Ventilator pressure estimation method and device, ventilator and medium

Abstract

The present application relates to the technical field of medical apparatus and instruments, and particularly relates to a ventilator pressure estimation method and device, a ventilator and a medium, which comprises the following steps: obtaining a respiratory airway signal of a target object in a preset period, determining a starting reference point and a terminal reference point of respiratory muscle pressure activity based on the respiratory airway signal, constructing a two-dimensional search grid according to the starting reference point and the terminal reference point of the respiratory muscle pressure activity, performing least square fitting solution on each candidate combination in the two-dimensional search grid by using a plurality of constraint conditions, obtaining a respiratory parameter estimation value corresponding to each candidate combination, and determining a respiratory muscle pressure estimation result according to the respiratory parameter estimation value corresponding to each candidate combination. It can be seen that the respiratory muscle activity interval is accurately divided by the time reference point, each candidate combination in the two-dimensional search grid is combined, and the plurality of constraint conditions are used for solution, so that the accuracy of the respiratory muscle pressure estimation is improved, the deviation of the estimation result is reduced, and the demand for accurate monitoring of the respiratory muscle pressure in the clinic is met.

Description

Technical Field

[0001] This invention relates to the field of medical device technology, and in particular to a method, device, ventilator, and medium for estimating ventilator pressure. Background Technology

[0002] With the continuous rise in the incidence of respiratory diseases, mechanical ventilation is an important means of respiratory support in intensive care and anesthesia. Respiratory muscle pressure (the pressure generated by the contraction of respiratory muscles, usually denoted as Pmus) is a key physiological parameter reflecting the patient's respiratory effort and the state of the ventilator-patient interaction. Accurate acquisition of respiratory muscle pressure helps clinicians adjust ventilator parameters in a timely manner, avoid respiratory muscle fatigue or over-support, and thus improve patient prognosis.

[0003] However, existing respiratory muscle pressure estimation methods in clinical practice mostly rely on end-inspiratory occlusion. But in complex scenarios, these methods do not fully consider the dynamic changes in airway mechanics during the patient's breathing process. When the patient's breathing effort is weak or there is interference in the airway signal, the final pressure estimation result is prone to large deviations, making it difficult to meet the clinical need for accurate monitoring of respiratory muscle pressure. Therefore, how to accurately estimate respiratory muscle pressure to meet the clinical need for accurate monitoring of respiratory muscle pressure is a technical problem that urgently needs to be solved. Summary of the Invention

[0004] Therefore, in order to address the above-mentioned technical problems, this invention provides a method, device, ventilator, and medium for estimating ventilator pressure, which can accurately estimate respiratory muscle pressure to meet the clinical need for precise monitoring of respiratory muscle pressure.

[0005] A first aspect of this application provides a method for estimating ventilator pressure, the method comprising: Acquire the respiratory airway signal of the target object within a preset period, and determine the time reference point of respiratory muscle pressure activity based on the respiratory airway signal, wherein the time reference point includes a start reference point and an end reference point.

[0006] A two-dimensional search grid is constructed based on the starting and ending reference points of the respiratory muscle pressure activity, wherein the two-dimensional search grid includes multiple candidate combinations.

[0007] By using multi-layered constraints, the least squares fitting solution is performed on each candidate combination in the two-dimensional search grid to obtain the estimated respiratory parameters corresponding to each candidate combination.

[0008] The respiratory muscle pressure estimation result is determined based on the estimated respiratory parameters corresponding to each of the candidate combinations.

[0009] A second aspect of this application provides a ventilator pressure estimation device, the ventilator pressure estimation device comprising: The acquisition module is used to acquire the respiratory airway signal of the target object within a preset period, and determine the time reference point of respiratory muscle pressure activity based on the respiratory airway signal, wherein the time reference point includes a start reference point and an end reference point.

[0010] A construction module is used to construct a two-dimensional search grid based on the start and end reference points of the respiratory muscle pressure activity, wherein the two-dimensional search grid includes multiple candidate combinations.

[0011] The solution module is used to perform least-squares fitting on each candidate combination in the two-dimensional search grid using multi-layer constraints to obtain the estimated respiratory parameters corresponding to each candidate combination.

[0012] The determination module is used to determine the respiratory muscle pressure estimation result based on the estimated respiratory parameters corresponding to each of the candidate combinations.

[0013] Thirdly, a ventilator is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the ventilator pressure estimation method as described in the first aspect.

[0014] Fourthly, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program that, when executed by a processor, implements the ventilator pressure estimation method as described in the first aspect.

[0015] In summary, this invention provides a method, device, ventilator, and medium for estimating ventilator pressure. It acquires the respiratory airway signal of a target subject within a preset period and determines the time reference points for respiratory muscle pressure activity based on these signals. The time reference points include a start reference point and an end reference point. A two-dimensional search grid is constructed based on these start and end reference points, comprising multiple candidate combinations. Least squares fitting is performed on each candidate combination within the two-dimensional search grid using multi-level constraints to obtain estimated respiratory parameters for each combination. The respiratory muscle pressure estimation result is then determined based on these estimated values. Therefore, this application accurately divides the respiratory muscle activity range using time reference points, combines this with candidate combinations within a two-dimensional search grid, and utilizes multi-level constraints to improve the accuracy of respiratory muscle pressure estimation, reduce estimation deviations, meet the clinical need for precise monitoring of respiratory muscle pressure, and provide reliable data support for the dynamic adjustment of ventilator parameters. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating a ventilator pressure estimation method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of a ventilator pressure estimation device according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of a ventilator provided in an embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] It should be understood that, when used in this specification and the appended claims, terms include indicating the presence of the described feature, integral, step, operation, element and / or component, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0020] It should also be understood that the terms used in this specification and the appended claims refer to any combination of one or more of the associated listed items and all possible combinations, and include such combinations.

[0021] As used in this specification and the appended claims, terms if can be interpreted in context as when... or once or in response to determination. Similarly, the phrase if determined or if matched to [described condition or event] can be interpreted in context as once determined or in response to determination or once matched to [described condition or event] or in response to matching to [described condition or event].

[0022] Furthermore, in the description of this invention and the appended claims, the terms first, second, third, etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance.

[0023] References to one or more embodiments described in this specification mean that a particular feature, structure, or characteristic described in connection with that embodiment is included in one or more embodiments of the invention. Therefore, phrases appearing in different parts of this specification as referring to one embodiment, some embodiments, some other embodiments, and others do not necessarily refer to the same embodiment, but rather mean one or more, but not all, embodiments, unless otherwise specifically emphasized. The terms include, comprise, have, and variations thereof mean including but not limited to, unless otherwise specifically emphasized.

[0024] It should be understood that the sequence number of each step in the following embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0025] To illustrate the technical solution of the present invention, specific embodiments are described below.

[0026] See Figure 1 This is a flowchart illustrating a ventilator pressure estimation method according to an embodiment of the present invention, as shown below. Figure 1 As shown, the ventilator pressure estimation method can be implemented through the following steps.

[0027] S201: Acquire the respiratory airway signal of the target object within a preset period, and determine the time reference point of respiratory muscle pressure activity based on the respiratory airway signal, wherein the time reference point includes a start reference point and an end reference point.

