Lung abnormality assessment method and device

Abstract

The invention relates to the technical field of medical image analysis and artificial intelligence, in particular to a lung abnormality assessment method and device based on multi-dimensional data, which not only respectively construct assessment sub-models for lung shadows (nodules and lumps) of different sizes, but also solve the problem of insufficient precision caused by size difference of lesions in unified model assessment, and improve the assessment accuracy of the lung abnormality. In addition, through integration of methylation data and clinical feature data, key indexes are screened out in a single-factor and multi-factor analysis mode, and the robustness and accuracy of the evaluation model are improved. Most importantly, quantitative evaluation of the lung abnormality is realized in a real sense, human experience interference is reduced, objective reference basis is provided for clinical diagnosis and treatment, doctors can be assisted in evaluating the lung abnormality state more effectively and accurately before an operation or treatment, and the diagnosis and treatment accuracy is improved. The efficiency and the accuracy of clinicians in accurately judging illness states and determining operations and treatment schemes are greatly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of medical image analysis and artificial intelligence, in particular to a lung abnormality evaluation method and device based on multi-dimensional data. BACKGROUND

[0002] The benign and malignant differentiation of lung abnormalities (such as lung nodules, lung masses, etc.) is a key link in clinical diagnosis and treatment. At present, clinical evaluation mainly relies on biomarker detection and imaging feature analysis, but there are obvious limitations: when using a certain biomarker (such as SHOX2, RASSF1A methylation index) alone, it is difficult to balance specificity and sensitivity; there are significant differences in key evaluation indicators corresponding to lung shadows of different sizes (nodules and masses), and unified model evaluation may lead to errors; traditional evaluation methods rely on physician experience and are highly subjective and inconsistent.

[0003] In the prior art, some schemes attempt to combine clinical data and imaging features to construct an evaluation model, but do not consider the influence of lesion size on the correlation of indicators, and the model has limited generalization ability. In addition, the integration and analysis of methylation data and clinical data in the model training process are insufficient, and it is difficult to fully exploit the synergistic value of multi-dimensional data. For example, the method, device, equipment and readable storage medium for assisting lung cancer risk assessment with publication number CN118888129A construct a model based on methylation superposition score and clinical data, but do not distinguish the size difference between lung nodules and lung masses, resulting in homogenization of feature weights for different sizes of lesions. In addition, since the imaging features (such as shadow size) are not explicitly combined to dynamically select sub-models, the evaluation accuracy is not high, and the output results have large errors when assisting in clinical judgment.

[0004] Therefore, there is an urgent need for a quantitative evaluation method that can accurately evaluate different sizes of lung shadows and integrate multi-source data. SUMMARY

[0005] The present application aims to solve the problems of insufficient evaluation accuracy, lack of differentiation of lesion size, and insufficient integration of multi-source data in existing lung abnormality evaluation methods, and provides a lung abnormality evaluation method and device.

[0006] To achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0007] A lung abnormality evaluation method, comprising the following steps:

[0008] 1) Obtain user information of the subject to be tested, the user information including lung image information;

[0009] 2) Determine a sub-evaluation model suitable for the lung condition of the subject to be tested according to the lung image information and an evaluation model that can be used for lung abnormality judgment of the subject to be tested;

[0010] 3) using the user information of the to-be-tested object and the adapted sub-evaluation model, performing lung abnormality evaluation on the to-be-tested object, and outputting an evaluation result.

[0011] Preferably, in step 2), the manner of determining the sub-evaluation model adapted to the lung condition of the to-be-tested object according to the lung image information and the evaluation model capable of being used for lung abnormality judgment of the to-be-tested object specifically comprises:

[0012] 2-1) extracting lung shadow size data of the to-be-tested object from the lung image information;

[0013] 2-2) selecting a corresponding sub-evaluation model from the evaluation model based on the lung shadow size data.

[0014] Preferably, the evaluation model comprises a nodule evaluation sub-model and a mass evaluation sub-model, and in step 2-2), the manner of selecting a corresponding sub-evaluation model from the evaluation model based on the lung shadow size data specifically comprises:

[0015] 2-2-1) if the lung shadow size does not exceed a preset threshold, taking the nodule evaluation sub-model as the sub-evaluation model adapted to the lung condition of the to-be-tested object;

[0016] 2-2-2) if the lung shadow size exceeds the preset threshold, taking the mass evaluation sub-model as the sub-evaluation model adapted to the lung condition of the to-be-tested object.

[0017] Preferably, the construction manner of the evaluation model specifically comprises:

[0018] S-1) taking various lung abnormality patients and healthy people as samples, collecting user information corresponding to a plurality of samples of various types, forming a lung abnormality evaluation data set, the lung abnormality evaluation data set containing lung image information, methylation data and clinical feature data of the samples;

[0019] S-2) dividing the lung abnormality evaluation data set into a training sample set and a verification data set;

[0020] S-3) performing single-factor analysis on the training sample set by using the verification data set to obtain an initial analysis index;

[0021] S-3) screening the initial analysis index to obtain key indexes with statistical significance;

[0022] S-4) performing multi-factor regression analysis on the key indexes and the verification data set to generate a comprehensive analysis result;

[0023] S-5) constructing an evaluation model according to the lung shadow type of the sample and the comprehensive analysis result.

[0024] Preferably, the training sample set comprises a methylation training subset and a clinical training subset, the initial analysis index comprises a methylation feature index and a clinical feature index, and in step S-3), the training sample set is subjected to single factor analysis by using the verification data set to obtain the initial analysis index, and the manner of obtaining the initial analysis index specifically comprises the following steps.

[0025] S-3-1) performing single factor analysis on the methylation training subset based on the verification data set to obtain a methylation feature index;

[0026] S-3-2) performing single factor analysis on the clinical training subset based on the verification data set to obtain a clinical feature index.

[0027] Preferably, in step S-5), the manner of constructing the evaluation model according to the lung shadow type of the sample and the comprehensive analysis result specifically comprises the following steps.

[0028] S-5-1) grouping the comprehensive analysis result according to the lung shadow type to obtain nodule group analysis data and mass group analysis data;

[0029] S-5-2) screening a nodule group core index according to the odds ratio data of the nodule group analysis data;

[0030] S-5-3) screening a mass group core index according to the odds ratio data of the mass group analysis data;

[0031] S-5-4) constructing a nodule evaluation sub-model and a mass evaluation sub-model based on the nodule group core index and the mass group core index respectively, and combining to form the evaluation model.

[0032] The application further provides a lung abnormality evaluation device, which comprises a data acquisition module for acquiring and storing user information, a model selection module for matching sub-evaluation model types, and an evaluation execution module for outputting evaluation results, and the data input ends of the model selection module and the evaluation execution module are connected with the data acquisition module, so that the model selection module can select a suitable sub-evaluation model in the evaluation execution module according to lung image information in the user information and output the result of evaluating the lung abnormality of a patient.

