Lung disease auxiliary analysis method based on exhaled VOC and related equipment

Abstract

The invention discloses a lung disease auxiliary analysis method based on exhaled VOC (volatile organic compounds) and related equipment. According to the method, a gas classification model and a chromatography-mass spectrometry analysis spectrogram of exhaled gas of a target object can be obtained; determining a classification marker for reflecting the lung function condition; identifying a characteristic peak corresponding to each classification marker from the chromatography-mass spectrometry analysis spectrogram, and determining the marker peak area of each classification marker; and determining whether the exhaled gas indicates that the target object suffers from the lung disease or not by utilizing a gas classification model based on the peak area of each marker, and determining the lung disease category corresponding to the exhaled gas when determining that the target object suffers from the lung disease. Therefore, according to the method, the expired gas can be classified by utilizing the gas classification model and combining the peak areas of the plurality of classification markers capable of representing the lung function condition in the expired gas, so that the data analysis difficulty is reduced, the reliability of the classification result is improved, and a reliable diagnosis direction is provided for lung diagnosis.

Description

Technical Field

[0001] The present application relates to the field of information processing technologies, and more specifically, to a method and related device for assisting in the analysis of lung diseases based on exhaled VOCs. Background Art

[0002] Currently, the identification of lung diseases mainly relies on imaging methods, pulmonary function tests, invasive detections, etc. However, the above methods have systematic defects: First, the imaging features of early lung cancer overlap significantly with those of benign lung nodules, COPD and other diseases (for example, the misdiagnosis rate of differentiating the malignancy of ground-glass nodules reaches 35%), and the specificity of LDCT for nodules <6mm is insufficient (false positive rate >20%); Second, the diagnosis of diseases such as PRISm depends on a standardized pulmonary function instrument, and the lack of equipment in primary hospitals leads to a missed diagnosis rate exceeding 60%; Third, diseases such as pulmonary embolism and pulmonary hypertension are often misdiagnosed due to atypical symptoms (such as isolated cough), and the specificity of serum markers such as D-dimer in tumor patients is only 58%. In addition, the coexistence of multiple diseases (such as COPD complicated with lung cancer) is more likely to cause detection blind spots, and the clinical misjudgment rate is as high as 40%.

[0003] Based on this, in order to avoid the above problems, before performing the above-mentioned lung disease identification, a technical solution that provides a reference direction for lung disease diagnosis can be introduced, and then the lung disease diagnosis can be performed according to the reference direction, so as to reduce the misjudgment rate. Summary of the Invention

[0004] In view of this, the present application provides a method and related device for assisting in the analysis of lung diseases based on exhaled VOCs, which are used to provide a reference direction for lung diagnosis.

[0005] In order to achieve the above object, the following solutions are proposed:

[0006] A method for assisting in the analysis of lung diseases based on exhaled VOCs, comprising:

[0007] Obtaining a gas classification model and a gas chromatography-mass spectrometry analysis spectrum of the exhaled gas corresponding to a target object;

[0008] Determining a plurality of classification markers for reflecting the lung function status;

[0009] Identifying the characteristic peaks corresponding to each classification marker from the gas chromatography-mass spectrometry analysis spectrum, and determining the marker peak area of each classification marker;

[0010] Using the gas classification model based on the marker peak areas of each classification marker to determine whether the exhaled gas indicates that the target object has a lung disease, and when it is determined that the target object has a lung disease, determining the category of the lung disease corresponding to the exhaled gas.

[0011] Optionally, the obtaining of the gas classification model includes:

[0012] Obtain a plurality of training samples, and divide each training sample into a training set and a test set. Among them, each training sample includes an annotation label indicating the corresponding pulmonary physiological state and the training peak area of each training marker;

[0013] Construct a logistic regression model;

[0014] Use the training set and the test set to adjust the parameters of the logistic regression model until the logistic regression model meets the stopping condition. The finally obtained logistic regression model is the gas classification model.

[0015] Optionally, the obtaining of the plurality of training samples includes:

[0016] Obtain the chromatographic-mass spectrometric training spectrograms of the exhaled gases corresponding to different pulmonary physiological states, and obtain the training peak area of each training marker from each chromatographic-mass spectrometric training spectrogram;

[0017] Use recursive feature elimination combined with cross-validation method and random forest model to perform feature screening on the training peak areas of different pulmonary physiological states, and calculate the feature scores of different training peak areas;

[0018] Sort each training peak area according to the feature score from large to small, retain the first N training peak areas, and return to execute the steps of using recursive feature elimination combined with cross-validation method and random forest model to perform feature screening on the training peak areas of different pulmonary physiological states and calculate the feature scores of different training peak areas until each training peak area meets the screening conditions;

[0019] The training peak areas corresponding to the same exhaled gas form a training sample.

[0020] Optionally, the determination of a plurality of classification markers for reflecting the pulmonary function status includes:

[0021] Collect the chromatographic-mass spectrometric spectrograms formed by the exhaled gases in different pulmonary states;

[0022] Calculate the peak retention time, peak vertex mass spectrum, combined peak area and peak quantitative signal-to-noise ratio of each characteristic peak in each chromatographic-mass spectrometric spectrogram;

[0023] Based on the peak retention time and peak vertex mass spectrum of each characteristic peak, identify the characteristic peaks corresponding to the same gas marker in different chromatographic-mass spectrometric spectrograms;

[0024] Construct a peak area matrix containing the peak areas of the combined peaks corresponding to each characteristic peak, and a signal-to-noise ratio matrix containing the signal-to-noise ratios of the peaks corresponding to each characteristic peak, where the elements in the same column of the peak area matrix and the signal-to-noise ratio matrix belong to the same gas marker;

[0025] Filter each combined peak area in the peak area matrix based on the signal-to-noise ratio matrix;

[0026] Perform a differential analysis on the filtered peak area matrix and calculate the characterization values of the lung function status of different gas markers;

[0027] Based on the characterization values of the lung function status of each gas marker, screen multiple classification markers for reflecting the lung function status.

[0028] Optionally, the identifying of the characteristic peaks corresponding to the same gas marker in different gas chromatography-mass spectrometry spectra based on the peak retention time and the peak apex mass spectrum of each characteristic peak includes:

[0029] Select one of the gas chromatography-mass spectrometry spectra as an anchor sample, and sequentially use each characteristic peak in the anchor sample as an anchor peak;

[0030] Based on the peak retention time and the peak apex mass spectrum of the anchor peak, perform characteristic peak matching on the anchor peak, and identify the characteristic peaks belonging to the same gas marker as the anchor peak from other gas chromatography-mass spectrometry spectra;

[0031] If there are target characteristic peaks in other gas chromatography-mass spectrometry spectra that are different from the gas markers of all the characteristic peaks in the anchor sample, then sequentially use each target characteristic peak as an anchor peak, and return to execute the step of performing characteristic peak matching on the anchor peak based on the peak retention time and the peak apex mass spectrum of the anchor peak, and identifying the characteristic peaks belonging to the same gas marker as the anchor peak from other gas chromatography-mass spectrometry spectra until all the characteristic peaks in all the gas chromatography-mass spectrometry spectra are matched.

