A method for measuring respiratory function parameters based on quantitative analysis of respiratory wave spectrum

Abstract

The present invention discloses a method for measuring respiratory function parameters based on quantitative analysis of respiratory wave spectrum, comprising the following steps: constructing a closed respiratory sensing system based on a digital mask to collect respiratory waves, and performing spectrum analysis to obtain relevant characteristic parameters; based on the closed respiratory sensing system, performing respiratory wave spectrum analysis to obtain respiratory wave spectrum markers related to personal lung health and disease; constructing an Ai prediction model based on machine learning or deep learning; during measurement, collecting natural respiratory waves of different states of the tested individual, and testing them using the trained Ai prediction model to achieve lung function parameter prediction; providing an Ai-assisted diagnosis prediction result report and transmitting it to a mobile device terminal. The present invention adopts the above-mentioned method for measuring respiratory function parameters based on quantitative analysis of respiratory wave spectrum, realizes non-invasive, convenient and efficient auxiliary diagnosis of lung function, and creates a simple, direct, fast and universal method for measuring respiratory function and auxiliary diagnosis of COPD.

Description

Technical Field

[0001] The present invention relates to the technical field of quantitative detection of respiratory function, and in particular to a method for measuring respiratory function parameters based on quantitative analysis of respiratory wave spectrum. Background Art

[0002] Lung diseases, including chronic respiratory diseases (CRDs) and lung cancer, pose a major health challenge worldwide. However, existing lung disease detection mainly relies on conventional pulmonary function tests (PFTs), but traditional pulmonary function testers still have many shortcomings: low sensitivity, inability to detect weak nasal breathing activities, inability to detect continuously, and the accuracy of test results is affected by the tester's cooperation and operation accuracy. These limitations not only increase logistics and operating costs, but also limit the widespread use of PFTs in home and community settings.

[0003] For the detection of autonomous respiratory waves, various respiratory-related sensors that have emerged in recent years mainly focus on capturing respiratory signals by measuring changes in temperature, humidity, and airflow during breathing. By monitoring changes in respiratory wave characteristics, abnormal respiratory processes such as fever, asthma symptoms, apnea, and other abnormal breathing patterns are qualitatively detected. However, in-depth quantitative research on respiratory waves as "waves" and their interactions with the respiratory system has not been carried out, and quantitative measurement of lung function test-related indicators based entirely on respiratory waves has not yet been achieved.

[0004] Respiratory wave spectra and spectrum decoding are essential for the effective detection of lung diseases. However, this requires that the measurement of spontaneous respiratory waves and waveforms is quantitative and free of external interference. Quantitative respiratory waveforms can reflect detailed pathological information and can play an important role in disease diagnosis through in-depth analysis of the waveforms. Researchers are using artificial intelligence models to classify respiratory patterns or groups of patients with lung diseases. But these models usually directly apply machine learning algorithms to process raw respiratory waveforms. As a black box method, this method lacks interpretability, making it difficult to elucidate the association between waveforms and diseases.

[0005] In addition, the current method of predicting lung function mainly relies on conventional respiratory function instruments for testing. Conventional respiratory function tests require a high degree of cooperation between patients and doctors, otherwise accurate results cannot be obtained. In addition, conventional respiratory function testing technology has the following contraindications: 1) patients in severe asthma attacks; 2) patients with severe chronic obstructive pulmonary disease; 3) patients with uncontrolled hypertension; 4) patients who have received chest, abdominal and ophthalmology in the past 3 months; 5) patients with heart disease (angina pectoris, myocardial infarction) in the past 3 months; 6) patients hospitalized for heart disease in the past month; 7) patients with severe hemoptysis in the past month; 8) patients with stroke in the past month; 9) patients with aortic aneurysm and surgery; 10) patients with severe hyperthyroidism; 11) patients with epileptic seizures who need drug treatment; patients with infarction, malignant arrhythmia, etc.

[0006] Therefore, it is urgent to develop a technology and system that can quantitatively measure, is convenient, and does not require professional guidance, and can be measured with just normal breathing. Applying it to respiratory health and auxiliary diagnosis of COPD will be of great significance to human health. Summary of the invention

[0007] The purpose of the present invention is to provide a method for measuring respiratory function parameters based on quantitative analysis of respiratory wave spectrum. The digital mask system can quantitatively collect respiratory waves, and combine spectrum analysis with Ai algorithm to achieve non-invasive, convenient and efficient auxiliary diagnosis of lung function and diseases; it not only makes up for the shortcomings of traditional lung function detection methods, but also expands its application scenarios, so that it can be widely used in various environments such as homes, communities and hospitals.

