An online analysis system based on hydrogen methane breath monitoring

Abstract

The application discloses an online analysis system based on hydrogen and methane breath monitoring, and belongs to the technical field of breath monitoring. The online analysis system comprises a zero point calibration module, a raw signal acquisition module, a data processing module, a signal processing module, a signal compensation module, a secondary zero point calibration module and a baseline correction module. The zero point calibration module is used for acquiring a first collection environment background carbon dioxide concentration. The raw signal acquisition module is used for acquiring a breath sample. The data processing module is used for processing raw signal data flow. The signal processing module is used for outputting a filtered concentration signal. The signal compensation module is used for obtaining compensated hydrogen and methane concentrations. The secondary zero point calibration module is used for acquiring a reference carbon dioxide zero point. The baseline correction module is used for baseline correction. The model establishment module is used for outputting a trend judgment. The online hydrogen and methane breath analysis system reaches the technical effect of maintaining high concentration accuracy under the conditions of breath volume fluctuation and sensor nonlinearity, and solves the measurement deviation problem caused by volume change, nonlinear response and carbon dioxide zero point drift in the prior art.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of respiratory monitoring, and in particular to an online analysis system based on hydrogen-methane breath monitoring. BACKGROUND

[0002] With the growing demand for intestinal microecological research and non-invasive diagnosis of gastrointestinal functional diseases, hydrogen-methane breath test has become an important means for non-invasive assessment of conditions such as carbohydrate malabsorption and small intestinal bacterial overgrowth. In the prior art, the mainstream breath analysis system combines a flow sensor with an electrochemical or infrared gas sensor to collect and calculate the concentrations of hydrogen, methane and carbon dioxide in the breath to determine the metabolic state of the subject.

[0003] In the prior art, a linear dilution model and a fixed zero-point calibration strategy are generally used. When the subject's breath volume fluctuates or the sensor exhibits nonlinear response in the high concentration zone, the measurement results are prone to significant deviations. At the same time, the carbon dioxide zero-point drift is not compensated in real time, resulting in baseline drift of hydrogen and methane concentrations, which further affects the diagnostic accuracy. SUMMARY

[0004] The embodiments of the present application provide an online analysis system based on hydrogen-methane breath monitoring, which solves the problem of hydrogen and methane measurement deviation caused by breath volume fluctuation, sensor nonlinear response and carbon dioxide zero-point drift in the prior art, and realizes high-precision real-time compensation and dynamic baseline correction of breath concentration.

[0005] The embodiments of the present application provide an online analysis system based on hydrogen-methane breath monitoring, which includes a zero-point calibration module for obtaining and recording the first collection environment background carbon dioxide concentration as zero point A;

[0006] An original signal acquisition module is configured to obtain the breath sample of the subject and obtain an original signal data stream containing hydrogen, methane and carbon dioxide concentrations.

[0007] A data processing module is configured to process the original signal data stream to generate a denoised time domain signal.

[0008] A signal processing module is configured to perform state estimation and smoothing processing based on the denoised time domain signal, and output a filtered concentration signal.

[0009] The step of performing state estimation and smoothing processing based on the denoised time domain signal and outputting a filtered concentration signal includes:

[0010] The received denoised time domain signal is used to initialize a state vector and an error covariance matrix.

[0011] performing a recursive calculation cycle based on each time point in the denoised time domain signal;

[0012] In the prediction phase of the calculation cycle, the state vector and error covariance matrix of the previous time are projected forward by a preset system process model to obtain the prior state estimation and prior error covariance of the current time;

[0013] In the update phase of the calculation cycle, the Kalman gain is obtained;

[0014] Based on the denoised time domain signal, the denoised signal value is obtained, the denoised signal value is combined with the prior state estimation, and the Kalman gain is weighted and corrected to obtain the posterior state estimation of the current time;

[0015] The error covariance matrix is updated, and the output posterior state estimation sequence is the filtered concentration signal;

[0016] A signal compensation module is configured to obtain a flow integral exhalation volume value and perform compensation on the filtered concentration signal to obtain compensated hydrogen and methane concentrations;

[0017] The step of performing compensation on the filtered concentration signal to obtain the compensated hydrogen and methane concentrations includes:

[0018] Based on the filtered concentration signal, the filtered hydrogen and methane concentration signals are extracted;

[0019] Based on the obtained flow integral exhalation volume value and the known calibration volume of the quantitative sampling loop of the device, the dilution ratio coefficient is obtained through a preset gas dilution physical model.

