Postoperative pulmonary complication early warning system and method

Abstract

The invention belongs to the technical field of data processing and machine dynamic learning, and provides a postoperative pulmonary complication early warning system and method.According to the technical scheme, a classic respiratory system model is converted into a respiratory dynamics model, and based on the respiratory dynamics model, a pulmonary complication early warning system is established; the non-stationary data change is processed by introducing forgetting factor processing in recursive least square, and the parameter vector of the system is dynamically updated; acquiring a respiration signal sampling point, performing noise reduction on the acquired respiration signal sampling point to obtain a respiration signal after noise reduction, and performing reconstruction based on the respiration signal after noise reduction to obtain a reconstructed respiration track; carrying out dynamic learning on the reconstructed breathing trajectory to carry out modeling, and carrying out visual display on a breathing dynamics result obtained after modeling to generate a breathing dynamics diagram; and obtaining a corresponding early warning strategy according to the abnormal characteristics of the respiratory dynamics diagram. And early-stage accurate early warning of postoperative pulmonary complications can be realized.

Description

Technical Field

[0001] This invention belongs to the field of data processing and machine dynamic learning technology, and particularly relates to a postoperative pulmonary complication early warning system and method. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Postoperative pulmonary complications are a major contributing factor to increased postoperative mortality, prolonged hospital stays, and higher healthcare costs, particularly in patients undergoing thoracic or abdominal surgery and elderly patients. These complications include atelectasis and pneumonia. Therefore, early detection and intervention are crucial.

[0004] Currently, commonly used clinical detection methods mainly rely on imaging techniques (such as chest X-rays and CT scans) and arterial blood gas analysis. While imaging examinations can provide visual information about lung structure, they have the following problems: First, repeated radiation exposure may increase the long-term health risks for patients (especially children and pregnant women); second, imaging examinations require patients to be transported to specific equipment, making bedside real-time monitoring impossible; third, the interpretation of imaging results is easily affected by subjective factors, especially in the early stages of subtle lesions, which may lead to missed diagnoses. Although arterial blood gas analysis can reflect oxygenation status, its invasive procedures may cause pain, bleeding, or infection, and the sampling frequency is limited, making it difficult to capture dynamic changes.

[0005] The main diagnostic methods for upper airway obstruction are nasal endoscopy and nasopharyngeal radiography. Nasal endoscopy can be affected by the subjective bias of the viewer, and nasopharyngeal radiography involves a certain amount of radiation and is difficult to fully assess the patient's respiratory dynamics.

[0006] Some non-invasive monitoring technologies, such as electrical impedance tomography (EIT) and respiratory plethysmography (RPP), are being explored, but their clinical application still faces challenges: the former has low resolution and is easily affected by pleural effusion, while the latter is sensitive to motion artifacts and has difficulty distinguishing between pathological and physiological respiratory fluctuations. In addition, traditional machine learning models often lag or overfit when processing postoperative respiratory signals due to data non-stationarity (such as sudden changes in parameters during anesthesia recovery), failing to meet the needs of real-time early warning. Summary of the Invention

[0007] To address at least one of the technical problems mentioned above, this invention provides a postoperative pulmonary complication early warning system and method that can accurately capture the patient's respiratory trajectory, reduce the impact of early data on the current patient's respiratory data, has strong noise resistance, and can meet the needs of real-time early warning.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: A first aspect of the present invention provides a postoperative pulmonary complication early warning system, comprising: The parameter dynamic update module is configured to: convert the classic respiratory system model into a respiratory dynamics model, introduce the forgetting factor in recursive least squares to handle non-stationary data changes, dynamically update the parameter vector of the respiratory system, construct a fluctuation criterion based on the parameter vector of the respiratory system, and adjust the clinical threshold of parameter fluctuation amplitude according to the patient's state data. The breathing estimation and reconstruction module is configured to: acquire breathing signal sampling points, denoise the acquired breathing signal sampling points to obtain denoised breathing signals, and reconstruct the breathing trajectory based on the denoised breathing signals. The dynamics graph generation module is configured to: dynamically learn and model the reconstructed respiratory trajectory, visualize the modeled results, and generate a respiratory dynamics graph; The early warning module is configured to obtain corresponding early warning strategies based on abnormal features of the respiratory dynamics diagram.