[0008] To achieve the above object, the present invention provides a method for quantitatively measuring respiratory waves and analyzing respiratory function parameters based on spectrum, comprising the following steps:

[0009] Step S1: Based on the flexible respiratory sensor, a closed respiratory sensing digital mask + cloud platform system is constructed, and personal parameter information including gender, age, weight, height, and whether or not to smoke is input; and the respiratory waves of the subjects are collected in three movement modes of sitting, standing, and walking;

[0010] Step S2, performing respiratory wave spectrum analysis on the respiratory data acquired in step S11 and step S12, obtaining relevant spectrum characteristic parameters, and obtaining spectrum markers related to personal lung health and disease;

[0011] Step S3, combining the above-mentioned relevant characteristic parameters with the lung function parameters measured in the hospital and the corresponding measurement results to form a training set, and training the Ai prediction model based on artificial intelligence machine learning or deep learning; wherein, the machine learning or deep learning model applied to the Ai prediction model training specifically includes: decision tree, random forest, support vector machine, k nearest neighbor, naive Bayes, neural network, convolutional neural network, recurrent neural network, long short-term memory network, generative adversarial network, reinforcement learning;

[0012] Step S4, during the Ai prediction model training process, the prediction of each lung function parameter, forced expiratory volume in 1 second FEV1, forced vital capacity FVC, FEV1 / FVC ratio and predicted value percentage FEV1%pred, is achieved by training an independent Ai prediction model;

[0013] Step S5, in actual application, the personal parameter information is combined with the respiratory waves of the three motion states and the obtained respiratory wave spectrum parameters of the subject, and input into the trained Ai prediction model to obtain relevant lung function parameters, including the forced expiratory volume in 1 second FEV1, forced vital capacity FVC, FEV1 / FVC ratio and predicted value percentage FEV1%pred;

[0014] Step S6: Give a prediction result report of Ai-assisted diagnosis and transmit it to the mobile device terminal.

[0015] Preferably, in step S1, the digital mask includes a PC plastic shell, an electronic module, a breathing sensor, a wireless module, a temperature sensor, medical silicone and a PC plastic body.

[0016] Preferably, based on a closed breathing sensing digital mask + cloud platform system, the fluctuations of respiratory gas generated by breathing are converted into electrical signal waves through flexible sensors to perform quantitative and continuous measurement of autonomous breathing waves.

[0017] Preferably, when collecting respiratory waves, there are three forms of exercise, including sitting for 4 minutes, standing for 4 minutes and walking for 6 minutes; when doing the above exercises, avoid talking, coughing and other abnormal breathing movements.

[0018] Preferably, the spontaneous breathing waveform includes a respiratory volume waveform, a vital capacity waveform, and an airflow rate waveform; the spontaneous breathing wave is measured by a flexible sensor, specifically including measuring the waveform, amplitude, frequency, temperature, and humidity of the breathing that changes over time, and capturing the breathing signal through airflow changes during the breathing process.

[0019] Preferably, in step S2, the respiratory data acquired in step S11 and step S12 are subjected to respiratory wave spectrum analysis to obtain relevant spectrum characteristic parameters, and spectrum markers related to personal lung health and disease are obtained. The specific process is as follows:

[0020] Step S21, taking the breathing data of the tester when sitting still, standing, and walking fast as input, processing and extracting the features of the breathing data pairs, detecting the waveform information of the peak, trough, and maximum flow rate point of the breathing data, and calculating and obtaining the relevant feature parameters of the breathing data of the three actions;

[0021] Among them, the relevant characteristic parameters include frequency, amplitude, inspiration-expiration ratio, i.e. the ratio of inspiration time to expiration time, Δt / t, i.e. the ratio of the time from the beginning of expiration to the maximum expiration flow rate to the total time of this breath;

[0022] Step S22: Based on the closed respiratory sensing digital mask + cloud platform system, synchronous extrusion wavelet analysis and HHT conversion are used to perform respiratory wave spectrum analysis to obtain spectrum markers related to personal lung health and disease.

[0023] Preferably, in step S3, the above-mentioned relevant characteristic parameters are combined with the lung function parameters provided by the hospital and the corresponding measurement results to form a training set, and an Ai prediction model is constructed based on machine learning or deep learning of artificial intelligence; wherein the relevant characteristic parameters and clinical data are used as input features of the Ai prediction model;

[0024] During the model training process, machine learning or deep learning is used for training, and the frequency, amplitude, inhalation-exhalation ratio, Δt / t calculated from the breathing data of the test subjects when sitting, standing, and walking fast, as well as the test subjects' age, height, and smoking history are used as feature inputs; RandomForestRegressor, XGBRegressor, and GradientBoostingRegr essor are used as base learners of the integrated model; and grid search is used to find the optimal parameters of each model.