[0020] A likelihood function is constructed, and a prior probability distribution is set, and the posterior probability distribution is obtained by combining the likelihood function and the prior probability distribution using the Bayesian inference rule;

[0021] The maximum probability value is extracted from the posterior probability distribution as the compensated hydrogen and methane concentrations;

[0022] A secondary zero point calibration module is configured to collect and record the second collection environment background carbon dioxide concentration as a zero point B to obtain a corrected reference carbon dioxide zero point;

[0023] A baseline correction module is configured to perform baseline correction on the compensated hydrogen and methane concentrations based on the corrected reference carbon dioxide zero point to obtain the final accurate concentration value;

[0024] The step of performing baseline correction on the compensated hydrogen and methane concentrations based on the corrected reference carbon dioxide zero point to obtain the final accurate concentration value includes:

[0025] The carbon dioxide zero point drift is obtained by calculating the difference between the corrected reference carbon dioxide zero point and the fixed reference carbon dioxide zero point based on the corrected reference carbon dioxide zero point by presetting a fixed reference carbon dioxide zero point value in the initial calibration stage;

[0026] The calculated carbon dioxide zero point drift is taken as an input feature through a preset multivariate transfer relationship model to obtain the correction amounts required for the hydrogen concentration and the methane concentration, respectively.

[0027] The final accurate hydrogen concentration value and the final accurate methane concentration value of the current detection are obtained based on the hydrogen concentration value after compensation minus the corresponding hydrogen correction amount and the methane concentration value after compensation minus the corresponding methane correction amount.

[0028] The model establishing module is configured to establish a time series prediction model based on a plurality of groups of historical final accurate concentration values, and output a trend judgment of future hydrogen and methane concentration changes.

[0029] Further, the step of obtaining the exhaled breath sample of the subject includes:

[0030] The hydrogen sensor, the methane sensor, and the carbon dioxide sensor arranged in the sampling channel continuously output electrical signals, and the electrical signals are converted into digital sampling values at a fixed sampling frequency;

[0031] The digital sampling values are written into the ring buffer area in time sequence to form the original signal data stream;

[0032] The time stamp corresponding to each sampling point is recorded synchronously during the writing process in the ring buffer area, and the time stamp corresponds to the digital sampling value one by one;

[0033] The original signal data stream includes a hydrogen digital sampling value sequence, a methane digital sampling value sequence, and a carbon dioxide digital sampling value sequence, and the three sequences have the same length and are time-aligned;

[0034] When the ring buffer area is full, the three sequences are packaged into a data frame, and the data frame is attached with a frame serial number and a starting time stamp.

[0035] Further, the step of obtaining the flow integral exhalation volume value includes:

[0036] The flow sensor located in the exhalation airflow channel measures the instantaneous flow rate of the gas passing through the sensor at a fixed time interval, and records the flow rate measurement result at each time as a data point;

[0037] The instantaneous flow rate is continuously monitored, and when the measured instantaneous flow rate is continuously lower than a preset flow rate termination threshold for a period of time, it is determined that the current exhalation process is ended;

[0038] After the expiration process is determined to be over, all instantaneous flow rate data points recorded during the expiration process are retrieved, and all the instantaneous flow rate data points together form a complete flow rate time series;

[0039] A numerical integration operation is performed on the flow rate time series to obtain the total amount of gas passing through the sensor during the entire expiration process;

[0040] The total amount of gas passing through the sensor is defined as a flow-integrated expiration volume value, and the flow-integrated expiration volume value is stored.

[0041] Further, the likelihood function is a gas concentration observation model based on non-ideal mixing and sensor response characteristics, and the expected observation concentration is calculated by the following formula:

[0042] ;

[0043] wherein, is the expected observation concentration, is the real gas concentration to be estimated, is the actual measurement concentration value extracted from the filtered concentration signal, is the flow-integrated expiration volume value obtained, is a dilution ratio coefficient, is the residual gas concentration in the quantitative sampling loop, is the sensor non-linear response coefficient.

