A cardiopulmonary resuscitation system and method based on non-invasive prediction of coronary artery perfusion pressure

Abstract

The present disclosure belongs to the field of cardiopulmonary resuscitation technology, and specifically relates to a cardiopulmonary resuscitation system and method based on non-invasive prediction of coronary perfusion pressure, comprising: an acquisition module configured to acquire electrocardiogram (ECG) signals and photoplethysmography (PEP) signals; an extraction module configured to decompose the acquired signals into different frequency bands based on discrete wavelet transform and extract features of the acquired signals; and a prediction module configured to construct a non-invasive prediction model based on a genetic algorithm, using information complexity as a fitness function, based on the extracted signal features and real-time coronary perfusion pressure, to obtain a predicted value of the coronary perfusion pressure and complete non-invasive prediction of the coronary perfusion pressure. The present disclosure predicts coronary perfusion pressure based on ECG signals and pulse signals, adjusts the compression depth of cardiopulmonary resuscitation in real time based on the obtained predicted value of the coronary perfusion pressure, and improves the robustness and adaptability of closed-loop automatic chest compression control.

Description

Technical Field

[0001] The present disclosure belongs to the technical field of cardiopulmonary resuscitation, and in particular relates to a cardiopulmonary resuscitation system and method based on non-invasive prediction of coronary artery perfusion pressure. Background Art

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

[0003] Cardiac arrest is a malignant disease in which the heart suddenly stops pumping blood, causing interruption of systemic circulation, loss of breathing, and loss of consciousness. Cardiopulmonary resuscitation (CPR) can maintain a certain level of blood circulation in patients with cardiac arrest and is a key rescue method for cardiac arrest. However, manual chest compressions require a lot of physical strength from rescuers, and it is difficult to meet the standard requirements for compressions. Although some automatic chest compression devices on the market can perform compressions according to standard requirements, they cannot make adaptive adjustments based on the patient's physiological condition. Summary of the Invention

[0004] To address the above-mentioned problems, the present disclosure proposes a cardiopulmonary resuscitation system and method based on non-invasive prediction of coronary perfusion pressure. Coronary perfusion pressure (CPP) is predicted based on electrocardiogram (ECG) signals and pulse signals. The compression depth of cardiopulmonary resuscitation is adjusted in real time according to the obtained predicted CPP value, thereby improving the robustness and adaptability of closed-loop automatic chest compression control.

[0005] According to some embodiments, a first solution of the present disclosure provides a non-invasive prediction system for coronary artery perfusion pressure, which adopts the following technical solutions:

[0006] A non-invasive prediction system for coronary artery perfusion pressure, comprising:

[0007] an acquisition module configured to acquire an electrocardiogram signal and a photoplethysmography signal;

[0008] an extraction module configured to decompose the acquired signal into different frequency bands based on discrete wavelet transform and extract features of the acquired signal;

[0009] The prediction module is configured to construct a non-invasive prediction model based on the genetic algorithm according to the extracted signal features and coronary perfusion pressure, with information complexity as the fitness function, to obtain the predicted value of coronary perfusion pressure and complete the non-invasive prediction of coronary perfusion pressure.

[0010] As a further technical limitation, in the extraction module, the signal obtained by the acquisition module is decomposed into different frequency bands through discrete wavelet transform, and the detail coefficients and signal characteristics of different frequency bands are calculated respectively. The signal characteristics of different frequency bands are obtained according to the different detail coefficients obtained, thereby completing the feature extraction of the obtained signal.

[0011] As a further technical limitation, in the prediction module, the invasive arterial pressure and the invasive right atrial pressure are monitored in real time, and the coronary perfusion pressure at the moment is obtained based on the amplitude of the trough point of the monitored invasive arterial pressure and the invasive right atrial pressure at the corresponding moment. The coronary perfusion pressure obtained at the moment is the real-time coronary perfusion pressure, and the non-invasive prediction model is trained based on the obtained real-time coronary perfusion pressure.

[0012] As a further technical limitation, in the prediction module, the information complexity score of the population is calculated during each iteration, and the end criterion is that the number of iterations reaches a set value; after the iteration ends, the population with the smallest information complexity score during the iteration is found as the result of feature selection.

[0013] Furthermore, the performance of the model on the test set is calculated using mean absolute error, mean square error, root mean square error, goodness of fit, and information complexity score; the smaller the mean absolute error, mean square error, and root mean square error, the better the model performance; the closer the goodness of fit is to 1, the better the model performance; the smaller the information complexity score, the better the model performance.