[0025] Preferably, the training set and test set data are allocated in a ratio of 5:5 to 8:2, including respiratory waveforms and extracted related characteristic parameters, and clinical data of various populations, specifically lung function and physical information measured in the hospital; lung function parameters include forced expiratory volume in 1 second FEV1, forced vital capacity FVC, FEV1 / FVC ratio and predicted value percentage FEV1%pred.

[0026] Therefore, the present invention adopts the above-mentioned method for measuring respiratory function parameters based on quantitative analysis of respiratory wave spectrum. Based on the digital mask system, the respiratory wave can be quantitatively collected, and the spectrum analysis and Ai algorithm are combined to achieve non-invasive, convenient and efficient auxiliary diagnosis of lung function and diseases. This innovative technical means combined with Ai not only makes up for the shortcomings of traditional lung function testing methods, but also expands its application scenarios, enabling it to be widely used in various environments such as homes, communities and hospitals. Future research will continue to explore the spectral characteristics of respiratory waves in depth, and as clinical data accumulates, it is planned to verify its applicability in different populations in larger-scale clinical trials.

[0027] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 is an exploded view of the digital mask of the present invention;

[0029] Figure 2 It is a closed breathing sensing digital mask + cloud platform system and breathing waveform in an embodiment of the present invention;

[0030] Figure 3 are characteristic parameters extracted from the respiratory waveform of the present invention and their meanings;

[0031] Figure 4 It is an overview of the present invention's system for quantitatively monitoring respiratory waveforms based on Ai;

[0032] Figure 5 It is a comparison of the closed respiratory sensing digital mask + cloud platform system of the embodiment of the present invention and the conventional "small lung" detection equipment in measuring respiratory function parameters; wherein, (a) is a comparison of FEV1; (b) is a comparison of FVC;

[0033] Figure 6 It is a comparison of the maximum value deviation between the respiratory waveform AI prediction and the conventional measurement deviation in the embodiment of the present invention;

[0034] Figure 7 This is a comparison of the closed respiratory sensing digital mask + cloud platform system of the embodiment of the present invention and the hospital's "big lung" detection equipment in measuring respiratory function parameters; wherein, the dark blue dots are female volunteers, and the orange dots are male volunteers; (a) is a comparison of the measured values ​​of FEV1; (b) is a comparison of the measured values ​​of FVC; (c) is a comparison of the measured values ​​of FEV1%pred; (d) is a comparison of the measured values ​​of FEV1 / FVC;

[0035] Figure 81. It is a comparison of the measured values ​​of lung function parameters between the closed respiratory sensing digital mask + cloud platform system and the hospital "big lung" detection equipment in the embodiment of the present invention; wherein, the dark blue dots are non-COPD volunteers, and the orange dots are COPD patients; (a) is a comparison of the measured values ​​of FEV1%pred; (b) is a comparison of the measured values ​​of FEV1 / FVC. DETAILED DESCRIPTION

[0036] The technical solution of the present invention is further described below through the accompanying drawings and embodiments.

[0037] The present invention provides a method for measuring respiratory function parameters based on respiratory wave spectrum quantitative analysis, comprising the following steps:

[0038] Step S1: Based on the flexible respiratory sensor, a closed respiratory sensing digital mask + cloud platform system is constructed, and personal parameter information, including gender, age, weight, height, and whether or not to smoke, is input; the respiratory waves of the subjects are collected in three movement modes: sitting, standing, and walking.

[0039] like Figure 1 As shown in the figure, the digital mask includes a PC plastic shell, an electronic module, a respiratory sensor, a wireless module, a temperature sensor, medical silicone and a PC plastic body. The closed respiratory sensing system based on the digital mask converts the respiratory gas fluctuations generated by breathing into electrical signal waves through a flexible sensor to perform quantitative and continuous measurement of autonomous respiratory waves. Figure 2 As shown. The flexible sensor has the advantages of being small and sensitive, easy to wear, fast response, and capable of continuous accurate / quantitative measurement. The closed respiratory sensing system is designed to prevent the respiratory wave from being affected by external airflow, which would interfere with the respiratory waveform and affect the quantitative analysis of the respiratory wave.

[0040] When collecting respiratory waves, there are three types of exercise, including sitting for 4 minutes, standing for 4 minutes and walking for 6 minutes. When doing the above exercises, avoid abnormal breathing movements such as talking and coughing.

[0041] Spontaneous breathing waveforms include respiratory volume waveforms, vital capacity waveforms, airflow rate waveforms, etc. These waveforms are crucial for evaluating the patient's respiratory function. Spontaneous breathing waves are measured through flexible sensors, including measuring the waveform, amplitude, frequency, temperature, humidity, etc. of breathing that changes over time, as well as capturing respiratory signals through airflow changes during breathing.

