CN-122000044-A - Postoperative pulmonary complications early warning system and postoperative pulmonary complications early warning method

CN122000044ACN 122000044 ACN122000044 ACN 122000044ACN-122000044-A

Abstract

The invention belongs to the technical field of data processing and machine dynamic learning, and provides a postoperative pulmonary complications early warning system and method, wherein the technical scheme is that a classical respiratory system model is converted into a respiratory dynamics model, and forgetting factors in recursive least square are introduced to process non-stationary data changes based on the respiratory dynamics model, so that parameter vectors of the system are dynamically updated; the method comprises the steps of obtaining a respiratory signal sampling point, carrying out noise reduction on the obtained respiratory signal sampling point to obtain a noise-reduced respiratory signal, carrying out reconstruction on the noise-reduced respiratory signal to obtain a reconstructed respiratory track, carrying out dynamic learning on the reconstructed respiratory track to carry out modeling, carrying out visual display on respiratory dynamics results obtained after modeling to generate a respiratory dynamics map, and obtaining a corresponding early warning strategy according to abnormal characteristics of the respiratory dynamics map. Can realize early accurate early warning of postoperative pulmonary complications.

Inventors

SONG JUNLIN
LI WEI
WU WEIMING
ZHAI QIAN
LIU ZHENGQIN
LI XIAOHE
WEI DONG
SHEN QI
BAI ZHENFENG
SONG ZHANGWEI

Assignees

山东大学

Dates

Publication Date: 20260508
Application Date: 20251226

Claims (10)