[0009] Furthermore, in the parameter dynamic update module, a forgetting factor from recursive least squares is introduced to handle non-stationary data changes, dynamically updating the parameter vector, as follows: , , , in, Let k be the parameter vector of the respiratory system at time k. Let k be the parameter vector of the respiratory system at time k-1. Here is the gain matrix. Let be the system state vector. For external output, Let be the error covariance matrix of the previous time step, and λ be the forgetting factor. The value is set differently depending on the patient's stage of life.

[0010] Furthermore, in the parameter dynamic update module, a fluctuation criterion is constructed based on the parameter vector of the respiratory system, and the clinical threshold for parameter fluctuation amplitude is adjusted according to the patient's state data, including: The parameter fluctuation criterion is constructed as follows: ,in, For parameter fluctuation range, This is the clinical threshold. Determine the patient's condition stage and adjust the clinical threshold for parameter fluctuation based on the patient's condition data. This includes: when the patient is in an unstable period, [the following can be done] Lower the alert level to improve its sensitivity; raise the alert level when the situation is stable. At this point, the sensitivity decreases.

[0011] Furthermore, in the respiratory estimation and reconstruction module, the step of acquiring respiratory signal sampling points based on the respiratory dynamics model, denoising the acquired respiratory signal sampling points to obtain a denoised respiratory signal, and reconstructing the reconstructed respiratory trajectory based on the denoised respiratory signal includes: Respiratory signal data within a sliding window is obtained by filling respiratory signal sampling points into a sliding window. Apply Kalman filtering to the data within the window to calculate the state estimate. and prediction signal ; Combined with prediction signals Calculate the observation noise variance in real time within a sliding window. process noise variance

[0012] Combined with observation noise variance process noise variance The signal-to-noise ratio (SNR) of the current signal is calculated using the signal within the SW-KF window. The current SNR is compared with a preset SNR threshold to determine the signal quality. The corresponding gain mode is then switched to perform noise reduction, resulting in a noise-reduced breathing signal. Combined with state estimates The reconstructed breathing trajectory is obtained by reconstructing the breathing signal after noise reduction.

[0013] Furthermore, in the breathing estimation and reconstruction module, the step of comparing the current signal-to-noise ratio with a preset signal-to-noise ratio threshold, determining the signal quality, switching to the corresponding gain mode, and performing noise reduction includes: If the current signal-to-noise ratio (SNR) Preset signal-to-noise ratio threshold (SNR) thigh The first gain matrix is ​​used. Quickly track sudden respiratory abnormalities; If the current signal-to-noise ratio (SNR) Preset signal-to-noise ratio threshold S The second gain matrix is ​​used. Increase the process covariance of SW-KF based on process noise variance. Adjust the Q matrix, where the diagonal elements are the calculated process noise variance, and the off-diagonal elements are set to 0.

[0014] Furthermore, in the dynamics graph generation module, the step of dynamically learning the reconstructed respiratory trajectory to perform respiratory dynamics modeling includes: Using the reconstructed breathing trajectory as input, an RBF neural network is chosen as the approximation tool. Combined with the weight update formula, the model prioritizes the latest data, achieving dynamic learning through updates. The weight update formula is as follows: , , in, The learning state vector at time k is constructed from the reconstructed breathing trajectory in step 2, including key features such as breathing airflow and airflow change rate. express k The learning state vector at time +1 For gain, The model predicts the state at time k. This represents the difference between the actual breathing trajectory state and the model-predicted state at time k. This represents the weights of the RBF network at time k+1. It is the first k The weights of the RBF network at time step. This represents the model's predicted state at time k+1. RBF networks are used to approximate the intrinsic dynamics of the system. The value, The regression vector is inside the Gaussian function, which is the embedded function of the RBF neural network.

