Membrane lung and ventilator end co discrimination monitoring method and system

Abstract

The present application relates to the technical field of critical medical respiratory circulation support monitoring, in particular to a membrane lung and respirator end CO distinguishing monitoring method and system. The present application collects membrane lung and respirator related signals through deploying a heterogeneous sensor network composed of multi-modal sensing array and carries out space-time calibration, develops a dynamic decoupling algorithm combining a blood vessel tree-alveolar coupling model, deduces a pure mechanical blood flow impact component at the membrane lung end and a blood flow oscillation component at the respirator end, then utilizes a parallel filter group with parameter self-correction function and a reference signal to dynamically update the cutoff frequency to realize complete separation of the CO signal frequency domain at both ends, finally eliminates common mode interference through multi-sensor fusion and distinguishes the contribution degree of the anatomical dead space and the membrane lung effective perfusion area by means of a joint matrix, effectively solving the problem that the prior art cannot accurately distinguish the membrane lung end and the respirator end CO.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of critical medical respiratory circulation support monitoring, in particular to a membrane lung and breathing machine end CO distinguishing monitoring method and system. BACKGROUND

[0002] In the respiratory circulation support treatment of critical patients, a membrane lung (ECMO) and a breathing machine are often used in combination. The membrane lung mainly helps patients complete gas exchange and blood circulation assistance through extracorporeal circulation to relieve the burden on the heart and lungs; the breathing machine delivers gas to the patient's lungs through the airway to maintain the ventilation function of the lungs, and the two work together to ensure the stability of the patient's vital signs. The cardiac output (CO) is a key indicator for evaluating the patient's circulation function and the effect of device support, wherein the CO of the membrane lung end can reflect the processing efficiency of the membrane lung on the blood flow, the matching degree of gas exchange and blood flow, and the CO of the breathing machine end is directly related to the patient's own heart pumping capacity and the coordination state of the lung ventilation and blood flow. Accurate acquisition of the CO data of the two ports is an important basis for doctors to adjust the parameters of the membrane lung and the breathing machine and judge the changes in the patient's condition.

[0003] At present, the CO monitoring technology suitable for the combined treatment scene of the membrane lung and the breathing machine generally has the problem of being unable to accurately distinguish and monitor the CO of the membrane lung end and the breathing machine end. Because of the interaction of blood flow and gas when the membrane lung and the breathing machine are working, the change of the blood flow supply of the membrane lung will affect the airway pressure of the breathing machine end, and the adjustment of the ventilation parameters of the breathing machine will also indirectly change the blood flow distribution of the membrane lung end, which makes it difficult for the existing monitoring system to separate the CO signal only from the membrane lung end and the CO signal only related to the breathing machine end. The final CO monitoring data is the fusion result of the information of the two ends. Such fused data cannot accurately judge whether the actual working efficiency of the membrane lung meets the treatment needs, nor can it accurately evaluate the real state of the patient's own respiratory system and circulation system under the support of the breathing machine, which may cause the doctor to be unable to timely and accurately adjust the treatment scheme, and adversely affect the optimization of the treatment effect and the stability of the patient's condition. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides a membrane lung and breathing machine end CO distinguishing monitoring method and system, which solves the problem that the prior art cannot accurately distinguish the cardiac output of the membrane lung end and the breathing machine end in the combined treatment scene of the membrane lung and the breathing machine.

[0005] To achieve the above purpose, the present application is implemented by the following technical scheme: a membrane lung and breathing machine end CO distinguishing monitoring method, comprising:

[0006] S1, deploy a heterogeneous sensor network at the membrane lung input end, oxygenator output end, and breathing machine loop branch point through a multi-modal sensor array, collect membrane lung-specific blood pump pulsatile waveform and chest cavity movement harmonic characteristics under the driving of the breathing machine, and perform time and space calibration on the heterogeneous sensing data to form a double-channel original signal pool;

[0007] S2, develop a dynamic decoupling algorithm based on a blood vessel tree-alveolar coupling model, and establish a phase difference equation of the membrane lung roller pump speed and the breathing machine positive pressure cycle;

[0008] S3, by real-time tracking of blood pump impeller angular acceleration and breathing valve opening and closing timing, the pure mechanical blood flow impact component at the membrane lung end and the blood flow oscillation component induced by the chest cavity volume change at the breathing machine end are reversely deduced;

[0009] S4, design a parallel filter bank with parameter self-correction function, and respectively introduce the membrane lung end high-frequency turbulent noise and the breathing machine end low-frequency ventilation disturbance into the corresponding Wiener filter channel;

[0010] S5, use the reference signals generated by the membrane lung blood flow predictor and the breathing machine tidal volume estimator to dynamically update the cutoff frequency of each channel, and ensure that the cardiac output signals at the membrane lung end and the breathing machine end are completely separated in the frequency domain;

[0011] S6, eliminate common mode interference through multi-sensor fusion, and use the blood vessel impedance-airway resistance joint matrix to distinguish the contribution degree of the anatomic dead space and the membrane lung effective perfusion area, and finally output the double-channel cardiac output data, which includes the real-time absolute value of the cardiac output at the membrane lung end, the real-time absolute value of the cardiac output at the breathing machine end, and the respective source proportion atlas.