[0032] Optionally, the performing a differential analysis on the filtered peak area matrix and calculating the characterization values of the lung function status of different gas markers includes:

[0033] Normalize each combined peak area in the filtered peak area matrix;

[0034] Judge whether the elements in the same column of the normalized peak area matrix meet the normal distribution condition. If so, use an independent T-test to calculate the status test value of the corresponding column; if not, use a rank sum test to calculate the status test value of the corresponding column;

[0035] Use orthogonal partial least squares discriminant analysis to calculate the multi-factor characterization values of each column element in the normalized peak area matrix.

[0036] Optionally, screening multiple classification markers for reflecting the pulmonary function status from the characterization values of the pulmonary function status based on each gas marker, including:

[0037] Regarding the gas marker whose corresponding status verification value is less than the preset verification threshold and whose corresponding multi-factor characterization value exceeds the preset characterization threshold as one of the classification markers.

[0038] An auxiliary analysis device for pulmonary diseases based on exhaled VOCs, comprising:

[0039] An acquisition module, configured to acquire a gas classification model and a chromatogram-mass spectrometry analysis spectrum of the exhaled gas corresponding to a target object;

[0040] A determination module, configured to determine multiple classification markers for reflecting the pulmonary function status;

[0041] An identification module, configured to identify the characteristic peak corresponding to each classification marker from the chromatogram-mass spectrometry analysis spectrum and determine the marker peak area of each classification marker;

[0042] A classification module, configured to use the gas classification model to determine whether the exhaled gas indicates that the target object has a pulmonary disease based on the marker peak areas of each classification marker, and when it is determined that the target object has a pulmonary disease, determine the category of the pulmonary disease corresponding to the exhaled gas.

[0043] An auxiliary analysis device for pulmonary diseases based on exhaled VOCs, comprising a memory and a processor;

[0044] The memory is used to store programs;

[0045] The processor is configured to execute the program to implement each step of the above-mentioned auxiliary analysis method for pulmonary diseases based on exhaled VOCs.

[0046] A readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, each step of the above-mentioned auxiliary analysis method for pulmonary diseases based on exhaled VOCs is implemented.

[0047] As can be seen from the above technical solution, the method for assisting in the analysis of lung diseases based on exhaled VOCs provided by this application can obtain a gas classification model and the chromatogram-mass spectrometry analysis spectrum of the exhaled gas corresponding to the target object; determine a plurality of classification markers for reflecting the lung function status; identify the characteristic peaks corresponding to each classification marker from the chromatogram-mass spectrometry analysis spectrum, and determine the marker peak area of each classification marker; based on this, this application can extract the marker peak areas of a plurality of classification markers with a higher degree of relevance to the lung function status from a large amount of information in the chromatogram-mass spectrometry analysis spectrum, simplify the redundant information into quantifiable and analyzable peak areas, while ensuring the reliability of classification, simplify the complexity of gas classification, and avoid the interference of redundant information on the classification result; subsequently, this application can use the gas classification model to determine whether the exhaled gas indicates that the target object has a lung disease based on the marker peak areas of each classification marker, and when it is determined that the target object has a lung disease, determine the category of the lung disease corresponding to the exhaled gas; based on this, this application can combine the gas classification model and the peak areas of each marker to complete gas classification, determine whether the category corresponding to the exhaled gas is that of healthy people or a certain lung disease category, provide a reliable reference direction for lung diagnosis, and reduce the misdiagnosis rate and delay rate; at the same time, since the marker peak area is proportional to the content of the corresponding classification marker, therefore, during the process of gas classification in this application, it is possible to classify comprehensively based on the components of the exhaled gas and the content of different components, fully considering the differences in the component content of different lung diseases, thereby improving the reliability of the classification result of this application. It can be seen that this application can use the gas classification model, combined with the peak areas of a plurality of classification markers that can characterize the lung function status in the exhaled gas, to classify the exhaled gas, while reducing the difficulty of data analysis and improving the reliability of the classification result, providing a reliable diagnosis direction for lung diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0049] Figure 1 It is a flowchart of a method for assisting in the analysis of lung diseases based on exhaled VOCs disclosed in an embodiment of the present application;

[0050] Figure 2 It is a structural block diagram of a device for assisting in the analysis of lung diseases based on exhaled VOCs disclosed in an embodiment of the present application;

[0051] Figure 3 This is a hardware structure block diagram of an auxiliary analysis device for lung diseases based on exhaled VOCs disclosed in an embodiment of the present application. Specific implementation manners

[0052] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0053] The embodiment of the present application provides an auxiliary analysis method for lung diseases based on exhaled VOCs. This auxiliary analysis method for lung diseases based on exhaled VOCs can be applied to various exhaled gas classification systems or auxiliary diagnosis systems, and can also be applied to various computer terminals or intelligent terminals. The execution subject can be the processor or server of the computer terminal or intelligent terminal.

[0054] Next, in combination with Figure 1 The auxiliary analysis method for lung diseases based on exhaled VOCs of the present application will be introduced in detail, including the following steps:

[0055] Step S1: Obtain a gas classification model and a gas chromatography-mass spectrometry analysis spectrum of the exhaled gas corresponding to the target object.

[0056] Specifically, it can be checked whether there is a gas classification model for providing a reference direction for lung diagnosis; if so, the gas classification model can be directly retrieved; if not, the logistic regression model can be trained to obtain the gas classification model.

[0057] The target object can be an authorized user who provides exhaled gas.

[0058] A gas chromatography-mass spectrometry instrument or a liquid chromatography-mass spectrometry instrument can be used to obtain a gas chromatography-mass spectrometry analysis spectrum of the exhaled gas corresponding to the target object.

[0059] Step S2: Determine multiple classification markers for reflecting the lung function status.