[0033] Preferably, the model selection module comprises:

[0034] a size extraction unit for extracting lung shadow size data of a to-be-tested object from the lung image information;

[0035] a sub-model determination unit for selecting a corresponding sub-evaluation model from the evaluation model based on the lung shadow size data.

[0036] The application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method when executing the computer program.

[0037] The application further provides a computer readable storage medium, which stores a computer program, wherein the computer program is executed by a processor to implement the steps of the method.

[0038] The application has the following beneficial effects:

[0039] 1) The application models the differences in invasiveness between nodules below 30 mm and masses above 30 mm (i.e., different sizes of lung shadows - nodules and masses are respectively constructed to evaluate sub-models), which solves the problem of insufficient precision caused by differences in lesion size in a unified model;

[0040] That is, the application dynamically selects sub-models (lung nodule model vs. mass model) by lesion size (preset threshold 30 mm), models different key indicators (such as CEA + methylation for the nodule group, lobulation + bronchial obstruction + methylation for the mass group) for different lesion sizes, and solves the problem of insufficient adaptability of a unified model to different lesion sizes.

[0041] 2) The application not only integrates methylation data and clinical feature data, but also uses lung image information (especially shadow size) as the core basis for model selection, realizes deep collaboration of "image-molecule-clinical" multidimensional data, filters key indicators through single-factor and multi-factor analysis, and improves the robustness and accuracy of the evaluation model;

[0042] 3) The application realizes quantitative evaluation of lung abnormalities, reduces human experience interference, and provides an objective reference for clinical diagnosis and treatment.

[0043] Glossary

[0044] Lung abnormalities: In radiological examination (such as X-ray, CT), lung shadows, lung masses and lung nodules are terms describing lung abnormal lesions. Lung shadows, which are areas of increased density in radiology, may have blurred or unclear boundaries, and generally refer to high-density shadows on images with variable shapes and sizes, and may be inflammatory, edematous, or neoplastic.

[0045] Preset threshold: The "preset threshold" in the application is actually set according to the concepts of nodules and masses. In radiological examination (such as X-ray, CT), lung shadows, lung masses and lung nodules are terms describing lung abnormal lesions. Lung nodules and lung masses are two relative concepts:

[0046] Pulmonary nodule: a focal round or round-like high-density shadow with a diameter ≤ 30 mm. It can be solid, ground glass (GGO) or partially solid.

[0047] Pulmonary mass: a focal lesion with a diameter > 30 mm, usually with clear boundaries. The malignancy is higher (about 60%-80%). BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 Flowchart of the lung abnormality evaluation method in the present application;

[0049] Figure 2 Structural diagram of the functional modules in the lung abnormality evaluation device of the present application;

[0050] Figure 3 Application environment diagram of the lung abnormality evaluation method in the embodiment of the present application, the terminal communicates with the server, and the server executes the evaluation method.

[0051] Figure 4 Flowchart of the lung abnormality evaluation method in the embodiment of the present application, which shows the steps of obtaining information, determining a sub-model, and obtaining an evaluation score.

[0052] Figure 5 Training flowchart of the target evaluation model in the embodiment of the present application, including steps of obtaining samples, analyzing and screening, and constructing a model.

[0053] Figure 6 Single factor analysis result diagram in the embodiment of the present application, which shows the correlation of each index of the pulmonary nodule group and the pulmonary mass group with the benign and malignant shadows.

[0054] Figure 7 ROC curve diagram in the embodiment of the present application, A is the ROC curve of the pulmonary nodule sub-model, and B is the ROC curve of the pulmonary mass sub-model, showing that the model containing methylation indexes has higher accuracy.

[0055] Figure 8 Structural block diagram of the lung abnormality evaluation device in the embodiment of the present application, containing an obtaining module, a determining module and an evaluation module.

[0056] Figure 9 Internal structure diagram of the computer device in the embodiment of the present application, including a processor, a memory and other components. DETAILED DESCRIPTION

[0057] As Figure 2As shown, the present application provides a lung abnormality evaluation device, which comprises a data acquisition module for collecting and storing user information, a model selection module for matching sub-evaluation model types, an evaluation execution module for outputting evaluation results, and the data input ends of the model selection module and the evaluation execution module are connected with the data acquisition module, so that the model selection module can select a suitable sub-evaluation model in the evaluation execution module according to the lung image information in the user information, and output the results of the lung abnormality evaluation of the patient.

[0058] Notably, the model selection module comprises a size extraction unit and a sub-model determination unit, the size extraction unit is used for extracting lung shadow size data of the to-be-tested object from the lung image information, and the sub-model determination unit is used for selecting a corresponding sub-evaluation model from the evaluation model based on the lung shadow size data.

[0059] In the present application, the method for lung abnormality evaluation by using the lung abnormality evaluation device is as follows Figure 1 As shown, the method specifically comprises the following steps:

[0060] 1) obtaining user information of a to-be-tested object, wherein the user information comprises lung image information;

[0061] 2) determining a sub-evaluation model suitable for the lung condition of the to-be-tested object according to the lung image information and an evaluation model capable of being used for lung abnormality judgment of the to-be-tested object, wherein the evaluation model comprises a nodule evaluation sub-model and a mass evaluation sub-model, and the determination is performed in the following manner:

[0062] 2-1) extracting lung shadow size data of the to-be-tested object from the lung image information;

[0063] 2-2) selecting a corresponding sub-evaluation model from the evaluation model based on the lung shadow size data in the following manner:

[0064] 2-2-1) if the lung shadow size does not exceed a preset threshold, taking the nodule evaluation sub-model as the sub-evaluation model suitable for the lung condition of the to-be-tested object;

[0065] 2-2-2) if the lung shadow size exceeds the preset threshold, taking the mass evaluation sub-model as the sub-evaluation model suitable for the lung condition of the to-be-tested object.

[0066] 3) performing lung abnormality evaluation on the to-be-tested object by using the user information of the to-be-tested object and the suitable sub-evaluation model, and outputting the evaluation results.

[0067] Notably, the evaluation model is established in the following manner:

[0068] S-1)collecting various types of lung abnormality patients and healthy people as samples, collecting a plurality of user information corresponding to each type of sample, forming a lung abnormality evaluation data set, the lung abnormality evaluation data set containing lung image information, methylation data and clinical feature data of the sample;

[0069] S-2)dividing the lung abnormality evaluation data set into a training sample set and a validation data set, the training sample set including a methylation training subset and a clinical training subset;

[0070] S-3)performing single factor analysis on the training sample set using the validation data set to obtain initial analysis indicators, the initial analysis indicators including methylation feature indicators and clinical feature indicators;

[0071] S-3)screening the initial analysis indicators in the following manner to obtain key indicators with statistical significance:

[0072]

[0073] S-3-1)performing single factor analysis on the methylation training subset based on the validation data set to obtain methylation feature indicators;

[0074] S-3-2)performing single factor analysis on the clinical training subset based on the validation data set to obtain clinical feature indicators.