[0042] Step S2: Perform respiratory wave spectrum analysis on the respiratory data acquired in step S11 and step S12 to obtain relevant spectrum characteristic parameters and derive spectrum markers related to personal lung health and disease.

[0043] Step S21, taking the breathing data of the tester when sitting, standing, and walking fast as input, processing and feature extraction of the breathing data pairs are performed, and the frequency (Hz), amplitude, inhalation-exhalation ratio (ratio of inhalation time to exhalation time), Δt / t (ratio of time from the beginning of exhalation to the maximum exhalation flow rate to the total breathing time) of the three action breathing data are calculated by detecting the waveform information such as the peak, trough, and maximum flow rate point of the breathing data, such as Figure 3 As shown, the information obtained by the above calculation is used as relevant characteristic parameters for subsequent use.

[0044] Step S22: Based on the closed respiratory sensing digital mask + cloud platform system, synchronous extrusion wavelet analysis and HHT conversion are used to perform spectrum analysis of the respiratory wave to obtain spectrum markers related to the respiratory wave and personal lung health and disease.

[0045] Step S3, combining the above-mentioned relevant characteristic parameters with the lung function parameters provided by the hospital and the corresponding measurement results to form a training set, and constructing an Ai prediction model based on artificial intelligence machine learning or deep learning, such as an integrated regression algorithm; wherein the relevant characteristic parameters and clinical data are used as input features of the Ai prediction model; lung function parameters include FEV1, FVC, FEV1 / FVC and FEV1 / Pred.

[0046] The training set and test set data are allocated in a ratio of 5:5 to 8:2, including respiratory waveforms and extracted related feature parameters, and clinical data of various populations, specifically lung function and physical information measured in the hospital.

[0047] FEV1 (forced expiratory volume in 1 second): The volume of gas that the subject can exhale in the first second of forced exhalation.

[0048] FEV1%pred (predicted value percentage): refers to the percentage of FEV1 relative to the predicted reference value of normal people of the same age, gender, height and race.

[0049] FVC (Forced Vital Capacity): The maximum volume of air that a person can exhale when they exhale completely and forcefully.

[0050] FEV1 / FVC ratio: It indicates the percentage of vital capacity that is expressed in the first second.

[0051] During the model training process, the frequency, amplitude, inhalation-exhalation ratio, Δt / t calculated from the breathing data of the tester's sitting, standing, and brisk walking movements, as well as the tester's age, height, and smoking history are used as feature inputs. RandomForestRegressor, XGBRe gressor, and GradientBoostingRegressor are used as base learners for the integrated model. Grid search is used to find the best parameters for each model. Among them, the machine learning or deep learning models used for Ai prediction model training include but are not limited to: decision tree, random forest, support vector machine, k nearest neighbor, naive Bayes, neural network, convolutional neural network, recurrent neural network, long short-term memory network, generative adversarial network, and reinforcement learning.

[0052] Step S4: During the Ai prediction model training process, the Ai prediction model is tested with a test set, and different natural breathing waves of the tested individuals are collected and input into the Ai prediction model for testing. The stacking method is used, and the multivariate linear regression model is used as a meta-learner to combine the outputs of the basic learners and learn the best way to integrate these results; wherein, the method is applied to an independent training model to predict each of the four lung function parameters: FEV1, FVC, FEV1 / FVC and FEV1 / Pred.

[0053] Step S5: In actual application, personal parameters are combined with the respiratory waves of the three motion states and the obtained respiratory wave spectrum parameters of the subject, and input into the trained Ai prediction model to obtain relevant lung function parameters, and predict the lung function parameters FEV1, FVC, FEV 1 / FVC and FEV1 / Pred. Among them, personal parameters include gender, age, weight, smoking history, etc.

[0054] Step S6: Give an Ai-assisted diagnosis result report and transmit it to the mobile device terminal.

[0055] like Figure 4 As shown, the present invention is a method for measuring respiratory function parameters based on quantitative analysis of respiratory wave spectrum, which has the functions of quantitative detection of respiratory waves, spectrum analysis and Ai auxiliary diagnosis, and can accurately quantify the lung health status through normal natural breathing, and can obtain the lung health status of the subject without "blowing", but only with normal and natural breathing; collect the breathing data of the subject in different motion states, extract the spectrum features and combine with Ai analysis to realize lung function detection and auxiliary diagnosis.