[0014] Furthermore, during the model training process, each iteration will have an information complexity score. Within a limited number of iterations, when the complexity of the model and the prediction accuracy reach a relative balance, the minimum value of the information complexity will appear; that is, the minimum value of the information complexity will be obtained.

[0015] According to some embodiments, a second solution of the present disclosure provides a non-invasive prediction method for coronary artery perfusion pressure, using the following technical solutions:

[0016] A non-invasive prediction method for coronary artery perfusion pressure, comprising:

[0017] Acquire ECG signals and photoplethysmography signals;

[0018] Decomposing the acquired signal into different frequency bands based on discrete wavelet transform to extract the features of the acquired signal;

[0019] According to the extracted signal features and coronary perfusion pressure, a non-invasive prediction model was constructed based on the genetic algorithm with information complexity as the fitness function to obtain the predicted value of coronary perfusion pressure and complete the non-invasive prediction of coronary perfusion pressure.

[0020] According to some embodiments, a third solution of the present disclosure provides a cardiopulmonary resuscitation system based on non-invasive prediction of coronary artery perfusion pressure, which adopts the following technical solutions:

[0021] A cardiopulmonary resuscitation system based on non-invasive prediction of coronary artery perfusion pressure, comprising:

[0022] A non-invasive prediction module for coronary artery perfusion pressure, configured to adopt the non-invasive prediction system for coronary artery perfusion pressure as described in the first solution to obtain a predicted value of coronary artery perfusion pressure;

[0023] a control module configured to adjust and optimize compression parameters through PID control according to the obtained coronary artery perfusion pressure prediction value and the coronary artery perfusion pressure;

[0024] The cardiopulmonary resuscitation module is configured to adjust the cardiopulmonary resuscitation compression depth according to the obtained compression parameters to complete the adaptive adjustment of cardiopulmonary resuscitation.

[0025] As a further technical limitation, in the control module, the current coronary perfusion pressure value obtained is subtracted from the set coronary perfusion pressure standard value to obtain the current error; if the output of the PID controller is positive, the compression depth of the cardiopulmonary resuscitation device is controlled to increase; if the output of the PID controller is negative, the compression depth is controlled to decrease. The larger the output value, the faster the increase and decrease speed, thereby realizing the adjustment of the compression parameters.

[0026] As a further technical limitation, in the cardiopulmonary resuscitation module, based on the error between the real-time predicted coronary perfusion pressure value and the set coronary perfusion pressure reference value, combined with the PID controller, the compression depth of the cardiopulmonary resuscitation device is adaptively adjusted to achieve the gradual convergence of the predicted coronary perfusion pressure value and the set coronary perfusion pressure value, thereby completing the adjustment of adaptive cardiopulmonary resuscitation.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] The present invention uses a CPP prediction model based on machine learning, and uses non-invasive electrocardiogram (ECG) and photoplethysmography (PPG) signals to predict CPP, thereby providing in-depth guidance on compression depth and frequency, and achieving personalized compression.

[0029] The present invention discloses a hybrid machine learning model that uses a genetic algorithm (GA) and information complexity (ICOMP) to select features during the CPP prediction model training process, thereby reducing the model complexity while maintaining a high model training effect.

[0030] The present invention discloses a method for adaptively regulating cardiopulmonary resuscitation by using a PID controller to automatically adjust the PID control parameters of an automatic chest compressor according to a predicted value of coronary perfusion pressure, so that the entire chest compression closed-loop control system achieves the best resuscitation effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The accompanying drawings, which constitute a part of the present disclosure, are used to provide a further understanding of the present disclosure. The exemplary embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation to the present disclosure.

[0032] Figure 1 is a flow chart of the non-invasive prediction system for coronary artery perfusion pressure in the first embodiment of the present disclosure;

[0033] Figure 2 is a flow chart of the genetic algorithm in the first embodiment of the present disclosure;

[0034] Figure 3 is a flowchart of the feature subset and model selection process in the first embodiment of the present disclosure;

[0035] Figure 4 Schematic diagram of a cardiopulmonary resuscitation system based on non-invasive prediction of coronary artery perfusion pressure in the third embodiment of the present disclosure;

[0036] Figure 5 is a flowchart of cardiopulmonary resuscitation in Example 3 of the present disclosure;

[0037] Figure 6 is a structural diagram of the PID control system in the third embodiment of the present disclosure;

[0038] Among them, 1. Blood oxygen saturation probe; 2. First ECG electrode; 3. Second ECG electrode; 4. Third ECG electrode; 5. Fourth ECG electrode; 6. Fifth ECG electrode; 7. Automatic chest compression cardiopulmonary resuscitation device; 8. Physiological signal monitoring device; 9. Physiological signal analysis device. DETAILED DESCRIPTION

[0039] The present disclosure will be further described below with reference to the accompanying drawings and embodiments.

