Method and system for nitric oxide data analysis for bronchial asthma assessment

Abstract

The present application relates to a method and system for analyzing one nitrogen oxide data for bronchial asthma evaluation. The method standardizes the collection of exhaled nitrogen monoxide and alveolar nitrogen monoxide, omics data and clinical information of patients, and constructs multi-dimensional characteristics after data cleaning and fusion, including inflammation level index, inflammation area entropy, inflammation synergy index, glucocorticoid response factor, infection-smoking synergy factor and genetic-symptom distribution index. Based on the explainable boosting machine EBM model, the initial risk assessment is output, and finally the asthma risk probability is generated through a weighted integrated formula, realizing three-level stratification: low / medium / high risk. The system dynamically correlates treatment decisions, such as increasing drug dosage and monitoring frequency when risk is upgraded, and optimizing drug regimen when risk is downgraded. This scheme breaks through the limitations of traditional single marker static analysis, integrates anatomic-immune-pharmacological multi-mechanism, significantly improves the evaluation accuracy and clinical applicability, reduces the acute attack rate and guides precise classification treatment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical diagnosis, in particular to a nitric oxide data analysis method and system for bronchial asthma evaluation. BACKGROUND

[0002] There are significant limitations in the precise evaluation of bronchial asthma: existing methods analyze exhaled nitric oxide in isolation, ignoring the value of alveolar nitric oxide in locating small airway inflammation, resulting in a high rate of missed diagnosis of small airway lesions; traditional models rely on a single clinical symptom or omics data, and cannot quantify the interactive effects of smoking history, recent infection and genetic background; mainstream tools use static threshold to adjust treatment, which is difficult to respond to dynamic changes in inflammation, resulting in ineffective treatment for some patients. Emerging technologies have introduced gene expression to improve accuracy, but gene detection is costly and invasive, and the black box nature of prediction models hinders the tracing of clinical decisions, and treatment plans lack dynamic correlation with environmental trigger events. It is urgent to develop an evaluation system that integrates spatial analysis of gas markers, dynamic environmental response and interpretable decisions to achieve a fundamental upgrade in asthma management paradigm. SUMMARY

[0003] In view of the deficiencies of the prior art, the present application provides a nitric oxide data analysis method and system for bronchial asthma evaluation.

[0004] To achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0005] The nitric oxide data analysis method for bronchial asthma evaluation comprises the following steps:

[0006] S1, collecting exhaled nitric oxide, alveolar nitric oxide, omics data and clinical information of the patient;

[0007] S2, data cleaning, normalization and missing value processing, and fusion of processed exhaled nitric oxide and alveolar nitric oxide, omics data and clinical information;

[0008] S3, constructing the following features:

[0009] Inflammation level index, inflammation region entropy, inflammation synergy index, glucocorticoid response factor, infection-smoking synergy factor, genetic-symptom distribution index;

[0010] S4, dividing the training set and the test set;

[0011] S5, using an explainable boosting machine EBM as a classifier, and using a decision tree as a base learner to iteratively train the model;

[0012] S6, applying the trained EBM model to output the initial evaluation value of bronchial asthma;

[0013] S7, constructing a final bronchial asthma evaluation formula, substituting the initial bronchial asthma evaluation value and the characteristics into the final bronchial asthma evaluation formula to obtain a bronchial asthma evaluation grade.

[0014] In an implementation method of the present application, the S1 includes the following specific contents:

[0015] S1, data collection: collecting exhaled nitric oxide and alveolar nitric oxide data of the patient, using standardized detection methods and equipment to ensure the accuracy and reliability of the data, and collecting other omics data of the patient, including blood routine test data, proteomics data. These data can be obtained through corresponding detection techniques and equipment, and the clinical information of the patient, such as age, gender, smoking history, allergy history, family asthma history, symptom manifestation, lung function index, inhaled corticosteroid use, whether there is a recent respiratory tract infection, etc., is collected as auxiliary analysis data;

[0016] In an implementation method of the present application, the data preprocessing in the S2 step includes the following specific contents:

[0017] S21, data preprocessing: preprocessing the collected multi-omics data, including data cleaning, normalization, missing value processing, etc., removing noise and outliers in the data to ensure the quality and consistency of the data;

[0018] S22, data fusion, fusing the preprocessed exhaled nitric oxide and alveolar nitric oxide data, other omics data and clinical information, constructing a multi-omics data set, which can use different data fusion methods, such as feature splicing based method, model fusion based method, etc., to integrate data from different sources together, further analyzing and mining the fused data, extracting key features, and reducing the dimension and complexity of the data;

[0019] In an implementation method of the present application, the S3 includes the following specific contents:

[0020] S31, for precise quantification of the spatial distribution characteristics of bronchial asthma, combining the exhaled nitric oxide and alveolar nitric oxide data in the data set to construct the following hierarchical features: inflammation level index, inflammation level index The derivation steps of the inflammation level index are as follows: standardizing the exhaled nitric oxide by the reference exhaled nitric oxide concentration of healthy people to obtain the exhaled nitric oxide weight adjustment term , and then correcting the alveolar nitric oxide by eosinophil count to calculate the alveolar nitric oxide-eosinophil logarithmic enhancement term The exhaled nitric oxide weight adjustment term is multiplied by the alveolar nitric oxide-eosinophil logarithmic enhancement term to obtain an inflammation level index evaluation formula, and the specific expression is , wherein is the exhaled nitric oxide concentration, is the exhaled nitric oxide concentration of a healthy population, is the alveolar nitric oxide concentration, is the alveolar nitric oxide baseline value, is the exhaled nitric oxide weight adjustment factor, is the eosinophil count, is the correction coefficient of the alveolar nitric oxide to the eosinophil;

[0021] S32, in order to understand the distribution of the inflammation area, an inflammation area entropy is constructed, and the inflammation area entropy The specific derivation steps are as follows: the exhaled nitric oxide and the alveolar nitric oxide are modeled to obtain , the single-point entropy of the exhaled nitric oxide and the alveolar nitric oxide is calculated respectively , and the two single-point entropies are integrated into the area entropy When , the inflammation is highly concentrated in a single area, and when , the inflammation is evenly distributed.

[0022] S33, in order to capture the nonlinear synergistic effect of multi-source data, the following interaction features are designed: inflammation synergy index, the specific derivation steps are as follows: the age attenuation term after sigmoid transformation is multiplied by the Th2 basic synergy term, and the allergic saturated interaction term , , wherein is the age-dependent attenuation factor, is the weight coefficient of the allergic history, is a binary variable, is the serum total immunoglobulin concentration;

[0023] glucocorticoid response factor, the specific derivation process is as follows: the exhaled nitric oxide response attenuation term is calculated, multiplied by the drug dose and the inflammation change rate to obtain the glucocorticoid response factor, and the expression is , wherein is the daily dose of inhaled glucocorticoid, is the change rate of , which is calculated by a sliding window: ;

[0024] Infection-smoking synergistic factor, the specific derivation process is: smoke destroys pulmonary vascular homeostasis through three mechanisms: cumulative toxic effect: polycyclic aromatic hydrocarbons in smoke tar continuously damage bronchial microvascular endothelial cells, making the vascular basement membrane thickened and the fragility increased. Hemodynamic stress: carbon monoxide (COHb>5%) significantly reduces the oxygen carrying capacity of red blood cells by forming carboxyhemoglobin (arterial oxygen partial pressure decreases by 28%), triggering compensatory pulmonary artery contraction. The formula for assessing the risk of smoking-induced pulmonary vascular damage is: wherein, is the cumulative exposure of tobacco, and the pathological transformation of pulmonary hemorrhage to asthma exacerbation: physical obstruction + iron deposition, is the carboxyhemoglobin saturation, is the mean pulmonary arterial pressure, is the pulmonary arterial hypertension marker, and the formula for assessing the risk of post-hemorrhagic inflammation explosion is: wherein is the amount of bleeding, is the serum iron concentration, and the dynamic integration of the "smoking-hemorrhage-asthma" cascade chain, which is not linear accumulation, but there are a two-phase amplification node and a critical mutation threshold: .

[0025] Genetic-symptom distribution index, the specific derivation process is: lung function defect term multiplied by symptom accumulation determined by genetic risk switch, and finally logarithmically transformed, , wherein, four core symptoms ( cough, wheeze, chest tightness, nighttime choking), 0-3 according to severity, weighting coefficient, predicted value based on age and height.

[0026] In an implementation method of the present application, the S4 includes the following specific contents:

[0027] The data is divided into a training set and a test set, and divided according to a certain proportion. The training set is used for model training, and the test set is used for evaluating the performance of the model.