[0028] In one implementation, the respiratory airway signals mainly include airway pressure, gas flow rate, and gas volume. The airway pressure signal is acquired by a pressure sensor placed at the Y-tube of the ventilator, reflecting the pressure change at the patient's airway opening. The airway flow rate signal is acquired by a differential pressure flow sensor or a thermal flow sensor, reflecting the instantaneous flow velocity of gas entering and leaving the patient's lungs. The gas volume signal can be obtained by integrating the instantaneous flow velocity, reflecting the total gas volume entering and leaving the lungs in each respiratory cycle. The coordinated acquisition of these three types of signals can comprehensively cover the changes in mechanical characteristics during the respiratory process. It should be noted that the sampling frequency of the above signals is synchronized with the control cycle of the ventilator, typically not lower than 100Hz, to ensure the temporal resolution of subsequent time reference point detection and parameter fitting. Furthermore, the waveform of the respiratory airway signal is parameterized using a sinusoidal basis function. The preset cycle can cover one or more consecutive respiratory cycles; for example, it can be set to continuously acquire data for several respiratory cycles to ensure that complete inspiratory and expiratory phases are included. Respiratory muscle pressure (Pmus) refers to the pressure generated by the contraction of the respiratory muscle group, especially the diaphragm. According to the respiratory system motion equation, airway pressure is the superposition of elastic pressure, resistance pressure, positive end-expiratory pressure, and respiratory muscle pressure. However, in spontaneously breathing patients, Pmus cannot be directly measured non-invasively and must be obtained through indirect estimation. The initial reference point corresponds to the moment when inspiratory muscle activity begins, and the termination reference point corresponds to the moment when inspiratory muscle activity ends. Accurate definition of these two time reference points is fundamental to subsequent parameter fitting. In practice, the acquired respiratory airway signals are first preprocessed using bandpass filtering and baseline removal to eliminate interference from environmental noise and sensor drift. Then, a signal feature detection algorithm is used to determine the two aforementioned time reference points.

[0029] Furthermore, based on the preprocessed respiratory airway signals, the system determines the time reference points for respiratory muscle pressure activity. These time reference points are used to accurately divide the time axis into the intervals of active contraction of the patient's respiratory muscles generating pressure and the intervals of respiratory muscle relaxation and passive lung recoil. Specifically, the starting reference point marks the moment when the respiratory muscles begin to actively exert force and the characteristic changes in airway pressure driven by the respiratory muscles appear; the ending reference point marks the moment when the active exertion of the respiratory muscles ends and the system enters the fully passive exhalation phase. The time period between the starting and ending reference points is the active segment of respiratory muscle pressure activity, and the time period from the ending reference point to the start of the next inspiration is the passive exhalation segment. Accurately determining these two time reference points is a prerequisite for subsequently distinguishing between the active and passive segments and applying different constraints for parameter fitting. Through these steps, accurate acquisition of respiratory airway signals and precise division of the respiratory muscle activity time intervals are achieved, laying a reliable time benchmark for subsequent two-dimensional search mesh construction and multi-layer constraint parameter solving, thereby ensuring the accuracy and robustness of the entire respiratory muscle pressure estimation process.

[0030] S202: Construct a two-dimensional search grid based on the starting and ending reference points of the respiratory muscle pressure activity, wherein the two-dimensional search grid includes multiple candidate combinations.

[0031] In one implementation, a two-dimensional search grid is constructed based on the starting and ending reference points of the respiratory muscle pressure activity, with each grid node corresponding to a set of candidate combinations. The two dimensions of this two-dimensional search grid correspond to two key uncertainties affecting the accuracy of respiratory muscle pressure estimation. These could be, for example, the time offset of the starting reference point relative to the detection result and the time offset of the ending reference point relative to the detection result, or candidate values ​​for airway resistance parameters and elastic resistance parameters. The specific dimension selection is determined based on the actual algorithm design and is not limited here. In each dimension, the system expands outwards by several discrete values ​​at a preset step size, centered on the detected reference point or initial parameter estimate, forming a sequence of candidate values ​​for that dimension. The candidate value sequences of the two dimensions are combined using a Cartesian product to obtain all candidate combinations in the two-dimensional search grid. Each candidate combination corresponds to a set of time offsets or parameter values ​​to be verified.

[0032] Furthermore, the granularity of the two-dimensional search grid is set by a trade-off between the number of candidate values ​​and the step size in each dimension, based on computational resources and accuracy requirements. A smaller step size and more candidate values ​​result in higher search resolution for parameter fitting, but also increase computational cost; a larger step size and fewer candidate values ​​result in faster computation, but may miss the optimal solution. This grid resolution can be flexibly adjusted according to hardware performance, balancing estimation accuracy and computational cost as needed. Each candidate undergoes a complete constraint fitting once, and the optimal candidate is selected through comprehensive scoring. Through these steps, a two-dimensional search space covering key uncertainties is constructed, transforming the respiratory muscle pressure estimation problem into an optimization problem of finding the optimal solution among multiple candidate combinations. This provides a structured search framework for subsequent multi-layered constrained least squares solutions, facilitating robust approximation of true respiratory parameters under conditions of noise and parameter uncertainty.

[0033] S203: Using multi-layer constraints, perform least squares fitting to solve for each candidate combination in the two-dimensional search grid to obtain the estimated respiratory parameters corresponding to each candidate combination.

[0034] In one implementation, for any candidate combination, the signal segment between the candidate start time and the candidate end time is taken as the fitting window, and a least squares objective function is constructed based on the respiratory system motion equation. The respiratory parameters to be estimated include elastic resistance, airway resistance, and respiratory muscle pressure waveform parameters. The multi-layered constraints include physiological boundary box constraints, passive segment independence constraints, full-window passive lower bound constraints, time constant coupling constraints, and waveform penalty constraints. These constraints collectively define the feasible region of the solution space. Among them, the physiological boundary box constraint is a≤0, b∈(−∞,+∞), and its R and E boundaries are set differently according to patient type (adult / child / infant). The specific range can refer to the recommended values ​​in medical literature and be adjusted in combination with clinical practice. The passive segment independence constraint is to take several points at the end of inspiration for independent estimation, apply tight constraints under static conditions, and tighten the boundary with weighted least squares under dynamic conditions. The full-window passive lower bound constraint is to take an appropriate proportion (such as 80%) of the full-window 2×2 pure passive solution as the R / E lower bound. The bidirectional coupling constraint is to cross-tighten the expiratory segment time constant box constraint and trigger matrix dimensionality reduction and resolution when there is a serious deviation. The waveform penalty constraint is to perform sampling checks and endpoint derivatives, and add a quadratic penalty term when violated, and iteratively tighten it. The five types of constraints each perform different physiological verification functions: the physiological boundary box constraint limits parameters such as elastic resistance and airway resistance to known physiologically reasonable ranges, preventing the fitting results from deviating from the actual mechanical characteristics of the human body; the passive segment independent constraint ensures that respiratory muscle pressure is zero during periods of clear muscle activity in the expiratory or inspiratory phases, avoiding misestimation of spurious muscle activity; the full-window passive lower bound constraint prevents the fitting of non-physiological negative pressure baseline drift, ensuring that the respiratory muscle pressure waveform is always above a reasonable baseline; the time constant coupling constraint uses the expiratory time constant to cross-validate the inspiratory parameters, using the time consistency of respiratory system mechanics to test the internal self-consistency of parameter estimation; and the waveform penalty constraint suppresses drastic non-physiological fluctuations in the estimated waveform, making the respiratory muscle pressure waveform smooth and consistent with the physiological characteristics of muscle contraction.