[0012] Further, the deployment of a heterogeneous sensor network at the membrane lung input end, the oxygenator output end, and the breathing machine loop branch point through a multi-modal sensor array comprises:

[0013] S11, deploy a first optical fiber blood flow sensor at the membrane lung input end;

[0014] S12, deploy a second optical fiber blood flow sensor and an ultrasonic Doppler probe at the oxygenator output end;

[0015] S13, deploy a differential air pressure sensor at the breathing machine loop branch point;

[0016] S14, the time and space calibration of the heterogeneous sensing data is realized through a Kalman filtering algorithm based on data fusion.

[0017] Further, the development of a dynamic decoupling algorithm based on a blood vessel tree-alveolar coupling model to establish a phase difference equation of the membrane lung roller pump speed and the breathing machine positive pressure cycle comprises:

[0018] S21, construct a dynamic impedance model of the blood vessel tree, the dynamic impedance model comprising blood flow inertia parameters and blood vessel wall elasticity parameters;

[0019] S22, construct a gas exchange kinetics model of the alveoli, the gas exchange kinetics model comprising ventilation pressure and lung compliance parameters;

[0020] S23, establish a physical coupling equation of the membrane lung roller pump driving blood flow pulsation and the ventilator positive pressure ventilation inducing chest pressure change;

[0021] S24, determine the mathematical expression of the influence of the membrane lung roller pump speed change on the blood flow dynamics in the ventilator positive pressure cycle through the physical coupling equation.

[0022] Further, the pure mechanical blood flow impact component at the membrane lung end and the blood flow oscillation component induced by the chest volume change at the ventilator end are reversely deduced by tracking the blood pump impeller angular acceleration and the respiratory valve opening and closing timing in real time, including:

[0023] S31, based on the continuous measurement value of the blood pump impeller angular acceleration, calculate the instantaneous flow rate change rate of the membrane lung blood flow pumping, and generate the pure mechanical blood flow impact original component at the membrane lung end;

[0024] S32, based on the respiratory valve opening and closing timing and the tidal volume parameter set by the ventilator, combined with the lung compliance parameter, estimate the periodic change of the chest volume, and generate the blood flow oscillation original component induced by the chest volume change at the ventilator end;

[0025] S33, by a nonlinear least squares method, orthogonally decompose the pure mechanical blood flow impact original component at the membrane lung end and the blood flow oscillation original component at the ventilator end, and remove the aliasing signal caused by interaction.

[0026] Further, the parallel filter bank with parameter self-correction function is designed, including:

[0027] S41, construct a statistical model of high-frequency turbulent flow noise at the membrane lung end, the statistical model being based on the inherent vibration spectrum of the membrane lung blood pump;

[0028] S42, construct a statistical model of low-frequency ventilation disturbance at the ventilator end, the statistical model being based on the ventilation frequency and the tidal volume change characteristics;

[0029] S43, initialize the initial parameters of the Wiener filter channel according to the statistical model;

[0030] S44, during the system operation, dynamically adjust the filter coefficients of the Wiener filter channel according to the statistical characteristics of the real-time signal through the adaptive least mean square error algorithm.

[0031] Further, the reference signal generated by the membrane lung blood flow predictor and the ventilator tidal volume estimator is used to dynamically update the cutoff frequency of each channel, including:

[0032] S51, the membrane lung blood flow predictor predicts the blood flow at the membrane lung end in real time based on the membrane lung roller pump speed and the pipe resistance model;

[0033] S52, the ventilator tidal volume estimator estimates the tidal volume of the ventilator in real time based on the differential pressure sensor data and the lung mechanics model;

[0034] S53, according to the predicted membrane lung blood flow and the estimated ventilator tidal volume, the cutoff frequency of the membrane lung end wiener filtering channel and the ventilator end wiener filtering channel is calculated and updated through the preset dynamic frequency mapping rule.

[0035] Further, the contribution degree of the anatomic dead space and the membrane lung effective perfusion area is distinguished by using the vascular impedance-airway resistance joint matrix, including:

[0036] S61, a joint matrix containing anatomic dead space volume, alveolar perfusion volume and membrane lung blood flow perfusion volume is established;

[0037] The joint matrix is calibrated in real time through the parameters in the vascular tree-alveolar coupling model;

[0038] S62, the singular value decomposition method is used to separate the independent contribution degrees of the anatomic dead space, the alveolar effective perfusion area and the membrane lung effective perfusion area to the total cardiac output from the joint matrix.

[0039] The application also provides a membrane lung and ventilator end CO discrimination monitoring system, comprising:

[0040] A multi-modal sensing array module is used to deploy a heterogeneous sensor network at the membrane lung input end, the oxygenator output end and the ventilator loop branch point, collect the membrane lung specific blood pump pulsatile waveform and the thoracic cavity movement harmonic characteristics under the driving of the ventilator, and perform space-time calibration on the heterogeneous sensing data to form a double-channel raw signal pool;

[0041] A physiological coupling calculation module is used to develop a dynamic decoupling algorithm based on the vascular tree-alveolar coupling model, establish a phase difference equation of the membrane lung roller pump speed and the ventilator positive pressure cycle, and inversely deduce the pure mechanical blood flow impact component at the membrane lung end and the blood flow oscillation component induced by the thoracic cavity volume change at the ventilator end by tracking the blood pump impeller angular acceleration and the breathing valve opening and closing time sequence in real time;