[0075] S-4)performing multiple factor regression analysis on the key indicators and the validation data set to generate a comprehensive analysis result;

[0076] S-5)constructing an evaluation model according to the lung shadow type of the sample and the comprehensive analysis result in the following manner:

[0077]

[0078] S-5-1)grouping the comprehensive analysis result according to the lung shadow type to obtain nodule group analysis data and mass group analysis data;

[0079] S-5-2)screening nodule group core indicators according to the odds ratio data of the nodule group analysis data;

[0080] S-5-3)screening mass group core indicators according to the odds ratio data of the mass group analysis data;

[0081] S-5-4)constructing a nodule evaluation sub-model and a mass evaluation sub-model based on the nodule group core indicators and the mass group core indicators respectively to form the evaluation model.

[0082] In order to make the purpose, technical scheme and advantages of the present application clearer, the implementation examples are made according to the above-mentioned manner. Figures 3 to 9 ​​The application will be further described in detail with reference to the drawings. It should be understood that the specific embodiments described herein are intended to explain the application and are not intended to limit the application.

[0083] The lung abnormality evaluation method provided by the embodiment can be applied to the application environment as shown. Figure 3 The terminal 102 communicates with the server 104 through the network. The data storage system can store the data required by the server 104 for processing. The data storage system can be integrated on the server 104, or placed on the cloud or other network servers. The server 104 obtains user information of a user to be predicted; the user information includes lung image data; determines an abnormality evaluation sub-model according to the lung image data and a pre-trained target evaluation model; and performs lung abnormality evaluation on the user information according to the abnormality evaluation sub-model to obtain an evaluation score. The terminal 102 can be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, Internet of Things devices, and portable wearable devices. The Internet of Things device can be a smart speaker, a smart television, a smart air conditioner, a smart vehicle device, a projection device, etc. The portable wearable device can be a smart watch, a smart bracelet, a head-mounted device, etc. The head-mounted device can be a virtual reality (VR) device, an augmented reality (AR) device, smart glasses, etc. The server 104 can be a stand-alone physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0084] In an exemplary embodiment, as shown in Figure 4 , a lung abnormality evaluation method is provided. The method is applied to the server 104 in Figure 3 for example, and includes the following steps 202 to 206.

[0085] Among them:

[0086] Step 202, obtaining user information of a user to be predicted; the user information includes lung image data.

[0087] Among them, the user to be predicted refers to a patient who may have an abnormal state in the lung, and the user information specifically includes clinical data of the patient and methylation data of a minimally invasive bronchoalveolar lavage fluid sample. The minimally invasive bronchoalveolar lavage fluid sample refers to BALF (bronchoalveolar lavage fluid), and the clinical data includes basic information, blood test data, and lung image data. The basic information includes age, gender, and smoking history, etc. The blood test data is a blood tumor marker, including CEA, Cyfra21-1, NSE, ProGRP, and SCCA, etc.

[0088] It should be noted that the patient with abnormal lung state refers to the patient with lung shadow but the lung shadow is not determined to be benign or malignant, and the patient with abnormal lung state can refer to the patient with lung shadow which can be malignant, and the patient without abnormal lung state can refer to the patient with benign lung shadow.

[0089] It should be further noted that the BALF methylation data is obtained by BALF sampling after the bronchoscopic biopsy of the user to be predicted, and the BALF methylation data can adopt LungMe methylation (SHOX2 and RASSF1A methylation). LungMe methylation positive is obtained based on any one of the methylation positive of the growth and proliferation regulator and the typical tumor suppressor gene. The growth and proliferation regulator refers to SHOX2, which has the function of regulating growth and development. The tumor suppressor gene refers to RASSF1A, which is related to gene transduction, signal transduction, cytoskeleton, cell cycle, cell adhesion and apoptosis, and can be used for auxiliary diagnosis and treatment effect evaluation of lung cancer, and can also be used for predicting the prognosis of lung cancer patients.

[0090] In some embodiments, the comparison result is obtained by comparing the positive rates of SHOX2 and RASSF1A methylation in the bronchial lavage fluid (BALF) samples collected before and after the bronchoscopic biopsy of the user to be predicted. The comparison result indicates that the positive rate of SHOX2 and RASSF1A methylation in the BALF after biopsy is significantly higher than that before biopsy, that is, the BALF methylation data obtained by BALF sampling after bronchoscopic biopsy of the user to be predicted can significantly improve the detection sensitivity of lung abnormal state.

[0091] In some embodiments, the user information of the user to be predicted can be obtained by clinical sampling and detection, or can be obtained from the patient database, without being limited thereto.

[0092] The user information is obtained by the consent of the user to be predicted.

[0093] Step 204, determining an abnormality evaluation sub-model according to the lung image data and the pre-trained target evaluation model.

[0094] The lung image data is CT imaging signs, and the CT imaging signs include burr, lobulation, irregular edge, bronchial obstruction, bronchial truncation sign, vacuole sign, pleural traction and small blood vessel running, etc. CT (Computed Tomography) refers to electronic computed tomography.

[0095] In some embodiments, determining the abnormality evaluation sub-model according to the lung image data and the pre-trained target evaluation model comprises: determining lung shadow size data of the user to be predicted according to the lung image data; and determining the abnormality evaluation sub-model from the target evaluation model according to the lung shadow size data.

[0096] The lung shadow size data is data for representing the size of the lung shadow of the user to be predicted, and specifically can be the maximum diameter of the lung shadow or other data for measuring the size of the lung shadow, without being limited thereto.

[0097] In some embodiments, lung image analysis data is obtained by analyzing the lung image data, and the size of the lung shadow of the user to be predicted can be determined according to the lung image analysis data, so as to obtain the lung shadow size data.

[0098] In some embodiments, the abnormality evaluation sub-model of the lung shadow type corresponding to the lung shadow size data can be determined from the target evaluation model according to the lung shadow size data, so as to select the corresponding abnormality evaluation sub-model according to the size of the lesion, thereby improving the accuracy of the subsequent lung abnormality evaluation step.

[0099] In some embodiments, the target evaluation model comprises a target lung nodule evaluation model and a target lung mass evaluation model; and determining the abnormality evaluation sub-model from the target evaluation model according to the lung shadow size data comprises: if the lung shadow size data is less than or equal to a preset threshold, taking the target lung nodule evaluation model as the abnormality evaluation sub-model; and if the lung shadow size data is greater than the preset threshold, taking the target lung mass evaluation model as the abnormality evaluation sub-model.

[0100] The preset threshold can be set according to the specific lung nodule and lung mass grouping standard of the training sample in the pre-training process, and the preset threshold used in the pre-training stage and the actual application stage is unified.