[0015] Furthermore, in the early warning module, the step of obtaining a corresponding early warning strategy based on the abnormal characteristics of the respiratory dynamics diagram includes: If the image shows a gradual decrease in the peak inspiratory flow and a prolonged expiratory flow period, there is a possibility of postoperative pulmonary complications. If the airflow trajectory exhibits periodic oscillations, secretions may become trapped. If the graph curve rises sharply and then stabilizes, the judgment should be made in conjunction with the forgetting factor, the amplitude of parameter fluctuations, and the clinical threshold.

[0016] A second aspect of the present invention provides a method for early warning of postoperative pulmonary complications, comprising: The classic respiratory system model is converted into a respiratory dynamics model. The forgetting factor in recursive least squares is introduced to handle non-stationary data changes, the parameter vector of the respiratory system is dynamically updated, a fluctuation criterion is constructed based on the parameter vector of the respiratory system, and the clinical threshold of parameter fluctuation amplitude is adjusted according to the patient's state data. Acquire respiratory signal sampling points, denoise the acquired respiratory signal sampling points to obtain denoised respiratory signals, and reconstruct the reconstructed respiratory trajectory based on the denoised respiratory signals; The reconstructed breathing trajectory is dynamically learned to model respiratory dynamics, and the resulting respiratory dynamics are visualized to generate a respiratory dynamics graph. Corresponding early warning strategies are derived based on abnormal characteristics of respiratory dynamics diagrams.

[0017] A third aspect of the present invention provides a computer-readable storage medium.

[0018] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in a postoperative pulmonary complication early warning system as described above.

[0019] A fourth aspect of the present invention provides a computer device.

[0020] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, performs the steps in a postoperative pulmonary complication early warning system as described above.

[0021] Compared with the prior art, the beneficial effects of the present invention are: This invention relies on real-time analysis of respiratory airflow signals, without relying on radiation-emitting detection methods such as chest X-rays or CT scans. It introduces a forgetting factor from recursive least squares to handle non-stationary data changes, accurately captures the patient's breathing trajectory, and promptly alarms when abnormalities are detected. It reduces noise at respiratory signal sampling points and reconstructs the breathing trajectory based on the denoised respiratory signals, reducing the impact of early data on current patient respiratory data and enabling early and accurate warning of postoperative pulmonary complications.

[0022] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0023] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0024] Figure 1 This is a flowchart of a postoperative pulmonary complication early warning system provided in an embodiment of the present invention; Figure 2 This is a flowchart of a postoperative pulmonary complication early warning system provided in an embodiment of the present invention. Detailed Implementation

[0025] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0026] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0027] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0028] Example 1 like Figure 1 As shown, this embodiment provides a postoperative pulmonary complication early warning system, including: The parameter dynamic update module is configured to: convert the classic respiratory system model into a respiratory dynamics model, introduce the forgetting factor in recursive least squares to handle non-stationary data changes, dynamically update the respiratory dynamics parameter vector in combination with parameter fluctuation criteria, construct fluctuation criteria based on the respiratory system parameter vector, and adjust the clinical threshold of parameter fluctuation amplitude according to the patient's state data. The parameter dynamic update module converts the classic respiratory system model into a kinetic model, including: In this embodiment, based on the physical quantities in the classical respiratory system model (such as respiratory airflow), the static relationship is transformed into a dynamic evolution equation by injecting the time dimension, i.e., a kinetic model. The dynamic model is expressed as: (1), in, For breathing airflow, The rate of change of respiratory airflow This represents the internal state vector of the respiratory dynamics system. Represents respiratory dynamics function, Represents a respiratory system parameter vector. For external output, For the next breath airflow, The rate of change of respiratory airflow in the next moment. The sampling time is the interval between two adjacent sampling times. For a certain sampling time; In the parameter dynamic update module, a forgetting factor from recursive least squares is introduced to handle non-stationary data changes, dynamically updating the respiratory dynamics parameter vector, specifically including: Postoperatively, respiratory dynamic parameters (such as airflow resistance and lung compliance) may fluctuate dramatically in a short period of time due to factors such as anesthetic drug metabolism, airway edema, and changes in body position. (This will dynamically change with the degree of airway obstruction). However, the accumulation of historical data can lead to a decrease in the algorithm's sensitivity to new data ("data saturation"), making it unable to provide timely warnings. Therefore, a forgetting factor from Recursive Least Squares (RLS) is introduced to handle non-stationary data changes. Through "exponential forgetting" of old data, the model has stronger real-time tracking capabilities and anti-interference capabilities in dynamically changing respiratory scenarios after surgery.