[0044] Further, the multivariate transfer relationship model calculates the correction amount of a specific gas by the following formula:

[0045] ;

[0046] wherein, is the calculated concentration correction amount for the specific gas, is the carbon dioxide zero-point drift amount, is the first-order drift transfer coefficient, is the second-order drift transfer coefficient, is the compensated target gas concentration before baseline correction, is the concentration-related cross-coupling coefficient.

[0047] Further, based on multiple sets of historical final accurate concentration values, the steps of establishing a time series prediction model include:

[0048] Extracting the hydrogen final accurate concentration value sequence and the methane final accurate concentration value sequence corresponding to past expirations from the historical result database;

[0049] Based on the extracted final accurate concentration value sequence, initialize the level component and the trend component of the final accurate concentration value sequence;

[0050] By means of iterative calculation, starting from the first data point of the sequence, process point by point backward;

[0051] For each data point in the sequence, perform the update calculation of the level component and the trend component once;

[0052] When all historical data points are processed and the smoothing coefficient optimization is completed, the final level component, trend component and optimized smoothing coefficient together constitute a complete time series prediction model.

[0053] Further, the step of outputting the trend judgment of the future hydrogen and methane concentration changes comprises:

[0054] By means of the time series prediction model, obtain an input parameter representing a prediction time span;

[0055] For each future time point that needs to be predicted, perform a prediction calculation once;

[0056] Repeat the prediction calculation until the entire prediction time span is covered, thereby generating a sequence containing multiple future predicted concentration values.

[0057] Based on the generated future predicted concentration value sequence, obtain the overall slope of the future predicted concentration value sequence;

[0058] Based on the obtained overall slope, if the overall slope is positive, the trend of the future hydrogen and methane concentration changes is judged to be rising;

[0059] If the overall slope is negative, the trend of the future hydrogen and methane concentration changes is judged to be falling.

[0060] The one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0061] 1. By double-stage calibration of the environmental carbon dioxide zero point and dynamic compensation of the zero point drift amount by combining the baseline correction step, the accuracy of hydrogen and methane concentration measurement is significantly improved, thereby realizing high-precision detection of online breath concentration analysis under complex environmental changes, and effectively solving the problem of increased measurement error caused by environmental zero point drift in the prior art.

[0062] 2. By applying Kalman filtering and recursive state estimation to the original breath signal, the instantaneous concentration data is quickly denoised and smoothed, thereby outputting a high signal-to-noise ratio hydrogen and methane concentration sequence in real time, and thereby realizing rapid response and stable tracking of sudden breath fluctuations, greatly improving the response speed and stability of the online analysis system.

[0063] 3. By combining the dilution coefficient calculation based on the sampling volume with Bayesian inference to jointly correct the sensor sampling value, the compensation concentration value of the maximum a posteriori estimate is obtained, thereby realizing the accurate correction of the gas dilution effect and the nonlinear response of the sensor, which significantly improves the reliability and repeatability of the detection results. Attached Figure Description

[0064] Figure 1 This is a schematic diagram of an online analysis system based on hydrogen methane breath monitoring, provided as an embodiment of this application. Detailed Implementation

[0065] This application provides an online analysis system based on hydrogen and methane exhalation monitoring, which solves the problem of hydrogen and methane measurement deviation caused by the combined effects of exhalation volume fluctuation, sensor nonlinear response, and carbon dioxide zero-point drift in the prior art. By introducing a nonlinear dilution compensation model, hyperbolic tangent saturation correction, Bayesian drift compensation, and multivariate zero-point drift transfer model, high-precision real-time compensation and dynamic baseline correction of exhalation concentration are achieved.

[0066] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0067] like Figure 1 The diagram shown is a schematic diagram of an online analysis system based on hydrogen methane exhalation monitoring provided in an embodiment of this application. It includes: a zero-point calibration module, used to acquire and record the first ambient carbon dioxide concentration as zero point A in the closed state of the sampling inlet before the exhalation sample is collected;

[0068] The raw signal acquisition module is used to acquire the subject's exhaled breath sample in real time and obtain raw signal data streams containing the concentrations of hydrogen, methane, and carbon dioxide.