[0035] Furthermore, by iteratively solving the constrained least squares problem using the effective set method, the estimated respiratory parameters corresponding to the candidate combination are obtained. This process is performed separately for each candidate combination in the two-dimensional search grid, ensuring that all possible time window assumptions are fully evaluated. The effective set method is an optimization algorithm that handles box constraints (parameters have upper and lower bounds). By fixing the out-of-bounds parameters to the boundary values, a reduced-order submatrix is ​​constructed for the remaining free parameters and the solution is re-solved, transferring the coupling effect to the new optimal value. The direction of the Lagrange multiplier is used to determine whether the bounded constraint is released—if the multiplier points into the feasible region, it indicates that the constraint should not continue to be activated, and the solution is re-solved after release. Iteration continues until the KKT conditions (necessary conditions for optimality of the constrained optimization problem) are met. Through the above steps, accurate fitting of respiratory mechanics parameters within a physiologically reasonable range is achieved, thereby transforming non-invasive airway signals into clinically meaningful respiratory muscle pressure and resistance parameters. The introduction of multi-layer constraints effectively eliminates non-physical solutions, significantly improving the reliability and accuracy of parameter estimation, and providing a reliable basis for clinical decision-making.

[0036] S204: Determine the respiratory muscle pressure estimation result based on the estimated respiratory parameters corresponding to each of the candidate combinations.

[0037] In one implementation, after obtaining the estimated respiratory parameters for each candidate combination in the two-dimensional search grid, the system comprehensively evaluates the fitting quality of each candidate combination, selects the optimal candidate combination, and determines the final respiratory muscle pressure estimation result based on the estimated respiratory parameters corresponding to this candidate combination. Since different candidate combinations correspond to different time reference point offsets or initial parameter values, the fitted respiratory parameters and respiratory muscle pressure waveforms differ in accuracy and reasonableness. Therefore, multi-dimensional quality indicators are needed for screening and ranking.

[0038] Furthermore, the system calculates multiple quality assessment metrics for each candidate combination, including but not limited to the overall cost of the fitting residuals, the autocorrelation degree of the residual sequence, and the residual level of the passive expiratory segment. The fitting cost reflects the overall fit of the model to the full-window airway pressure signal; the residual autocorrelation reflects whether there are systematic components in the residuals that are not explained by the model; the lower the autocorrelation, the more adequate the model fit; the passive segment residual reflects the deviation between the pressure predicted by the model based on the kinetic parameters and the measured pressure during the passive expiratory segment when the respiratory muscles are fully relaxed; the smaller this deviation, the more accurate the passive segment parameter estimation. Based on these metrics, the system scores and ranks each candidate combination, selects the candidate combination with the best score as the final solution, and outputs the corresponding respiratory muscle pressure waveform as the estimation result. Through these steps, the system automatically identifies the optimal estimation result from multiple candidate solutions, avoiding local optima or overfitting problems that may be caused by a single fitting metric, ensuring that the final output respiratory muscle pressure estimation result is optimal, and providing a quantitative indicator that can be directly used for assessment and decision-making in clinical practice.

[0039] In summary, this invention provides a method, device, ventilator, and medium for estimating ventilator pressure. It acquires the respiratory airway signal of a target subject within a preset period and determines the time reference points for respiratory muscle pressure activity based on these signals. The time reference points include a start reference point and an end reference point. A two-dimensional search grid is constructed based on these start and end reference points, comprising multiple candidate combinations. Least squares fitting is performed on each candidate combination within the two-dimensional search grid using multi-level constraints to obtain estimated respiratory parameters for each combination. The respiratory muscle pressure estimation result is then determined based on these estimated values. Therefore, this application accurately divides the respiratory muscle activity range using time reference points, combines this with candidate combinations within a two-dimensional search grid, and utilizes multi-level constraints to improve the accuracy of respiratory muscle pressure estimation, reduce estimation deviations, meet the clinical need for precise monitoring of respiratory muscle pressure, and provide reliable data support for the dynamic adjustment of ventilator parameters.

[0040] In one embodiment, specifically in step S201, which involves determining the time reference point for respiratory muscle pressure activity based on the respiratory airway signal, the following steps are included: Acquire the airway pressure signal from the respiratory airway signal; The airway pressure signal is processed using a sliding window variance comparison algorithm, and the airway pressure signal is divided into a left half-window and a right half-window. Calculate the first standard deviation and the second standard deviation corresponding to the left half-window and the right half-window, respectively; When the first standard deviation of the left half-window is lower than the preset left standard deviation threshold, the second standard deviation of the right half-window is higher than the preset right standard deviation threshold, and the downward trend of the pressure at the endpoint of the right half-window exceeds the preset trend threshold, the corresponding window center is taken as a candidate reference point, and the starting reference point is determined from the candidate starting points.

[0041] In one implementation, the detection of the initial reference point is based on a sliding window variance comparison algorithm for airway pressure signals. First, the airway pressure signal is extracted from the respiratory airway signal, reflecting the change in pressure at the patient's airway opening over time. Before the patient's respiratory muscles begin active contraction, the airway pressure is mainly maintained by the ventilator's delivery pressure, resulting in a relatively stable signal with minimal fluctuations. When the respiratory muscles begin active contraction, the pressure they generate is superimposed on the airway pressure, causing characteristic fluctuations or a downward trend in the airway pressure signal, significantly increasing the signal dispersion. The sliding window variance comparison algorithm utilizes this change in signal statistical characteristics to capture the onset of respiratory muscle activity. Specifically, a sliding time window is set on the airway pressure signal, sliding point by point along the time axis with the current moment as the center. Each sliding window is evenly divided into two sub-windows: a left half-window corresponding to a period before the current moment, and a right half-window corresponding to a period after the current moment. The first standard deviation of the airway pressure signal in the left half-window and the second standard deviation of the airway pressure signal in the right half-window were calculated separately. The standard deviation is used to measure the degree of fluctuation of the signal within the sub-window. When the respiratory muscles have not yet started to move, the signals in both half-windows are relatively stable, and both standard deviations are at low levels. When the center of the sliding window just passes through the starting point of respiratory muscle activity, the left half-window is still in the stable segment before activity, and the first standard deviation is low, while the right half-window has entered the fluctuating segment after activity, and the second standard deviation increases significantly.

[0042] Furthermore, to avoid misjudging random noise spikes in the signal as starting reference points, the algorithm also introduces the pressure drop trend at the right half-window endpoint as an additional criterion. Specifically, the slope of pressure change at several sampling points near the right end of the window within the right half-window is calculated. If the slope is negative and its absolute value exceeds a preset trend threshold, it indicates a significant pressure drop trend at the right half-window endpoint, which is consistent with the physiological characteristic of airway pressure drop caused by active contraction of the respiratory muscles. Only when all three conditions are met simultaneously—the first standard deviation of the left half-window being lower than a preset left standard deviation threshold, the second standard deviation of the right half-window being higher than a preset right standard deviation threshold, and the pressure drop trend at the right half-window endpoint exceeding a preset trend threshold—is the center of the current sliding window marked as a candidate starting reference point. That is, the window center that simultaneously satisfies the following conditions—first standard deviation lower than a preset left standard deviation threshold (stable segment), second standard deviation of the right half-window being higher than a preset right standard deviation threshold (disturbance zone), and pressure drop trend at the right half-window endpoint exceeding a preset trend threshold (downward direction)—is listed as a candidate reference point, and then the one closest to the current PEEP is selected from the candidate reference points. Finally, the candidate point closest to the trigger time is selected from the candidate points within the tolerance range as the optimal reference point, and this optimal reference point is used as the starting reference point. If no candidate passes in three rounds, the system reverts to a PEEP stability reverse scan as a fallback strategy: backtracking step by step from near the trigger time in the historical direction, calculating the pressure deviation for each step, and if the pressure deviation is lower than the preset stability threshold for several consecutive steps (e.g., 4 steps), this area is considered a stable end-expiratory baseline area, and the starting point of this stable segment is taken as the starting reference point. It should be noted that the stability threshold and the number of consecutive steps are adjustable parameters, and this application does not impose any limitations on them. Multiple candidate starting reference points may be detected within a respiratory cycle, and the system selects the candidate point that appears earliest or has the highest comprehensive score as the final starting reference point. Through the above steps, by using a sliding window variance comparison combined with a multi-condition joint criterion of pressure decrease trend, accurate and robust detection of the starting point of respiratory muscle activity is achieved, effectively distinguishing between active respiratory muscle exertion and random noise interference, and providing a reliable starting time benchmark for the accurate division of the subsequent active and passive segments.