[0060] Specifically, it can be checked whether the names of each classification marker are pre-recorded. If so, the names of each classification marker can be directly retrieved to complete the determination of the classification marker. If not, each gas marker that can be used to reflect the pulmonary function status can be screened from each exhaled volatile organic compound (VOC) as each classification marker. Exhaled VOCs can include methanol, ethanol, 1-propanol, 1-butanol, acetaldehyde, n-valeraldehyde, furfural, 4-Hydroxyhexenal (chemical formula: C6H12O), n-hexanal, benzaldehyde, n-heptanal, n-octanal, n-nonanal, n-decanal, methyl acetate, ethyl acetate, butyl acetate, 2-Methylbutyl Acetate (chemical formula: C7H14O2), 2-methylfuran, 2,5-dimethylfuran, 2-Pentylfuran (chemical formula: C9H14O), n-propane, n-butane, 2-methylbutane, n-pentane, cyclohexane, n-hexane, methylcyclohexane, n-heptane, n-octane, Propylcyclohexane (chemical formula: C9H18), 2,4-dimethylheptane, n-nonane, n-decane, n-undecane, n-dodecane, isoprene, 1-heptene, styrene, α-pinene, limonene, benzene, 1,2-dichlorobenzene, toluene, p-xylene, ethylbenzene, 1,2,3-trimethylbenzene, 1-ethyl-4-methylbenzene, acetone, 2-butanone, cyclohexanone, methyl isobutyl ketone, acetophenone, acetonitrile, 1,2,3,4-tetrahydroisoquinoline, tolazoline, phenol, dimethyl sulfide, dimethyl disulfide, dimethyl trisulfide, and Methyl propyl sulfide (chemical formula: C4H10S), etc.

[0061] Step S3: Identify the characteristic peak corresponding to each classification marker from the chromatogram-mass spectrometry analysis spectrum, and determine the marker peak area of each classification marker.

[0062] Specifically, according to the peak retention time corresponding to each classification marker, the characteristic peak corresponding to each classification marker can be identified from the chromatogram-mass spectrometry analysis spectrum, and the peak area of the characteristic peak corresponding to each classification marker can be used as the marker peak area of the corresponding classification marker.

[0063] Step S4: Use the gas classification model to determine whether the exhaled gas indicates that the target object has a lung disease based on the marker peak areas of each classification marker, and when it is determined that the target object has a lung disease, determine the category of the lung disease corresponding to the exhaled gas.

[0064] Specifically, the name of each classification marker and its marker peak area can be correspondingly input into the gas classification model to obtain the classification result output by the gas classification model.

[0065] The classification result can indicate whether the corresponding exhaled gas is associated with a lung disease. When it is determined that the exhaled gas is associated with a lung disease, the classification result can also indicate the specific type of the lung disease.

[0066] As can be seen from the above technical solution, the method for assisting in the analysis of lung diseases based on exhaled VOC provided by the present application can obtain a gas classification model and a chromatogram-mass spectrometry analysis spectrum of the exhaled gas corresponding to the target object; determine a plurality of classification markers for reflecting the lung function status; identify the characteristic peaks corresponding to each classification marker from the chromatogram-mass spectrometry analysis spectrum, and determine the marker peak area of each classification marker; based on this, the present application can extract the marker peak areas of a plurality of classification markers with a higher degree of relevance to the lung function status from a large amount of information in the chromatogram-mass spectrometry analysis spectrum, simplify the redundant information into quantifiable and analyzable peak areas, while ensuring the reliability of classification, simplify the complexity of gas classification, and avoid the interference of redundant information on the classification result; subsequently, the present application can use the gas classification model to determine whether the exhaled gas indicates that the target object has a lung disease based on the marker peak areas of each classification marker, and when it is determined that the target object has a lung disease, determine the type of lung disease corresponding to the exhaled gas; based on this, the present application can combine the gas classification model and the marker peak areas of each marker to complete gas classification, determine whether the category corresponding to the exhaled gas is that of healthy people or a certain type of lung disease, provide a reliable reference direction for lung diagnosis, and reduce the misdiagnosis rate and delay rate; at the same time, since the marker peak area is directly proportional to the content of the corresponding classification marker, therefore, during the process of gas classification, the present application can classify by comprehensively considering the components of the exhaled gas and the contents of different components, fully taking into account the differences in the component contents of different lung diseases, thereby improving the reliability of the classification result of the present application. It can be seen that the present application can use the gas classification model, combined with the peak areas of a plurality of classification markers that can characterize the lung function status in the exhaled gas, to classify the exhaled gas, while reducing the difficulty of data analysis and improving the reliability of the classification result, providing a reliable diagnosis direction for lung diagnosis.

[0067] In some embodiments of the present application, the process of obtaining the gas classification model in step S1 is described in detail as follows:

[0068] S10. Obtain a plurality of training samples, and divide each training sample into a training set and a test set.

[0069] Specifically, a plurality of training samples including annotation labels indicating the corresponding lung physiological state and the training peak areas of each training marker can be obtained.

[0070] The physiological state of the lungs can include a healthy state, as well as lung disease states such as lung cancer state, benign pulmonary nodule state, chronic obstructive pulmonary disease state, interstitial lung disease state, asthma state, lung function impairment state with impaired retention ratio, bronchiectasis state, pulmonary embolism state, and pulmonary hypertension state.

[0071] The set of biomarker types composed of each training biomarker can be the same as the set of biomarker types composed of each classification biomarker.

[0072] Each training sample can be divided into a training set and a test set according to a ratio.

[0073] S11. Construct a logistic regression model.

[0074] Specifically, a logistic regression model can be constructed according to the following model definition:

[0075]

[0076] Among them, is the model prediction result, is the feature matrix, is the model weight matrix, is the model residual matrix. represents the model definition function. This model can use cross-entropy as the main body of the loss function. The cross-entropy calculation formula is as follows:

[0077]

[0078] Among them, is the set of labeled tags of the training samples, is the set of predicted disease states output by the model, is the labeled tag of the training sample i; is the predicted disease state corresponding to the training sample i.

[0079] In addition, in order to alleviate the phenomenon of model overfitting and improve the stability and generalization of the model, a model regularization term can be added to the loss function. The calculation method of the regularization term is as follows:

[0080]

[0081] Among them, w is one of the elements in the model weight matrix, is the adjustable coefficient of regularization.

[0082] To sum up, the calculation formula of the loss function is defined as follows:

[0083]

[0084] In the formula, λ is the regularization intensity parameter.

[0085] S12. Use the training set and the test set to adjust the parameters of the logistic regression model until the logistic regression model meets the stop condition. The finally obtained logistic regression model is the gas classification model.

[0086] Specifically, the gradient descent method can be combined to use the training set and the test set to adjust the weight parameters and hyperparameters of the logistic regression model until the logistic regression model converges. The finally obtained logistic regression model is the gas classification model.

[0087] The gradient descent method is defined as follows:

[0088]

[0089] Where is the model parameter at the th iteration; is the model parameter at the th iteration; is the loss function calculated from the model parameter at the th iteration, is the gradient descent learning rate; is the loss function with respect to the model parameter gradient.