[0028] In an implementation method of the present application, the S5 includes the following specific contents:

[0029] S51, using EBM as a classifier model, selecting a decision tree as a basic learner, EBM gradually trains each feature function in an additive manner, so that its contribution can be separated and explained, and the form of EBM is , wherein The EBM model outputs the bronchial asthma assessment results based on the input data of the subjects being assessed. Let i be the feature function corresponding to the i-th feature. Let i be the pairwise interactive feature function of the i-th and j-th features, and let j be the feature function. , Implemented using decision trees;

[0030] S52. Train the base learner and initialize the constant model. Where L is the loss function, For the actual data of the i-th patient, Let be a constant representing the initial predicted value, and the expression for the loss function is: ,in, These are the model coefficients. To set regularization parameters, the base learners are trained iteratively, and then combined to form a strong learner. In each iteration, a new weak learner is trained on the residuals of the current model to gradually reduce the loss of the entire model. The negative gradient is calculated. ,in, The model obtained from the previous iteration has a negative gradient. Indicate the direction of error using data Training decision trees Calculate step size Step length Control the model update magnitude and update the model based on the calculated step size; .

[0031] S53. Perform ANOVAF test on candidate feature pairs to screen them. The significant interaction features are used to train a decision tree, and the interaction function is initialized as a constant zero matrix. Based on the current model (including single feature terms), calculate the negative gradient and iterate. t During the round of iterations, the negative gradient is ,in The optimal step size is determined by line search for the current model output. Update the interaction function: .

[0032] In one implementation of the present invention, step S6 includes the following specific contents:

[0033] The trained assessment model is used to calculate the initial assessment value of bronchial asthma from all data of the user to be assessed. .

[0034] In one implementation of the present invention, step S7 includes the following specific contents:

[0035] constructing a total bronchial asthma assessment formula: wherein Dp is the final assessment result, represents a sigmoid function, represents a weight coefficient.

[0036] The application also provides a nitric oxide data analysis system for bronchial asthma assessment, which comprises a data acquisition module, a preprocessing module, a fusion dimension reduction module, a feature construction module, a model training module, and an assessment output module, which work cooperatively to realize the assessment of bronchial asthma.

[0037] The nitric oxide data analysis system for bronchial asthma assessment is realized based on the nitric oxide data analysis method for bronchial asthma assessment, and specifically comprises:

[0038] The data acquisition module is configured to acquire the exhaled nitric oxide concentration, alveolar nitric oxide concentration, blood routine test data, and proteomics data of a patient through a standardized detection device, and synchronously acquire a clinical information dataset, including age, smoking history, allergy history, lung function index, and glucocorticoid use dose.

[0039] The preprocessing module is configured to perform noise filtering and outlier removal on the collected exhaled nitric oxide, alveolar nitric oxide, and omics data, fill in missing values by using a multiple imputation method, and realize dimension unification of multi-source data by Z-score standardization.

[0040] The fusion dimension reduction module is configured to splice the preprocessed exhaled nitric oxide, alveolar nitric oxide, omics data, and clinical information at the feature level.

[0041] The feature construction module is configured to perform the following calculations: based on the exhaled nitric oxide reference value and the eosinophil count of healthy people, an inflammation level index is generated, the inflammation regional entropy is calculated to quantify the spatial aggregation degree of inflammation by modeling the probability distribution of exhaled nitric oxide concentration, a drug response factor is generated by combining the daily dose of glucocorticoids and the change rate of exhaled nitric oxide, an infection-smoking synergistic influence factor is calculated by fusing the number of smoking years, the respiratory tract infection state, and the alveolar nitric oxide constraint term, and a genetic-symptom distribution index is constructed by weighting the sum of lung function defects and symptoms.

[0042] The model training module is configured to: divide the data into a training set and a test set in a ratio of 7:3; use an explainable boosting machine (EBM) as a classifier and a decision tree as a base learner to iteratively train the model by using a gradient boosting algorithm; optimize the hyperparameters by using a grid search, and perform cross-validation by using the Matthews correlation coefficient (MCC) as an evaluation index.

[0043] The evaluation output module is configured to input the fusion features of the patient to be evaluated into the trained EBM model, and output a bronchial asthma risk score.

[0044] Compared with the prior art, the beneficial effects of the present application are as follows:

[0045] The present application significantly improves the accuracy and clinical applicability of bronchial asthma evaluation through multi-dimensional inflammation analysis technology and dynamic interaction effect modeling. Specifically, the inflammation level index, which is the first of its kind, integrates the standardized exhaled nitric oxide, the eosinophil correction, and the logarithmic enhancement mechanism, and combines the regional entropy quantification technology based on the probability distribution of alveolar nitric oxide, breaking through the limitations of traditional single threshold method, and breaking through to construct four types of medical mechanism-driven interaction features: the inflammation synergy index, which integrates age-dependent immune attenuation, Th2 pathway basic activity, and the saturation effect of allergic history and immunoglobulin, quantifies the synergistic amplification mechanism of Th2-type inflammation; the glucocorticoid response factor combines the drug dose and the dynamic change rate of inflammation, revealing the nonlinear attenuation law of drug anti-inflammatory efficacy; the infection-smoking synergistic effect factor is the product of smoking years and small airway function constraint term, which adds the infection-triggered and hormone-compensation effect, and describes the acute risk multiplication mechanism of environmental exposure; the genetic-symptom distribution index takes lung function defect as the base, multiplies the weighted sum of four core symptoms and is regulated by the genetic risk switch, realizing the spatial coupling modeling of genetic susceptibility and clinical symptoms. At the model architecture level, the interpretable boosting machine EBM is used as the core, and its additive form explicitly separates the contribution of each feature: the base feature function realizes single feature effect visualization by decision tree, and the pair-wise interaction function captures the synergy of cross-modal features, and through gradient boosting iteration, it provides feature-level decision basis while ensuring accuracy. The final output of the bronchial asthma evaluation value has three values verified by clinical trials: 1) stratified diagnosis value: scores 0-30 / 31-70 / 71-100 correspond to mild / moderate / severe asthma; 2) treatment guidance value; 3) risk warning value. The system integrates multi-source data acquisition, dynamic feature calculation, and interpretable AI decision-making, completes the whole process from detection to grading report in a short time, is compatible with mainstream medical equipment, and reduces the annual per capita medical expenditure. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 The present application is a whole process schematic diagram for the nitric oxide data analysis method for bronchial asthma evaluation;

[0047] Figure 2 The present application is a whole process schematic diagram for the nitric oxide data analysis method for bronchial asthma evaluation;

[0048] Figure 3 The present application is a whole process schematic diagram for the nitric oxide data analysis method for bronchial asthma evaluation; DETAILED DESCRIPTION

[0049] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments of the present application.

[0050] Embodiment 1

[0051] The present application provides an embodiment: as shown in Figure 1 and Figure 2 ,

[0052] The method for analyzing one nitrogen oxide data for bronchial asthma evaluation comprises the following specific steps:

[0053] S1, collecting exhaled nitrogen oxide, alveolar nitrogen oxide, omics data and clinical information of a patient;

[0054] It needs to be specifically explained in this embodiment that the data acquisition comprises the following specific contents:

[0055] The exhaled nitrogen oxide and alveolar nitrogen oxide data of the patient are collected, and a standardized detection method and equipment are used to ensure the accuracy and reliability of the data. At the same time, other omics data of the patient are collected, including blood routine detection data and proteomics data. These data can be obtained through corresponding detection technology and equipment. The clinical information of the patient, such as age, gender, smoking history, allergy history, family asthma history, symptom performance, lung function index, inhaled steroid use, and whether there is a recent respiratory tract infection, is collected as auxiliary analysis data;

[0056] S2, data cleaning, normalization and missing value processing, and fusion of the processed exhaled nitrogen oxide and alveolar nitrogen oxide, omics data and clinical information;

[0057] It needs to be specifically explained in this embodiment that the data preprocessing comprises the following specific contents:

[0058] Data preprocessing: the collected multi-omics data is preprocessed, including data cleaning, normalization, missing value processing and the like, to remove noise and abnormal values in the data and ensure the quality and consistency of the data;

[0059] Data fusion: the exhaled nitrogen oxide and alveolar nitrogen oxide data, other omics data and clinical information after preprocessing are fused to construct a multi-omics data set. Different data fusion methods can be used, such as feature splicing and model fusion, to integrate data from different sources together. The fused data is further analyzed and mined to extract key features and reduce the dimension and complexity of the data;

[0060] S3, constructing the following features:

[0061] Inflammation hierarchy index, inflammation regional entropy, inflammation synergy index, glucocorticoid response factor, infection-smoking synergy factor, genetic-symptom distribution index;

[0062] It is particularly pointed out in the embodiment that the construction feature includes the following specific contents:

[0063] In order to accurately quantify the spatial distribution characteristics of bronchial asthma, the exhaled nitric oxide and alveolar nitric oxide are combined by using the nitric oxide data in the data set to construct the following hierarchical features: inflammation hierarchy index, inflammation hierarchy index The derivation steps are as follows: the exhaled nitric oxide is standardized by the reference exhaled nitric oxide concentration of healthy people, and the standardized exhaled nitric oxide is subjected to power law transformation to obtain the exhaled nitric oxide and weight adjustment term , and then the alveolar nitric oxide is corrected by the eosinophil count, and the alveolar nitric oxide-eosinophil logarithmic enhancement term is calculated after the corrected alveolar nitric oxide is standardized , the weight adjustment term is multiplied by the alveolar nitric oxide-eosinophil logarithmic enhancement term to obtain the inflammation hierarchy index evaluation formula, and the specific expression is , wherein, is the exhaled nitric oxide concentration, is the exhaled nitric oxide reference value of healthy people, is the exhaled nitric oxide, is the alveolar nitric oxide baseline value, is the exhaled nitric oxide weight adjustment factor, preferably 0.35, is the eosinophil count, is the correction coefficient of alveolar nitric oxide to eosinophils, preferably 0.15, the index innovatively combines the airway anatomical structure and the non-invasive inflammation activity, and realizes the spatial accurate stratification of airway inflammation through the synergistic weighting of exhaled nitric oxide and alveolar nitric oxide;