[0069] The data processing module is used to process the original signal data stream using wavelet transform to generate a denoised time-domain signal.

[0070] The signal processing module is used to perform state estimation and smoothing processing based on the denoised time-domain signal using the Kalman filter algorithm, and output the filtered concentration signal.

[0071] A state vector is initialized using the received denoised time-domain signal, which describes the estimated gas concentration, and an error covariance matrix is ​​used to quantify the uncertainty of the state vector estimate.

[0072] Based on each time point in the denoised time-domain signal, a recursive calculation loop is executed;

[0073] In the prediction phase of the calculation cycle, the state vector and error covariance matrix of the previous time are projected forward by the preset system process model to obtain the prior state estimation and prior error covariance of the current time;

[0074] In the update phase of the calculation cycle, the Kalman gain is obtained, and the value of the gain depends on the prior error covariance and the covariance of the measurement noise, to balance the weight between the prior estimation and the new measurement value;

[0075] Based on the denoised time domain signal, the denoised signal value is obtained, the denoised signal value is combined with the prior state estimation, and the posterior state estimation of the current time is obtained by weighting correction through the Kalman gain;

[0076] The error covariance matrix is updated to reflect the reduced uncertainty due to the adoption of the new measurement value. This prediction and update cycle will run through the entire time domain signal sequence, and the final output posterior state estimation sequence is the filtered concentration signal.

[0077] A signal compensation module is configured to obtain a flow integral exhalation volume value, and perform Bayesian dilution compensation on the filtered concentration signal to correct the concentration dilution caused by insufficient exhalation volume, and obtain compensated hydrogen and methane concentrations.

[0078] The step of performing compensation on the filtered concentration signal to obtain the compensated hydrogen and methane concentrations comprises:

[0079] Based on the filtered concentration signal, the filtered hydrogen and methane concentration signals are extracted;

[0080] A preset gas dilution physical model is used, which describes the physical process of the actual exhaled gas mixed with the residual gas in the quantitative sampling ring inside the device. Based on the obtained flow integral exhalation volume value and the known calibration volume of the quantitative sampling ring of the device itself, a dilution ratio coefficient is obtained, which represents the actual proportion of the exhaled sample in the sampling ring.

[0081] A likelihood function is constructed, which represents the probability of observing the current filtered concentration signal under the condition of the given true concentration and the calculated dilution ratio coefficient. At the same time, a prior probability distribution is set, which represents the preliminary estimation of the true concentration before the measurement. By combining the likelihood function and the prior probability distribution, the Bayesian inference rule is applied to obtain the posterior probability distribution; this posterior probability distribution comprehensively describes the updated knowledge of the true gas concentration after considering the measurement value and the dilution effect.

[0082] The maximum probability value is extracted from the posterior probability distribution as the final result of the compensation for the dilution effect caused by insufficient exhalation, i.e. the compensated hydrogen and methane concentrations.

[0083] a secondary zero point calibration module, configured to, after sample analysis is completed, collect and record a second collected ambient background carbon dioxide concentration as a zero point B under the condition that the sampling inlet is closed, and calculate an arithmetic mean of the zero point A and the zero point B to obtain a corrected reference carbon dioxide zero point;

[0084] a baseline correction module, configured to, based on the corrected reference carbon dioxide zero point, perform baseline correction on the compensated hydrogen and methane concentrations to obtain final accurate concentration values;

[0085] the step of, based on the corrected reference carbon dioxide zero point, performing baseline correction on the compensated hydrogen and methane concentrations to obtain final accurate concentration values comprises:

[0086] a fixed reference carbon dioxide zero point value is preset through an initial calibration stage, and based on the corrected reference carbon dioxide zero point, which is obtained by calculating an arithmetic mean of the first collected ambient background carbon dioxide concentration and the second collected ambient background carbon dioxide concentration, a carbon dioxide zero point drift is obtained by calculating a difference between the corrected reference carbon dioxide zero point and the fixed reference carbon dioxide zero point;

[0087] a multivariate transfer relationship model is preset, the model describing a quantitative correlation between the carbon dioxide sensor zero point drift and the hydrogen sensor and methane sensor reading offsets, and the calculated carbon dioxide zero point drift is taken as an input feature to obtain correction amounts required for the hydrogen concentration and the methane concentration, respectively;

[0088] the final accurate hydrogen concentration value and the final accurate methane concentration value of the current detection are obtained by subtracting the corresponding hydrogen correction amount from the compensated hydrogen concentration value and subtracting the corresponding methane correction amount from the compensated methane concentration value.