[0056] This method is simple, non-invasive, accurate, and does not require professional operation. It performs well in COPD diagnosis by identifying and monitoring symptoms such as cough through Ai. It is worth noting that the spectrum analysis of the respiratory wave reveals the pathological information of the respiratory system. Through synchronous wavelet analysis, the spectrum characteristics of the respiratory wave are decoded to achieve auxiliary diagnosis of diseases such as COPD and lung cancer, providing a new auxiliary means for clinical diagnosis and risk assessment.

[0057] Example

[0058] 1. Introduction to data collection and statistical methods.

[0059] This example compares the lung function test results of the closed breathing sensing digital mask + cloud platform system proposed in the present invention with those of conventional "small lung" detection equipment and "large lung" detection equipment used in hospitals, and studies the differences in test results of different types of equipment, as well as the specificity of different groups and COPD / asthma disease monitoring, as well as auxiliary diagnosis.

[0060] 1.1 Data collection.

[0061] This embodiment selected 385 subjects from a university hospital in a certain region between 2022 and 2024 for examination. In order to ensure that the tested population is representative, the subjects were preliminarily screened. Among the testers, males and females each accounted for 50%. The age range of the subjects ranged from 20 to 90 years old, of which 77% were over 40 years old, which is consistent with the age distribution of the lung function test population in most hospitals. The subjects included healthy people and patients with chronic obstructive pulmonary disease (COPD), asthma and other lung diseases. Different types of lung diseases have different degrees of impact on lung function, especially COPD has a more significant impact on lung function. Therefore, the proportion of COPD patients in the total tested population is about 16%, which is similar to the national COPD prevalence of 13.7% in people over 40 years old.

[0062] 1.2 Method introduction.

[0063] The subjects used the closed respiratory sensing digital mask + cloud platform system (Ai spectrum prediction) proposed in the present invention to perform control measurements with conventional "small lung" detection equipment (Red Elephant measurement) and hospital "big lung" detection equipment (PowerCube-Body measurement), and all measurements were performed under the guidance of professionals. The hospital "big lung" detection equipment is regarded as the gold standard for lung function measurement. As mentioned above, the measurement accuracy of the subject is closely related to the doctor's guidance and the subject's cooperation. Therefore, all measurements, especially those of "small lungs" and "big lungs", are performed under the guidance of professional medical staff.

[0064] 1.3 Statistical methods

[0065] SPSS 17.0 statistical software was used for data analysis in error statistics and correlation analysis. The data were expressed as [cases (%)] and subjected to t-test; the count data were expressed as [cases (%)] and subjected to χ2 test. The criterion for statistical significance was P < 0.001. The calculation of all parameter prediction results was performed using Python v3.11.8 (Python software foundation) and Scikit-learn v1.1.3 (machine learning part).

[0066] 2. Measurement results.

[0067] The subjects used a closed respiratory sensing digital mask + cloud platform system, conventional "small lung" testing equipment and hospital "big lung" testing equipment to conduct synchronous measurement and comparison under the guidance of professionals.

[0068] 2.1 Comparison of the lung function parameter measurements of the closed breathing sensing digital mask + cloud platform system proposed in the present invention and conventional "small lung" detection equipment.

[0069] First, the measurement results of the closed breathing sensing digital mask + cloud platform system and the conventional "small lung" detection equipment were compared. In view of the limited measurement range of conventional "small lung" detection equipment, the measurement data of the two lung function parameters FEV1 and FVC were analyzed in detail.

[0070] Regarding the measurement accuracy, it should be noted that the results of individual tests may have certain errors. According to the latest standards, the error of FVC and FEV1 in three measurements should be less than 150 ml, while the error of the system proposed in this invention is controlled within 70 ml in five measurements, showing a high measurement accuracy. Figure 5 As shown in Table 1, the two devices showed high consistency in FEV1 and FVC measurements, with relative errors of 8.14% and 7.12% for FEV1 and FVC, respectively. In addition, the maximum value deviation predicted by Ai spectrum was within 10%, which was significantly lower than the conventional measurement value. Figure 6 shown.

[0071] Table 1 Error statistics and correlation analysis of lung function parameters measured by closed respiratory sensing digital mask + cloud platform system and conventional "small lung" detection equipment

[0072] Error Analysis Project FEV1(L) FVC(L) The average relative error percentage 8.14％ 7.12％ Measurement error 0.085±0.264 0.085±0.320 Pearson correlation coefficient 0.944 0.935 ICC3 0.944 0.933

[0073] Note: The above are the statistical results of measurement errors for 122 people.

[0074] 2.2 Comparison of lung function parameter measurements between different genders.