[0043] In one embodiment, specifically in step S201, which involves determining the time reference point for respiratory muscle pressure activity based on the respiratory airway signal, the following step is further included: Real-time feature extraction is performed on the respiratory airway signals to obtain multi-dimensional feature data; Within a preset sliding time window, dynamic zero-crossing detection and amplitude attenuation evaluation are performed on the multi-dimensional feature data to generate candidate termination reference points; The hysteresis threshold is dynamically calculated based on the historical distribution of the candidate termination reference points within a continuous respiratory cycle. The validity of the candidate termination reference points is verified based on the hysteresis threshold to obtain the termination reference points for respiratory muscle pressure activity.

[0044] In one implementation, real-time feature extraction is performed on the respiratory airway signals to obtain multi-dimensional feature data. This multi-dimensional feature data is extracted from airway pressure signals, airway flow signals, and gas volume signals, including but not limited to the rate of change of airway pressure, the zero-crossing state of airway flow, the peak value of volume, and its decay rate. These multi-dimensional features reflect the patient's transition from the inspiratory phase to the expiratory phase and the gradual relaxation of respiratory muscles from active contraction. When the respiratory muscles are still in the active contraction phase, the flow rate is positive and the tidal volume continues to rise, and the airway pressure exhibits active-driven characteristics. When the respiratory muscles gradually relax and enter the passive exhalation phase, the flow rate changes from positive to negative, the tidal volume begins to decrease from its peak, and the airway pressure gradually returns to kinetic equilibrium. Within a preset sliding time window, dynamic zero-crossing detection and amplitude decay assessment are performed on the above multi-dimensional feature data to generate candidate termination reference points. Dynamic zero-crossing detection monitors the moment when the airway flow signal changes from positive to negative, typically corresponding to the transition from the end of inspiration to the beginning of expiration. Amplitude decay assessment monitors the rate of tidal volume decline from its peak and the degree of decay of the active airway pressure component. When a zero-crossing is detected and both tidal volume and the active airway pressure component show a continuous decay trend, this moment is marked as a candidate termination reference point. Due to differences in breathing patterns among patients and potential interference in the signal, a single detection may generate multiple candidate termination reference points or false alarm points.

[0045] Furthermore, the hysteresis threshold is dynamically calculated based on the historical distribution of candidate termination reference points over consecutive respiratory cycles. The historical distribution refers to the statistical characteristics of the termination reference point positions over the most recent several respiratory cycles (e.g., 5 to 10 respiratory cycles), such as the mean, median, and standard deviation. These statistical characteristics reflect the stability of the patient's current breathing pattern. The time delay distribution of the occurrence time of the termination reference point relative to the peak inspiratory time over several past consecutive respiratory cycles is statistically analyzed, and the mean and standard deviation of this distribution are calculated. The upper and lower bounds of the hysteresis threshold are then dynamically determined based on the mean and standard deviation. The hysteresis threshold sets a time window with hysteresis characteristics for determining the termination reference point. Only when a candidate termination reference point falls within this time window and meets the amplitude decay condition is it considered valid. If the candidate point is earlier than the lower hysteresis bound, inspiration is considered not yet complete, constituting a false detection; if the candidate point is later than the upper hysteresis bound, the normal termination point has been missed, constituting a delayed detection. The validity of candidate termination reference points is verified based on the hysteresis threshold, eliminating abnormal candidate points that are too early or too late, ultimately yielding the termination reference point for respiratory muscle pressure activity. Through the above steps, candidate points are generated using dynamic zero-crossing detection and amplitude attenuation assessment based on multi-dimensional features, and validity is verified by combining the hysteresis threshold dynamically calculated from the historical distribution of continuous respiratory cycles. This achieves adaptive and robust detection of respiratory muscle activity termination points, effectively suppressing false positives and false negatives caused by signal fluctuations within a single respiratory cycle, ensuring the continuity and stability of the termination reference point across consecutive respiratory cycles, and providing a reliable time boundary for accurate reconstruction of respiratory muscle pressure waveforms.

[0046] In one embodiment, specifically step S203, which involves using multi-layered constraints to perform least-squares fitting on each candidate combination in the two-dimensional search grid to obtain the estimated respiratory parameters corresponding to each candidate combination, includes the following steps: The multi-layered constraint conditions include physiological bounding box constraints, passive segment independence constraints, full-window passive lower bound constraints, time constant coupling constraints, and waveform penalty constraints.

[0047] The physiological bounding box constraint is used to constrain each candidate combination in the two-dimensional search grid to obtain the first respiratory parameter constraint value corresponding to each candidate combination; By using the passive segment independent constraint to constrain each candidate combination in the two-dimensional search grid, the second breathing parameter constraint value corresponding to each candidate combination is obtained; The candidate combinations in the two-dimensional search grid are constrained by the full-window passive lower bound constraint to obtain the third breathing parameter constraint value corresponding to each candidate combination. The time constant coupling constraint is used to constrain each candidate combination in the two-dimensional search grid to obtain the fourth breathing parameter constraint value corresponding to each candidate combination; The waveform penalty constraint is used to constrain each candidate combination in the two-dimensional search grid to obtain the fifth breathing parameter constraint value corresponding to each candidate combination; The first, second, third, fourth, and fifth respiratory parameter constraint values ​​are solved by least squares fitting using the preset effective set method to obtain the estimated respiratory parameter values ​​corresponding to each candidate combination.

[0048] In one implementation, the multi-layered constraint conditions consist of five independent yet cooperative constraint layers. Each layer constrains and filters the respiratory parameter solutions. For each candidate combination in the two-dimensional search grid, least-squares fitting is performed using the multi-layered constraint conditions to obtain the estimated respiratory parameters for each candidate combination. The first layer is a physiological bounding box constraint. This constraint applies to each estimated respiratory parameter, including airway resistance, elastic resistance, and positive end-expiratory pressure (PEEP), setting upper and lower bounds that conform to common human respiratory physiology, forming a multi-dimensional parameter bounding box. For example, airway resistance should be within the reasonable range of normal human airway resistance, elastic resistance should be within the reasonable range of normal lung tissue elasticity, and PEEP should not be lower than the zero point corresponding to atmospheric pressure. Any parameter solution exceeding this bounding box is deemed invalid, thus eliminating anomalous solutions that clearly violate physiological common sense at the outermost layer. After constraining each candidate combination using the physiological bounding box constraint, the first respiratory parameter constraint value is obtained, i.e., the feasible set of parameters falling within the physiological bounding box. The second layer is the passive segment independent constraint. This constraint utilizes passive expiratory segment data after the termination reference point to independently estimate the respiratory mechanics parameters of the passive segment without relying on the active segment respiratory muscle pressure model. In the passive expiratory segment, the respiratory muscles are completely relaxed, and the respiratory muscle pressure is zero. The respiratory mechanics equation simplifies to a passive equation containing only airway resistance, elastic resistance, and positive end-expiratory pressure. At this point, the number of unknowns is reduced, and the passive segment parameters can be solved independently based on the pressure and flow data of the passive segment. The parameters obtained from the independent passive segment solution are used as the constraint benchmark, requiring that the parameters obtained from the full-window fitting remain consistent with this benchmark in the passive segment or that the deviation is within an allowable range, thereby avoiding interference from the active segment fitting on the passive segment parameters. After constraining each candidate combination using the passive segment independent constraint, the constraint values ​​for the second respiratory parameter are obtained.