[0089] a model establishment module, configured to establish a time series prediction model based on a plurality of groups of historical final accurate concentration values, and output a trend judgment of future hydrogen and methane concentration changes.

[0090] Further, the step of obtaining the exhaled sample of the subject in real time comprises:

[0091] the hydrogen sensor, the methane sensor, and the carbon dioxide sensor arranged in the sampling channel continuously output electric signals, and an analog-to-digital converter converts the electric signals into digital sampling values at a fixed sampling frequency;

[0092] The raw signal data stream is formed by writing the digital sample values in time sequence into a ring buffer, the buffer capacity is sufficient to cover a complete exhalation cycle, and the old data is overwritten by the new data when the buffer is full;

[0093] By synchronously recording the timestamp corresponding to each sampling point during the writing process in the ring buffer, the timestamp corresponds to the digital sample value one by one;

[0094] The raw signal data stream includes a sequence of hydrogen digital sample values, a sequence of methane digital sample values, and a sequence of carbon dioxide digital sample values, and the three sequences have the same length and are time-aligned;

[0095] When the ring buffer is full, the three sequences are packed into a data frame, the data frame is attached with a frame number and a starting timestamp, and the raw signal data stream is sent to the data processing module, and the raw signal data stream remains unchanged in time sequence and is not compressed during transmission.

[0096] Further, the step of obtaining the flow integral exhalation volume value comprises:

[0097] The flow sensor located in the exhalation airflow channel measures the instantaneous flow rate of the gas passing through the sensor at fixed time intervals, and records the flow rate measurement at each time as a data point;

[0098] The instantaneous flow rate is continuously monitored, and when the measured instantaneous flow rate is continuously below a preset flow rate termination threshold for a period of time, it is determined that the current exhalation process is ended;

[0099] After the exhalation process is determined to be ended, all the instantaneous flow rate data points recorded during the exhalation period are retrieved, and all the instantaneous flow rate data points together constitute a complete flow rate time sequence;

[0100] The flow rate time sequence is subjected to a numerical integral operation, which is obtained by multiplying the value of each flow rate data point by the sampling time interval, and then adding all the products. The sum of this addition operation, i.e. the total amount of gas passing through the sensor during the entire exhalation process;

[0101] The total amount of gas passing through the sensor is defined as the flow integral exhalation volume value, and the flow integral exhalation volume value is stored for subsequent steps to be called by the signal compensation module.

[0102] Further, the likelihood function is based on the gas concentration observation model of non-ideal mixing and sensor response characteristics, and the expected observation concentration is calculated by the following formula:

[0103] ;

[0104] wherein, is the expected observed concentration, representing the theoretically most likely concentration value to be measured by the sensor at a given true concentration and breath volume, is the true gas concentration to be estimated, which is the target variable of the Bayesian inference, is the actual measured concentration value extracted from the filtered concentration signal, is the obtained flow-integrated breath volume value, is the dilution ratio coefficient, is the residual gas concentration in the quantitative sampling loop, which is dynamically assigned using the corrected baseline carbon dioxide zero point or set to the ambient air concentration, is the sensor non-linear response coefficient.

[0105] Further, the multi-variable transfer relationship model calculates the correction amount for a specific gas by the following formula:

[0106] ;

[0107] wherein, is the calculated concentration correction amount for a specific gas ( or ), is the carbon dioxide zero point drift, i.e., the difference between the corrected baseline carbon dioxide zero point and the fixed reference carbon dioxide zero point, is the first-order drift transfer coefficient, describing the linear effect of the carbon dioxide zero point drift on the target gas concentration, determined through multi-variable calibration experiments, is the second-order drift transfer coefficient, describing the non-linear effect of the carbon dioxide zero point drift on the target gas concentration, used to correct the response relationship under large drift, also determined through calibration experiments, is the compensated target gas concentration before baseline correction, is a natural constant, is the concentration-dependent cross-coupling coefficient, which reflects the correlation between the influence degree of the zero point drift and the concentration level of the target gas. For example, under a high-concentration methane background, the drift of carbon dioxide can cause more significant reading errors. This coefficient is calibrated through drift experiments at different background concentrations.