[0101] In this embodiment, the preset threshold is 30 mm. That is, if the lung shadow size data is less than or equal to the preset threshold 30 mm, the lung shadow at this time is determined to be a lung nodule, and the target lung nodule evaluation model should be used as the abnormality evaluation sub-model to evaluate it. If the lung shadow size data is greater than the preset threshold 30 mm, the lung shadow at this time is determined to be a lung mass, and the target lung mass evaluation model should be used as the abnormality evaluation sub-model to evaluate it.

[0102] In this embodiment, the lung shadow is evaluated for lung abnormalities by selecting a corresponding model as an abnormality evaluation sub-model according to the lung shadow size data. Since the target evaluation model can be trained without using the lung shadow size data as variable data, the training amount of the target evaluation model can be effectively reduced, thereby reducing the waste of computer resources, and improving the prediction accuracy of each sub-model in the target evaluation model for the lung shadow of each grouping size.

[0103] In step 206, the user information is evaluated for lung abnormalities according to the abnormality evaluation sub-model to obtain an evaluation score.

[0104] The lung abnormality evaluation is an evaluation step for predicting the possibility of lung abnormality, and the evaluation score is only an intermediate result for assisting in judging the possibility of lung abnormality.

[0105] In this embodiment, the user information is input into the abnormality evaluation sub-model for lung abnormality evaluation, thereby obtaining an evaluation score that can assist in judging the possibility of lung abnormality.

[0106] In the lung abnormality evaluation method, user information of a user to be predicted is obtained, the user information includes lung image data, an abnormality evaluation sub-model is determined according to the lung image data and a pre-trained target evaluation model, and the user information is evaluated for lung abnormalities according to the abnormality evaluation sub-model to obtain an evaluation score. Therefore, different lung image data in the user information of the user to be predicted is combined with the pre-trained target evaluation model to determine the abnormality evaluation sub-model, and the user information is evaluated for lung abnormalities by using the abnormality evaluation sub-model, thereby obtaining the evaluation score. The evaluation score can more effectively and accurately evaluate the lung abnormality state of the user to be predicted before surgery or treatment, and the evaluation score can also be combined with other related lung abnormality data for joint evaluation, which can promote the determination of more accurate disease condition, surgical method and treatment plan before treatment by a clinician.

[0107] In an exemplary embodiment, as shown in Figure 5 The training steps of the target evaluation model include steps 302 to 310. Wherein:

[0108] In step 302, training sample data of a target user is obtained.

[0109] The target user can be a patient who may have an abnormal state in the lung, and the training sample data includes methylation training data and training sample data of the target user. The training sample data includes clinical features, tumor markers and CT imaging features of lung nodules and lung masses, and the methylation training data can be LungMe methylation indicators.

[0110] In some embodiments, the training sample data of the target user can be obtained from a patient database, wherein the training sample data of the target user is stored in the patient database after the target user agrees.

[0111] At step 304, the training sample data is subjected to single-factor analysis by the validation set data corresponding to the target user, to obtain a single-factor analysis result.

[0112] The validation set data corresponding to the target user refers to lung shadow benign and malignant data of the target user, which can be benign or malignant.

[0113] In this embodiment, the training sample data includes methylation training data and clinical training data; the single-factor analysis result includes methylation index data and clinical index data; the single-factor analysis of the training sample data by the validation set data corresponding to the target user includes single-factor analysis of the methylation training data according to the validation set data, to obtain the methylation index data, and single-factor analysis of the clinical training data according to the validation set data, to obtain the clinical index data.

[0114] In this embodiment, the single-factor logistic analysis of the methylation training data and the clinical training data in the training sample data by the validation set data corresponding to the target user obtains the single-factor analysis result including the methylation index data and the clinical index data, which can be used to screen and verify the risk factors related to the lung shadow benign and malignant data in the validation set data.

[0115] In this embodiment, please refer to Figure 6 The single-factor analysis result is obtained by single-factor logistic analysis of the clinical characteristics, tumor markers, CT imaging characteristics, and LungMe methylation index of the lung nodules and lung masses in the training sample data, as shown in Figure 6 , wherein the group where A is located refers to the lung nodule group, and the group where B is located refers to the lung mass group, Figure 6 A in the lung nodule group of the training sample data shows the correlation between the age, smoking history, CEA, Cyfra21-1, ProGRP, airway obstruction, and LungMe methylation index and the lung nodule benign and malignant data in the validation set data, and B in the lung mass group of the training sample data shows the correlation between the age, smoking history, CEA, Cyfra21-1, SCCA, lobulation, airway obstruction, and LungMe methylation index and the lung mass benign and malignant data in the validation set data. Figure 6

[0116] It should be noted that, Figure 6 ​A higher median odds ratio (OR) indicates a stronger correlation between the corresponding factor and the benign or malignant nature of lung shadows. Figure 6 The p-value is a test parameter used to determine the result of a hypothesis test. The smaller the p-value, the more significant the result.

[0117] Step 306: Filter the results of the univariate analysis to obtain significant index data;

[0118] The results of univariate analysis can be filtered using the P-value, and the results obtained from the filtering can be determined based on the size of the P-value and actual needs, but are not limited to this.

[0119] In some embodiments, please refer to Figure 6 It can be done Figure 6 The p-value was used to filter the results of the univariate analysis. The results were processed according to preset requirements to obtain significant index data. The preset requirement was p < 0.05.

[0120] It is worth noting that in univariate analysis (such as univariate screening or univariate regression analysis), the setting of the p-value threshold (significance level) usually depends on the research objective, disciplinary conventions, and tolerance for false positives (Type I error).

[0121] The preset requirement used in this invention is actually the classic threshold: P<0.05;

[0122] The most widely used standard corresponds to a 5% significance level (α = 0.05). This indicates statistical significance, but it should be noted that results may be unstable when the p-value is close to 0.05 (e.g., 0.04 vs 0.06).

[0123] Ronald Fisher proposed in the 20th century that a false positive rate of 5% is an acceptable equilibrium point.

[0124] Too strict a threshold (e.g., P<0.01) may miss true correlations (false negatives), while too lenient a threshold (e.g., P<0.1) may introduce too much noise (false positives).

[0125] In this embodiment, a univariate analysis is performed on the training sample data using the validation set data corresponding to the target user to obtain the univariate analysis results. The univariate analysis results are then filtered to identify risk factors related to abnormal lung conditions in the validation set data.

[0126] Step 308: Perform multivariate regression analysis based on the significant index data and validation set data to obtain the multivariate analysis results;

[0127] For example, a logistic regression model is used when performing a multi-factor regression analysis, specifically, it is a formula for calculating a probability P2, which can be used for medical diagnosis or risk assessment, for example:

[0128] predicting the probability of a patient having a certain disease (e.g., lung cancer).

[0129] comprehensive judgment combining CEA (tumor marker) and bronchial obstruction (imaging feature).