[0075] This group compared the results of the closed respiratory sensing digital mask + cloud platform system and the hospital's "big lung" detection equipment. When the closed respiratory sensing digital mask + cloud platform system measured the same subject five times, the deviation of FEV1 and FVC was about 70 ml, which was lower than the standard requirement that the deviation of three measurements should be less than 100 ml and 150 ml. Four common lung function parameters, FEV1, FVC, FEV1%pred and FEV1 / FVC, were selected for comparison. The test results are as follows Figure 7 shown.

[0076] Table 2 shows the error statistics of the lung function parameter measurement results of the closed breathing sensing digital mask + cloud platform system and the hospital "big lung" detection equipment. The results show that the two devices have good consistency in the measurement of these four parameters.

[0077] Table 2 Error statistics and correlation analysis of lung function parameters measured by closed respiratory sensing digital mask + cloud platform system and hospital "big lung" detection equipment

[0078]

[0079] Note: The above are the statistical results of measurement errors for 385 people.

[0080] 2.3 Comparison of respiratory function measurement parameters between different populations and COPD patients.

[0081] Among the 385 people tested, 17% of them had COPD, which is close to the 14% COPD prevalence rate among people over 40 years old in China. Therefore, the above population is more representative. In addition, some of the tested people also had asthma (not taking bronchodilators).

[0082] In order to further understand the distribution of respiratory parameters of different groups of people, different groups of people are distinguished by points of different colors, such as Figure 8 shown.

[0083] 3. Result analysis.

[0084] 3.1 Relationship between lung function and gender.

[0085] Figure 7 The data points in the figure are separated by gender: dark blue points represent female volunteers and orange points represent male volunteers. Compared with females, males have a wider distribution of FEV1 values, which is related to the following factors:

[0086] (1) Differences in lung capacity: Men generally have larger lung capacity and airways, so their absolute FEV1 values ​​are higher and have a wider distribution. Due to differences in body shape, height, and lung capacity, men's FEV1 values ​​show a wider distribution.

[0087] (2) Differences in the prevalence of COPD: Figure 7 As shown in (c) and (d) in Figure 1, the proportion of men with chronic obstructive pulmonary disease (FEV1 / FVC < 0.7) is higher than that of women. The reasons can be attributed to the following points:

[0088] (a) Airway size differences: Men generally have larger airways than women, but their alveoli may be less elastic. Although men have larger lung volumes, they are more likely to have airway closure or partial collapse due to the weaker elastic recoil of lung tissue, which may result in a lower FEV1 / FVC ratio during forced exhalation.

[0089] (b) Decreased lung elasticity: Men tend to experience more significant deterioration in lung elasticity as they age. This deterioration reduces FEV1 because the airway's ability to recoil during forced exhalation decreases, leading to air trapping in the lungs and a decrease in the FEV1 / FVC ratio.

[0090] (c) Differences in respiratory muscle strength: Although men generally have stronger respiratory muscles, airway resistance may increase during forced expiration due to larger airways. Larger airways experience greater resistance during forced expiration, which may also lead to a decrease in FEV1 and, therefore, a lower FEV1 / FVC ratio.

[0091] (d) Obesity and abdominal pressure: Men are more likely to accumulate abdominal fat than women, which may put additional pressure on the diaphragm and affect breathing efficiency. The accumulation of abdominal fat may limit the expansion of the lungs, thereby reducing the FEV1 / FVC ratio.

[0092] These factors work together to make the average FEV1 / FVC ratio of men generally lower than that of women. Women usually have a relative advantage in airway size and lung elasticity, so their FEV1 / FVC ratio is higher. This phenomenon is particularly evident in older age groups.

[0093] 3.2 Relationship between respiratory function measurement parameters and COPD and asthma in different populations.

[0094] The combination of the closed breathing sensing digital mask + cloud platform system and the hospital's "big lung" detection equipment shows unique advantages in measuring the lung function parameters of patients with chronic obstructive pulmonary disease (COPD). By comparing the lung function parameters measured by the two devices, especially the FEV1 / FVC ratio and FEV1%pred value, it can assist in predicting the difference between COPD and asthma.

[0095] In the diagnosis of COPD, the FEV1 / FVC ratio and FEV1%pred are two key indicators. According to the Global Initiative for Chronic Obstructive Lung Disease (GOLD) criteria, when the FEV1 / FVC ratio is less than 70% (i.e. 0.7) and the FEV1%pred is less than 80%, a preliminary diagnosis of COPD can be made.

[0096] Figure 8 The comparison of the lung function parameters FEV1%pred and FEV1 / FVC measured by the closed respiratory sensing digital mask + cloud platform system and the hospital's "big lung" detection equipment is shown. The two diagnostic marker values ​​are marked on Figure 8 By comparison, it can be seen that the FEV1%pred values ​​of most COPD patients are lower than 80%, which is consistent with the predicted value; similarly, the FEV1 / FVC ratio is also lower than 0.7, which meets the normal judgment standard.