[0064] In order to understand the distribution of inflammation area, the inflammation area entropy is constructed, and the inflammation area entropy The specific derivation steps are as follows: the exhaled nitric oxide and alveolar nitric oxide are subjected to probability modeling to obtain , the single-point entropy of exhaled nitric oxide and alveolar nitric oxide is calculated respectively , the two single-point entropies are combined into regional entropy , when , the inflammation is highly concentrated in a single area, and when When the inflammation is evenly distributed, the innovation value of this index lies in the information theory quantification of spatial inflammation heterogeneity, solving the problem that traditional technology only focuses on gene expression while ignoring spatial distribution. The logarithmic conversion further compresses the complex three-dimensional distribution into a [0, 1] operable index, enabling doctors to intuitively identify the inflammation diffusion pattern.

[0065] To capture the nonlinear synergistic effect of multi-source data, the following interaction features are designed: Inflammation synergy index, the specific derivation steps are: multiply the age attenuation term after sigmoid transformation with the Th2 basic synergy term, and add the allergic- saturated interaction term , , where is the age-dependent attenuation factor, is the weight coefficient of allergic history, preferably 0.28, is a binary variable, is the serum total immunoglobulin concentration. The advantage of this formula lies in the analysis of the multi-level synergy of Th2 pathway. The core biomarker of eosinophilic inflammation is integrated through the FeNO x EOS product term, and the age-dependent sigmoid attenuation factor accurately captures the difference in Th2 activity between adolescent asthma (β<0) and adult asthma (β>0).

[0066] Glucocorticoid response factor, the specific derivation process is: calculate response attenuation term , multiply the drug dose and the change rate of inflammation to get the glucocorticoid response factor, its expression is , where, is the daily dose of inhaled glucocorticoid, is the change rate of , which is calculated by a sliding window: The core advantage of this formula is to dynamically quantify the anti-inflammatory treatment efficiency. By introducing the time derivative of the inflammation level index, the inflammation regression rate under hormone therapy is captured in real time, solving the defect that traditional static models cannot reflect the timeliness of treatment. The design of the nonlinear inhibition term is particularly key: when FeNO>30ppb, the output value increases from 0.5 to 0.9, accurately simulating the "high inflammation state hormone resistance" phenomenon observed in clinical observation. The product structure of ICS dose realizes double regulation, encouraging sufficient medication to suppress inflammation, and warning of the dose efficiency decay under high inflammation.

[0067] Infection-smoking synergistic factor, the specific derivation process is: smoke destroys pulmonary vascular homeostasis through three mechanisms: tobacco smoke first directly damages bronchial microvascular endothelial cells through cumulative toxic effects, among which polycyclic aromatic hydrocarbons contained in tobacco tar can continuously induce mitochondrial DNA breakage of endothelial cells, causing the vascular basement membrane to abnormally thicken by 3 to 5 times and significantly increase its mechanical fragility; At the same time, hemodynamic stress as the second mechanism, when the carbon monoxide hemoglobin saturation COHb exceeds 5%, the arterial oxygen partial pressure can decrease by 28%, triggering a compensatory pulmonary artery contraction response; And the synergistic effect of infection and smoking constitutes the third damage mechanism, which is typically manifested as the replication efficiency of influenza virus is improved in the environment of high expression of TLR4 induced by nicotine. The formula for assessing the risk of smoking to pulmonary vascular damage is: wherein, is the cumulative exposure of tobacco, and the pathological transformation of pulmonary hemorrhage to asthma exacerbation: physical obstruction + iron deposition starts, is the carbon monoxide hemoglobin saturation, is the mean pulmonary arterial pressure, is the pulmonary arterial hypertension marker, and the formula for assessing the risk of inflammation outbreak after hemorrhage is: wherein is the amount of bleeding, is the serum iron concentration, and the dynamic integration of the "smoking-hemorrhage-asthma" cascade chain, which is not linear accumulation, but there are two-phase amplification nodes and critical mutation thresholds: The model accurately locates the intervention window of pulmonary vascular rupture and asthma acute attack, and provides dynamic navigation for clinical stratified intervention.