[0042] A dual-domain adaptive filtering module is used to design a parallel filter bank with a parameter self-correcting function, to introduce the membrane lung end high-frequency turbulent flow noise and the respirator end low-frequency ventilation disturbance into corresponding Wiener filtering channels respectively, and to use the reference signals generated by the membrane lung blood flow predictor and the respirator tidal volume estimator to dynamically update the cut-off frequency of each channel, so as to ensure that the membrane lung end cardiac output signal and the respirator end cardiac output signal are completely separated in the frequency domain;

[0043] A data fusion and contribution analysis module is used to eliminate common mode interference through multi-sensor fusion, and to distinguish the contribution of the anatomic dead space and the membrane lung effective perfusion area by using a vascular impedance-airway resistance joint matrix, and finally to output dual-channel cardiac output data, which includes the real-time absolute value of the membrane lung end cardiac output, the real-time absolute value of the respirator end cardiac output, and the respective source proportion atlas.

[0044] The application also provides a computing device comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to realize the membrane lung and respirator end CO differentiation monitoring method of any one of the above.

[0045] The application also provides a computer storage medium storing a computer program, which realizes the membrane lung and respirator end CO differentiation monitoring method of any one of the above when executed by a computer.

[0046] Compared with the prior art, the application has the following beneficial effects:

[0047] The application collects the membrane lung and respirator related signals by deploying a heterogeneous sensor network composed of a multi-modal sensor array and performs time and space calibration, develops a dynamic decoupling algorithm combined with a vascular tree-alveolar coupling model, deduces the membrane lung end pure mechanical blood flow impact component and the respirator end blood flow oscillation component, and then uses a parallel filter bank with a parameter self-correcting function and a reference signal to dynamically update the cut-off frequency to realize the complete separation of the CO signals at the two ends in the frequency domain, and finally eliminates common mode interference through multi-sensor fusion and distinguishes the contribution of the anatomic dead space and the membrane lung effective perfusion area by using a joint matrix, effectively solving the problem that the prior art cannot accurately distinguish the membrane lung end and respirator end CO, and being able to output dual-channel data containing the real-time absolute values of the CO at the two ends and the source proportion atlas, so that doctors can accurately grasp the membrane lung working efficiency and the patient's own circulation and respiration state, provide reliable basis for timely and accurate adjustment of treatment plan, and help to optimize patient treatment effect and disease stability. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 The method flowchart of the application;

[0049] Figure 2Flow chart of physiological signal decoupling and filtering process of the present application;

[0050] Figure 3 Iteration principle diagram of dynamic decoupling algorithm of the present application;

[0051] Figure 4 Double-channel CO output and clinical feedback schematic diagram of the present application. DETAILED DESCRIPTION

[0052] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0053] Please refer to Figures 1-4 The present application provides a membrane lung and ventilator end CO distinguishing monitoring method, comprising:

[0054] S1, deploying a heterogeneous sensor network through a multi-modal sensing array at the membrane lung input end, the oxygenator output end and the ventilator loop branch point, collecting the membrane lung specific blood pump pulsatile waveform and the chest cavity motion harmonic characteristics under the driving of the ventilator, and performing space-time calibration on the heterogeneous sensing data to form a double-channel original signal pool;

[0055] S2, developing a dynamic decoupling algorithm based on a blood vessel tree-alveolar coupling model, and establishing a phase difference equation of the membrane lung roller pump speed and the ventilator positive pressure cycle;

[0056] S3, by real-time tracking of the blood pump impeller angular acceleration and the respiratory valve opening and closing time sequence, the pure mechanical blood flow impact component at the membrane lung end and the blood flow oscillation component induced by the chest cavity volume change at the ventilator end are reversely deduced;

[0057] S4, designing a parallel filter bank with a parameter self-correction function, and introducing the membrane lung end high-frequency turbulent noise and the ventilator end low-frequency ventilation disturbance into the corresponding Wiener filtering channels respectively;

[0058] S5, using the reference signals generated by the membrane lung blood flow predictor and the ventilator tidal volume estimator, dynamically updating the cut-off frequency of each channel, and ensuring that the cardiac output signals at the membrane lung end and the cardiac output signals at the ventilator end are completely separated in the frequency domain;

[0059] S6, eliminating common-mode interference through multi-sensor fusion, and using a blood vessel impedance-airway resistance joint matrix to distinguish the contribution degree of the anatomic dead space and the membrane lung effective perfusion area, and finally outputting double-channel cardiac output data, which includes the real-time absolute value of the cardiac output at the membrane lung end, the real-time absolute value of the cardiac output at the ventilator end, and the respective source proportion atlas.