[0049] Furthermore, the third layer is a full-window passive lower bound constraint. This constraint requires that, within the entire analysis window, the passive pressure component, consisting of the passive component (airway resistance multiplied by flow rate and elastic resistance multiplied by volume), must not be lower than the lower bound of the measured airway pressure. In other words, as an active driving component, the respiratory muscle pressure should act in a direction that conforms to physiological laws, and there should be no situation where the passive pressure component exceeds the measured airway pressure, leading to an abnormal reversal in respiratory muscle pressure. This constraint ensures the physical meaning and physiological rationality of the respiratory muscle pressure waveform from a full-window perspective. After constraining each candidate combination using the full-window passive lower bound constraint, the constraint value of the third respiratory parameter is obtained. The fourth layer is a time constant coupling constraint, mainly including time constant box constraint cross-tightening and dimensionality reduction re-solution triggered by severe deviation. This constraint utilizes the physical law that the time constant of the respiratory system, i.e., the product of airway resistance and elastic resistance, should be consistent between the inspiratory and expiratory phases, to couple the constraint on the ratio of airway resistance to elastic resistance. The time constant of the respiratory system reflects the rate of change in intrapulmonary pressure or volume. Within the same respiratory cycle, the time constants of the inspiratory and expiratory phases should be at similar levels. Using the online estimated time constant of the expiratory phase as a reference, the product of airway resistance and elastic resistance obtained from the inspiratory phase is constrained to avoid decoupling anomalies where the two parameters increase at the expense of each other and the product deviates from the true time constant. After constraining each candidate combination using time constant coupling constraints, the constrained value of the fourth respiratory parameter is obtained.

[0050] Finally, the fifth layer is the waveform penalty constraint. This constraint imposes a penalty term on the morphological characteristics of the fitted respiratory muscle pressure waveform, suppressing high-frequency jitter, abrupt spikes, or non-physiological morphologies in the waveform. As a force generated by muscle contraction, respiratory muscle pressure should physiologically have a smooth, continuous waveform with specific rise and fall patterns, and should not exhibit violent fluctuations resembling high-frequency noise. The waveform penalty constraint encourages waveform smoothness by introducing a penalty term for the second derivative or total variation of the respiratory muscle pressure waveform into the objective function. This constraint is tightened through several iterations, and exits after all checkpoints satisfy the constraint, thus achieving a balance between fitting accuracy and waveform smoothness. After constraining each candidate combination using the waveform penalty constraint, the fifth respiratory parameter constraint value is obtained. The effective set method unifies the treatment of five types of constraints. Its core idea is to dynamically identify the currently effective set of constraints during the iteration process and perform unconstrained optimization within the feasible region defined by this set. When the iteration point touches a new constraint boundary, the effective set is updated until convergence to the optimal solution that satisfies all constraints. The five layers of constraints complement each other and progressively narrow the solution space. The physiological boundary box constraint provides the outermost feasible region, the passive segment independence constraint and the full-window passive lower bound constraint further restrict the parameter range, the time constant coupling constraint introduces the physical relationship between parameters, and the waveform penalty constraint optimizes the estimation results from the perspective of signal quality. Through the above steps, a five-layer constraint system is constructed, from physiological boundary to passive segment independence, from full-window lower bound to time constant coupling, and then to waveform morphology. By solving it uniformly through the effective set method, multi-dimensional and physiologically consistent constraint estimation of respiratory parameters is achieved, significantly improving parameter identifiability and solution stability, effectively suppressing noise amplification and abnormal solutions in the ill-conditioned inverse problem, and significantly improving the robustness and physiological rationality of respiratory parameter estimation.

[0051] In one embodiment, that is, constraining each candidate combination in the two-dimensional search grid using the time constant coupling constraint to obtain the fourth breathing parameter constraint value corresponding to each candidate combination, the following steps are included: Obtain the online estimated expiratory time constant of the target object during the expiratory cycle; The time constant of the expiratory phase is used to establish a time constant coupling constraint between airway resistance and elastic resistance, and the current time constant is calculated based on the time constant coupling constraint. Determine whether the deviation between the current time constant and the expiratory time constant exceeds a preset deviation threshold; If the deviation exceeds the preset deviation threshold, the ratio of the airway resistance to the elastic resistance is substituted into the pre-constructed respiratory parameter solution equation as an equality constraint, and the respiratory parameter solution equation is re-solved by dimensionality reduction using a matrix to obtain the fourth respiratory parameter constraint value corresponding to each candidate combination.

[0052] In one implementation, the expiratory time constant of the target object within the expiratory cycle is first estimated online for each candidate combination in a two-dimensional search grid. During passive expiration, the respiratory muscles are completely relaxed, and the mechanical behavior of the respiratory system can be approximated by a first-order linear model. The decay process of lung volume or airway pressure follows an exponential decay law, and this exponentially decaying time constant is the respiratory system's time constant, equal to the product of airway resistance and elastic resistance. Based on passive expiratory data after the termination reference point, the system estimates the expiratory time constant of the current respiratory cycle online by exponentially fitting the decay curve of tidal volume or airway pressure. Since the passive segment has few unknowns and a relatively high signal-to-noise ratio, this estimate has good reliability and can serve as a reference benchmark for constraining inspiratory parameters. The aforementioned expiratory time constant is used to establish a time constant coupling constraint between airway resistance and elastic resistance. The physical meaning of this coupling constraint is that within the same respiratory cycle, the product of airway resistance and elastic resistance obtained from the inspiratory fitting, i.e., the current time constant, should be consistent with or within the allowable deviation range of the independently estimated time constant for the expiratory segment. Based on this coupling constraint, the current time constant obtained by fitting the inhalation section parameters under the current candidate combination is calculated, which is the product of the current airway resistance estimate and the current elastic resistance estimate.

[0053] Further, it is determined whether the deviation between the current time constant and the expiratory time constant exceeds a preset deviation threshold. The degree of deviation can be measured by absolute difference or relative ratio, and the preset deviation threshold is set according to the clinically permissible range of time constant fluctuation. If the deviation between the current time constant and the expiratory time constant is within the preset deviation threshold, it indicates that the airway resistance and elastic resistance ratio under this candidate combination meets the time constant consistency constraint, and no further intervention is required. If the deviation exceeds the preset deviation threshold, it indicates that the airway resistance and elastic resistance under this candidate combination have decoupled abnormally, that is, the two parameters increase or decrease in opposite directions, causing the product to deviate from the true time constant. In this case, stronger equality constraints need to be applied for correction. When the deviation exceeds the preset deviation threshold, the ratio of airway resistance to elastic resistance is substituted into the pre-constructed respiratory parameter solving equation as an equality constraint. Specifically, since the expiratory time constant is equal to airway resistance multiplied by elastic resistance, the ratio of airway resistance to elastic resistance can be derived. This ratio is introduced into the respiratory parameter solving equation as an equality constraint, so that the two parameters of airway resistance and elastic resistance no longer change independently, but are coupled together through the time constant. After introducing equality constraints, the dimension of the unknowns in the original solution equations is reduced. The system performs dimension reduction and re-solution of the respiratory parameter solution equations through matrix transformations, such as using null space projection of the constraint matrix or elimination methods. The remaining free parameters are solved in the subspace that satisfies the time constant equality constraints, yielding the fourth respiratory parameter constraint values ​​corresponding to each candidate combination. Through the above steps, the airway resistance and elastic resistance in the inspiratory phase are coupled and constrained using the time constant of the reliable estimate in the expiratory phase. When the deviation is too large, the airway resistance and elastic resistance are forcibly corrected by substituting equality constraints and matrix dimension reduction and re-solution. This effectively avoids the decoupling anomaly of the two parameters, ensures the physical consistency of respiratory mechanics parameters between the inspiratory and expiratory phases, thereby improving the overall reliability and physiological rationality of respiratory parameter estimation and enhancing the ventilator's real-time monitoring capability of changes in the patient's respiratory mechanics.