[0040] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of the present disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present disclosure belongs.

[0041] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present disclosure. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0042] In the present disclosure, terms such as "upper", "lower", "left", "right", "front", "back", "vertical", "horizontal", "side", "bottom", etc. indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. They are merely relational words determined for the convenience of describing the structural relationships of the various parts or elements of the present disclosure, and do not specifically refer to any part or element in the present disclosure, and should not be understood as limitations on the present disclosure.

[0043] In this disclosure, terms such as "fixed connection," "connected," and "connection" should be interpreted broadly to mean a fixed connection, an integral connection, or a detachable connection; a direct connection or an indirect connection through an intermediary. Relevant researchers or technicians in this field may determine the specific meanings of these terms in this disclosure based on specific circumstances, and they should not be construed as limitations on this disclosure.

[0044] In the absence of conflict, the embodiments of the present disclosure and the features thereof may be combined with each other.

[0045] Example 1

[0046] Embodiment 1 of the present disclosure introduces a non-invasive prediction system for coronary artery perfusion pressure.

[0047] A non-invasive prediction system for coronary artery perfusion pressure, comprising:

[0048] an acquisition module configured to acquire an electrocardiogram signal and a photoplethysmography signal;

[0049] an extraction module configured to decompose the acquired signal into different frequency bands based on discrete wavelet transform and extract features of the acquired signal;

[0050] The prediction module is configured to construct a non-invasive prediction model based on the genetic algorithm according to the extracted signal features and coronary perfusion pressure, with information complexity as the fitness function, to obtain the predicted value of coronary perfusion pressure and complete the non-invasive prediction of coronary perfusion pressure.

[0051] like Figure 1As shown in the figure, CPP is predicted using ECG signals and pulse wave signals: after obtaining ECG and PPG signals, they are preprocessed and features are extracted; after feature selection, they enter the trained machine learning model, and finally the predicted CPP result is output to obtain the CPP prediction value.

[0052] In this embodiment, discrete wavelet transform (DWT) is used to extract signal features. The following is a detailed introduction using ECG signal as an example:

[0053] DWT decomposes a square-integrable x(t)∈V0 into detailed W j and approximate V j Number of sub-levels. In this framework, the projection to V j The approximate information of x(t) on scale j is recorded as:

[0054]

[0055] Among them, φ j,p is the proportional function, a j,p = <x(t),φ j,p (t)> is called the approximation coefficient, where <·> is the inner product operator. Similarly, projecting onto W j The detailed information of x(t) at scale j is defined as:

[0056]

[0057] in, is called the detail coefficient. Let h(n) and g(n) be low-pass filter and high-pass filter respectively, then the relationship between the wavelet and the scaling function defined in the equation can be written as:

[0058]

[0059]

[0060] Therefore, DWT decomposes the signal x(t) into approximate coefficients a through g(n) and h(n) j,p (low-frequency component) and detail coefficient d j,p (High frequency component) coefficient:

[0061]

[0062]

[0063] In the proposed feature extraction process, the initial ECG signal is divided into 3-second segments. After the segmentation process, it is decomposed into high-frequency bands (detail coefficients, d j) and low frequency band (approximate coefficient, a j ), up to the fifth level of decomposition (j=1, 2, ..., 5). Therefore, many significant features can be calculated from these sub-bands at each level.

[0064] In this embodiment, in order to reduce data redundancy, all detail coefficients (d j , j = 1, 2, ..., 5) and the approximation coefficients of the last decomposition level (level 5) are used in the feature extraction process. For each subband, various features can be calculated, such as curve length, energy, maximum, minimum, median, mean, entropy, range, kurtosis, skewness, trend, and number of zero crossings. These features can be calculated from the first detail coefficient (d1). This allows the same features to be calculated from other subbands.

[0065] Similarly, PPG can obtain features similarly through the above process.