[0060] Specifically, in the clinical scenario of membrane lung combined with ventilator treatment, first, a heterogeneous sensor network is built through a multi-modal sensor array, sensors are arranged at the input end of the membrane lung, the output end of the oxygenator, and the branch point of the ventilator loop, the pulsatile waveform of the membrane lung blood pump and the harmonic characteristics of the chest movement are collected, and then the spatio-temporal calibration of the heterogeneous sensor data is completed through a data fusion algorithm to form a double-channel original signal pool. When developing a dynamic decoupling algorithm based on the vascular tree-alveolar coupling model, a phase difference equation of the membrane lung roller pump speed and the positive pressure cycle of the ventilator is established in combination with the physiological parameters of the patient , where φ is the phase difference unit rad, k1 is the speed phase coefficient unit rad / (r / min), n is the membrane lung roller pump speed unit r / min, k2 is the positive pressure phase coefficient unit rad / kPa, P is the ventilator positive pressure unit kPa, and the interaction between the two is quantified through the equation. The blood pump impeller angular acceleration unit rad / s² and the breathing valve opening and closing timing are tracked in real time, and the pure mechanical blood flow impact component at the membrane lung end is derived according to the instantaneous flow rate change rate formula , where Qm' is the instantaneous flow rate change rate of the membrane lung unit L / min², is the angular acceleration flow conversion coefficient unit (L / min²) / (rad / s²); and the blood flow oscillation component at the ventilator end is derived in combination with the ventilator tidal volume and lung compliance. A parallel filter group with self-correcting parameters is designed, the noises at both ends are introduced into the corresponding Wiener filter channels, and the generated by the membrane lung blood flow predictor and the generated by the ventilator tidal volume estimator are used to dynamically adjust the cut-off frequency through the cut-off frequency update formula (membrane lung end) and (ventilator end), where is the cut-off frequency unit Hz, c and d are frequency mapping coefficients respectively Hz / (L / min) and Hz / L, is the membrane lung predicted blood flow unit L / min, is the ventilator estimated tidal volume unit L, to ensure complete separation in the frequency domain. Finally, common-mode interference is eliminated through multi-sensor fusion, and the contribution of the anatomic dead space and the membrane lung effective perfusion area is distinguished by means of the vascular impedance-airway resistance combined matrix, and the double-channel CO data is output. This way solves the problem that the prior art cannot distinguish the CO at both ends, and can accurately reflect the membrane lung efficiency and the patient's own circulation state.

[0061] In the embodiment, the deployment of the heterogeneous sensor network at the input end of the membrane lung, the output end of the oxygenator, and the branch point of the ventilator loop through the multi-modal sensor array comprises:

[0062] S11, deploying a first fiber-optic blood flow sensor at the input end of the membrane lung;

[0063] S12, deploying a second fiber-optic blood flow sensor and an ultrasonic Doppler probe at the output end of the oxygenator;

[0064] S13, deploying a differential pressure sensor at the branching point of the ventilator circuit;

[0065] S14, heterogeneous sensor data spatio-temporal calibration is achieved through a Kalman filtering algorithm based on data fusion.

[0066] Specifically, in the heterogeneous sensor network deployment, a first fiber-optic blood flow sensor is installed at the input end of the membrane lung to collect blood flow pulsation signals, a second fiber-optic blood flow sensor and an ultrasonic Doppler probe are simultaneously deployed at the output end of the oxygenator to double collect blood flow velocity and pulsation characteristics, and a differential pressure sensor is installed at the branching point of the ventilator circuit to capture airway pressure changes to reflect the harmonic characteristics of chest movement. After completing the sensor deployment, a Kalman filtering algorithm based on data fusion is used for spatio-temporal calibration of heterogeneous data. The calibration process follows the Kalman filtering state update formula , wherein is the calibrated data state vector at time k, is the state transition matrix, is the input matrix, is the sensor input parameter vector at time k-1, is the Kalman gain matrix at time k, is the sensor observation vector at time k, is the observation matrix. Through this algorithm, the deviation of different sensor data in time and space is eliminated, ensuring the synchronization and accuracy of the double-channel raw signal pool data. This embodiment improves the reliability of the sensor data and provides a high-quality data basis for subsequent CO signal processing.

[0067] In this embodiment, a dynamic decoupling algorithm is developed based on the blood vessel tree-alveolar coupling model, and a phase difference equation of the membrane lung roller pump speed and the ventilator positive pressure cycle is established, including:

[0068] S21, constructing a dynamic impedance model of the blood vessel tree, which includes blood flow inertia parameters and blood vessel wall elasticity parameters;

[0069] S22, constructing a gas exchange kinetics model of the alveoli, which includes ventilation pressure and lung compliance parameters;

[0070] S23, establishing a physical coupling equation of the membrane lung roller pump driving blood flow pulsation and the ventilator positive pressure ventilation inducing chest pressure change;

[0071] S24, determine the mathematical expression of the influence of the membrane lung roller pump speed change on the blood flow dynamics in the ventilator positive pressure cycle through the physical coupling equation.

[0072] Specifically, when constructing the vascular tree dynamic impedance model, the dynamic impedance formula is used , wherein Z is the vascular dynamic impedance unit kPa·min / L, R is the vascular resistance unit kPa·min / L, ω is the blood flow pulsation angular frequency unit rad / s, L is the blood flow inertia coefficient unit kPa·min² / L, and C is the vascular compliance unit L / kPa. The R, L and C parameters are obtained through clinical monitoring to fit the actual vascular state of the patient. When constructing the alveolar gas exchange dynamics model, the ventilation pressure and lung compliance parameters are combined to quantify the correlation between alveolar gas exchange and ventilation parameters. Based on the above two models, the physical coupling equation is established , wherein ΔP is the blood flow pressure change unit kPa, is the membrane lung blood flow unit L / min, D is the coupling coefficient unit kPa / kPa, is the ventilator positive pressure unit kPa, which describes the interaction between the membrane lung blood flow pulsation and the ventilator positive pressure induced thoracic pressure change through the equation, and further determines the mathematical expression of the influence of the membrane lung roller pump speed change on the blood flow dynamics in the ventilator positive pressure cycle. This method clearly quantifies the coupling relationship between the membrane lung and the ventilator, providing theoretical support for subsequent CO signal decoupling.