[0130] It should be noted that since the methylation index data has the largest correlation with the lung shadow benign and malignant data and accounts for a large proportion in the significant index data, in some embodiments, the multi-factor regression analysis is divided into multi-factor regression analysis containing methylation index data and multi-factor regression analysis not containing methylation index data.

[0131] In some embodiments, in the lung nodule group, the multi-factor analysis results obtained by the multi-factor regression analysis containing methylation index data are as shown in Table 1 below:

[0132] Table 1 Multi-factor analysis results containing methylation index data

[0133]

[0134] It should be noted that from Table 1 above, in the multi-factor analysis results containing methylation index data in the lung nodule group, the significant indicators with P value less than 0.05 in the multi-factor analysis results are screened out, and the OR value of LungMe methylation data is the highest, which is 20.191, followed by the OR value of CEA level, which is 1.050. Therefore, in the multi-factor analysis results containing methylation index data, the methylation index data shows the best tumor specificity, and LungMe methylation data and CEA are independent influencing indicators in lung abnormality evaluation.

[0135] In some embodiments, according to the significant indicators in the multi-factor analysis results in Table 1, a first target sub-evaluation model is constructed, and the expression of the first target sub-evaluation model P1 is as follows:

[0136] P1 = 1 / (1+exp(-(-2.36+0.231*CEA+3.005*LungMe methylation)));

[0137] Wherein, CEA represents the index data of CEA level, LungMe methylation represents the index data of whether LungMe methylation occurs, and it should be noted that the evaluation score P1 obtained after the risk assessment of the first target sub-evaluation model is used to help the clinician to assist in judging the benignity and malignancy of the lung shadow. For example, when P1>0.5, it can be considered that the probability of lung shadow being malignant is large, and when P1<0.5, it can be considered that the probability of lung shadow being benign is large.

[0138] The calculation formula of P1 is actually a Logistic regression model, which is used to calculate the probability P1 of an event (such as the risk of disease). The following explains the definition of each parameter in the formula:

[0139] (1) Intercept: -2.36

[0140] Definition: The log odds (logarithmic odds) when all independent variables (CEA and LungMe methylation) are 0.

[0141] Explanation: If CEA=0 and LungMe methylation=0, the log odds of the event occurring is -2.36.

[0142] Corresponding baseline probability:

[0143] P1=1 / (1+e2.36)≈0.086

[0144] (2) Regression coefficient of CEA: 0.231

[0145] Definition: For every 1 unit increase in CEA (carcinoembryonic antigen), the log odds increase by 0.231.

[0146] Explanation:

[0147] OR value (odds ratio): e0.231≈1.26, indicating that for every 1 unit increase in CEA, the odds of the event occurring increase by 26%.

[0148] Clinical significance: The higher the CEA level, the higher the risk of disease (assuming that CEA is positively correlated with disease).

[0149] (3) Regression coefficient of LungMe methylation: 3.005

[0150] Definition: For every 1 unit increase in LungMe methylation, the log odds increase by 3.005.

[0151] Explanation:

[0152] OR value: e3.005≈20.2, which means that the risk of lung cancer increases by about 20 times for every 1 unit increase in LungMe methylation.

[0153] Clinical significance: LungMe methylation may be a strong predictor and has a significant impact on risk.

[0154] In the present application, the above model is used for cancer risk prediction (such as lung cancer, lung abnormalities, etc.) in combination with the following indicators:

[0155] CEA: tumor marker, reflecting tumor burden.

[0156] LungMe methylation: epigenetic marker (such as the methylation level of a certain gene), which may be related to tumorigenesis.

[0157] In some embodiments, the multi-factor analysis results obtained by multi-factor regression analysis without methylation indicator data in the lung nodule group are shown in Table 2 below:

[0158] Table 2 Multi-factor analysis results without methylation indicator data

[0159]

[0160] It should be noted that from Table 2 above, in the multi-factor analysis results without methylation indicator data in the lung nodule group, the significant indicators with P value less than 0.05 in the multi-factor analysis results are CEA level and bronchial obstruction, and the OR values of CEA level and bronchial obstruction are 1.213 and 6.744, respectively. Therefore, in the multi-factor analysis results without methylation indicator data, CEA level and bronchial obstruction are significant indicators in lung abnormality evaluation.

[0161] In some embodiments, a second target sub-evaluation model is constructed according to the significant indicators in the multi-factor analysis results of Table 2, and the expression of the second target sub-evaluation model P2 is as follows:

[0162] P2 = 1 / (1 + exp(-(-1.611 + 0.193*CEA + 1.909*bronchial obstruction)));

[0163] Where CEA represents the indicator data of CEA level, and bronchial obstruction represents the indicator data of whether bronchial obstruction occurs. It should be noted that the evaluation score P2 obtained by the second target sub-evaluation model after risk evaluation is used to help clinicians assist in judging the benignity or malignancy of tumors. For example, when P2>0.5, it can be considered that the probability of lung shadow being malignant is large, and when P2<0.5, it can be considered that the probability of lung shadow being benign is large.

[0164] In some embodiments, the multi-factor analysis results of the lung mass group containing methylation index data are shown in Table 3 below:

[0165] Table 3 Multi-factor analysis results of the lung mass group containing methylation index data

[0166]

[0167] It should be noted that from Table 3, in the multi-factor analysis results of the lung mass group containing methylation index data, the significant indicators with P value less than 0.05 in the multi-factor analysis results are selected, including CEA level, bronchial obstruction, lobulation, Cyfra21-1 and LungMe methylation. Therefore, in the multi-factor analysis results of the lung mass group containing methylation index data, CEA level, bronchial obstruction, lobulation, Cyfra21-1 and LungMe methylation are significant indicators in lung abnormality evaluation.

[0168] In some embodiments, the third target sub-evaluation model is constructed according to the significant indicators in the multi-factor analysis results of Table 3, and the expression of the third target sub-evaluation model P3 is as follows:

[0169] P3 = 1 / (1+exp(-(-6.114+0.533*CEA+0.887*Cyfra21-1+2.442*lobulation+3.693*bronchial obstruction+4.981*LungMe methylation)));

[0170] Wherein, CEA represents the index data of CEA level, LungMe methylation represents the index data of whether LungMe methylation occurs, Cyfra21-1 represents the index data of Cyfra21-1 level, lobulation represents the index data of whether lobulation occurs, and bronchial obstruction represents the index data of whether bronchial obstruction occurs. It should be noted that the evaluation score P3 obtained by the third target sub-evaluation model after risk evaluation is used to help clinicians assist in judging the benign and malignant of the shadow. For example, when P3>0.5, it can be considered that the probability of lung shadow being malignant is large, and when P3<0.5, it can be considered that the probability of lung shadow being benign is large.