[0029] Recursively update parameter vector The calculation formula is: (2), in, This is the parameter vector of the respiratory system at time k-1, containing core parameters of respiratory dynamics (such as airflow resistance, lung compliance, etc.). This is the gain matrix, used to adjust the weighting of new observation data on parameter updates. The larger its value, the greater the influence of new data (…). (with prediction error) The stronger the corrective effect, This is the system state vector, corresponding to the internal state of the respiratory dynamics system. X1 represents respiratory airflow, and X2 represents the rate of change of respiratory airflow, which are the core variables describing the current respiratory state. Wherein, the gain matrix And error covariance matrix All updates include a forgetting factor. Specifically: (3), (4), in, Let be the error covariance matrix of the previous time step, reflecting the uncertainty of parameter estimation, and λ be the forgetting factor. The value is set differently depending on the patient's stage: during the anesthesia recovery period, when parameter fluctuations are large, the effect of amnesia is increased, i.e., the value is reduced. Value; during the stable period, the parameter fluctuates little, reducing the degree of forgetting, i.e., increasing Value (close to) =1); λ influences Indirect change Ultimately, it acts on formula (2). The update enables dynamic tracking of non-stationary data; In the parameter dynamic update module, parameter fluctuation criteria are constructed, and clinical thresholds for parameter fluctuation amplitude are adjusted based on the patient's status data. Specifically, this includes: Constructing parameter fluctuation criteria: (4), in, For parameter fluctuation range, This is the clinical threshold. In this implementation, The initial value was set at 0.22. Based on clinical respiratory physiology data, the parameter fluctuation ranges in healthy individuals and patients with atelectasis were compared. In healthy individuals, the fluctuation range of respiratory dynamic parameters (airflow resistance, lung compliance, etc.) is relatively stable, while the parameter fluctuations in patients with postoperative pulmonary complications such as atelectasis are significantly increased. 0.22 was determined through analysis of clinical data as a threshold that can preliminarily distinguish between normal physiological fluctuations and pathological abnormalities. When the parameter fluctuation exceeds 0.22, it suggests a potential risk of postoperative pulmonary complications.

[0030] Determine the patient's condition stage and adjust the clinical threshold for parameter fluctuation based on the patient's condition data. ; In this embodiment, the patient's state stages include the anesthesia recovery period and the stable period; During the recovery period from anesthesia: At this time, due to factors such as the metabolism of anesthetic drugs and respiratory edema, the patient's respiratory dynamic parameters fluctuate drastically, which is an unstable stage.

[0031] When in a stable phase: the patient's respiratory function gradually recovers, and parameter fluctuations tend to stabilize.

[0032] In this embodiment, the clinical threshold for parameter fluctuation is adjusted based on the patient's condition data. Specifically, it includes: When the patient is in an unstable period, Lowering the alert level improves the sensitivity of early warning systems. When in a stable period, adjust upwards. At this point, sensitivity decreases, and clinical observation is necessary to avoid missed diagnoses.

[0033] By real-time monitoring of patients The magnitude of the change determines the range of fluctuations, respiratory trajectory, and dynamic characteristics (such as a decrease in peak inspiratory velocity), and determines whether the patient is in an unstable or stable phase. Based on the judgment result, the clinical threshold δ is adjusted (δ is lowered during unstable phases to improve sensitivity, and δ is raised during stable phases to reduce sensitivity), so as to achieve dynamic adaptation between parameter updates and clinical warnings.

[0034] The respiratory estimation and reconstruction module is configured to: acquire respiratory signal sampling points based on the respiratory dynamics model, denoise the acquired respiratory signal sampling points to obtain the denoised respiratory signal, and reconstruct the reconstructed respiratory trajectory based on the denoised respiratory signal. A balance between noise immunity and tracking accuracy is achieved through sliding window Kalman filtering (SW-KF) preprocessing and dynamic gain switching strategies, which helps to reconstruct the breathing trajectory.