[0073] In the present embodiment, the pure mechanical blood flow impact component at the membrane lung end and the blood flow oscillation component induced by the thoracic volume change at the ventilator end are reversely deduced by tracking the blood pump impeller angular acceleration and the respiratory valve opening and closing timing in real time, including:

[0074] S31, based on the continuous measurement value of the blood pump impeller angular acceleration, calculate the instantaneous flow rate change rate of the membrane lung blood flow pumping, and generate the pure mechanical blood flow impact original component at the membrane lung end;

[0075] S32, based on the respiratory valve opening and closing timing and the tidal volume parameter set by the ventilator, and combined with the lung compliance parameter, estimate the periodic change of the thoracic volume, and generate the blood flow oscillation original component induced by the thoracic volume change at the ventilator end;

[0076] S33, by using the nonlinear least squares method, the pure mechanical blood flow impact original component at the membrane lung end and the blood flow oscillation original component at the ventilator end are orthogonally decomposed to remove the aliasing signal caused by the interaction.

[0077] Specifically, based on the continuous measurement value of the blood pump impeller angular acceleration, the instantaneous flow rate change rate formula is used The instantaneous flow rate change rate of the membrane lung blood flow pumping is calculated to generate a membrane lung end pure mechanical blood flow shock original component (wherein the parameter definitions are consistent with the foregoing). Based on the timing of the opening and closing of the breathing valve and the ventilator tidal volume parameter , in combination with the lung compliance , the periodic change in thoracic volume is estimated by the thoracic volume change formula , wherein ΔV is the thoracic volume change, e is the volume conversion coefficient (dimensionless), is the ventilator tidal volume, L, is the lung compliance, L / kPa, and a ventilator end blood flow oscillation original component is generated. The two-end original components are orthogonally decomposed by a nonlinear least squares method to eliminate the aliasing signals generated by the interaction, and the pure two-end blood flow components are obtained. This embodiment reduces the mutual interference of the two-end signals and improves the accuracy of blood flow component extraction.

[0078] In this embodiment, the design of the parallel filter bank with the parameter self-correction function comprises the following steps:

[0079] S41, a statistical model of the membrane lung end high-frequency turbulent flow noise is constructed, and the statistical model is based on the inherent vibration spectrum of the membrane lung blood pump;

[0080] S42, a statistical model of the ventilator end low-frequency ventilation disturbance is constructed, and the statistical model is based on the breathing frequency and the tidal volume change characteristics;

[0081] S43, the initial parameters of the Wiener filter channel are initialized according to the statistical model;

[0082] S44, in the system running process, the filter coefficients of the Wiener filter channel are dynamically adjusted according to the statistical characteristics of the real-time signals by the adaptive least mean square error algorithm.

[0083] Specifically, when the statistical model of the membrane lung end high-frequency turbulent flow noise is constructed, the noise distribution characteristics are determined based on the inherent vibration spectrum of the membrane lung blood pump; when the statistical model of the ventilator end low-frequency ventilation disturbance is constructed, the breathing frequency and the tidal volume change characteristics are combined. The initial parameters of the Wiener filter channel are initialized according to the two statistical models, and the filter coefficients are adjusted by the adaptive least mean square error algorithm in the system running process, and the filter coefficient updating formula is followed , wherein is the filter coefficient at k+1, is the filter coefficient at k, u is the adaptive step (dimensionless), is the filter error at k, V, is the input signal at k, V. The coefficients are dynamically adjusted by the formula to ensure that the filter channel can accurately filter the corresponding noise and disturbance. This way realizes the adaptive optimization of the filter and continuously maintains good filtering effect.

[0084] In the embodiment, the reference signal generated by the membrane lung blood flow predictor and the ventilator tidal volume estimator is used to dynamically update the cutoff frequency of each channel, including:

[0085] S51, the membrane lung blood flow predictor predicts the blood flow at the membrane lung end in real time based on the membrane lung roller pump speed and the pipe resistance model;

[0086] S52, the ventilator tidal volume estimator estimates the tidal volume of the ventilator in real time based on the differential pressure sensor data and the lung mechanics model;

[0087] S53, according to the predicted membrane lung blood flow and the estimated ventilator tidal volume, the cutoff frequency of the membrane lung end Wiener filtering channel and the ventilator end Wiener filtering channel is calculated and updated through the preset dynamic frequency mapping rule.

[0088] Specifically, the membrane lung blood flow predictor is based on the membrane lung roller pump speed n and the pipe resistance , the blood flow prediction formula is used to predict the blood flow at the membrane lung end in real time, wherein is the predicted blood flow unit L / min, f is the speed flow coefficient unit L / (r·min), n is the roller pump speed unit r / min, g is the resistance correction coefficient (dimensionless), is the pipe resistance unit kPa·min / L. The ventilator tidal volume estimator is based on the pressure change collected by the differential pressure sensor and the lung compliance , the tidal volume estimation formula is used to estimate the tidal volume in real time, wherein is the estimated tidal volume unit L, h is the air pressure tidal volume coefficient (dimensionless), is the airway pressure change unit kPa, is the lung compliance unit L / kPa. According to and , the cutoff frequency of the two-end Wiener filtering channel is calculated and updated through the preset dynamic frequency mapping rule to ensure the frequency domain separation effect. This embodiment realizes the dynamic adaptation of the cutoff frequency, further improving the CO signal separation accuracy.