[0090] It can be seen from the above technical solution that this embodiment provides an optional way to obtain a gas classification model. Through the above method, the logistic regression model can be trained by combining the test set and the training set to obtain the gas classification model.

[0091] Furthermore, experiments show that the true positive rate of the gas classification model trained by the above method in this application exceeds 0.85, and the classification ability is very excellent.

[0092] In some embodiments of this application, the process of obtaining multiple training samples in step S10 is described in detail as follows:

[0093] S100. Obtain the chromatographic-mass spectrometric training spectrograms of the exhaled gas corresponding to different pulmonary physiological states, and obtain the training peak areas of each training marker from each chromatographic-mass spectrometric training spectrogram.

[0094] Specifically, the chromatographic-mass spectrometric analysis spectrograms of the exhaled gas provided by the authorized population with different pulmonary physiological states can be obtained as the corresponding chromatographic-mass spectrometric training spectrograms.

[0095] Multiple training markers for indicating the pulmonary physiological state can be determined.

[0096] Each training marker can be each classification marker.

[0097] The training peak area corresponding to each training marker can be obtained from each chromatographic training spectrogram in sequence.

[0098] S101. Using recursive feature elimination combined with cross-validation method and random forest model, perform feature screening on the training peak areas of different pulmonary physiological states, and calculate the feature scores of different training peak areas.

[0099] Specifically, random forest modeling and evaluation can be performed based on all training peak areas to obtain a baseline performance score;

[0100] Each training peak area is removed in sequence, and the performance score is re-modeled and evaluated.

[0101] The following calculation expression can be used to calculate the mean decrease in impurity score mdi of each training peak area as the feature score corresponding to the training peak area:

[0102]

[0103] In the formula, is the impurity score of the th node of the random forest, is the impurity score of the parent node of the th node, is the total number of training peak areas currently included in the th node, is the total number of training peak areas of a certain class label in the th node, is the total number of training peak areas included in the dataset, is the total number of nodes of the random forest.

[0104] S102. Sort each training peak area according to the feature scores from large to small, retain the top N training peak areas, and return to execute step S101 until each training peak area meets the screening conditions.

[0105] Specifically, multiple training peak areas with lower feature scores can be removed from each training peak area, and the feature scores of each training peak area are recalculated. After removal, return to execute step S101 until the number of each training peak area is lower than the preset lower limit.

[0106] S103. The training peak areas corresponding to the same exhaled gas form a training sample.

[0107] Specifically, all training peak areas corresponding to the same exhaled gas can be combined to form a training sample.

[0108] To further improve the model performance, training samples with the highest feature scores can be used to perform hyperparameter search based on cross-validation on a logistic regression model. The hyperparameters involved in the search include, but are not limited to: model regularization coefficient and strength, number of learning iterations, learning rate, and learning rate change method, etc.

[0109] As can be seen from the above technical solution, this embodiment provides an optional way to obtain multiple training samples. Through the above method, this application can repeatedly execute the process of calculating feature scores, eliminating training peak areas with low scores by combining recursive feature elimination with the cross-validation method and the random forest model, thereby improving the representation ability of the finally generated training samples for the pulmonary physiological state, and thus improving the training efficiency and prediction reliability of the gas classification model.

[0110] In some embodiments of the present application, the process of step S2, determining multiple classification markers for reflecting the pulmonary function status, is described in detail as follows:

[0111] S20. Collect the chromatogram-mass spectrometry (GC-MS) spectra formed by exhaled gases in different pulmonary states.

[0112] Specifically, each pulmonary state may include a healthy state and multiple pulmonary disease states.

[0113] The chromatogram-mass spectrometry analysis spectrum corresponding to each exhaled gas can be obtained as the corresponding GC-MS spectrum.

[0114] Among them, each GC-MS spectrum may be the same as or different from each GC-MS training spectrum.

[0115] S21. Calculate the peak retention time, peak vertex mass spectrum, combined peak area, and peak quantitative signal-to-noise ratio of each feature peak in each GC-MS spectrum.

[0116] Specifically, all the peak retention times of each GC-MS spectrum can be combined to form a peak retention time set, and the peak retention time sets of each GC-MS spectrum can form a peak retention time matrix.

[0117] All the peak vertex mass spectra of each GC-MS spectrum can be combined to form a spectrum set, and the spectrum sets of each GC-MS spectrum can form a spectrum matrix.

[0118] All the combined peak areas of each GC-MS spectrum can be combined to form a peak area set, and the peak area sets of each GC-MS spectrum can form a peak area matrix.

[0119] All the peak quantitative signal-to-noise ratios of each GC-MS spectrum can be combined to form a signal-to-noise ratio set, and the signal-to-noise ratio sets of each GC-MS spectrum can form a peak quantitative matrix.

[0120] Different characteristic peaks in the same exhaled gas correspond to different gas markers.

[0121] S22. Based on the peak retention time and the peak apex mass spectrum of each characteristic peak, identify the characteristic peaks corresponding to the same gas marker in different gas chromatography-mass spectrometry (GC-MS) spectra.

[0122] Specifically, considering that the peak retention times of the same marker are the same and those of different markers are different, it is possible to identify whether different characteristic peaks correspond to the same gas marker by the peak retention time.

[0123] At the same time, considering that each peak apex mass spectrum is a two-dimensional array, the abscissa of the two-dimensional array is the mass-to-charge ratio, representing the ratio of the mass of an ion to its charge number, reflecting the distribution of ions with different masses in the molecule of the corresponding gas marker in the corresponding exhaled gas; the ordinate of the two-dimensional array is the relative signal intensity of the ion current, representing the signal intensity generated by ions with different mass-to-charge ratios on the detector. Therefore, the peak apex mass spectrum can reflect the relative abundance of the corresponding gas marker in the corresponding exhaled gas.

[0124] Based on this, it is possible to identify the characteristic peaks with a similar relationship in different GC-MS spectra based on the peak retention time and the peak apex mass spectrum of each characteristic peak.

[0125] S23. Construct a peak area matrix containing the integrated peak areas corresponding to each characteristic peak, and a signal-to-noise ratio matrix containing the peak quantitative signal-to-noise ratios corresponding to each characteristic peak, where the elements in the same column of the peak area matrix and the signal-to-noise ratio matrix belong to the same gas marker.

[0126] Specifically, according to the similarity relationship between each characteristic peak in different GC-MS spectra, the elements of the integrated peak matrix and the peak quantitative matrix can be adjusted respectively, and the elements corresponding to the same gas marker can be adjusted to the same column elements to obtain the peak area matrix and the signal-to-noise ratio matrix respectively.