[0171] In some embodiments, the multi-factor analysis results of the lung mass group not containing methylation index data are shown in Table 4 below:

[0172] Table 4 Multi-factor analysis results of the lung mass group not containing methylation index data

[0173]

[0174] It should be noted that, according to Table 4, in the multi-factor analysis result of the lung mass group without including the methylation index data, the significant indicators with P value less than 0.05 screened out in the multi-factor analysis result include: CEA level, bronchial obstruction, lobulation, and Cyfra21-1. Therefore, in the multi-factor analysis result of the lung mass group including the methylation index data, CEA level, bronchial obstruction, lobulation, and Cyfra21-1 are significant indicators in the lung abnormality evaluation.

[0175] In some embodiments, a fourth target sub-evaluation model is constructed according to the significant indicators in the multi-factor analysis result of Table 1, and the expression of the fourth target sub-evaluation model P4 is as follows:

[0176] P4 = 1 / (1+exp(-(-4.209+0.427*CEA+0.795*Cyfra21-1+2.724*lobulation+3.515*bronchial obstruction)));

[0177] Wherein, CEA represents the index data of CEA level, Cyfra21-1 represents the index data of Cyfra21-1 level, lobulation represents the index data of whether lobulation occurs, and bronchial obstruction represents the index data of whether bronchial obstruction occurs. It should be noted that the evaluation score P4 obtained after risk evaluation by the fourth target sub-evaluation model is used to help clinicians assist in judging the benignity or malignancy of the shadow. For example, when P4>0.5, it can be considered that the probability of the lung shadow being malignant is large, and when P4<0.5, it can be considered that the probability of the lung shadow being benign is large.

[0178] Step 310, constructing a target evaluation model according to the lung shadow type of the target user and the multi-factor analysis result.

[0179] Wherein, the lung shadow type is determined by the shadow diameter size of the target user.

[0180] In some embodiments, constructing a target evaluation model according to the lung shadow type of the target user and the multi-factor analysis result includes: dividing the multi-factor analysis result according to the lung shadow type to obtain a nodular group analysis result and a mass group analysis result; performing index screening according to the odds ratio data of the nodular group analysis result to obtain nodular group key index data; performing index screening according to the odds ratio data of the mass group analysis result to obtain mass group key index data; and constructing a target evaluation model according to the nodular group index data and the mass group index data.

[0181] In some embodiments, the target user is divided into different data groups through the lung shadow type, and the multi-factor analysis result is divided into multiple group results, and the corresponding type of analysis result data is screened from the multi-factor analysis result respectively, and a target sub-evaluation model of each data group is constructed, and then the target sub-evaluation models are combined to obtain the target evaluation model.

[0182] It should be noted that when the lung shadow type is a nodule, the nodule group key indicator data corresponding to the nodule group in the multi-factor analysis result is obtained, and a target lung nodule evaluation model of the nodule group is constructed according to the analysis result data. Similarly, a target lung mass evaluation model is constructed and combined to obtain a target evaluation model.

[0183] The nodule group key indicator data refers to the most important indicator data in the multi-factor analysis result corresponding to the nodule group. For example, in the multi-factor analysis result containing methylation indicator data, the methylation indicator data shows the best tumor specificity and is the most important indicator in lung abnormality evaluation. The CEA level is second. Therefore, a target lung nodule evaluation model containing two indicators of LungMe methylation and CEA level can be constructed. Similarly, the construction process of the target lung mass evaluation model can be obtained.

[0184] In this embodiment, by constructing a target evaluation model according to the nodule type and multi-factor analysis result of a target user, the training amount of the target evaluation model can be effectively reduced without using the nodule type as a training parameter, and the evaluation accuracy of each evaluation sub-model for each grouped data is improved, thereby improving the evaluation accuracy of the target evaluation model.

[0185] In order to more clearly understand the scheme of the present application, an exemplary embodiment is combined herein Figure 7 for illustration, as follows:

[0186] In an exemplary embodiment, a 70-year-old male patient is used as a user to be predicted, and the user information of the male patient includes the following contents: a history of smoking, a CT image taken on admission diagnoses that there is a 21 mm x 18 mm shadow in the right middle lobe of the lung, the shadow edge has a burr sign, has lobulation, has pleural traction, the edge is regular, has no bronchial obstruction, has no bronchial obstruction, has no air bubble sign, has no small blood vessel passing through, etc. The RASSF1A methylation detection result is positive, the SHOX2 methylation detection is negative, the CEA level is 9.41 ng / mL, the NSE level is 25.57 ng / mL, the CYFRA21-1 level is 17.23 ng / mL, the ProGRP level is 54.79 ng / mL, the SCC level is 1.4 ng / mL, the alveolar lavage fluid brush sheet finds abnormal cells, the bronchoscopic biopsy finds cancer cells, and finally identifies it as stage IV lung adenocarcinoma. The identification result is used as verification data in this embodiment.

[0187] First, based on the patient's lung imaging data, the maximum diameter of the lung shadow (21 mm) was determined to be less than 30 mm, belonging to the small nodule group. Therefore, the abnormal assessment sub-models P1 and P2 were determined from the pre-trained target assessment model, as shown below. P1 is the abnormal assessment sub-model trained using methylation results, while P2 is the abnormal assessment sub-model trained without methylation results. Since the RASSF1A methylation result is positive, it is recorded as 1; the patient has no bronchial obstruction, so it is recorded as 0. Based on the equations corresponding to models P1 and P2, the user information was used to assess lung abnormalities, resulting in the following assessment scores:

[0188] P1=1 / (1+exp(-(-2.36+0.231*9.41+3.005*1)))=0.944;

[0189] P2=1 / (1+exp(-(-1.611+0.193*9.41+1.909*0)))=0.551;

[0190] It should be noted that the assessment scores calculated by the two models are P-values, both of which are greater than 0.5, indicating that the patient's lung abnormality has a high probability of being malignant; P1 is closer to 1, which means that the model including methylation is relatively more accurate.

[0191] In this exemplary embodiment, when validating the pre-trained target evaluation model, relevant validation data can be obtained by training an existing ROC-AUC model in combination with diagnosed patient data.

[0192] Specifically, the ROC-AUC model can obtain data through ROC curves, such as... Figure 7 As shown, Figure 7 Provided for this embodiment, wherein, Figure 7 Figure A shows the ROC curve of the target lung nodule sub-evaluation model in the lung abnormality assessment method, and Figure B shows the ROC curve of the target lung mass sub-evaluation model. Figure 7 Specifically, it also includes:

[0193] Figure 7 (A): ROC curves of the target lung nodule evaluation model with and without LungMe methylation. Figure 7 (B): ROC curves of target lung mass evaluation models with and without LungMe methylation.

[0194] like Figure 7As shown in the figure, the AUC of the ROC curve of the target lung nodule sub-evaluation model containing LungMe methylation and the target lung nodule sub-evaluation model not containing LungMe methylation is 0.873 and 0.763 respectively, and thus we conclude that the accuracy of the target lung nodule sub-evaluation model containing methylation is 87.3%, and the accuracy of the target lung nodule sub-evaluation model not containing LungMe methylation is 76.3%.