[0035] Postoperative respiratory signals exhibit non-stationary characteristics (such as abrupt changes in noise intensity during anesthesia recovery). Traditional Kalman filtering (KF) suffers from reduced filtering effectiveness due to its reliance on fixed noise statistical characteristics. Therefore, improvements are achieved through a sliding window mechanism and dynamic updates of noise covariance. The respiratory estimation reconstruction module specifically includes the following steps: Step 1: Obtain respiratory signal data within the sliding window by filling the respiratory signal sampling points, specifically including: In this embodiment, L respiratory signal sampling points are obtained. Initialize the sliding window and fill it with the initial L respiratory signal sampling points. ; When initializing the sliding window, only the latest L sampling points within the window are used for filtering, and old data outside the window are discarded. This mechanism avoids the misleading effect of early steady-state data on the current noisy environment. For example, when sudden noise occurs during the anesthesia recovery period, only the latest signal is considered.

[0036] Step 2: Apply Kalman filtering to the data within the window and calculate the state estimate. and prediction signal ; Step 3: Combine the predicted signal Calculate the observation noise variance in real time within a sliding window. process noise variance : (5), (6), in, The variance of the process noise at the previous time step. For the first i Respiratory signals collected at all times (such as raw data from airflow sensors). The forgetting factor is used to suppress the interference of outliers on the covariance and ensure that noise statistics are updated in real time.

[0037] In this embodiment, The value range is 0.9-0.99.

[0038] Step 4: Slide the window forward as new data is input, remove the oldest data point, add the latest data point, and repeat steps 2-3 to obtain the real-time updated state estimate, predicted signal, observation noise variance, and process noise variance. Step 5: Combine observation noise variance process noise variance The signal-to-noise ratio (SNR) of the current signal is calculated using the signal within the SW-KF window. The current SNR is compared with a preset SNR threshold to determine the signal quality. The corresponding gain mode is then switched to perform noise reduction, resulting in a noise-reduced breathing signal. The signal-to-noise ratio (SNR) is calculated using the signal within the SW-KF window, and is expressed as follows: (7), in, The mean of the signal within the sliding window. Furthermore, the high-gain observer matrix presents a trade-off between "tracking speed" and "noise resistance," which is addressed by automatically switching between high and low gain modes as follows: The current signal-to-noise ratio is compared with the preset signal-to-noise ratio threshold to determine the signal quality, and then the corresponding gain mode is switched to obtain the noise-reduced breathing signal. Specifically, when SNR Preset signal-to-noise ratio threshold (SNR) thigh High-gain matrix is ​​used Quickly track sudden respiratory abnormalities; When SNR Preset signal-to-noise ratio threshold S Using a low-gain matrix Furthermore, it increases the process covariance of the SW-KF, allowing for a larger estimation error tolerance and suppressing noise amplification under weak signals. Specifically, it is based on the process noise variance. Adjust the Q matrix, where the Q matrix is ​​composed of diagonal elements that are the calculated process noise variance, and off-diagonal elements that are set to 0. in, and Optimized in advance through clinical trials, these technologies address the requirements of "rapid tracking" and "strong noise resistance," respectively. The following methods were used to determine their suitability: Respiratory signals were collected from patients in different postoperative states (anesthesia recovery, stable period, and complication onset period), and clinical diagnostic results were recorded simultaneously. Through simulation debugging and multi-center clinical trials, parameters were optimized to meet the requirements of "tracking error <= 5%" and "noise suppression rate >= 30%".