[0127] S24. Filter each integrated peak area in the peak area matrix based on the signal-to-noise ratio matrix.

[0128] Specifically, considering the differences such as fluctuations in the detection system and changes in the sampling environment of exhaled gas, the concentrations of not all gas markers exceed the lower limit of analysis. Therefore, to ensure the effectiveness of numerical analysis, gas markers and samples with too low response ratios can be excluded.

[0129] On this basis, since the peak quantitative signal-to-noise ratio can be the ratio of the peak height of the corresponding characteristic peak to the baseline height, it can reflect the degree of noise interference of the corresponding gas biomarker. That is, the numerical value of the peak quantitative signal-to-noise ratio is directly proportional to the signal intensity and reliability of the gas biomarker. Therefore, the signal-to-noise ratio matrix can be used to remove the combined peak areas of gas biomarkers with relatively low reliability in the peak area matrix, improving the stability of differential screening.

[0130] The filtering method is as follows:

[0131]

[0132] In the formula, is the filtering function; is the peak area matrix; is the signal-to-noise ratio matrix; is the minimum signal-to-noise ratio, ; is the lower limit of the response ratio of the gas biomarker, is the lower limit of the response ratio of the GC-MS chromatogram; is the peak area matrix after low signal-to-noise ratio filtering; is the number of remaining GC-MS chromatograms after elimination; is the number of remaining gas biomarkers after elimination.

[0133] Since the lower limit of the response ratio is usually not 1, the peak area matrix after elimination of too low response ratios may still contain gas biomarkers with low responses. To retain the numerical differences between groups and ensure the effectiveness of numerical analysis, the remaining gas biomarkers with low responses can be filled according to the grouping of lung states. Referring to the consistent standards of bioinformatics analysis, half of the non-low-response gas biomarkers within the group can be used for filling.

[0134] S25. Perform a differential analysis on the filtered peak area matrix and calculate the lung function status characterization values of different gas biomarkers.

[0135] Specifically, the combined peak area can be the area enclosed by the corresponding characteristic peak to the baseline, and its size usually has a linear relationship with the concentration of the corresponding gas biomarker in the corresponding exhaled gas. Therefore, the combined peak areas of the same gas biomarker can be integrated to calculate the lung function status characterization values of different gas biomarkers.

[0136] S26. Based on the lung function status characterization values of each gas biomarker, screen multiple classification biomarkers for reflecting the lung function status.

[0137] Specifically, each gas biomarker with a lung function status characterization value exceeding the threshold can be used as multiple classification biomarkers.

[0138] As can be seen from the above technical solutions, this embodiment provides an optional method for determining multiple classification markers for reflecting the pulmonary function status. Through the above method, multiple dimensions can be integrated to gradually analyze the characterization ability of each gas marker for the pulmonary function status, further ensuring the particularity and reliability of each classification marker.

[0139] In some embodiments of the present application, the process of step S22, identifying the characteristic peaks corresponding to the same gas marker in different chromatogram - mass spectrometry spectra based on the peak retention time and the peak - vertex mass spectrometry of each characteristic peak, is described in detail as follows:

[0140] S220: Select one of the chromatogram - mass spectrometry spectra as the anchor sample, and sequentially use each characteristic peak in the anchor sample as the anchor peak.

[0141] Specifically, one of the chromatogram - mass spectrometry spectra can be randomly selected as the anchor sample, and each characteristic peak in the anchor sample is sequentially selected as the anchor peak.

[0142] S221: Based on the peak retention time and the peak - vertex mass spectrometry of the anchor peak, perform characteristic peak matching on the anchor peak to identify the characteristic peaks belonging to the same gas marker as the anchor peak from other chromatogram - mass spectrometry spectra.

[0143] Specifically, a matching function can be combined to calculate the peak retention score and the mass spectrometry score between the anchor peak and each characteristic peak, and the similarity score is calculated by integrating the peak retention score and the mass spectrometry score. The characteristic peaks corresponding to the similarity scores exceeding the similarity threshold are confirmed as the characteristic peaks belonging to the same gas marker as the anchor peak.

[0144] The matching function is as follows:

[0145]

[0146] In the formula, SimScore is the similarity score; RTsim is the peak retention score; MSSim is the mass spectrometry score; is the peak retention time of the anchor peak, is the peak - vertex mass spectrometry spectrum of the anchor peak; is the peak retention time of characteristic peak i, is the peak - vertex mass spectrometry spectrum of characteristic peak i; is the upper limit of the retention time difference, which can be used to control the maximum retention time difference. For example, ; is the upper limit of the mass spectrometry similarity difference, which can be used to control the maximum mass spectrometry difference. For example, ; is the weighting for balancing the contribution degrees of the retention time similarity and the mass spectrometry similarity. To ensure the same contribution to the final similarity, .

[0147] The similarity threshold can be 0.75.

[0148] S222. If there are target characteristic peaks in the other chromatogram-mass spectrometry combined spectrograms that are different from all the gas markers of the characteristic peaks in the anchor sample, then each target characteristic peak is taken as an anchor peak in turn, and step S221 is returned for execution until all the characteristic peaks in all the chromatogram-mass spectrometry combined spectrograms are completely matched.

[0149] Specifically, the characteristic peaks that are not matched with other characteristic peaks can be taken as target characteristic peaks, and each characteristic peak is selected as an anchor peak in turn, and step S221 is returned for execution until all the characteristic peaks of the same gas marker are identified.

[0150] It can be seen from the above technical solution that this embodiment provides an optional method for identifying the characteristic peaks corresponding to the same gas marker in different chromatogram-mass spectrometry combined spectrograms based on the peak retention time and the peak apex mass spectrometry of each characteristic peak. Through the above method, the similarity calculation can be performed on pairwise characteristic peaks by combining the peak retention time and the peak apex mass spectrometry, so as to identify each characteristic peak corresponding to the same gas marker.

[0151] In some embodiments of the present application, step S25, the process of performing differential analysis on the filtered peak area matrix and calculating the lung function status characterization values of different gas markers is described in detail as follows:

[0152] S250. Normalize each combined peak area in the filtered peak area matrix.

[0153] Specifically, the following logarithmic transformation is used to reduce the differences between the combined peak areas:

[0154]

[0155] Among them, is the filtered peak area matrix; is the peak area matrix after logarithmic transformation.

[0156] In order to eliminate the differences, the internal standard method with the acetone peak area as the internal standard and the Z-Score method can be combined for normalization, and the method is as follows:

[0157]

[0158]

[0159] In the formula, is the array of combined peak areas of gas marker i after logarithmic transformation; is the peak area of the acetone peak in is the combined peak integral array corresponding to gas marker i after internal standard method processing; is the peak area matrix after internal standard method processing; is the peak area array corresponding to a gas marker i after normalization; is the peak area array of gas marker i after internal standard method processing; is the peak area matrix corresponding to a gas marker i after normalization; mean(v i ) is the average value; std(v i ) is the standard deviation.