[0054] In one embodiment, the step of using a preset effective set method to perform least squares fitting on the first, second, third, fourth, and fifth respiratory parameter constraint values ​​to obtain the estimated respiratory parameters corresponding to each candidate combination includes the following steps: Based on the first respiratory parameter constraint value, the second respiratory parameter constraint value, the third respiratory parameter constraint value, the fourth respiratory parameter constraint value, and the fifth respiratory parameter constraint value, construct the objective function for least squares fitting; The objective function is solved iteratively using the effective set method, and the Lagrange multipliers on the feasible region boundary are calculated in each iteration. Based on the Lagrange multipliers on the feasible region boundary of each iteration, update the parameter estimates for the current iteration until the iteration termination condition is met, and obtain the respiratory parameter estimates for each candidate combination.

[0055] In one implementation, a constrained least-squares fitting objective function is constructed based on the first to fifth respiratory parameter constraint values ​​obtained from five layers of constraints. The core of this objective function is the sum of squared residuals between the model-predicted airway pressure and the measured airway pressure. Simultaneously, waveform penalty constraints are superimposed on the objective function as regularization penalty terms. The objective function is iteratively solved using the effective set method, an optimization algorithm that handles box constraints (parameters with upper and lower bounds). By fixing out-of-bounds parameters to boundary values, a reduced-order submatrix is ​​constructed for the remaining free parameters and the solution is recalculated, transferring the coupling effect to the new optimal value. The direction of the Lagrange multiplier determines whether to release the bounded constraint—if the multiplier points inside the feasible region, the constraint should not be activated, and the solution is recalculated after release. The Lagrange multiplier reflects the marginal impact of the corresponding constraint on the objective function, i.e., the influence of the constraint tightness on the optimal solution. For inequality constraints, if the corresponding Lagrange multiplier is negative, it means that relaxing the constraint can further reduce the objective function value, and the constraint should not be kept in the effective set and should be removed from the effective set; if the Lagrange multiplier is positive, it means that the constraint does have a limiting effect and should be kept in the effective set.

[0056] Furthermore, by updating the parameter estimates of the current iteration based on the Lagrange multipliers on the feasible region boundary of each iteration until the iteration termination condition is met, the KKT optimality condition is satisfied. Specifically, a one-dimensional line search is performed along the search direction obtained from solving the subproblem to determine the maximum step size that can be advanced without violating any ineffective constraints. The parameter estimates are then updated along this step size. If a new inequality constraint is encountered during the advancement, it is added to the effective set. After each update, the Lagrange multipliers are recalculated, and it is determined whether the effective set needs adjustment. This iterative process is repeated until the iteration termination condition is met. Finally, when the termination condition is met, the parameter estimates of the current iteration are output as the respiratory parameter estimates corresponding to the candidate combination. Through the above steps, an accurate and stable solution to the multi-constraint least squares problem can be achieved, ensuring that the final respiratory parameter estimates simultaneously satisfy all physiological constraints and minimize the fitting residuals, thus providing a reliable parameter solution guarantee for the accurate estimation of respiratory muscle pressure.

[0057] In one embodiment, specifically step S204, which involves determining the respiratory muscle pressure estimation result based on the estimated respiratory parameters corresponding to each of the candidate combinations, includes the following steps: Based on the estimated respiratory parameters corresponding to each candidate combination, calculate the fitting cost, residual autocorrelation, and passive segment residual for each candidate combination; If the fitting cost is within a preset tolerance range of candidate combinations, then a comprehensive score is calculated based on the residual autocorrelation and the passive segment residual. The parameters corresponding to the candidate combination with the best comprehensive score are selected as the optimal respiratory parameter estimates. Based on the optimal respiratory parameter estimates, the respiratory muscle pressure estimation results are determined.

[0058] In one implementation, three quality assessment metrics are calculated for each candidate combination based on the estimated respiratory parameters corresponding to each combination: fit cost, residual autocorrelation, and passive segment residuals. The fit cost is the sum of squared residuals between the model-predicted airway pressure and the measured airway pressure within the entire analysis window, reflecting the overall fit of the model to the full-window signal; a lower fit cost indicates higher overall fitting accuracy. The residual autocorrelation is the first-order autocorrelation coefficient of the residual sequence or a combination of several orders of autocorrelation, reflecting whether there are systematic temporal structures in the residuals that are not explained by the model. If the model has adequately fitted the effective components of the signal, the residuals should approximate white noise, with the autocorrelation coefficient close to zero. If the residual autocorrelation is significantly non-zero, it indicates that the model has missed some systematic components and the fit is insufficient. The passive segment residual is the sum of squares or mean square error of the residuals between the model-predicted pressure and the measured pressure in the passive expiratory segment after the termination reference point. It reflects the accuracy of the estimation of respiratory mechanics parameters in the passive segment. Since the respiratory muscle pressure in the passive segment is zero and the model is simpler, the smaller the passive segment residual, the more accurate the passive segment parameters are, which also indirectly reflects the reliability of the full window parameters.

[0059] Further, a preliminary screening is performed using fitting cost, retaining only candidate combinations whose fitting costs fall within a preset tolerance range to proceed to the next stage. The preset tolerance range is based on the minimum fitting cost among all candidate combinations, allowing combinations within a certain multiple or absolute increment of this benchmark to pass the preliminary screening. Candidate combinations whose fitting costs exceed the tolerance range have significantly insufficient overall fitting accuracy and can be directly eliminated without further evaluation. Candidate combinations that pass the preliminary screening have acceptable overall fitting accuracy; at this point, the difference in fitting costs is insufficient to distinguish between superior and inferior combinations, requiring further refined evaluation using residual autocorrelation and passive segment residuals. For candidate combinations that pass the preliminary screening, a comprehensive score is calculated based on residual autocorrelation and passive segment residuals. The comprehensive score is a weighted sum of the normalized residual autocorrelation and passive segment residuals, with weights set according to the importance of each indicator in clinical application. For example, the absolute value of the residual autocorrelation and the root mean square of the passive segment residuals can be normalized to the 0-1 interval, and then weighted according to preset weights to obtain the comprehensive score. A lower comprehensive score indicates that the candidate combination performs better in terms of model sufficiency and passive segment accuracy. Other reasonable scoring functions can also be used, and there are no specific limitations. The parameter corresponding to the candidate combination with the best overall score (i.e., the lowest score value) is selected as the optimal respiratory parameter estimate. Based on this optimal respiratory parameter estimate, the respiratory muscle pressure waveform is calculated using the respiratory mechanics equation, and this waveform is determined as the final respiratory muscle pressure estimation result. This respiratory muscle pressure estimation result can be further used for clinical applications such as ventilator-patient synchronization assessment, ventilation support level adjustment, and respiratory muscle fatigue monitoring. Through the above steps, a multi-dimensional, two-stage selection of the optimal candidate combination is achieved, avoiding overfitting or local optima that may be caused by a single fitting index. This ensures that the final output respiratory muscle pressure estimation result achieves optimal overall accuracy, model adequacy, and passive segment consistency, thereby significantly improving the clinical reliability and practical value of respiratory muscle pressure estimation.