[0108] Further, based on multiple sets of historical final accurate concentration values, the steps of establishing a time series prediction model include:

[0109] Extracting the sequence of hydrogen final accurate concentration values and the sequence of methane final accurate concentration values corresponding to past breaths from the historical result database; the sequence is for a specific gas, such as hydrogen or methane.

[0110] Based on the extracted final concentration value sequence, the level component and the trend component of the final concentration value sequence are initialized; the level component represents the baseline value of the sequence, while the trend component represents the growth or decrease rate of the sequence over time.

[0111] By means of iterative calculation, starting from the first data point of the sequence, each point is processed backward;

[0112] For each data point in the sequence, the update calculation of the level component and the trend component is performed once;

[0113] The update of the level component is completed through a weighted average process, which combines the value of the current data point and the sum of the level component and the trend component at the previous time. The update of the trend component is also completed through a weighted average process, which combines the difference between the current updated level component and the level component at the previous time, and the trend component at the previous time. The weight factor, i.e. the smoothing coefficient, used in the two weighted average processes is determined by an optimization algorithm. The goal of the algorithm is to find a set of smoothing coefficients that minimizes the total error of one-step prediction based on historical data.

[0114] When all historical data points are processed and the smoothing coefficient is optimized, the final level component, trend component and optimized smoothing coefficient together constitute a complete time series prediction model.

[0115] Further, the step of outputting the trend judgment of the future hydrogen and methane concentration changes includes:

[0116] Through the time series prediction model, the model contains the latest level component, the latest trend component and the optimized smoothing coefficient. An input parameter representing the prediction time span is obtained; the parameter defines the number of future time points that need to be predicted.

[0117] For each future time point that needs to be predicted, a prediction calculation is performed. The calculation is completed by adding the latest level component to an increment, which is the product of the latest trend component and the step distance of the future time point relative to the current time.

[0118] Repeat the prediction calculation until the entire prediction time span is covered, thereby generating a sequence containing multiple future predicted concentration values.

[0119] Based on the generated future predicted concentration value sequence, the system analyzes the sequence to determine its overall evolution direction. The analysis process includes comparing the predicted values in the later period of the sequence with the predicted values in the earlier period, obtaining the overall slope of the future predicted concentration value sequence;

[0120] Based on the acquired overall slope, if the overall slope is positive, the trend of the future hydrogen and methane concentration change is determined to be rising;

[0121] If the overall slope is negative, the trend of the future hydrogen and methane concentration change is determined to be falling.

[0122] Those skilled in the art will appreciate that embodiments of the present application can be readily used as software, hardware, or a combination of software and hardware. In one embodiment, the present application can be implemented in software and can be stored on a computer readable medium, which can include random access memory (RAM), read only memory (ROM), magnetic disk or optical disk, or the like. The software implementation can comprise one or more computer program components embodied on one or more computer readable media. The computer readable medium can be resident within a computing device, external to the computing device, or distributed across multiple computing devices.

[0123] The present application is described in relation to flow diagrams and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the present application. It is understood that each flow and / or block in the flow diagrams and / or block diagrams, and combinations of flows and / or blocks in the flow diagrams and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processing device, or other programmable data processing device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing device, create means for implementing the functions specified in the flow diagrams and / or block diagrams block or blocks. Figure 1 The flow diagram and / or block diagram in which each flow and / or block includes the functions described in that flow and / or block. Figure 1 The apparatus with the function specified in the flow diagram and / or block diagram block or blocks.

[0124] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the flow diagrams and / or block diagrams flow or flows and / or block or blocks. Figure 1 The flow diagram and / or block diagram in which each flow and / or block includes the functions described in that flow and / or block. Figure 1 The apparatus with the function specified in the flow diagram and / or block diagram block or blocks.