[0160] S251. Determine whether the elements in the same column of the normalized peak area matrix satisfy the normal distribution condition. If so, use the independent T-test to calculate the status check value of the corresponding column; if not, use the rank sum test to calculate the status check value of the corresponding column.

[0161] Specifically, it can be determined whether each combined peak integral in each column element after normalization corresponds to a healthy state. When each combined peak integral corresponding to the healthy state in the same column element satisfies the normal distribution and each combined peak integral corresponding to the non-healthy state satisfies the normal distribution, it is determined that the elements in the corresponding column satisfy the normal distribution condition. The independent T-test can be used to calculate the p-value as the status check value, or the p-value can be converted to the false discovery rate by the Benjamini-Hochberg method and the false discovery rate can be used as the status check value of the corresponding column.

[0162] If each combined peak integral corresponding to the healthy state in the same column element does not satisfy the normal distribution and / or each combined peak integral corresponding to the non-healthy state does not satisfy the normal distribution, it is determined that the elements in the corresponding column do not satisfy the normal distribution condition. The rank sum test can be used to calculate the p-value as the status check value, or the p-value can be converted to the false discovery rate by the Benjamini-Hochberg method and the false discovery rate can be used as the status check value of the corresponding column.

[0163] The Benjamini-Hochberg method is as follows:

[0164]

[0165] Among them, is the Benjamini-Hochberg method conversion formula, is any element in the set of p-values to be converted, is the total number of elements in the p-value set, is the ascending sorting order of the element to be converted in the p-value set, is the converted false discovery rate.

[0166] S252. Use orthogonal partial least squares discriminant analysis to calculate the multi-factor characterization values of each column element in the normalized peak area matrix.

[0167] Specifically, in order to explore the internal relationship between different gas markers, orthogonal partial least squares discriminant analysis can be tried to calculate the multi-factor characterization values of each column element in the normalized peak area matrix.

[0168] Orthogonal partial least squares discriminant analysis includes matrix decomposition of the normalized peak area matrix and the one-hot grouping label matrix, orthogonal extraction of the main components, and establishment of a discriminant model, as follows:

[0169]

[0170] Among them, is the normalized peak area matrix, is the one-hot grouping label matrix, is the score matrix of the normalized peak area matrix, is the score matrix of the one-hot grouping label matrix, is the loading matrix of the normalized peak area matrix, is the loading matrix of the one-hot grouping label matrix, is the residual matrix of the normalized peak area matrix, is the residual matrix of the one-hot grouping label matrix, is the orthogonalized score matrix of the normalized peak area matrix, is the orthogonalized score matrix of the one-hot grouping label matrix, is the weight matrix of the discriminant model, is the residual matrix of the discriminant model, is the multi-factor characterization value of all combined peak products in the normalized peak area matrix, is the number of principal components of OPLS-DA, is the direction of the score matrix of the normalized peak area matrix, is the orthogonalized score matrix of the normalized peak area matrix, is the loading matrix of the one-hot grouping label matrix.

[0171] It can be seen from the above technical solutions that this embodiment provides an optional method for calculating the characterization values of the lung function status of different gas markers. Through the above method, the characterization values of the lung function status calculated in this application include two dimensions: the status verification value and the multi-factor characterization value, which further improves the reliability of the characterization values of the lung function status.

[0172] In some embodiments of the present application, the process of step S26, screening a plurality of classification markers for reflecting the pulmonary function status based on the pulmonary function status characterization values of each gas marker, is described in detail as follows:

[0173] S260: Use a gas marker whose corresponding status verification value is less than a preset verification threshold and whose corresponding multi-factor characterization value exceeds the preset characterization threshold as one of the classification markers.

[0174] Specifically, the verification threshold can be 0.05, and the multi-factor characterization value can be 1.

[0175] A gas marker whose corresponding status verification value is less than 0.05 and whose corresponding multi-factor characterization value is greater than 1 can be used as one of the classification markers.

[0176] It can be seen from the above technical solutions that this embodiment provides an optional method for screening a plurality of classification markers for reflecting the pulmonary function status based on the pulmonary function status characterization values of each gas marker. Through the above method, the accuracy of the classification markers can be further improved.

[0177] After experimental verification, for the auxiliary analysis method of pulmonary diseases based on exhaled VOCs proposed in this application, the area under the working characteristic curve (AUROC) is 0.987, the F1 value is 0.923, the accuracy is 0.918, the sensitivity is 0.914, and the specificity is 0.922.

[0178] Next, in combination with Figure 2 The auxiliary analysis device for pulmonary diseases based on exhaled VOCs provided in this application will be introduced in detail below. The auxiliary analysis device for pulmonary diseases based on exhaled VOCs provided below can be compared with the auxiliary analysis method of pulmonary diseases based on exhaled VOCs provided above.

[0179] See Figure 2 It can be found that the auxiliary analysis device for pulmonary diseases based on exhaled VOCs may include:

[0180] An acquisition module 10, configured to acquire a gas classification model and a gas chromatography-mass spectrometry analysis spectrum of the exhaled gas corresponding to the target object;

[0181] A determination module 20, configured to determine a plurality of classification markers for reflecting the pulmonary function status;

[0182] An identification module 30, configured to identify the characteristic peaks corresponding to each classification marker from the gas chromatography-mass spectrometry analysis spectrum and determine the marker peak area of each classification marker;

[0183] A classification module 40, configured to determine whether the exhaled gas indicates that the target object has a lung disease based on the peak areas of the classification markers in the gas classification model, and when it is determined that the target object has a lung disease, determine the category of the lung disease corresponding to the exhaled gas.

[0184] Further, the acquisition module 10 may include:

[0185] A training sample acquisition unit, configured to acquire a plurality of training samples and divide each training sample into a training set and a test set, where each training sample includes an annotated label indicating the corresponding lung physiological state and the training peak areas of each training marker;

[0186] A logistic regression model construction unit, configured to construct a logistic regression model;

[0187] A logistic regression model training unit, configured to use the training set and the test set to adjust the parameters of the logistic regression model until the logistic regression model meets the stop condition, and the finally obtained logistic regression model is the gas classification model.