[0060] Please see Figure 2 , Figure 2 This is a schematic diagram of a ventilator pressure estimation device provided in an embodiment of the present invention. This ventilator pressure estimation device corresponds one-to-one with the ventilator pressure estimation method described in the above embodiments. Please refer to [link / reference] for details. Figure 1 as well as Figure 1 The relevant descriptions in the corresponding embodiments are shown below. For ease of explanation, only the parts relevant to this embodiment are shown. See also... Figure 2 The ventilator pressure estimation device 20 includes: an acquisition module 21, a construction module 22, a solution module 23, and a determination module 24.

[0061] The acquisition module 21 is used to acquire the respiratory airway signal of the target object within a preset period, and determine the time reference point of respiratory muscle pressure activity based on the respiratory airway signal, wherein the time reference point includes a start reference point and an end reference point. Construction module 22 is used to construct a two-dimensional search grid based on the start reference point and end reference point of the respiratory muscle pressure activity, wherein the two-dimensional search grid includes multiple candidate combinations; Solver module 23 is used to perform least squares fitting on each of the candidate combinations in the two-dimensional search grid using multi-layer constraints to obtain the estimated respiratory parameters corresponding to each candidate combination. The determining module 24 is used to determine the respiratory muscle pressure estimation result based on the estimated respiratory parameters corresponding to each of the candidate combinations.

[0062] Optionally, the acquisition module 21 described above is specifically used for: Acquire the airway pressure signal from the respiratory airway signal; The airway pressure signal is processed using a sliding window variance comparison algorithm, and the airway pressure signal is divided into a left half-window and a right half-window. Calculate the first standard deviation and the second standard deviation corresponding to the left half-window and the right half-window, respectively; When the first standard deviation of the left half-window is lower than the preset left standard deviation threshold, the second standard deviation of the right half-window is higher than the preset right standard deviation threshold, and the downward trend of the pressure at the endpoint of the right half-window exceeds the preset trend threshold, the corresponding window center is taken as a candidate reference point, and the starting reference point is determined from the candidate starting points.

[0063] Optionally, the acquisition module 21 described above is further used for: Real-time feature extraction is performed on the respiratory airway signals to obtain multi-dimensional feature data; Within a preset sliding time window, dynamic zero-crossing detection and amplitude attenuation evaluation are performed on the multi-dimensional feature data to generate candidate termination reference points; The hysteresis threshold is dynamically calculated based on the historical distribution of the candidate termination reference points within a continuous respiratory cycle. The validity of the candidate termination reference points is verified based on the hysteresis threshold to obtain the termination reference points for respiratory muscle pressure activity.

[0064] Optionally, the solution module 23 described above is specifically used for: The physiological bounding box constraint is used to constrain each candidate combination in the two-dimensional search grid to obtain the first respiratory parameter constraint value corresponding to each candidate combination; By using the passive segment independent constraint to constrain each candidate combination in the two-dimensional search grid, the second breathing parameter constraint value corresponding to each candidate combination is obtained; The candidate combinations in the two-dimensional search grid are constrained by the full-window passive lower bound constraint to obtain the third breathing parameter constraint value corresponding to each candidate combination. The time constant coupling constraint is used to constrain each candidate combination in the two-dimensional search grid to obtain the fourth breathing parameter constraint value corresponding to each candidate combination; The waveform penalty constraint is used to constrain each candidate combination in the two-dimensional search grid to obtain the fifth breathing parameter constraint value corresponding to each candidate combination; The first, second, third, fourth, and fifth respiratory parameter constraint values ​​are solved by least squares fitting using the preset effective set method to obtain the estimated respiratory parameter values ​​corresponding to each candidate combination.

[0065] Optionally, the solution module 23 described above is also used for: Obtain the online estimated expiratory time constant of the target object during the expiratory cycle; The time constant of the expiratory phase is used to establish a time constant coupling constraint between airway resistance and elastic resistance, and the current time constant is calculated based on the time constant coupling constraint. Determine whether the deviation between the current time constant and the expiratory time constant exceeds a preset deviation threshold; If the deviation exceeds the preset deviation threshold, the ratio of the airway resistance to the elastic resistance is substituted into the pre-constructed respiratory parameter solution equation as an equality constraint, and the respiratory parameter solution equation is re-solved by dimensionality reduction using a matrix to obtain the fourth respiratory parameter constraint value corresponding to each candidate combination.

[0066] Optionally, the solution module 23 described above is also used for: Based on the first respiratory parameter constraint value, the second respiratory parameter constraint value, the third respiratory parameter constraint value, the fourth respiratory parameter constraint value, and the fifth respiratory parameter constraint value, construct the objective function for least squares fitting; The objective function is solved iteratively using the effective set method, and the Lagrange multipliers on the feasible region boundary are calculated in each iteration. Based on the Lagrange multipliers on the feasible region boundary of each iteration, update the parameter estimates for the current iteration until the iteration termination condition is met, and obtain the respiratory parameter estimates for each candidate combination.

[0067] Optionally, the aforementioned determining module 24 is specifically used for: Based on the estimated respiratory parameters corresponding to each candidate combination, calculate the fitting cost, residual autocorrelation, and passive segment residual for each candidate combination; If the fitting cost is within a preset tolerance range of candidate combinations, then a comprehensive score is calculated based on the residual autocorrelation and the passive segment residual. The parameters corresponding to the candidate combination with the best comprehensive score are selected as the optimal respiratory parameter estimates. Based on the optimal respiratory parameter estimates, the respiratory muscle pressure estimation results are determined.

[0068] It should be noted that the information interaction and execution process between the above-mentioned units are based on the same concept as the method embodiments of the present invention. For details on their specific functions and technical effects, please refer to the method embodiments section, which will not be repeated here.

[0069] Figure 3 This is a schematic diagram of the structure of a ventilator provided in an embodiment of the present invention. Figure 3 As shown, the ventilator of this embodiment includes: at least one processor ( Figure 3 Only one is shown in the diagram), a memory, and a computer program stored in the memory and executable on at least one processor, which, when executed by the processor, implements the steps in any of the above embodiments of the ventilator pressure estimation method.

[0070] The ventilator may include, but is not limited to, a processor and memory. Those skilled in the art will understand that... Figure 3 This is merely an example of a ventilator and does not constitute a limitation on ventilators. A ventilator may include more or fewer components than shown in the illustration, or combine certain components, or different components, such as a network interface, display screen, and input system.

[0071] In one embodiment, a computer-readable storage medium is provided that, when the instructions in the computer-readable storage medium are executed by a processor in a ventilator, enables the ventilator to perform the steps of any embodiment of the ventilator pressure estimation method disclosed in this invention, which will not be repeated here. The computer-readable storage medium may be non-volatile or volatile.