[0068] Genetic-symptom distribution index, the specific derivation process is: lung function defect term multiplied by symptom accumulation Determine whether to count through genetic risk switch, and finally perform logarithmic transformation, , wherein, is the four core symptoms cough, wheeze, chest tightness, night suffocation), 0-3 according to severity, is the weight coefficient, preferably , For age-height-based prediction values, the innovation of this formula is to convert family genetic risk into a synergistic amplifier of symptoms-lung function. The design of family history as a genetic switch is in line with the medical consensus that "genetic background is a necessary but not sufficient condition for the development of asthma", ensuring that the synergistic effect of symptoms and lung function is activated only when genetic risk exists. The symptom score item highlights the clinical value through differentiated weights, and the introduction of the logarithmic function cleverly solves the problem of product explosion (such as symptom full score + lung function defect 30% → 1.8), compresses the characteristic value to the clinically interpretable range of [0, 3], while retaining the nonlinear amplification effect of genetic risk on symptom load.

[0069] S4, dividing the training set and the test set;

[0070] It needs to be specifically pointed out in this embodiment that the division of the training set and the test set includes the following specific content: dividing the data into a training set and a test set, and dividing according to a certain proportion, and in this embodiment, a proportion of 70%-30% is adopted. The training set is used for training the model, and the test set is used for evaluating the performance of the model.

[0071] S5, using an interpretable boosting machine EBM as a classifier, and using a decision tree as a base learner to iteratively train the model;

[0072] It needs to be specifically pointed out in this embodiment that the training of the interpretable model includes the following specific content: using EBM as a classifier model, and selecting a decision tree as a base learner, EBM gradually trains each feature function in an additive manner, so that its contribution can be separated and explained, and the form of EBM is , wherein is the bronchial asthma evaluation result output by the EBM model for the input data of the to-be-evaluated person, is a feature function corresponding to the i-th feature, is a pair of interactive feature function of the i-th and j-th features, the feature function, , is realized by a decision tree;

[0073] Training the base learner, initializing the constant model , wherein L is a loss function, is the actual data of the i-th patient, is a constant representing an initial prediction value, and the expression of the loss function is , wherein, is a model coefficient, is a regularization parameter, the base learner is iteratively trained, and the base learners are combined to form a strong learner, in each iteration, the new weak learner will be trained for the residual of the current model to gradually reduce the loss of the entire model, and the negative gradient is calculated, , wherein, negative gradient of the model obtained in the last iteration error direction, using data training decision tree , calculating step size step size controlling the amplitude of model update, updating the model according to the calculated step size; .

[0074] performing ANOVA F test on candidate feature pairs, screening significant interaction features, training a decision tree from the screened features, and initializing the interaction function as a constant zero matrix , based on the current model (including single feature items) to calculate the negative gradient for iteration, in the first t iteration, the negative gradient is , wherein is the output of the current model, and the optimal step size is determined by line search , updating the interaction function: .

[0075] S6, applying the trained EBM model to output the initial bronchial asthma evaluation value;

[0076] In this embodiment, it should be noted that the S7 includes the following specific contents:

[0077] using the trained evaluation model to calculate all data of the user to be evaluated to obtain the initial bronchial asthma evaluation value , constructing a final bronchial asthma evaluation formula, substituting the initial bronchial asthma evaluation value and each feature into the final bronchial asthma evaluation formula to obtain the bronchial asthma evaluation grade.

[0078] In this embodiment, it should be noted that the S7 includes the following specific contents:

[0079] constructing a total bronchial asthma evaluation formula: , wherein Dp is the final evaluation result, represents a sigmoid function, represents a weight coefficient, and is preferably .

[0080] Embodiment 2

[0081] As shown in Figure 3 , the present embodiment provides a nitric oxide data analysis system for bronchial asthma evaluation, which specifically includes:

[0082] The data acquisition module is configured to obtain the exhaled nitric oxide concentration, alveolar nitric oxide concentration, blood routine test data and proteomics data of the patient through a standardized detection device, and synchronously acquire a clinical information dataset, including age, smoking history, allergy history, lung function index and glucocorticoid use dose; the preprocessing module is configured to perform noise filtering and abnormal value elimination on the collected exhaled nitric oxide and alveolar nitric oxide and omics data, fill in the missing values by using a multiple imputation method, and realize dimension unification of the multi-source data by Z-score standardization; the fusion dimension reduction module is configured to splice the preprocessed exhaled nitric oxide and alveolar nitric oxide, omics data and clinical information at a feature level; the feature construction module is configured to perform the following calculations: based on the exhaled nitric oxide reference value and the eosinophil count of healthy people, an inflammation level index is generated, the inflammation regional entropy is calculated to quantify the spatial aggregation degree of inflammation by modeling the probability distribution of alveolar nitric oxide concentration; a drug response factor is generated by combining the daily dose of glucocorticoids and the change rate of exhaled nitric oxide; an infection-smoking synergistic influence factor is calculated by combining the number of years of smoking, the respiratory infection state and the alveolar nitric oxide constraint term; a genetic-symptom distribution index is constructed by the lung function defect term and the symptom weighted product; the model training module is configured: the data is divided into a training set and a test set in a ratio of 7:3; the explainable boosting machine EBM is used as a classifier, and the decision tree is used as a base learner, and the model is iteratively trained by the gradient boosting algorithm; the grid search is used to optimize the hyperparameters, and the cross-validation is performed by taking the Matthews correlation coefficient MCC as the evaluation index; the evaluation output module is configured to input the fusion features of the patient to be evaluated into the trained EBM model, and output a bronchial asthma risk score.

[0083] Embodiment 3

[0084] The embodiment provides an electronic device, comprising a processor and a memory, wherein the memory has a computer program that can be called by the processor.

[0085] The processor executes the above-mentioned nitric oxide data analysis method for bronchial asthma evaluation by calling the computer program stored in the memory.

[0086] The electronic device can have great differences due to different configurations or performances, and can include one or more processors (Central Processing Units, CPUs) and one or more memories, wherein the memory stores at least one computer program, which is loaded and executed by the processor to realize the infectious disease risk prediction method based on detection sharing network provided by the above-mentioned method embodiment. The electronic device can also include other components for realizing the functions of the device, for example, the electronic device can also have a wired or wireless network interface and an input and output interface, etc., so as to input and output data. This embodiment will not be described here.

[0087] Embodiment 4

[0088] In this embodiment, a computer readable storage medium is provided, and the computer readable storage medium stores a computer program capable of performing erasing and writing operations. When the computer program is run on a computer device, the computer device is caused to perform the above-mentioned method for analyzing one-oxidized nitrogen data for evaluating bronchial asthma.

[0089] The embodiments of the present application are described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the difference from other embodiments.

[0090] The system and medium provided by the embodiments of the present application are one-to-one corresponding to the method. Therefore, the system and medium also have similar beneficial technical effects to the method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the system and medium will not be described here.

[0091] It should be clear to those skilled in the art that the embodiments of the present application can be presented in the form of a method, a system or a computer program product. Therefore, the present application can be in the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can be implemented in the form of a computer program product on one or more computer usable storage media (such as disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0092] The present application is described with reference to flowcharts and / or block diagrams of the methods, devices (systems) and computer program products of the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device, so as to generate a machine. By executing these instructions by the processor of the computer or other programmable data processing device, an apparatus for implementing the functions specified by one or more flows in the flowcharts or one or more blocks in the block diagrams can be generated.

[0093] These computer program instructions can also be stored in a computer readable storage medium, which can direct the computer or other programmable data processing device to work in a specific manner. Thus, the instructions stored in the computer readable storage medium can generate a manufactured product containing instruction apparatus, which can implement the functions specified by one or more flows in the flowcharts or one or more blocks in the block diagrams.

[0094] In one typical arrangement, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory. The memory can include non-persistent memory such as volatile random access memory (RAM); also can include non-volatile memory, such as read-only memory (ROM), or flash memory (flash RAM). The memory is an example of computer-readable media.

[0095] Computer-readable media includes permanent and non-permanent, movable and non-movable media that can store information through any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), and other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape magnetic disk storage or other magnetic storage devices, and any other non-transmission medium that can be used to store information accessible to a computing device. According to the definition herein, computer-readable media does not include transitory media, such as modulated data signals and carriers.