[0188] Further, the training sample acquisition unit may include:

[0189] A first training sample acquisition subunit, configured to acquire the chromatogram-mass spectrometry training spectra of the exhaled gas corresponding to different lung physiological states, and acquire the training peak areas of each training marker from each chromatogram-mass spectrometry training spectrum;

[0190] A second training sample acquisition subunit, configured to use recursive feature elimination combined with cross-validation method and random forest model to perform feature screening on the training peak areas of different lung physiological states, and calculate the feature scores of different training peak areas;

[0191] A third training sample acquisition subunit, configured to sort each training peak area according to the feature scores from large to small, retain the top N training peak areas, and call the second training sample acquisition subunit until each training peak area meets the screening condition;

[0192] A fourth training sample acquisition subunit, configured to form a training sample from each training peak area corresponding to the same exhaled gas.

[0193] Further, the determination module 20 may include:

[0194] A chromatogram-mass spectrometry combined spectrum acquisition unit, configured to acquire the chromatogram-mass spectrometry combined spectra formed by the exhaled gas in different lung states;

[0195] A peak retention time calculation unit, configured to calculate the peak retention time, peak vertex mass spectrum, combined peak area, and peak quantitative signal-to-noise ratio of each characteristic peak in each chromatogram-mass spectrometry combined spectrum;

[0196] A characteristic peak recognition unit, configured to recognize characteristic peaks corresponding to the same gas marker in different gas chromatography-mass spectrometry (GC-MS) spectra based on the peak retention time and the peak apex mass spectrum of each characteristic peak;

[0197] A peak area matrix construction unit, configured to construct a peak area matrix including the peak areas corresponding to each characteristic peak and a signal-to-noise ratio matrix including the signal-to-noise ratios of peak quantification corresponding to each characteristic peak, wherein the elements in the same column of the peak area matrix and the signal-to-noise ratio matrix belong to the same gas marker;

[0198] A signal-to-noise ratio matrix utilization unit, configured to filter each peak area in the peak area matrix based on the signal-to-noise ratio matrix;

[0199] A difference analysis unit, configured to perform difference analysis on the filtered peak area matrix and calculate the characterization values of the pulmonary function status of different gas markers;

[0200] A classification marker screening unit, configured to screen a plurality of classification markers for reflecting the pulmonary function status based on the characterization values of the pulmonary function status of each gas marker.

[0201] Further, the characteristic peak recognition unit may include:

[0202] A first characteristic peak recognition subunit, configured to select one of the GC-MS spectra as an anchor sample and sequentially use each characteristic peak in the anchor sample as an anchor peak;

[0203] A second characteristic peak recognition subunit, configured to perform characteristic peak matching on the anchor peak based on the peak retention time and the peak apex mass spectrum of the anchor peak, and recognize characteristic peaks belonging to the same gas marker as the anchor peak from other GC-MS spectra;

[0204] A third characteristic peak recognition subunit, configured to, if there are target characteristic peaks in other GC-MS spectra that are different from the gas markers of all the characteristic peaks in the anchor sample, sequentially use each target characteristic peak as an anchor peak and call the second characteristic peak recognition subunit until all the characteristic peaks in all the GC-MS spectra are matched.

[0205] Further, the difference analysis unit may include:

[0206] A first difference analysis subunit, configured to normalize each peak area in the filtered peak area matrix;

[0207] A second difference analysis subunit, configured to determine whether the elements in the same column of the normalized peak area matrix satisfy the normal distribution condition. If so, an independent T-test is used to calculate the status verification value of the corresponding column; if not, a rank sum test is used to calculate the status verification value of the corresponding column.

[0208] A third difference analysis subunit, configured to perform orthogonal partial least squares discriminant analysis to calculate the multi-factor characterization value of each column element in the normalized peak area matrix.

[0209] Furthermore, the classification biomarker screening unit may include:

[0210] A status verification value comparison subunit, configured to use the gas biomarker whose corresponding status verification value is less than the preset verification threshold and whose corresponding multi-factor characterization value exceeds the preset characterization threshold as one of the classification biomarkers.

[0211] The auxiliary analysis device for lung diseases based on exhaled VOC provided by the embodiments of the present application can be applied to the auxiliary analysis device for lung diseases based on exhaled VOC, such as a PC terminal, a cloud platform, a server, and a server cluster, etc. Optionally, Figure 3 shows the hardware structure block diagram of the auxiliary analysis device for lung diseases based on exhaled VOC. Refer to Figure 3 , the hardware structure of the auxiliary analysis device for lung diseases based on exhaled VOC may include: at least one processor 1, at least one communication interface 2, at least one memory 3, and at least one communication bus 4;

[0212] In the embodiments of the present application, the number of the processor 1, the communication interface 2, the memory 3, and the communication bus 4 is at least one, and the processor 1, the communication interface 2, and the memory 3 complete mutual communication through the communication bus 4;

[0213] The processor 1 may be a central processing unit CPU, or a specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present invention, etc.;

[0214] The memory 3 may include a high-speed RAM memory, and may also include a non-volatile memory, such as at least one disk memory;

[0215] Wherein, the memory stores a program, and the processor can call the program stored in the memory. The program is used for:

[0216] Obtain a gas classification model and a chromatogram-mass spectrometry analysis spectrum of the exhaled gas corresponding to the target object;

[0217] Determine a plurality of classification markers for reflecting the condition of lung function;

[0218] Identify the characteristic peaks corresponding to each classification marker from the chromatogram-mass spectrometry analysis spectrum, and determine the marker peak area of each classification marker;

[0219] Use the gas classification model to determine whether the exhaled gas indicates that the target object has a lung disease based on the marker peak areas of the respective classification markers, and when it is determined that the target object has a lung disease, determine the category of lung disease corresponding to the exhaled gas.

[0220] Optionally, the refinement function and expansion function of the program can be referred to the above description.

[0221] The embodiment of the present application further provides a readable storage medium, which can store a program suitable for execution by a processor, and the program is used for:

[0222] Obtain a gas classification model and a chromatogram-mass spectrometry analysis spectrum of the exhaled gas corresponding to the target object;

[0223] Determine a plurality of classification markers for reflecting the condition of lung function;

[0224] Identify the characteristic peaks corresponding to each classification marker from the chromatogram-mass spectrometry analysis spectrum, and determine the marker peak area of each classification marker;

[0225] Use the gas classification model to determine whether the exhaled gas indicates that the target object has a lung disease based on the marker peak areas of the respective classification markers, and when it is determined that the target object has a lung disease, determine the category of lung disease corresponding to the exhaled gas.

[0226] Optionally, the refinement function and expansion function of the program can be referred to the above description.

[0227] Finally, it should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover a non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of additional identical elements in the process, method, article or device including the element.

[0228] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. For the same or similar parts among the various embodiments, reference can be made to each other.