[0066] The prediction model's label is determined using the CPP value calculated from the ABP and RAP signals. The ABP signal trough for each segment is found and the RAP value at the corresponding time point is subtracted from the amplitude at that point. The CPP value for each segment is calculated as the average CPP value, which is used as the prediction model's label. Cross-validation is performed on the split data, followed by feature subset and model selection. Finally, the optimal model is selected based on the calculated ICOMP score and other performance metrics.

[0067] The information complexity criterion ICOMP is designed to penalize the increased complexity of dealing with the correlations between estimated parameters, rather than simply penalizing unnecessary parameters. For both univariate and multivariate models, the structure of ICOMP is:

[0068]

[0069] Where n is the number of samples, and the error variance is It is estimated by the root mean square error between the actual group label and the predicted group label (the actual group label is the CPP value obtained through experiment, and the predicted group label is the CPP value predicted by the model. The error variance is obtained by calculating the root mean square error between the two). is a stable and smooth convex estimator of the model parameter vector and covariance matrix:

[0070]

[0071]

[0072]

[0073] Where p is rank, λ i for The eigenvalues ​​of is the average

[0074] By stabilizing and smoothing the estimator of the convex sum covariance matrix, it is able to overcome ill-conditioned covariance matrices and minimize the error in addition to selecting the optimal kernel function.

[0075] To reduce the dimensionality of the system input and the complexity of the model, this embodiment uses a new feature subset selection method and integrates this process into the training process of the machine learning model. Basically, the feature subset selection method used is based on the genetic algorithm (GA), in which ICOMP is used as the fitness function.

[0076] The GA generates a certain number of feature subsets based on the population size in each generation. These subsets are created using specialized operators such as crossover, mutation, migration, and selection. Machine learning models are estimated on datasets corresponding to these feature subsets. The performance of the estimated models with different feature subsets is scored. These scores indicate the contribution of the feature subsets to the goodness of fit. Furthermore, they can be used to create new feature subsets for the next generation during the regeneration process.

[0077] The operators in the GA loop are as follows Figure 2 As shown, the detailed steps are:

[0078] a. Create an initial population at random, including a subset of features called chromosomes, which only contain binary numbers {0,1}. For example, if feature F j If it is included in the feature subset, the binary number is 1, otherwise it is 0.

[0079] b. Train any machine learning model for all subsets (chromosomes) separately and score them according to the ICOMP criteria.

[0080] c. Send some of the highest-scoring favorable subsets (nearly 1%) to the next generation.

[0081] d. Use heuristics or random operators, such as random universal sampling, to collect a subset of favored traits (sires) into the mating pool. (This process is called selection)

[0082] e. Apply the mutation and crossover process to the selected fathers and create a new subset of traits (subset) for the next generation. (Stages c, d, and e are called the reproduction process).

[0083] f. If one of the stopping criteria is met, terminate the program; otherwise, return to the third stage.

[0084] g. Maintain optimal functionality.

[0085] To ensure that the best feature subset and regression model are obtained, the process in this embodiment can be repeated multiple times. In essence, the training process with feature selection may also estimate several alternative models. To select the best model among these alternatives, certain unique performance statistics can be considered, such as mean absolute error, mean square error, root mean square error and goodness of fit, as well as ICOMP criteria and other criteria. The flowchart of feature subset and model selection is shown in FIG. Figure 3 shown.

[0086] The criterion for feature subset selection is the minimum ICMP score. The model selection criterion is the best performance on the test set. The models available include support vector regression (kernel functions include linear kernel, quadratic kernel, cubic kernel, radial kernel, and Gaussian kernel), ensemble learning (using adaptive boosting (Adaboost) hybridized with weak learners such as decision trees and K-nearest neighbors), and logistic regression.

[0087] During each iteration, the ICOMP score of the population is calculated. The termination criterion is that the number of iterations reaches the set value. After the end, the population with the smallest ICOMP score during the iteration is found as the result of feature selection.

[0088] Calculate the model's performance on the test set using mean absolute error, mean square error, root mean square error, goodness of fit, and ICMP score. Theoretically, smaller mean absolute error, mean square error, and root mean square error indicate better model performance; closer to 1 is the goodness of fit, better model performance; and smaller ICMP score indicates better model performance. The optimal model is determined based on these results and serves as the model selection result.