[0039] Step 6: Combine state estimates The reconstructed breathing trajectory is obtained by reconstructing the denoised breathing signal, and is represented as follows: (8), in, Indicates system state x The estimated value, It is the first k The implicit, unobservable states within the respiratory system at all times include the transitional nature of airflow or tissue movement in respiratory mechanics, and the time difference between respiratory effort and output airflow. It is the implicit state of the respiratory system at time k+1. It is a 2×2 state transition matrix used to describe the natural evolution of the respiratory state. It is a 2×1 input gain matrix used to quantify the driving force of respiratory effort on the state. It is a 1×2 output observation matrix used to establish the mapping between the state and the observed signal, and it is a fixed value. It is a 1×1 direct transfer matrix describing how breathing effort (such as during a sudden cough or choking, where the effort instantaneously changes airflow) directly affects the observed output without passing through the internal state. Because this type of scenario has a low probability and a short-lived impact, the matrix is ​​designed as follows: D do =[0.05], a small constant value, which preserves the instantaneous effect while avoiding excessive amplification of interference. The SW-KF module is used to process the original signal. y(k) The signal is filtered and then output with noise reduction. By appropriately selecting the parameters in the sampled data observer, and assuming the observer gain and sampling frequency are sufficiently high, converges to x A sufficiently small neighborhood, and cyclically B in a finite time. do The gain parameter in K high and K low Dynamic changes within.

[0040] The dynamics graph generation module is configured to: dynamically learn the reconstructed respiratory trajectory to model respiratory dynamics, visualize the modeled respiratory dynamics results, and generate a respiratory dynamics graph; In the dynamics graph generation module, the reconstructed respiratory trajectory is dynamically learned to model respiratory dynamics, specifically including: In this embodiment, the reconstructed respiratory trajectory output from the respiratory estimation reconstruction module is used as the modeling input. This trajectory has been denoised by SW-KF to retain the real respiratory dynamics characteristics. The RBF neural network is selected as the approximation tool. The RBF function has the characteristics of strong local approximation ability and fast convergence speed, which is suitable for processing the nonlinear dynamic characteristics of respiratory signals. Through the weight update formula, the model prioritizes the latest data to achieve dynamic learning.

[0041] The weight update formula is expressed as: (9), (10) in, The learning state vector at time k is constructed from the reconstructed breathing trajectory in step 2, including key features such as breathing airflow and airflow change rate. express k The learning state vector at time +1, 0 < <1 represents gain. The model predicts the state at time k. This represents the difference between the actual breathing trajectory state and the model-predicted state at time k. This represents the weights of the RBF network at time k+1. It is the first k The weights of the RBF network at time step. This represents the model's predicted state at time k+1. RBF networks are used to approximate the intrinsic dynamics of the system. The value, The regression vector is inside the Gaussian function, which is the embedding function of the RBF neural network; The final modeling function is expressed as: (11), in, It is a modeled dynamic function represented by a constant RBF network. The time period after the transient learning process is The time-varying weight vector of the RBF network, It is the modeling error around trajectory iteration. It is a very small positive constant representing the modeling accuracy. This indicates the end time of a certain dynamic learning cycle. This indicates the start time of a certain dynamic learning iteration. For ease of labeling, we will... Recorded as , It is the regression vector of the Gaussian function. These are neurons that are regularly distributed in the input space. In the dynamics diagram generation module, the results obtained after modeling are visualized and a respiratory dynamics diagram is generated. In this embodiment, the respiratory dynamic function modeled based on the dynamic learning of the sampled data observer is approximately as follows: , in, It is the respiratory dynamics function along the inspiratory airflow trajectory. A locally accurate neural network approximation is achieved. Here, respiratory dynamics trajectories are utilized. This model represents the respiratory dynamics along the airflow trajectory, generating a respiratory dynamics graph. The respiratory signal is divided into four stages, which can be mapped one by one onto the respiratory dynamics graph.

[0042] The early warning module is configured to obtain corresponding early warning strategies based on abnormal features of the respiratory dynamics diagram; Specifically, if the image shows a gradual decrease in the peak inspiratory flow and a prolonged expiratory flow period, there is a possibility of postoperative pulmonary complications. If the airflow trajectory exhibits periodic oscillations, secretions may become trapped. If the graph curve rises sharply and then stabilizes, the judgment should be made in conjunction with the forgetting factor, the amplitude of parameter fluctuations, and the clinical threshold.

[0043] For example, adjusting the forgetting factor to <0.8 (indicating that the data is in a non-stationary state) will... Adjust to 0.15, if If the value is greater than 0.15 for 10 consecutive minutes, the patient needs to be closely monitored.