[0195] Optionally, the AUC of the ROC curve of the target lung nodule sub-evaluation model containing LungMe methylation and the target lung nodule sub-evaluation model not containing LungMe methylation is 0.991 and 0.932 respectively, and thus we conclude that the accuracy of the target lung nodule sub-evaluation model containing methylation is 99.1%, and the accuracy of the target lung nodule sub-evaluation model not containing LungMe methylation is 93.2%.

[0196] In combination with specific external hospital validation set data, the above two pre-trained target lung nodule sub-evaluation models and target lung mass sub-evaluation models are used to verify the accuracy of the models, and 30 lung nodule patients and 30 lung mass patients are used to verify the accuracy of the models.

[0197] The user information includes at least one of the basic information, blood test data, minimally invasive BALF methylation data and lung image data of the user to be predicted. The basic information includes age, gender, smoking history, etc.; the blood test data is blood tumor markers, including CEA, Cyfra21-1, NSE, ProGRP and SCCA, etc.; the BALF methylation data is LungMe methylation, wherein LungMe is SHOX2 methylation combined with RASSF1A methylation; the lung image data is CT imaging signs, including burr, lobulation, irregular edge, bronchial obstruction, bronchial truncation sign, vacuole sign, pleural traction and small blood vessel traversal, etc.; the user to be predicted can be a lung cancer patient or a non-lung cancer patient, and can be a lung nodule or a lung mass. The user information is obtained by user consent.

[0198] Optionally, the user information of the 30 lung nodule users can be presented in the form of Table 5, as follows:

[0199] Table 5 Target lung nodule sub-evaluation model verification table

[0200]

[0201]

[0202] In Table 5, the benign and malignant results in Table 5 are obtained according to the evaluation score, 1 represents malignant, 0 represents benign; 1 represents methylation positive, 0 represents methylation negative; 1 represents bronchial obstruction, 0 represents unobstructed; 1 represents model prediction of malignant, 0 represents prediction of benign.

[0203] Among them, when the evaluation score is greater than or equal to 0.5, the patient is marked as malignant, and when the evaluation score is less than 0.5, the patient is marked as benign.

[0204] As shown in Table 5, among the 30 patients with lung nodules, the accuracy of model 1 is 86.7% (26 / 30), and the accuracy of model 2 is 73.3% (22 / 30). The results are consistent with the ROC results.

[0205] Alternatively, the information of the 30 lung tumor users can be presented in the form of Table 6, as follows:

[0206] Table 6 Target lung tumor sub-evaluation model verification table

[0207]

[0208]

[0209] In Table 6, the benign and malignant results in Table 6 are obtained according to the evaluation score, 1 represents malignant, 0 represents benign; 1 represents methylation positive, 0 represents methylation negative; 1 represents lobulation, 0 represents unlobulation; 1 represents bronchial obstruction, 0 represents unobstructed; 1 represents model prediction of malignant, 0 represents prediction of benign.

[0210] Among them, when the evaluation score is greater than or equal to 0.5, the patient is marked as malignant, and when the evaluation score is less than 0.5, the patient is marked as benign.

[0211] As shown in Table 6, among the 30 patients with lung nodules, the accuracy of model 3 is 96.7% (29 / 30), and the accuracy of model 4 is 83.3% (25 / 30).

[0212] Therefore, the accuracy of the lung cancer benign and malignant model containing methylation (model 1 and model 3) is higher than that of the model not containing methylation (model 2 and model 4), so the model containing methylation is selected for lung cancer benign and malignant identification in this embodiment.

[0213] It should be understood that although the steps in the flowcharts involved in the embodiments described above are shown in sequence according to the arrows, the steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, the execution of the steps is not strictly limited in sequence, and the steps can be executed in other orders. Moreover, at least some of the steps in the flowcharts involved in the embodiments described above can include multiple steps or stages, which are not necessarily executed at the same time but can be executed at different times, and the execution of the steps or stages is not necessarily sequential but can be performed alternately or alternately with at least part of other steps or stages.

[0214] Based on the same inventive concept, the embodiments also provide a lung abnormality evaluation device for implementing the lung abnormality evaluation method described above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more lung abnormality evaluation device embodiments provided below can refer to the limitations of the lung abnormality evaluation method described above, which will not be repeated here.

[0215] In an exemplary embodiment, as shown in Figure 8 A lung abnormality evaluation device 600 is provided, comprising: an acquisition module 601, a determination module 602, and an evaluation module 603, wherein:

[0216] The acquisition module 601, i.e., the data acquisition module, is configured to acquire user information of a to-be-predicted user; the user information comprises: lung image data;

[0217] The determination module 602, i.e., the model selection module, is configured to determine an abnormality evaluation sub-model according to the lung image data and a pre-trained target evaluation model;

[0218] The evaluation module 603, i.e., the target evaluation model (or evaluation execution module), is configured to perform lung abnormality evaluation on the user information according to the abnormality evaluation sub-model to obtain an evaluation score.

[0219] In some embodiments, the determination module 602 is further configured to determine lung nodule size data of the to-be-predicted user according to the lung image data; and determine the abnormality evaluation sub-model from the target evaluation model according to the lung nodule size data.

[0220] In some embodiments, the target evaluation model comprises: a target lung nodule evaluation model and a target lung mass evaluation model; and the determination module 602 is further configured to, if the lung nodule size data is less than or equal to a preset threshold, take the target lung nodule evaluation model as the abnormality evaluation sub-model; and if the lung nodule size data is greater than the preset threshold, take the target lung mass evaluation model as the abnormality evaluation sub-model.

[0221] In some embodiments, the lung abnormality evaluation device further comprises a training module configured to obtain training sample data of the target user, perform single-factor analysis on the training sample data based on the corresponding verification set data of the target user to obtain single-factor analysis results, filter the single-factor analysis results to obtain significant indicator data, perform multi-factor regression analysis based on the significant indicator data and the verification set data to obtain multi-factor analysis results, and construct a target evaluation model based on the nodule type of the target user and the multi-factor analysis results.

[0222] In some embodiments, the training sample data comprises methylation training data and clinical training data, and the single-factor analysis results comprise methylation indicator data and clinical indicator data. The training module is further configured to perform single-factor analysis on the methylation training data based on the verification set data to obtain the methylation indicator data, and perform single-factor analysis on the clinical training data based on the verification set data to obtain the clinical indicator data.

[0223] In some embodiments, the training module is further configured to divide the multi-factor analysis results based on the nodule type to obtain nodule group analysis results and mass group analysis results, perform indicator filtering based on odds ratio data of the nodule group analysis results to obtain nodule group key indicator data, perform indicator filtering based on odds ratio data of the mass group analysis results to obtain mass group key indicator data, and construct the target evaluation model based on the nodule group indicator data and the mass group indicator data.