[0097] However, from Figure 8 It can be observed from (a) and (b) that within these predicted values, some subjects are not COPD patients, especially some asthma patients (dark blue dots). It is worth noting that the lung function parameters of some asthma patients also fall into the range of normal people (light blue dots), indicating that the breathing characteristics of asthma patients are often in a transitional state between normal people and COPD patients.

[0098] Specifically:

[0099] 1. Normal people: FEV1%pred is usually higher than 80%, indicating normal lung function; FEV1 / FVC ratio is generally above 75%, indicating normal airflow without obstruction.

[0100] 2. COPD patients: FEV1%pred is significantly reduced, and may be lower than 80%, 50% or even 30% (corresponding to mild, moderate, severe and extremely severe COPD, respectively). The FEV1 / FVC ratio is lower than 70% for a long time. Even with the use of bronchodilators, the improvement effect is limited, showing the characteristics of persistent and irreversible airflow limitation, and the dyspnea gradually worsens.

[0101] 3. Asthma patients: In the asymptomatic period, FEV1%pred is close to normal or only slightly decreased (usually above 80%), but in the attack period, FEV1%pred may drop significantly. The FEV1 / FVC ratio is close to normal in the asymptomatic period, but may be lower than 70% in the attack period, but it can be significantly improved after inhalation of bronchodilators, showing reversibility.

[0102] from Figure 8As can be seen in (b), when the FEV1 / FVC ratio is below 0.70, the measurement values ​​of most COPD patients are concentrated below the average value of the measurement values ​​of the closed breathing sensing digital mask + cloud platform system and the hospital's "big lung" detection equipment. In other words, even if the FEV1 / FVC ratio meets the diagnostic criteria for COPD, the FEV1 / FVC value measured by the closed breathing sensing digital mask + cloud platform system is often slightly lower than the measurement value of the hospital's "big lung" detection equipment. Taking this into account may help improve the accuracy of COPD diagnosis.

[0103] 4. Conclusion

[0104] The closed breathing sensing digital mask + cloud platform system proposed in the present invention is highly consistent with the conventional "small lung" detection equipment and the hospital "big lung" detection equipment in the measurement results of lung function parameters, indicating that the detection method proposed in the present invention has excellent applicability, accuracy and reliability in people aged 20 to 90 years old and individuals with normal and abnormal lung function. When comparing lung function parameters of different populations, age, gender differences and some diseases (such as patients with chronic obstructive pulmonary disease (COPD)) will affect the statistical results.

[0105] In particular, the comparison of the FEV1%pred and FEV1 / FVC average values ​​between the closed breathing sensing digital mask + cloud platform system and the hospital's "big lung" detection equipment significantly improved the diagnostic accuracy of COPD. Generally, COPD can only be diagnosed with high accuracy when the FEV1 / FVC ratio is lower than 70% (0.7) and this value is below the measured average value of the closed breathing sensing digital mask + cloud platform system and the hospital's "big lung" detection equipment.

[0106] Therefore, the present invention adopts the above-mentioned method for measuring respiratory function parameters based on quantitative analysis of respiratory wave spectrum. Based on the digital mask system, the respiratory wave can be quantitatively collected, and the spectrum analysis and Ai algorithm are combined to achieve non-invasive, convenient and efficient auxiliary diagnosis of lung function and diseases. This innovative technical means combined with Ai not only makes up for the shortcomings of traditional lung function testing methods, but also expands its application scenarios, enabling it to be widely used in various environments such as homes, communities and hospitals. Future research will continue to explore the spectral characteristics of respiratory waves in depth, and as clinical data accumulates, it is planned to verify its applicability in different populations in larger-scale clinical trials.

[0107] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for measuring respiratory function parameters based on quantitative analysis of respiratory wave spectrum, characterized in that: The following steps are involved: Step S1: Based on the flexible respiratory sensor, a closed respiratory sensing digital mask + cloud platform system is constructed, and personal parameter information including gender, age, weight, height, and whether or not to smoke is input; and the respiratory waves of the subjects are collected in three movement modes of sitting, standing, and walking; Step S2, performing respiratory wave spectrum analysis on the respiratory data obtained in step S1, obtaining relevant spectrum characteristic parameters, and obtaining spectrum markers related to personal lung health and disease; Step S3, combining the above-mentioned relevant characteristic parameters with the lung function parameters measured in the hospital and the corresponding measurement results to form a training set, and training the Ai prediction model based on artificial intelligence machine learning or deep learning; wherein, the machine learning or deep learning model applied to the Ai prediction model training specifically includes: decision tree, random forest, support vector machine, k nearest neighbor, naive Bayes, neural network, convolutional neural network, recurrent neural network, long short-term memory network, generative adversarial network, reinforcement learning; Step S4, during the Ai prediction model training process, the prediction of each lung function parameter, forced expiratory volume in 1 second FEV1, forced vital capacity FVC, FEV1 / FVC ratio and predicted value percentage FEV1%pred, is achieved by training an independent Ai prediction model; Step S5, in actual application, the personal parameter information is combined with the respiratory waves of the three motion states and the obtained respiratory wave spectrum parameters of the subject, and input into the trained Ai prediction model to obtain relevant lung function parameters, including the forced expiratory volume in 1 second FEV1, forced vital capacity FVC, FEV1 / FVC ratio and predicted value percentage FEV1%pred; Step S6: Give a prediction result report of Ai-assisted diagnosis and transmit it to the mobile device terminal.