[0044] Example 2 This embodiment provides a method for early warning of postoperative pulmonary complications, including the following steps: The classic respiratory system model is converted into a respiratory dynamics model. The forgetting factor in recursive least squares is introduced to handle non-stationary data changes, the parameter vector of the respiratory system is dynamically updated, a fluctuation criterion is constructed based on the parameter vector of the respiratory system, and the clinical threshold of parameter fluctuation amplitude is adjusted according to the patient's state data. Acquire respiratory signal sampling points, denoise the acquired respiratory signal sampling points to obtain denoised respiratory signals, and reconstruct the reconstructed respiratory trajectory based on the denoised respiratory signals; The reconstructed breathing trajectory is dynamically learned to model respiratory dynamics, and the modeling results are visualized to generate a respiratory dynamics diagram. Corresponding early warning strategies are derived based on abnormal characteristics of respiratory dynamics diagrams.

[0045] It should be noted that the specific implementation of the postoperative pulmonary complication early warning method in this embodiment of the invention is similar to the specific implementation of the postoperative pulmonary complication early warning system in this embodiment of the invention. Please refer to the description in the method section for details. In order to reduce redundancy, it will not be repeated here.

[0046] Example 3 This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the postoperative pulmonary complication early warning method described above.

[0047] Example 4 This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the postoperative pulmonary complication early warning method described above.

[0048] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0049] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0050] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0051] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1The steps of the function specified in one or more boxes.

[0052] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0053] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A postoperative pulmonary complication early warning system, characterized in that, include: The parameter dynamic update module is configured to: convert the classic respiratory system model into a respiratory dynamics model; based on the respiratory dynamics model, introduce the forgetting factor in recursive least squares to handle non-stationary data changes; dynamically update the parameter vector of the respiratory system; construct a fluctuation criterion based on the parameter vector of the respiratory system; and adjust the clinical threshold of parameter fluctuation amplitude according to the patient's state data. The respiratory estimation and reconstruction module is configured to: acquire respiratory signal sampling points based on the respiratory dynamics model, denoise the acquired respiratory signal sampling points to obtain the denoised respiratory signal, and reconstruct the reconstructed respiratory trajectory based on the denoised respiratory signal. The dynamics graph generation module is configured to: dynamically learn and model the reconstructed respiratory trajectory, visualize the respiratory dynamics results obtained after modeling, and generate a respiratory dynamics graph; The early warning module is configured to obtain corresponding early warning strategies based on abnormal features of the respiratory dynamics diagram.

2. The postoperative pulmonary complication early warning system as described in claim 1, characterized in that, In the parameter dynamic update module, a forgetting factor from recursive least squares is introduced to handle non-stationary data changes, and the parameter vector is dynamically updated, as follows: ， ， ， in, Let k be the parameter vector of the respiratory system at time k. Let k be the parameter vector of the respiratory system at time k-1. Here is the gain matrix. Let be the system state vector. For external output, Let be the error covariance matrix of the previous time step, and λ be the forgetting factor. The value is set differently depending on the patient's stage of life.

3. The postoperative pulmonary complication early warning system as described in claim 1, characterized in that, In the parameter dynamic update module, fluctuation criteria are constructed based on the parameter vectors of the respiratory system, and the clinical threshold for parameter fluctuation amplitude is adjusted according to the patient's state data, including: The parameter fluctuation criterion is constructed as follows: ,in, For parameter fluctuation range, This is the clinical threshold. Determine the patient's condition stage and adjust the clinical threshold for parameter fluctuation based on the patient's condition data. This includes: when the patient is in an unstable period, [the following can be done] Lower the alert level to improve its sensitivity; raise the alert level when the situation is stable. At this point, the sensitivity decreases.