[0089] This embodiment uses a hybrid machine learning model of genetic algorithm (GA) and information complexity (ICOMP) to select features during the CPP prediction model training process, while maintaining a high model training effect while reducing model complexity.

[0090] Example 2

[0091] The second embodiment of the present disclosure introduces a non-invasive prediction method for coronary artery perfusion pressure, which adopts the non-invasive prediction system for coronary artery perfusion pressure introduced in the first embodiment.

[0092] A non-invasive prediction method for coronary artery perfusion pressure, comprising:

[0093] Acquire ECG signals and photoplethysmography signals;

[0094] Decomposing the acquired signal into different frequency bands based on discrete wavelet transform to extract the features of the acquired signal;

[0095] According to the extracted signal features and real-time coronary perfusion pressure, a non-invasive prediction model was constructed based on the genetic algorithm with information complexity as the fitness function to obtain the predicted value of coronary perfusion pressure and complete the non-invasive prediction of coronary perfusion pressure.

[0096] The detailed steps are the same as those of the non-invasive prediction system for coronary artery perfusion pressure provided in Example 1 and will not be repeated here.

[0097] Example 3

[0098] The third embodiment of the present disclosure introduces a cardiopulmonary resuscitation system based on non-invasive prediction of coronary artery perfusion pressure.

[0099] A cardiopulmonary resuscitation system based on non-invasive prediction of coronary artery perfusion pressure, characterized by comprising:

[0100] A non-invasive prediction module for coronary artery perfusion pressure, which is configured to adopt the non-invasive prediction system for coronary artery perfusion pressure in the first embodiment to obtain a predicted value of coronary artery perfusion pressure;

[0101] a control module configured to adjust and optimize compression parameters through PID control according to the obtained predicted coronary artery perfusion pressure value and the real-time coronary artery perfusion pressure;

[0102] The cardiopulmonary resuscitation module is configured to adjust the cardiopulmonary resuscitation compression depth according to the obtained compression parameters to complete the adaptive adjustment of cardiopulmonary resuscitation.

[0103] like Figure 4 As shown, this embodiment monitors the patient's blood oxygen saturation in real time through the blood oxygen saturation probe 1, obtains ABP and RAP under the action of the first ECG electrode 2, the second ECG electrode 3, the third ECG electrode 4, the fourth ECG electrode 5 and the fifth ECG electrode 6, and obtains the photoplethysmogram signal and the ECG signal through the physiological signal monitoring device 8; under the action of the physiological signal analysis device 9, based on the real-time ECG and PPG signals, the CPP value obtained by the prediction model is used to judge the compression quality and feed it back to the cardiopulmonary resuscitation instrument.

[0104] like Figure 5 As shown in the figure, after the patient is found, the patient's heart rhythm is monitored. If the patient's heart rhythm is normal, no operation is required. If the patient's heart rhythm is abnormal, the cardiopulmonary resuscitation device is placed on the patient and activated. Throughout the cardiopulmonary resuscitation process, the patient's physiological signals are monitored in real time, the CPP value is obtained through the prediction model, and the compression parameters are adjusted through the PID control system until the patient is successfully resuscitated.

[0105] This embodiment uses a PID control system to automatically adjust the compression depth through closed-loop control. Figure 6 As shown in the figure, the compression depth d(t) of the cardiopulmonary resuscitation instrument is adjusted by the output u(t) of the PID system, and the compression depth is adjusted, thereby adjusting the physiological state, forming an adaptive cardiopulmonary resuscitation system, which can maximize the robustness and adaptability of the closed-loop automatic chest compression control. After obtaining the current CPP value c(t), it is subtracted from the set CPP standard value r(t) to obtain the current error e(t). The output u(t) of the PID controller is positive, which controls the compression depth d(t) of the cardiopulmonary resuscitation instrument to increase; if it is negative, it controls the compression depth to decrease.

[0106] The foregoing description is merely a preferred embodiment of the present disclosure and is not intended to limit the present disclosure. Those skilled in the art will readily appreciate that various modifications and variations are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present disclosure shall be included within the scope of protection of the present disclosure.

[0107] Although the above describes the specific implementation methods of the present disclosure in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present disclosure. Those skilled in the art should understand that on the basis of the technical solution of the present disclosure, various modifications or variations that can be made by those skilled in the art without creative work are still within the scope of protection of the present disclosure.