2. The method for measuring respiratory function parameters based on respiratory wave spectrum quantitative analysis according to claim 1, characterized in that: In step S1, the digital mask includes a PC plastic shell, an electronic module, a breathing sensor, a wireless module, a temperature sensor, medical silicone and a PC plastic body.

3. A method for measuring respiratory function parameters based on respiratory wave spectrum quantitative analysis according to claim 2, characterized in that: Based on the closed respiratory sensing digital mask + cloud platform system, the fluctuations of respiratory gas generated by breathing are converted into electrical signal waves through flexible sensors to perform quantitative and continuous measurement of spontaneous respiratory waves.

4. The method for measuring respiratory function parameters based on respiratory wave spectrum quantitative analysis according to claim 3, characterized in that: When collecting respiratory waves, there are three forms of exercise, including sitting for 4 minutes, standing for 4 minutes and walking for 6 minutes; when doing the above exercises, avoid talking, coughing and other abnormal breathing movements.

5. The method for measuring respiratory function parameters based on respiratory wave spectrum quantitative analysis according to claim 4, characterized in that: Spontaneous breathing waveforms include respiratory volume waveforms, vital capacity waveforms, and airflow rate waveforms. Spontaneous breathing waves are measured by flexible sensors, including measuring the waveform, amplitude, frequency, temperature, and humidity of breathing that changes over time, and capturing respiratory signals through airflow changes during the breathing process.

6. The method for measuring respiratory function parameters based on respiratory wave spectrum quantitative analysis according to claim 5, characterized in that: In step S2, the respiratory data obtained in step S11 and step S12 are subjected to respiratory wave spectrum analysis to obtain relevant spectrum characteristic parameters and obtain spectrum markers related to personal lung health and disease. The specific process is as follows: Step S21, taking the breathing data of the tester when sitting still, standing, and walking fast as input, processing and extracting the features of the breathing data pairs, detecting the waveform information of the peak, trough, and maximum flow rate point of the breathing data, and calculating and obtaining the relevant feature parameters of the breathing data of the three actions; Among them, the relevant characteristic parameters include frequency, amplitude, inspiration-expiration ratio, i.e. the ratio of inspiration time to expiration time, Δt / t, i.e. the ratio of the time from the beginning of expiration to the maximum expiration flow rate to the total time of this breath; Step S22: Based on the closed respiratory sensing digital mask + cloud platform system, synchronous extrusion wavelet analysis and HHT conversion are used to perform respiratory wave spectrum analysis to obtain spectrum markers related to personal lung health and disease.

7. The method for measuring respiratory function parameters based on respiratory wave spectrum quantitative analysis according to claim 6, characterized in that: In step S3, the above-mentioned relevant characteristic parameters are combined with the lung function parameters provided by the hospital and the corresponding measurement results to form a training set, and an Ai prediction model is constructed based on machine learning or deep learning of artificial intelligence; wherein the relevant characteristic parameters and clinical data are used as input features of the Ai prediction model; During the model training process, machine learning or deep learning is used for training, and the frequency, amplitude, inhalation-exhalation ratio, Δt / t calculated from the tester's breathing data when sitting, standing, and walking fast, as well as the tester's age, height, and smoking history are used as feature inputs; RandomForestRegressor, XGBRegressor, and GradientBoostingRegr essor are used as base learners of the integrated model; and grid search is used to find the optimal parameters of each model.

8. The method for measuring respiratory function parameters based on respiratory wave spectrum quantitative analysis according to claim 6, characterized in that: The training set and test set data are allocated in a ratio of 5:5 to 8:2, including respiratory waveforms and extracted related feature parameters, and clinical data of various populations, specifically lung function and physical information measured in the hospital; lung function parameters include forced expiratory volume in 1 second FEV1, forced vital capacity FVC, FEV1 / FVC ratio and predicted value percentage FEV1%pred.

Images