[0229] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. The various embodiments of the present application can be combined with each other. Therefore, the present application will not be limited to these embodiments shown herein, but rather will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An auxiliary analysis method for lung diseases based on exhaled VOCs, characterized in that, Including: Obtaining a gas classification model and a gas chromatography-mass spectrometry analysis spectrum of the exhaled gas corresponding to a target object; Determining a plurality of classification markers for reflecting the pulmonary function status; Identifying the characteristic peaks corresponding to each classification marker from the gas chromatography-mass spectrometry analysis spectrum and determining the marker peak area of each classification marker; Using the gas classification model to determine whether the exhaled gas indicates that the target object has a lung disease based on the marker peak areas of the respective classification markers, and when it is determined that the target object has a lung disease, determining the category of the lung disease corresponding to the exhaled gas.

2. The auxiliary analysis method for lung diseases based on exhaled VOC according to claim 1, wherein The obtaining of the gas classification model includes: Obtaining a plurality of training samples and dividing each training sample into a training set and a test set, where each training sample includes a labeled tag indicating the corresponding pulmonary physiological state and the training peak areas of each training marker; Constructing a logistic regression model; Using the training set and the test set to adjust the parameters of the logistic regression model until the logistic regression model meets the stop condition, and the finally obtained logistic regression model is the gas classification model.

3. The auxiliary analysis method for lung diseases based on exhaled VOC according to claim 2, wherein The obtaining of the plurality of training samples includes: Obtaining the gas chromatography-mass spectrometry training spectra of the exhaled gas corresponding to different pulmonary physiological states and obtaining the training peak areas of each training marker from each gas chromatography-mass spectrometry training spectrum; Using recursive feature elimination combined with cross-validation and a random forest model to perform feature screening on the training peak areas of different pulmonary physiological states and calculating the feature scores of different training peak areas; Sorting each training peak area according to the feature scores from large to small, retaining the top N training peak areas, and returning to execute the steps of using recursive feature elimination combined with cross-validation and a random forest model to perform feature screening on the training peak areas of different pulmonary physiological states and calculating the feature scores of different training peak areas until each training peak area meets the screening conditions; Each training peak area corresponding to the same exhaled gas forms a training sample.

4. The auxiliary analysis method for lung diseases based on exhaled VOC according to claim 1, wherein The determining of the plurality of classification markers for reflecting the pulmonary function status includes: Collecting gas chromatography-mass spectrometry spectra formed by the exhaled gas in different pulmonary states; Calculating the peak retention time, the peak vertex mass spectrum, the combined peak area, and the peak quantitative signal-to-noise ratio of each feature peak in each gas chromatography-mass spectrometry spectrum; Based on the peak retention time and the peak vertex mass spectrum of each feature peak, identifying the feature peaks corresponding to the same gas marker in different gas chromatography-mass spectrometry spectra; Constructing a peak area matrix containing the combined peak areas corresponding to each feature peak and a signal-to-noise ratio matrix containing the peak quantitative signal-to-noise ratios corresponding to each feature peak, where the elements in the same column of the peak area matrix and the signal-to-noise ratio matrix belong to the same gas marker; Filtering each combined peak area in the peak area matrix based on the signal-to-noise ratio matrix; Performing a difference analysis on the filtered peak area matrix and calculating the pulmonary function status characterization values of different gas markers; Based on the pulmonary function status characterization values of each gas marker, screening a plurality of classification markers for reflecting the pulmonary function status.

5. The auxiliary analysis method for lung diseases based on exhaled VOC according to claim 4, wherein Identifying characteristic peaks corresponding to the same gas biomarker in different gas chromatography-mass spectrometry (GC-MS) spectra based on the peak retention time and peak vertex mass spectrum of each characteristic peak, including: Selecting one of the GC-MS spectra as the anchor sample, and sequentially taking each characteristic peak in the anchor sample as the anchor peak; Based on the peak retention time and peak vertex mass spectrum of the anchor peak, performing characteristic peak matching on the anchor peak, and identifying characteristic peaks belonging to the same gas biomarker as the anchor peak from other GC-MS spectra; If there are target characteristic peaks in other GC-MS spectra that are different from the gas biomarkers of all the characteristic peaks in the anchor sample, then sequentially taking each target characteristic peak as the anchor peak, and returning to execute the step of performing characteristic peak matching on the anchor peak based on the peak retention time and peak vertex mass spectrum of the anchor peak, and identifying characteristic peaks belonging to the same gas biomarker as the anchor peak from other GC-MS spectra, until all the characteristic peaks of all the GC-MS spectra are completed in matching.

6. The method for auxiliary analysis of lung diseases based on exhaled VOC according to claim 4, characterized in that Performing differential analysis on the filtered peak area matrix, and calculating the lung function status characterization values of different gas biomarkers, including: Normalizing each combined peak area in the filtered peak area matrix; Judging whether the elements in the same column of the normalized peak area matrix satisfy the normal distribution condition. If so, using an independent T-test to calculate the status test value of the corresponding column; if not, using a rank sum test to calculate the status test value of the corresponding column; Performing orthogonal partial least squares discriminant analysis to calculate the multi-factor characterization values of each column element in the normalized peak area matrix.

7. The auxiliary analysis method for lung diseases based on exhaled VOC according to claim 6, characterized in that, Based on the lung function status characterization values of each gas biomarker, screening multiple classification biomarkers for reflecting the lung function status, including: Taking the gas biomarker whose corresponding status test value is less than the preset test threshold and whose corresponding multi-factor characterization value exceeds the preset characterization threshold as one of the classification biomarkers.

8. An auxiliary analysis device for lung diseases based on exhaled VOCs, characterized in that, Including: An acquisition module, configured to acquire a gas classification model and a gas chromatography-mass spectrometry analysis spectrum of the exhaled gas corresponding to the target object; A determination module, configured to determine multiple classification biomarkers for reflecting the lung function status; An identification module, configured to identify the characteristic peaks corresponding to each classification biomarker from the gas chromatography-mass spectrometry analysis spectrum, and determine the biomarker peak area of each classification biomarker; A classification module, configured to use the gas classification model to determine whether the exhaled gas indicates that the target object has a lung disease based on the biomarker peak areas of each classification biomarker, and when it is determined that the target object has a lung disease, determine the category of the lung disease corresponding to the exhaled gas.

9. An auxiliary analysis device for lung diseases based on exhaled VOCs, characterized in that, Including a memory and a processor; The memory is configured to store a program; The processor is configured to execute the program to implement each step of the method for assisting in analyzing lung diseases based on exhaled VOCs as described in any one of claims 1-7.

10. A readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements each step of the method for assisting in analyzing lung diseases based on exhaled VOCs as described in any one of claims 1-7.

Images