for classifying pulmonary nodules as described in any one of claims 1 to 4.

Images