[0089] In the embodiment, the contribution degree of the anatomic dead space and the membrane lung effective perfusion area is distinguished by using the vascular impedance-airway resistance joint matrix, including:

[0090] S61, a joint matrix containing anatomic dead space volume, alveolar perfusion volume and membrane lung blood flow perfusion volume is established;

[0091] The joint matrix is calibrated in real time through the parameters in the blood vessel tree-alveolar coupling model;

[0092] S62, separate the independent contribution of the anatomic dead space, the alveolar effective perfusion area and the membrane lung effective perfusion area to the total cardiac output from the joint matrix by singular value decomposition method.

[0093] Specifically, a joint matrix containing anatomic dead space volume , alveolar perfusion volume and membrane lung perfusion volume is established , wherein is the anatomic dead space volume unit L, is the alveolar perfusion volume unit L, is the membrane lung perfusion volume unit L. The joint matrix M is calibrated using real-time parameters in the vascular tree-alveolar coupling model (such as vascular compliance, lung compliance) to ensure that the matrix reflects the current physiological state. The calibrated matrix is decomposed by singular value decomposition method, and the decomposition formula is followed , wherein U is a left orthogonal matrix, is a diagonal singular value matrix, is a right orthogonal matrix transpose, and the independent contribution of , , to the total cardiac output is separated by decomposition. This embodiment enables doctors to clearly understand the influence of each region on CO, and helps disease judgment and device parameter adjustment.

[0094] The application also provides a membrane lung and ventilator end CO distinction monitoring system, comprising:

[0095] A multi-modal sensing array module is used to deploy a heterogeneous sensor network at the membrane lung input end, the oxygenator output end and the ventilator loop branch point, to collect the membrane lung specific blood pump pulsatile waveform and the chest cavity motion harmonic characteristics under the driving of the ventilator, and to perform spatio-temporal calibration on the heterogeneous sensing data to form a double-channel raw signal pool;

[0096] A physiological coupling calculation module is used to develop a dynamic decoupling algorithm based on the vascular tree-alveolar coupling model, to establish a phase difference equation of the membrane lung roller pump speed and the ventilator positive pressure cycle, and to inversely deduce the pure mechanical blood flow impact component at the membrane lung end and the blood flow oscillation component induced by the chest cavity volume change at the ventilator end by real-time tracking of the blood pump impeller angular acceleration and the breathing valve opening and closing timing;

[0097] A dual-domain adaptive filtering module is used to design a parallel filter bank with parameter self-correction function, to respectively introduce the membrane lung end high-frequency turbulent noise and the ventilator end low-frequency ventilation disturbance into the corresponding Wiener filtering channels, and to dynamically update the channel cutoff frequency using the reference signals generated by the membrane lung blood flow predictor and the ventilator tidal volume estimator, to ensure that the membrane lung end cardiac output signal and the ventilator end cardiac output signal are completely separated in the frequency domain.

[0098] The data fusion and contribution analysis module is used to eliminate common-mode interference through multi-sensor fusion and to distinguish the contribution of the anatomical dead space and the effective perfusion area of ​​the membrane lung using the vascular impedance-airway resistance joint matrix. Finally, it outputs dual-channel cardiac output data, which includes the real-time absolute value of cardiac output at the membrane lung end, the real-time absolute value of cardiac output at the ventilator end, and the respective source proportion map.

[0099] Specifically, the multimodal sensor array module deploys heterogeneous sensors to acquire signals, performs spatiotemporal calibration using the Kalman filter formula, and forms a dual-channel raw signal pool. The physiological coupling solution module is based on a vascular tree-alveolar coupling model, using a phase difference equation... Establish a correlation between membrane lung and ventilator parameters, combined with and The blood flow components at both ends are derived. A parallel filter bank is designed using a dual-domain adaptive filtering module, through... Adjust the filter coefficients, using and The generated reference signal is used to update the cutoff frequency, achieving frequency domain separation. The data fusion and contribution analysis module eliminates common-mode interference through multi-sensor fusion, leveraging... and The system differentiates the contribution of anatomical dead space and the effective perfusion area of ​​the membrane lung, outputting dual-channel CO data. It stably achieves differentiated CO monitoring at both ends, providing reliable support for clinical treatment.

[0100] The present invention also provides a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are to be invoked and executed by the processing component to implement the membrane lung and ventilator-side CO differentiation monitoring method described above.

[0101] Specifically, the computing device includes a processing component and a storage component, the storage component storing computer instructions written based on the method of this invention. When the processing component invokes the instructions, it first controls the multimodal sensor array to acquire data, and then uses the Kalman filter formula... Complete data calibration; then based on and Perform dynamic decoupling, combined with and Deducing blood flow components; then through Adjust the filter coefficients, using and Update cutoff frequency; finally approved The device analyzes contribution and outputs dual-channel CO data. It efficiently performs monitoring methods, meeting the needs of real-time clinical monitoring.

[0102] The application also provides a computer storage medium storing a computer program, which, when executed by a computer, implements the membrane lung and ventilator end CO distinguishing monitoring method.