[0125] These computer program instructions can also be loaded onto a computer or other programmable data processing device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer implemented process such that the instructions which execute on the computer or other programmable device provide steps for implementing the flow diagrams and / or block diagrams flow or flows and / or block or blocks. Figure 1 The flow diagram and / or block diagram in which each flow and / or block includes the functions described in that flow and / or block. Figure 1 The apparatus with the function specified in the flow diagram and / or block diagram block or blocks.

[0126] While the preferred embodiments of the application have been described, additional variations and modifications can be made to these embodiments by those skilled in the art once they have the benefit of the present disclosure without departing from the spirit and scope of the application. Accordingly, it is intended that the appended claims include all such modifications and variations as fall within the scope of the present application.

[0127] It is apparent that those skilled in the art can make various changes and modifications to the application without departing from the spirit and scope of the application. It is therefore intended that the present application cover all such changes and modifications that are within its scope.

Claims

1. An online analytical system based on hydrogen methane breath monitoring, characterized in that, The method comprises the following steps: A zero calibration module is used to obtain and record the first collected ambient background carbon dioxide concentration as zero point A; An original signal collection module is used to obtain the exhaled sample of the subject and obtain the original signal data stream containing the concentrations of hydrogen, methane and carbon dioxide; A data processing module is used to process the original signal data stream to generate a denoised time domain signal; A signal processing module is used to perform state estimation and smoothing processing based on the denoised time domain signal, and output a filtered concentration signal; The step of performing state estimation and smoothing processing based on the denoised time domain signal and outputting a filtered concentration signal comprises: Initializing a state vector and an error covariance matrix through the received denoised time domain signal; Performing a recursive calculation loop based on each time point in the denoised time domain signal; In the prediction stage of the calculation loop, the state vector and the error covariance matrix of the previous moment are projected forward through a preset system process model to obtain the prior state estimation and the prior error covariance of the current moment; In the update stage of the calculation loop, the Kalman gain is obtained; Based on the denoised time domain signal, the denoised signal value is obtained, the denoised signal value is combined with the prior state estimation, and the posterior state estimation of the current moment is obtained through weighted correction by the Kalman gain; The error covariance matrix is updated, and the output posterior state estimation sequence is the filtered concentration signal; A signal compensation module is used to obtain the flow integral exhaled volume value and perform compensation on the filtered concentration signal to obtain the compensated hydrogen and methane concentrations; The step of performing compensation on the filtered concentration signal to obtain the compensated hydrogen and methane concentrations comprises: Based on the filtered concentration signal, the filtered hydrogen and methane concentration signals are extracted; Based on the obtained flow integral exhaled volume value and the known calibration volume of the quantitative sampling loop of the device itself, the dilution ratio coefficient is obtained through a preset gas dilution physical model; A likelihood function is constructed, and a prior probability distribution is set, and the posterior probability distribution is obtained by combining the likelihood function and the prior probability distribution through the Bayesian inference rule; The maximum probability value is extracted from the posterior probability distribution as the compensated hydrogen and methane concentrations; A secondary zero calibration module is used to collect and record the second collected ambient background carbon dioxide concentration as zero point B to obtain a corrected reference carbon dioxide zero point; A baseline correction module is used to perform baseline correction on the compensated hydrogen and methane concentrations based on the corrected reference carbon dioxide zero point to obtain the final accurate concentration value; The step of performing baseline correction on the compensated hydrogen and methane concentrations based on the corrected reference carbon dioxide zero point to obtain the final accurate concentration value comprises: Through the preset fixed reference carbon dioxide zero point value in the initial calibration stage, the difference between the corrected reference carbon dioxide zero point and the fixed reference carbon dioxide zero point is calculated based on the corrected reference carbon dioxide zero point to obtain the carbon dioxide zero point drift. The calculated zero-point drift of carbon dioxide is taken as an input feature through a preset multivariate transfer relationship model to obtain correction amounts required for hydrogen concentration and methane concentration respectively; The final accurate hydrogen concentration value and the final accurate methane concentration value of the current detection are obtained by subtracting the corresponding hydrogen correction amount from the compensated hydrogen concentration value and subtracting the corresponding methane correction amount from the compensated methane concentration value; A model establishing module is configured to establish a time series prediction model based on a plurality of groups of historical final accurate concentration values, and output a trend judgment of future hydrogen and methane concentration changes.