[0224] In the lung abnormality evaluation device described above, user information of a user to be predicted is obtained, the user information comprises lung image data, an abnormality evaluation sub-model is determined based on the lung image data and a pre-trained target evaluation model, and lung abnormality evaluation is performed on the user information based on the abnormality evaluation sub-model to obtain an evaluation score. Therefore, different lung image data in the user information of the user to be predicted is combined with the pre-trained target evaluation model to determine the abnormality evaluation sub-model, and the abnormality evaluation sub-model is used to perform lung abnormality evaluation on the user information, thereby obtaining the evaluation score. The evaluation score can more effectively and accurately evaluate the lung abnormality state of the user to be predicted before surgery or treatment, and the evaluation score can also be combined with other related lung abnormality data for joint evaluation, thereby achieving the effect of promoting the clinician to make more accurate disease judgment, surgical method, and treatment plan before treatment.

[0225] Each module in the lung abnormality evaluation device described above can be realized by software, hardware, and combinations thereof, in whole or in part. Each module described above can be embedded in or independent of a processor in a computer device in hardware form, or can be stored in a memory in a computer device in software form, so as to be called and executed by a processor to perform the operations corresponding to each module.

[0226] In an example embodiment, a computer device, which can be a server, is provided, and an internal structure diagram of the computer device can be as shown in FIG. 1. Figure 9 The computer device includes a processor, a memory, an input / output interface (I / O), and a communication interface. The processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The database of the computer device is configured to store a target evaluation model. The input / output interface of the computer device is configured to exchange information between the processor and external devices. The communication interface of the computer device is configured to communicate with terminals outside through a network connection. The computer program is executed by the processor to implement a lung abnormality evaluation method.

[0227] Those skilled in the art can understand that Figure 9 The structure shown in the above embodiment is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the diagram, or combine certain components, or have a different arrangement of components.

[0228] In an example embodiment, a computer device is provided, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the lung abnormality evaluation method.

[0229] In an example embodiment, a computer readable storage medium is provided, which stores a computer program. The computer program is executed by a processor to implement the steps of the lung abnormality evaluation method.

[0230] In an example embodiment, a computer program product is provided, which includes a computer program. The computer program is executed by a processor to implement the steps of the lung abnormality evaluation method.

[0231] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use, and processing of the related data need to comply with relevant regulations.

[0232] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, database or other medium used in the embodiments provided by the present application can include at least one of non-volatile memory and volatile memory. The non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. The volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, the RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The database involved in the embodiments provided by the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided by the present application can be a general processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., without being limited thereto.

[0233] The above only describes the preferred embodiments of the present application, and is not intended to limit the present application. Those skilled in the art can make changes to the present application without departing from the spirit of the present application, and such changes are within the scope of the present application.

Claims

1. A method for assessing lung abnormalities, characterized in that, Includes the following steps: 1) Obtain user information of the subject to be tested, wherein the user information includes lung imaging information; 2) Based on lung imaging information and assessment models that can be used to judge lung abnormalities in the test subjects, determine sub-assessment models that are suitable for the lung conditions of the test subjects. 3) Using the user information of the test subject and the adapted sub-evaluation model, the lung abnormalities of the test subject are evaluated and the evaluation results are output.

2. The method according to claim 1, characterized in that, In step 2), the method for determining a sub-assessment model that is suitable for the lung condition of the test subject based on lung imaging information and an assessment model that can be used to judge lung abnormalities in the test subject specifically includes: 2-1) Extract lung shadow size data of the subject from lung imaging information; 2-2) Select the corresponding sub-evaluation model from the evaluation model based on the lung shadow size data.

3. The method according to claim 2, characterized in that, The assessment model includes a nodule assessment sub-model and a mass assessment sub-model. In step 2-2), the method of selecting the corresponding sub-assessment model from the assessment model based on the lung shadow size data specifically includes: 2-2-1) If the size of the lung shadow does not exceed the preset threshold, the nodule assessment sub-model will be used as the sub-assessment model that is adapted to the lung condition of the subject. 2-2-2) If the size of the lung shadow exceeds the preset threshold, the mass assessment sub-model will be used as the sub-assessment model that is adapted to the lung condition of the subject.

4. The method according to any one of claims 1-3, characterized in that, The construction method of the evaluation model specifically includes: S-1) Using patients with various lung abnormalities and healthy individuals as samples, collect user information corresponding to several types of samples to form a lung abnormality assessment dataset. This lung abnormality assessment dataset includes lung imaging information, methylation data and clinical characteristic data of the samples. S-2) divides the lung abnormality assessment dataset into a training sample set and a validation dataset; S-3) Perform univariate analysis on the training sample set using the validation dataset to obtain initial analysis indicators; S-3) Screen the initial analysis indicators to obtain key indicators with statistical significance; S-4) Combine key indicators with validation datasets to perform multivariate regression analysis and generate comprehensive analysis results; S-5) An evaluation model is constructed based on the lung shadow type of the sample and the comprehensive analysis results.

5. The method according to claim 4, characterized in that, The training sample set includes a methylation training subset and a clinical training subset. The initial analysis indicators include methylation feature indicators and clinical feature indicators. In step S-3), the initial analysis indicators are obtained by performing a univariate analysis on the training sample set using the validation dataset, specifically including: S-3-1) Based on the validation dataset, a univariate analysis was performed on the methylation training subset to obtain methylation feature indicators; S-3-2) Based on the validation dataset, a univariate analysis was performed on the clinical training subset to obtain clinical characteristic indicators.

6. The method according to claim 4, characterized in that, In step S-5), the method for constructing the evaluation model based on the type of lung shadow in the sample and the comprehensive analysis results specifically includes: S-5-1) The comprehensive analysis results were grouped according to the type of lung shadow to obtain the analysis data of the nodule group and the mass group; S-5-2) Based on the ratio data of the nodule group analysis data, the core indicators of the nodule group are selected; S-5-3) Based on the ratio data of the mass group analysis data, the core indicators of the mass group are screened out; S-5-4) Based on the core indicators of the nodule group and the core indicators of the mass group, a nodule assessment sub-model and a mass assessment sub-model are constructed respectively, and combined to form an assessment model.

7. A lung abnormality assessment device, characterized in that, It includes a data acquisition module for collecting and storing user information, a model selection module for matching sub-assessment model types, and an assessment execution module for outputting assessment results. The data input terminals of the model selection module and the assessment execution module are both connected to the data acquisition module, enabling the model selection module to select the appropriate sub-assessment model in the assessment execution module based on the lung imaging information in the user information, and output the results of the assessment of the patient's lung abnormalities.

8. The apparatus according to claim 7, characterized in that, The model selection module includes: The size extraction unit is used to extract the size data of the lung shadow of the subject from lung imaging information; The sub-model determination unit is used to select the corresponding sub-evaluation model from the evaluation models based on the lung shadow size data.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images