4. The postoperative pulmonary complication early warning system as described in claim 1, characterized in that, In the respiratory estimation and reconstruction module, the process of acquiring respiratory signal sampling points based on a respiratory dynamics model, denoising the acquired respiratory signal sampling points to obtain a denoised respiratory signal, and reconstructing the reconstructed respiratory trajectory based on the denoised respiratory signal includes: Respiratory signal data within a sliding window is obtained by filling respiratory signal sampling points into a sliding window. Apply Kalman filtering to the data within the window to calculate the state estimate. and prediction signal ; Combined with prediction signals Calculate the observation noise variance in real time within a sliding window. process noise variance Combined with observation noise variance process noise variance The signal-to-noise ratio (SNR) of the current signal is calculated using the signal within the SW-KF window. The current SNR is compared with a preset SNR threshold to determine the signal quality. The corresponding gain mode is then switched to perform noise reduction, resulting in a noise-reduced breathing signal. Combined with state estimates The reconstructed breathing trajectory is obtained by reconstructing the breathing signal after noise reduction.

5. A postoperative pulmonary complication early warning system as described in claim 4, characterized in that, In the breathing estimation and reconstruction module, the step of comparing the current signal-to-noise ratio (SNR) with a preset SNR threshold, determining the signal quality, switching to the corresponding gain mode, and performing noise reduction includes: If the current signal-to-noise ratio (SNR) Preset signal-to-noise ratio threshold (SNR) thigh The first gain matrix is ​​used. Quickly track sudden respiratory abnormalities; If the current signal-to-noise ratio (SNR) Preset signal-to-noise ratio threshold S The second gain matrix is ​​used. Increase the process covariance of SW-KF based on process noise variance. Adjust the Q matrix, where the diagonal elements are the calculated process noise variance, and the off-diagonal elements are set to 0.

6. The postoperative pulmonary complication early warning system as described in claim 1, characterized in that, In the dynamics graph generation module, the step of dynamically learning the reconstructed respiratory trajectory to model respiratory dynamics includes: Using the reconstructed breathing trajectory as input, an RBF neural network is chosen as the approximation tool. Combined with the weight update formula, the model prioritizes the latest data, achieving dynamic learning through updates. The weight update formula is as follows: ， ， in, The learning state vector at time k is constructed from the reconstructed breathing trajectory in step 2, including key features such as breathing airflow and airflow change rate. express k The learning state vector at time +1 For gain, The model predicts the state at time k. This represents the difference between the actual breathing trajectory state and the model-predicted state at time k. This represents the weights of the RBF network at time k+1. It is the first k The weights of the RBF network at time step. This represents the model's predicted state at time k+1. RBF networks are used to approximate the intrinsic dynamics of the system. The value, The regression vector is inside the Gaussian function, which is the embedded function of the RBF neural network.

7. A postoperative pulmonary complication early warning system as described in claim 1, characterized in that, In the early warning module, the corresponding early warning strategy obtained based on the abnormal characteristics of the respiratory dynamics diagram includes: If the image shows a gradual decrease in the peak inspiratory flow and a prolonged expiratory flow period, there is a possibility of postoperative pulmonary complications. If the airflow trajectory exhibits periodic oscillations, secretions may become trapped. If the graph curve rises sharply and then stabilizes, the judgment should be made in conjunction with the forgetting factor, the amplitude of parameter fluctuations, and the clinical threshold.

8. A method for early warning of postoperative pulmonary complications, characterized in that, Includes the following steps: The classic respiratory system model is converted into a respiratory dynamics model. The forgetting factor in recursive least squares is introduced to handle non-stationary data changes, the parameter vector of the respiratory system is dynamically updated, a fluctuation criterion is constructed based on the parameter vector of the respiratory system, and the clinical threshold of parameter fluctuation amplitude is adjusted according to the patient's state data. Acquire respiratory signal sampling points, denoise the acquired respiratory signal sampling points to obtain denoised respiratory signals, and reconstruct the reconstructed respiratory trajectory based on the denoised respiratory signals; The reconstructed breathing trajectory is dynamically learned to model respiratory dynamics, and the resulting respiratory dynamics are visualized to generate a respiratory dynamics graph. Corresponding early warning strategies are derived based on abnormal characteristics of respiratory dynamics diagrams.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the steps in the postoperative pulmonary complication early warning method as described in claim 8.

10. A computer device, 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 program, it implements the steps in the postoperative pulmonary complication early warning method as described in claim 8.

Images