[0103] Specifically, the computer storage medium stores a computer program corresponding to the monitoring method. When the computer reads and executes the program, it first receives multi-modal sensor data and calibrates the data through a Kalman filtering formula; then decouples based on a phase difference equation and a physical coupling equation, and deduces blood flow components in combination with a flow and volume formula; then optimizes the filtering effect through a filtering coefficient update formula and adjusts the cutoff frequency using a blood flow and tidal volume prediction formula; finally, it analyzes the contribution degree through joint matrix singular value decomposition and outputs the two-end CO data. The medium facilitates program deployment and migration, and improves the applicability of the method.

[0104] To sum up, the application collects membrane lung and ventilator related signals by deploying a heterogeneous sensor network composed of a multi-modal sensing array and performs time and space calibration, develops a dynamic decoupling algorithm in combination with a vascular tree-alveolar coupling model, deduces the pure mechanical blood flow impact component at the membrane lung end and the blood flow oscillation component at the ventilator end, and then uses a parallel filter group with a parameter self-correction function and a reference signal to dynamically update the cutoff frequency to achieve complete separation of the two-end CO signal in the frequency domain. Finally, through multi-sensor fusion, common-mode interference is eliminated, and with the help of joint matrix, the contribution degree of anatomical dead space and membrane lung effective perfusion area is distinguished. The application effectively solves the problem that the prior art cannot accurately distinguish the membrane lung end and the ventilator end CO, can output double-channel data containing real-time absolute values and source proportion atlas of the two-end CO, enables doctors to accurately grasp the membrane lung working efficiency and the patient's own circulation and breathing state, provides a reliable basis for timely and accurate adjustment of the treatment scheme, and helps to optimize the patient's treatment effect and disease stability.

[0105] It should be noted that, in this text, relational terms such as first and second are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "contain" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device.

[0106] Although embodiments of the application have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and alterations can be made thereto without departing from the principles and spirit of the application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for differentiating and monitoring CO levels between membrane lung and ventilator-mounted systems, characterized in that, include: S1. A heterogeneous sensor network is deployed at the membrane lung input end, oxygenator output end and ventilator loop branch point through a multimodal sensor array to collect the membrane lung-specific blood pump pulsation waveform and the harmonic characteristics of thoracic cavity motion under ventilator drive, and the heterogeneous sensor data is spatiotemporally calibrated to form a dual-channel raw signal pool. S2. Develop a dynamic decoupling algorithm based on the vascular tree-alveolar coupling model and establish the phase difference equation between the membrane lung roller pump speed and the positive pressure cycle of the ventilator; S3. By tracking the angular acceleration of the blood pump impeller and the opening and closing sequence of the breathing valve in real time, the purely mechanical blood flow impact component at the membrane lung end and the blood flow oscillation component induced by the change in thoracic cavity volume at the ventilator end are deduced in reverse. S4. Design a parallel filter bank with parameter self-correction function to import the high-frequency turbulence noise at the membrane lung end and the low-frequency ventilation disturbance at the ventilator end into the corresponding Wiener filter channels respectively. S5. Using the reference signals generated by the membrane lung blood flow predictor and the ventilator tidal volume estimator, the cutoff frequency of each channel is dynamically updated to ensure that the membrane lung end-cardiac output signal and the ventilator end-cardiac output signal are completely separated in the frequency domain. S6. Common-mode interference is eliminated by multi-sensor fusion, and the contribution of anatomical dead space and membrane lung effective perfusion area is distinguished by the vascular impedance-airway resistance joint matrix. Finally, dual-channel cardiac output data is output, which includes the real-time absolute value of cardiac output at the membrane lung end, the real-time absolute value of cardiac output at the ventilator end, and the respective source proportion map.

2. The method for distinguishing and monitoring CO at the membrane lung and ventilator end according to claim 1, characterized in that, The deployment of a heterogeneous sensor network at the membrane lung inlet, oxygenator outlet, and ventilator loop branch points via a multimodal sensor array includes: S11. Deploy the first fiber optic blood flow sensor at the inlet end of the membrane lung; S12. Deploy a second fiber optic blood flow sensor and an ultrasonic Doppler probe at the output end of the oxygenator; S13. Deploy differential pressure sensors at the branch points of the ventilator loop; S14. Spatiotemporal calibration of heterogeneous sensor data is achieved through a Kalman filter algorithm based on data fusion.

3. The method for distinguishing and monitoring CO at the membrane lung and ventilator end according to claim 1, characterized in that, The dynamic decoupling algorithm developed based on the vascular tree-alveolar coupling model establishes the phase difference equation between the membrane lung roller pump speed and the positive pressure cycle of the ventilator, including: S21. Construct a dynamic impedance model of the vascular tree, which includes blood flow inertia parameters and vascular wall elastic parameters. S22. Construct a gas exchange dynamics model of the alveoli, which includes ventilation pressure and lung compliance parameters; S23. Establish the physical coupling equation between blood flow pulsation driven by membrane lung roller pump and changes in intrathoracic pressure induced by positive pressure ventilation of ventilator; S24. Using physical coupling equations, determine the mathematical expression of the effect of membrane lung roller pump speed change on hemodynamics during the positive pressure cycle of the ventilator.