[0072] The processor referred to can be a CPU, but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0073] The memory includes readable storage media, internal memory, etc., wherein the internal memory can be the RAM of the sputum suction device, providing an environment for the operation of the operating system and computer-readable instructions stored in the readable storage media. The readable storage media can be the hard drive of the sputum suction device, or in some embodiments, it can be an external storage device of the sputum suction device, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the sputum suction device. Furthermore, the memory can include both internal storage units and external storage devices of the sputum suction device. The memory is used to store the operating system, cooperative applications, bootloader, data, and other programs, such as program code for computer programs. The memory can also be used to temporarily store data that has been output or will be output.

[0074] Those skilled in the art will understand that implementing all or part of the processes in the methods of the above embodiments can be accomplished by a computer program instructing related hardware. This computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0075] Those skilled in the art will understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this invention. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here. If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium.

[0076] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for estimating ventilator pressure, characterized in that, include: Acquire the respiratory airway signal of the target object within a preset period, and determine the time reference point of respiratory muscle pressure activity based on the respiratory airway signal, wherein the time reference point includes a start reference point and an end reference point; A two-dimensional search grid is constructed based on the starting and ending reference points of the respiratory muscle pressure activity, wherein the two-dimensional search grid includes multiple candidate combinations; By using multi-layered constraints, the least squares fitting solution is performed on each candidate combination in the two-dimensional search grid to obtain the estimated respiratory parameters corresponding to each candidate combination. The respiratory muscle pressure estimation result is determined based on the estimated respiratory parameters corresponding to each of the candidate combinations.

2. The ventilator pressure estimation method as described in claim 1, characterized in that, The determination of the time reference point for respiratory muscle pressure activity based on the respiratory airway signals includes: Acquire the airway pressure signal from the respiratory airway signal; The airway pressure signal is processed using a sliding window variance comparison algorithm, and the airway pressure signal is divided into a left half-window and a right half-window. Calculate the first standard deviation and the second standard deviation corresponding to the left half-window and the right half-window, respectively; When the first standard deviation of the left half-window is lower than the preset left standard deviation threshold, the second standard deviation of the right half-window is higher than the preset right standard deviation threshold, and the downward trend of the pressure at the endpoint of the right half-window exceeds the preset trend threshold, the corresponding window center is taken as a candidate reference point, and the starting reference point is determined from the candidate starting points.

3. The ventilator pressure estimation method as described in claim 1, characterized in that, The step of determining the time reference point for respiratory muscle pressure activity based on the respiratory airway signal further includes: Real-time feature extraction is performed on the respiratory airway signals to obtain multi-dimensional feature data; Within a preset sliding time window, dynamic zero-crossing detection and amplitude attenuation evaluation are performed on the multi-dimensional feature data to generate candidate termination reference points; The hysteresis threshold is dynamically calculated based on the historical distribution of the candidate termination reference points within a continuous respiratory cycle. The validity of the candidate termination reference points is verified based on the hysteresis threshold to obtain the termination reference points for respiratory muscle pressure activity.

4. The ventilator pressure estimation method as described in claim 1, characterized in that, The multi-layered constraint conditions include physiological bounding box constraints, passive segment independence constraints, full-window passive lower bound constraints, time constant coupling constraints, and waveform penalty constraints. The least-squares fitting solution is performed on each candidate combination in the two-dimensional search grid using these multi-layered constraint conditions to obtain the estimated respiratory parameters corresponding to each candidate combination, including: The physiological bounding box constraint is used to constrain each candidate combination in the two-dimensional search grid to obtain the first respiratory parameter constraint value corresponding to each candidate combination; By using the passive segment independent constraint to constrain each candidate combination in the two-dimensional search grid, the second breathing parameter constraint value corresponding to each candidate combination is obtained; The candidate combinations in the two-dimensional search grid are constrained by the full-window passive lower bound constraint to obtain the third breathing parameter constraint value corresponding to each candidate combination. The time constant coupling constraint is used to constrain each candidate combination in the two-dimensional search grid to obtain the fourth breathing parameter constraint value corresponding to each candidate combination; The waveform penalty constraint is used to constrain each candidate combination in the two-dimensional search grid to obtain the fifth breathing parameter constraint value corresponding to each candidate combination; The first, second, third, fourth, and fifth respiratory parameter constraint values ​​are solved by least squares fitting using the preset effective set method to obtain the estimated respiratory parameter values ​​corresponding to each candidate combination.

5. The ventilator pressure estimation method as described in claim 4, characterized in that, The step of constraining each candidate combination in the two-dimensional search grid using the time constant coupling constraint to obtain the fourth breathing parameter constraint value corresponding to each candidate combination includes: Obtain the online estimated expiratory time constant of the target object during the expiratory cycle; The time constant of the expiratory phase is used to establish a time constant coupling constraint between airway resistance and elastic resistance, and the current time constant is calculated based on the time constant coupling constraint. Determine whether the deviation between the current time constant and the expiratory time constant exceeds a preset deviation threshold; If the deviation exceeds the preset deviation threshold, the ratio of the airway resistance to the elastic resistance is substituted into the pre-constructed respiratory parameter solution equation as an equality constraint, and the respiratory parameter solution equation is re-solved by dimensionality reduction using a matrix to obtain the fourth respiratory parameter constraint value corresponding to each candidate combination.

6. The ventilator pressure estimation method as described in claim 4, characterized in that, The method of using a preset effective set approach to perform least-squares fitting on the first, second, third, fourth, and fifth respiratory parameter constraint values ​​to obtain estimated respiratory parameter values ​​for each candidate combination includes: Based on the first respiratory parameter constraint value, the second respiratory parameter constraint value, the third respiratory parameter constraint value, the fourth respiratory parameter constraint value, and the fifth respiratory parameter constraint value, construct the objective function for least squares fitting; The objective function is solved iteratively using the effective set method, and the Lagrange multipliers on the feasible region boundary are calculated in each iteration. Based on the Lagrange multipliers on the feasible region boundary of each iteration, update the parameter estimates for the current iteration until the iteration termination condition is met, and obtain the respiratory parameter estimates for each candidate combination.

7. The ventilator pressure estimation method as described in claim 1, characterized in that, The step of determining the respiratory muscle pressure estimation result based on the estimated respiratory parameters corresponding to each of the candidate combinations includes: Based on the estimated respiratory parameters corresponding to each candidate combination, calculate the fitting cost, residual autocorrelation, and passive segment residual for each candidate combination; If the fitting cost is within a preset tolerance range of candidate combinations, then a comprehensive score is calculated based on the residual autocorrelation and the passive segment residual. The parameters corresponding to the candidate combination with the best comprehensive score are selected as the optimal respiratory parameter estimates. Based on the optimal respiratory parameter estimates, the respiratory muscle pressure estimation results are determined.

8. A ventilator pressure estimation device, characterized in that, include: The acquisition module is used to acquire the respiratory airway signal of the target object within a preset period, and determine the time reference point of respiratory muscle pressure activity based on the respiratory airway signal, wherein the time reference point includes a start reference point and an end reference point. A construction module is used to construct a two-dimensional search grid based on the start and end reference points of the respiratory muscle pressure activity, wherein the two-dimensional search grid includes multiple candidate combinations; The solution module is used to perform least squares fitting on each of the candidate combinations in the two-dimensional search grid using multi-layer constraints, so as to obtain the estimated respiratory parameters corresponding to each candidate combination. The determination module is used to determine the respiratory muscle pressure estimation result based on the estimated respiratory parameters corresponding to each of the candidate combinations.

9. A ventilator, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the ventilator pressure estimation method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the ventilator pressure estimation method as described in any one of claims 1 to 7.

Images