Claims

1. A nitric oxide data analysis system for assessing bronchial asthma, characterized in that, It includes the following modules: a data acquisition module, configured to acquire patients' exhaled nitric oxide concentration, alveolar nitric oxide concentration, routine blood test data and proteomics data through standardized testing equipment, and simultaneously collect clinical information datasets, including age, smoking history, allergy history, lung function indicators and glucocorticoid dosage; The preprocessing module is configured to perform noise filtering and outlier removal on the collected exhaled nitric oxide, alveolar nitric oxide and omics data, fill missing values ​​with multiple imputation, and achieve dimensionality unification of multi-source data through Z-score standardization; the fusion and dimensionality reduction module is configured to perform feature-level splicing on the preprocessed exhaled nitric oxide, alveolar nitric oxide, omics data and clinical information. The feature construction module is configured to perform the following calculations: Based on the reference values ​​of exhaled nitric oxide and eosinophil count in healthy individuals, an inflammation hierarchy index is generated; by modeling the probability distribution of exhaled nitric oxide concentration, the entropy of the inflammatory region is calculated to quantify the spatial aggregation of inflammation; a glucocorticoid response factor is generated by combining the daily dose of glucocorticoids with the rate of change of exhaled nitric oxide; an infection-smoking synergistic influence factor is calculated by fusing years of smoking, respiratory infection status, and alveolar nitric oxide constraints; and a genetic-symptom distribution index is constructed by weighting the pulmonary function deficit term and symptoms. The model training module is configured to: divide the data into training and testing sets in a 7:3 ratio; use the Interpretable Enhancement Machine (EBM) as the classifier and the decision tree as the base learner, iteratively training the model using a gradient boosting algorithm; optimize hyperparameters using grid search, and perform cross-validation using the Matthews correlation coefficient (MCC) as the evaluation metric; and the evaluation output module is configured to input the fused features of the patient to be evaluated into the trained EBM model and output a bronchial asthma risk score. The feature construction module also includes constructing an inflammation synergy index; The inflammation level index construction includes the following steps: combining exhaled nitric oxide and alveolar nitric oxide in the data set to construct the following hierarchical features: inflammation level index, inflammation level index The derivation steps are as follows: standardizing exhaled nitric oxide by the reference exhaled nitric oxide concentration of healthy people, performing power-law transformation on the standardized exhaled nitric oxide to obtain an exhaled nitric oxide weight adjustment term , and then correcting alveolar nitric oxide by eosinophil count, standardizing the corrected alveolar nitric oxide to calculate an alveolar nitric oxide-eosinophil logarithmic enhancement term , multiplying the exhaled nitric oxide weight adjustment term and the alveolar nitric oxide-eosinophil logarithmic enhancement term to obtain an inflammation level index evaluation formula, and the specific expression is , wherein is the exhaled nitric oxide concentration, is the exhaled nitric oxide reference value of healthy people, is the alveolar nitric oxide, is the alveolar nitric oxide baseline value, is the exhaled nitric oxide weight adjustment factor, is the eosinophil count, is the correction coefficient of alveolar nitric oxide by eosinophils; probability modeling is performed on exhaled nitric oxide and alveolar nitric oxide to obtain , and the single-point entropy of exhaled nitric oxide and alveolar nitric oxide is calculated respectively , the two single-point entropies are integrated into a regional entropy , wherein k is any of the exhaled nitric oxide concentration and the alveolar nitric oxide concentration; the inflammation synergy index construction includes the following steps: multiplying the sigmoid-transformed age attenuation term and the Th2-based synergy term, adding the allergic saturation interaction term , , wherein is the age-dependent attenuation factor, is the allergic history weight coefficient, is a binary variable, is the serum total immunoglobulin concentration.

2. The nitric oxide data analysis system for assessing bronchial asthma according to claim 1, characterized in that, The glucocorticoid response factor is specifically derived as follows: calculating the exhaled nitric oxide response decay term , multiplying the drug dose and the rate of change of inflammation to obtain the glucocorticoid response factor, which is expressed as , wherein is the daily dose of inhaled glucocorticoid, is the rate of change of , wherein is the inflammation level index, which is calculated by a sliding window: .

3. The nitric oxide data analysis system for assessing bronchial asthma according to claim 2, characterized in that, The specific derivation process of the infection-smoking synergistic influencing factor is as follows: Smoke disrupts pulmonary vascular homeostasis through a three-pronged mechanism, and the formula for assessing the risk of pulmonary vascular damage from smoking is: ,in, The cumulative tobacco exposure contributes to the pathological transformation of pulmonary hemorrhage into asthma. Carboxyhemoglobin saturation. Mean pulmonary artery pressure, As a marker of pulmonary hypertension, the formula for assessing the risk of post-hemorrhagic inflammatory flare-up is: ,in This refers to the amount of bleeding. This represents serum iron concentration.

4. The nitric oxide data analysis system for assessing bronchial asthma according to claim 3, characterized in that, The specific derivation process of the genetic-symptom distribution index is as follows: Lung function deficit item Multiplication of symptoms Whether to include a sample is determined by a genetic risk switch, and finally a logarithmic transformation is performed. ,in, The core symptom is assigned a score of 0-3 based on its severity. These are the weighting coefficients. These are predicted values ​​based on age and height.

5. The nitric oxide data analysis system for assessing bronchial asthma according to claim 4, characterized in that, The final formula for assessing bronchial asthma is: ,in This represents the sigmoid function. These are the initial assessment values ​​for bronchial asthma. These are the weighting coefficients. This is the result of a risk assessment for post-bleeding inflammatory outbreaks.

Images