4. The method for distinguishing and monitoring CO at the membrane lung and ventilator end according to claim 3, characterized in that, The method involves real-time tracking of the blood pump impeller angular acceleration and the opening and closing sequence of the breathing valve to deduce, inversely, the purely mechanical blood flow impact component at the membrane lung end and the blood flow oscillation component induced by changes in thoracic cavity volume at the ventilator end, including: S31. Based on the continuous measurement of the angular acceleration of the blood pump impeller, calculate the instantaneous flow rate change rate of the membrane lung blood flow pump and generate the original component of the pure mechanical blood flow impact at the membrane lung end. S32. Based on the opening and closing timing of the breathing valve and the tidal volume parameters set by the ventilator, combined with the lung compliance parameters, estimate the periodic changes in thoracic volume and generate the original components of blood flow oscillation induced by the changes in thoracic volume at the ventilator end. S33. Using the nonlinear least squares method, the original components of the pure mechanical blood flow impact at the membrane lung end and the original components of the blood flow oscillation at the ventilator end are orthogonally decomposed to remove the aliasing signal caused by the interaction.

5. The method for distinguishing and monitoring CO at the membrane lung and ventilator end according to claim 1, characterized in that, The design of the parallel filter bank with parameter self-correction function includes: S41. Construct a statistical model for high-frequency turbulent noise at the membrane lung end. The statistical model is based on the inherent vibration spectrum of the membrane lung blood pump. S42. Construct a statistical model of low-frequency ventilation disturbances at the ventilator end. The statistical model is based on the characteristics of changes in respiratory rate and tidal volume. S43. Initialize the initial parameters of the Wiener filter channel according to the statistical model; S44. During system operation, the filter coefficients of the Wiener filter channel are dynamically adjusted based on the statistical characteristics of the real-time signal using an adaptive minimum mean square error algorithm.

6. The method for distinguishing and monitoring CO at the membrane lung and ventilator end according to claim 1, characterized in that, The reference signals generated by the membrane lung blood flow predictor and the ventilator tidal volume estimator are used to dynamically update the cutoff frequency of each channel, including: S51, The membrane lung blood flow predictor is based on the membrane lung roller pump speed and pipeline resistance model to predict the blood flow at the membrane lung end in real time. S52, The tidal volume estimator of the ventilator estimates the tidal volume of the ventilator in real time based on differential pressure sensor data and lung mechanics model; S53. Based on the predicted membrane lung blood flow and the estimated ventilator tidal volume, calculate and update the cutoff frequencies of the membrane lung Wiener filter channel and the ventilator Wiener filter channel through a preset dynamic frequency mapping rule.

7. The method for distinguishing and monitoring CO at the membrane lung and ventilator end according to claim 1, characterized in that, The contribution of the vascular impedance-airway resistance joint matrix to distinguish the effective perfusion area of ​​anatomical dead space and membrane lung includes: S61. Establish a joint matrix that includes anatomical dead space volume, alveolar perfusion volume, and membrane lung blood flow perfusion volume; The joint matrix is ​​calibrated in real time using parameters from the vascular tree-alveolar coupling model. S62. Using singular value decomposition, the independent contributions of the anatomical dead space, the effective alveolar perfusion region, and the effective membrane lung perfusion region to total cardiac output are separated from the joint matrix.

8. A membrane lung and ventilator-based CO differentiation monitoring system, characterized in that, include: The multimodal sensor array module is used to deploy a heterogeneous sensor network at the membrane lung input end, oxygenator output end and ventilator loop branch point to collect the membrane lung-specific blood pump pulsation waveform and the harmonic characteristics of thoracic cavity motion under ventilator drive, and to perform spatiotemporal calibration on heterogeneous sensor data to form a dual-channel raw signal pool. The physiological coupling solution module is used to develop a dynamic decoupling algorithm based on the vascular tree-alveolar coupling model, establish the phase difference equation between the speed of the membrane lung roller pump and the positive pressure cycle of the ventilator, and derive the pure mechanical blood flow impact component at the membrane lung end and the blood flow oscillation component induced by the change in thoracic cavity volume at the ventilator end by tracking the blood pump impeller angular acceleration and the opening and closing sequence of the breathing valve in real time. The dual-domain adaptive filtering module is used to design a parallel filter bank with parameter self-correction function. It imports the high-frequency turbulence noise at the membrane lung end and the low-frequency ventilation disturbance at the ventilator end into the corresponding Wiener filter channels respectively. It also uses the reference signals generated by the membrane lung blood flow predictor and the ventilator tidal volume estimator to dynamically update the cutoff frequency of each channel, ensuring that the cardiac output signal at the membrane lung end and the cardiac output signal at the ventilator end are completely separated in the frequency domain. The data fusion and contribution analysis module is used to eliminate common-mode interference through multi-sensor fusion and to distinguish the contribution of the anatomical dead space and the effective perfusion area of ​​the membrane lung using the vascular impedance-airway resistance joint matrix. Finally, it outputs dual-channel cardiac output data, which includes the real-time absolute value of cardiac output at the membrane lung end, the real-time absolute value of cardiac output at the ventilator end, and the respective source proportion map.

9. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement the membrane lung and ventilator-side CO differentiation monitoring method as described in any one of claims 1-7.

10. A computer storage medium, characterized in that, The device contains a computer program that, when executed by a computer, implements the method for distinguishing and monitoring CO at the membrane lung and ventilator end as described in any one of claims 1-7.

Images