1.A post-operative pulmonary complications pre-warning system, comprising: The parameter dynamic updating module is configured to convert a classical respiratory system model into a respiratory dynamics model, introduce forgetting factor processing in recursive least square to process nonstationary data change based on the respiratory dynamics model, dynamically update a parameter vector of a respiratory system, construct a fluctuation criterion based on the parameter vector of the respiratory system, and adjust a parameter fluctuation amplitude clinical threshold according to state data of a patient; The respiratory estimation reconstruction module is configured to acquire respiratory signal sampling points based on a respiratory dynamics model, perform noise reduction on the acquired respiratory signal sampling points to obtain noise-reduced respiratory signals, and reconstruct based on the noise-reduced respiratory signals to obtain a reconstructed respiratory track; the dynamic diagram generating module is configured to dynamically learn and model the reconstructed respiratory track, visually display respiratory dynamic results obtained after modeling, and generate a respiratory dynamic diagram; the early warning module is configured to obtain a corresponding early warning strategy according to abnormal characteristics of the respiratory dynamics diagram.
2. The post-operative pulmonary complications pre-warning system of claim 1, wherein the parameter dynamic update module introduces forgetting factor processing in recursive least squares to handle non-stationary data changes, the dynamic update parameter vector expressed as: , , , Wherein, the Is the parameter vector of the respiratory system at time k, Is the respiratory system parameter vector at time k-1, In the form of a gain matrix, As a system state vector of the system, For the external output it is possible to provide, For the error covariance matrix of the last moment, lambda is the forgetting factor, Different values are set according to different periods of the patient.
3. The post-operative pulmonary complications pre-warning system of claim 1, wherein the parameter dynamic updating module constructs a fluctuation criterion based on a parameter vector of the respiratory system, adjusts a parameter fluctuation range clinical threshold according to patient status data, and comprises: the construction parameter fluctuation criterion is as follows: , wherein, For the amplitude of the fluctuation of the parameter, Is a clinical threshold; Determining a patient status phase, adjusting a parameter fluctuation range clinical threshold according to patient status data Includes that when the patient is in an unstable period, the patient can be treated Down-regulating to raise early warning sensitivity, up-regulating when it is in stable period At this time, the sensitivity is lowered.
4. The post-operative pulmonary complications pre-warning system of claim 1, wherein in the respiratory estimation reconstruction module, the respiratory signal sampling points are obtained based on the respiratory dynamics model, noise is reduced on the obtained respiratory signal sampling points to obtain a noise-reduced respiratory signal, and the reconstructed respiratory track is obtained based on the noise-reduced respiratory signal, and the post-operative pulmonary complications pre-warning system comprises: Filling respiratory signal sampling points based on the sliding window to obtain respiratory signal data in the sliding window; Applying Kalman filtering to data within a window, calculating a state estimate And predicting a signal ; Binding predicted signals Real-time calculation of observed noise variance within sliding window Sum of process noise variance Combining observed noise variances Sum of process noise variance Calculating the signal-to-noise ratio of the current signal by utilizing the signal in the SW-KF window, comparing the current signal-to-noise ratio with a preset signal-to-noise ratio threshold, judging the signal quality, switching to a corresponding gain mode, and carrying out noise reduction to obtain a noise-reduced respiratory signal; combining state estimates And reconstructing based on the noise-reduced respiratory signal to obtain a reconstructed respiratory track.
5. The post-operative pulmonary complications pre-warning system of claim 4, wherein in the respiratory estimation reconstruction module, comparing a current signal-to-noise ratio with a preset signal-to-noise ratio threshold, determining a signal quality, switching to a corresponding gain mode, and performing noise reduction, includes: if the current signal-to-noise ratio SNR A preset SNR threshold SNR thigh , which adopts a first gain matrix Rapidly tracking sudden respiratory abnormalities; if the current signal-to-noise ratio SNR Preset signal-to-noise threshold S Using a second gain matrix Increasing the process covariance of the SW-KF based on the process noise variance The Q matrix is adjusted, wherein the Q matrix is composed of diagonal elements that are calculated process noise variances and non-diagonal elements that are set to 0.
6. The post-operative pulmonary complications pre-warning system of claim 1, wherein in the dynamic map generation module, the dynamically learning the reconstructed respiratory track for respiratory dynamics modeling includes: Taking the reconstructed respiratory track as input, selecting RBF neural network as approximation tool, and combining weight updating formula to make the model pay priority to the latest data, realizing updating dynamic learning, wherein the weight updating formula is as follows: , , Wherein, the The learning state vector representing the kth moment is constructed by the respiratory track reconstructed in the step 2, and comprises key characteristics such as respiratory airflow, airflow change rate and the like, Represents the learning state vector at time k +1, In order for the gain to be a function of, The state is predicted for the model at the kth time, As the difference between the actual respiratory track state at time k and the model predicted state, The weight of the RBF network at time k +1, Is the weight of the RBF network at time k, Represents the model predictive state at time k +1, Is RBF network used to approximate the system internal power Is used as a reference to the value of (a), The regression vector inside the gaussian function, which is an embedded function of the RBF neural network.
7. The post-operative pulmonary complications early warning system according to claim 1, wherein, in the early warning module, the early warning strategy is obtained according to abnormal characteristics of the respiratory kinetic image, and the post-operative pulmonary complications early warning system comprises: If the peak value of the inspiratory phase airflow is slowly reduced and the expiratory phase airflow period is prolonged, the postoperative pulmonary complications are likely to occur; If the airflow track oscillates periodically, secretion retention may occur; if the image curve rises steeply, then the image curve is stable, and judgment is carried out by combining the forgetting factor, the parameter fluctuation range and the clinical threshold value.
8. The postoperative pulmonary complications early warning method is characterized by comprising the following steps of: converting a classical respiratory system model into a respiratory dynamics model, introducing forgetting factor processing in recursive least square to process non-stationary data change, dynamically updating a respiratory system parameter vector, constructing a fluctuation criterion based on the respiratory system parameter vector, and adjusting a parameter fluctuation range clinical threshold according to state data of a patient; Acquiring a respiratory signal sampling point, denoising the acquired respiratory signal sampling point to obtain a denoised respiratory signal, and reconstructing based on the denoised respiratory signal to obtain a reconstructed respiratory track; Dynamically learning the reconstructed respiratory track to carry out respiratory dynamics modeling, and carrying out visual display on respiratory dynamics results obtained after modeling to generate a respiratory dynamics diagram; And obtaining a corresponding early warning strategy according to the abnormal characteristics of the respiratory dynamics diagram.
9. A computer readable storage medium having stored thereon a computer program, which when executed by a processor, implements the steps of a post-operative pulmonary complications pre-warning method according to claim 8.
10. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor performs the steps of a post-operative pulmonary complications pre-warning method according to claim 8.