2. An online analytical system based on hydrogen methane breath monitoring as claimed in claim 1, wherein, The step of obtaining the exhalation sample of the subject in real time includes: The hydrogen sensor, the methane sensor, and the carbon dioxide sensor continuously output electrical signals, and the electrical signals are converted into digital sampling values at a fixed sampling frequency; The digital sampling values are written into the ring buffer area in time sequence to form the original signal data stream; The time stamp corresponding to each sampling point is recorded synchronously during the writing process in the ring buffer area, and the time stamp corresponds to the digital sampling value one by one; The original signal data stream includes a hydrogen digital sampling value sequence, a methane digital sampling value sequence, and a carbon dioxide digital sampling value sequence, and the three sequences have the same length and are time-aligned; When the ring buffer area is full, the three sequences are packaged into a data frame, and the data frame is attached with a frame serial number and a starting time stamp.

3. The online analytical system based on hydrogen methane breath monitoring as claimed in claim 1, wherein, The step of obtaining the flow-integrated exhalation volume value includes: The flow sensor in the exhalation airflow channel measures the instantaneous flow rate of the gas passing through the sensor at fixed time intervals and records the flow rate measurement result at each time as a data point; The instantaneous flow rate is continuously monitored, and when the measured instantaneous flow rate is below a preset flow rate termination threshold for a continuous period of time, it is determined that the current exhalation process is ended; After the exhalation process is determined to be ended, all instantaneous flow rate data points recorded during the exhalation are retrieved, and all the instantaneous flow rate data points collectively constitute a complete flow rate time sequence; Numerical integration operation is performed on the flow rate time sequence to obtain the total amount of gas passing through the sensor during the entire exhalation process; The total amount of gas passing through the sensor is defined as the flow-integrated exhalation volume value, and the flow-integrated exhalation volume value is stored.

4. The online analytical system based on hydrogen methane breath monitoring as claimed in claim 1, wherein, Likelihood function is a gas concentration observation model based on non-ideal mixing and sensor response characteristics, whose expected observation concentration is calculated by the following equation: ； wherein, is the desired observed concentration, is the true gas concentration to be estimated, is the actual measured concentration value extracted from the filtered concentration signal, is the obtained flow integrated exhalation volume value, is the dilution ratio coefficient, is the residual gas concentration in the quantitative sampling loop, is the sensor non-linear response coefficient.

5. The online analytical system based on hydrogen methane breath monitoring as claimed in claim 1, wherein, The multivariate transfer relationship model calculates the correction amount of a specific gas by the following formula: ； wherein is a calculated concentration correction for a particular gas, is a carbon dioxide zero point drift, is a first order drift transfer coefficient, is a second order drift transfer coefficient, is a compensated target gas concentration before baseline correction, is a concentration dependent cross coupling coefficient.

6. The online analytical system based on hydrogen methane breath monitoring as claimed in claim 1, wherein, The step of establishing a time series prediction model based on a plurality of groups of historical final accurate concentration values includes: The hydrogen final accurate concentration value sequence and the methane final accurate concentration value sequence corresponding to past exhalations are extracted from the historical result database; Based on the extracted final accurate concentration value sequences, the level component and the trend component of the final accurate concentration value sequence are initialized; Starting from the first data point of the sequence, the processing is performed point by point backward through iterative calculation; For each data point in the sequence, the update calculation of the level component and the trend component is performed once; After all historical data points are processed and the smoothing coefficient is optimized, the final level component, the final trend component, and the optimized smoothing coefficient collectively constitute a complete time series prediction model.

7. The online analytical system based on hydrogen methane breath monitoring as claimed in claim 1, wherein, The step of outputting the trend judgment of the future hydrogen and methane concentration changes comprises: obtaining, by the time series prediction model, input parameters representing a prediction time span; performing a prediction calculation once for each future time point that needs to be predicted; repeating the prediction calculation until the entire prediction time span is covered, thereby generating a sequence comprising a plurality of future predicted concentration values; obtaining an overall slope of the sequence of future predicted concentration values based on the generated sequence of future predicted concentration values; based on the obtained overall slope, if the overall slope is positive, judging the trend of the future hydrogen and methane concentration changes as rising; and if the overall slope is negative, judging the trend of the future hydrogen and methane concentration changes as falling.

Images