Description

Postoperative pulmonary complications early warning system and postoperative pulmonary complications early warning method Technical Field The invention belongs to the technical field of data processing and machine dynamic learning, and particularly relates to a postoperative pulmonary complications early warning system and method. Background The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art. Postoperative pulmonary complications are one of the major factors that lead to increased mortality in postoperative patients, prolonged hospitalization and increased medical costs. Particularly, the traditional Chinese medicine composition is more remarkable in chest and abdomen operations and elderly patients, including atelectasis, pneumonia and the like. Early discovery and intervention is therefore of paramount importance. Currently, the detection methods commonly used in clinic mainly rely on imaging techniques (such as chest X-ray and CT) and arterial blood gas analysis. Although imaging examination can provide visual information of lung structures, the problems are that firstly, repeated radiation exposure can increase long-term health risks of patients (especially children and pregnant women), secondly, imaging examination requires the patients to be transported to specific equipment and cannot realize bedside real-time monitoring, thirdly, image result interpretation is easily influenced by subjective factors, and especially diagnosis omission is possible in early stage of fine lesions. Although arterial blood gas analysis can reflect oxygenation status, invasive procedures can cause pain, bleeding or infection, and sampling frequency is limited, which makes it difficult to capture dynamic changes. The diagnosis of the cause of the upper respiratory tract obstruction is mainly a nasal endoscope and a nasopharyngeal radiological examination, the nasal endoscope is influenced by subjective prejudice of a viewer, a certain amount of radiation exists in the nasopharyngeal radiological examination, and the respiratory dynamics change of a patient is difficult to comprehensively evaluate. Some non-invasive monitoring techniques, such as electrical impedance tomography and respiratory induction plethysmography, are increasingly being explored, but their clinical application is still challenged by the fact that the former is low in resolution and susceptible to pleural effusion, while the latter is sensitive to motion artifacts and difficult to distinguish between pathological and physiological respiratory fluctuations. In addition, when the traditional machine learning model processes postoperative respiratory signals, the model is lagged or overfitted due to data non-stationarity (such as parameter mutation in anesthesia recovery period) frequently, and the real-time early warning requirement cannot be met. Disclosure of Invention In order to solve at least one technical problem in the background art, the invention provides a postoperative pulmonary complications early warning system and method, which can accurately capture the respiratory track of a patient, can reduce the influence of early data on the respiratory data of the current patient, have strong noise immunity and can meet the real-time early warning requirement. In order to achieve the above purpose, the present invention adopts the following technical scheme: a first aspect of the present invention provides a post-operative pulmonary complications pre-warning system comprising: The parameter dynamic updating module is configured to convert a classical respiratory system model into a respiratory dynamics model, introduce forgetting factor processing in recursive least square to process nonstationary data change, dynamically update a respiratory system parameter vector, construct a fluctuation criterion based on the respiratory system parameter vector, and adjust a parameter fluctuation range clinical threshold according to state data of a patient; the respiratory estimation reconstruction module is configured to acquire respiratory signal sampling points, perform noise reduction on the acquired respiratory signal sampling points to obtain noise-reduced respiratory signals, and reconstruct based on the noise-reduced respiratory signals to obtain a reconstructed respiratory track; The dynamic diagram generating module is configured to dynamically learn and model the reconstructed respiratory track, visually display the modeled result and generate a respiratory dynamic diagram; the early warning module is configured to obtain a corresponding early warning strategy according to abnormal characteristics of the respiratory dynamics diagram. Further, in the parameter dynamic updating module, forgetting factor processing in recursive least square is introduced to process non-stationary data change, and the parameter vector is dynamically updated, which is expressed