Claims

1. A non-invasive prediction system for coronary artery perfusion pressure, characterized in that: include: an acquisition module configured to acquire an electrocardiogram signal and a photoplethysmography signal; an extraction module configured to decompose the acquired signal into different frequency bands based on discrete wavelet transform and extract features of the acquired signal; A prediction module is configured to construct a non-invasive prediction model based on a genetic algorithm according to the extracted signal features and coronary artery perfusion pressure, using information complexity as a fitness function, to obtain a predicted value of the coronary artery perfusion pressure, and to complete the non-invasive prediction of the coronary artery perfusion pressure; Information complexity standard ICOMP It aims to penalize the increased complexity of evaluating the correlation between estimated parameters, rather than simply penalizing unnecessary parameters, for both univariate and multivariate models. ICOMP The structure is: in, n is the number of samples, the error variance is estimated by the root mean square error between the actual group labels and the predicted group labels, is a stable and smooth convex estimator of the model parameter vector and covariance matrix: in, yes rank, for The eigenvalues ​​of is the average value.

2. The non-invasive prediction system for coronary artery perfusion pressure as claimed in claim 1, characterized in that: In the extraction module, the signal obtained by the acquisition module is decomposed into different frequency bands through discrete wavelet transform, and the detail coefficients and signal characteristics of different frequency bands are calculated respectively. The signal characteristics of different frequency bands are obtained according to the different detail coefficients obtained, thereby completing the feature extraction of the obtained signal.

3. The non-invasive prediction system for coronary artery perfusion pressure as claimed in claim 1, characterized in that: In the prediction module, the invasive arterial pressure and the invasive right atrial pressure are monitored in real time. The coronary perfusion pressure at the moment is obtained based on the amplitude of the trough point of the monitored invasive arterial pressure and the invasive right atrial pressure at the corresponding moment. The coronary perfusion pressure obtained at the moment is the real-time coronary perfusion pressure, and the non-invasive prediction model is trained based on the obtained real-time coronary perfusion pressure.

4. The non-invasive prediction system for coronary artery perfusion pressure as claimed in claim 1, characterized in that: In the prediction module, the information complexity score of the population is calculated during each iteration, and the end criterion is that the number of iterations reaches a set value; after the iteration ends, the population with the smallest information complexity score during the iteration is found as the result of feature selection.

5. The non-invasive prediction system for coronary artery perfusion pressure as claimed in claim 4, characterized in that: The performance of the model on the test set is calculated using mean absolute error, mean square error, root mean square error, goodness of fit, and information complexity score; the smaller the mean absolute error, mean square error, and root mean square error, the better the model performance; the closer the goodness of fit is to 1, the better the model performance; the smaller the information complexity score, the better the model performance.

6. The non-invasive prediction system for coronary artery perfusion pressure according to claim 4, characterized in that: During the model training process, each iteration will have an information complexity score. Within a limited number of iterations, when the complexity of the model and the prediction accuracy reach a relative balance, the minimum value of the information complexity will appear; that is, the minimum value of the information complexity will be obtained.

7. A non-invasive prediction method for coronary artery perfusion pressure, characterized in that: include: Acquire ECG signals and photoplethysmography signals; Decomposing the acquired signal into different frequency bands based on discrete wavelet transform to extract the features of the acquired signal; According to the extracted signal features and coronary artery perfusion pressure, a non-invasive prediction model is constructed based on a genetic algorithm with information complexity as the fitness function to obtain the predicted value of coronary artery perfusion pressure and complete the non-invasive prediction of coronary artery perfusion pressure. Information complexity standard ICOMP It aims to penalize the increased complexity of evaluating the correlation between estimated parameters, rather than simply penalizing unnecessary parameters, for both univariate and multivariate models. ICOMP The structure is: in, n is the number of samples, the error variance is estimated by the root mean square error between the actual group labels and the predicted group labels, is a stable and smooth convex estimator of the model parameter vector and covariance matrix: in, yes rank, for The eigenvalues ​​of is the average value.

8. A cardiopulmonary resuscitation system based on non-invasive prediction of coronary artery perfusion pressure, characterized in that: include: A non-invasive prediction module for coronary artery perfusion pressure, configured to adopt the non-invasive prediction system for coronary artery perfusion pressure according to any one of claims 1 to 6 to obtain a predicted value of coronary artery perfusion pressure; a control module configured to adjust and optimize compression parameters through PID control according to the obtained coronary artery perfusion pressure prediction value and the coronary artery perfusion pressure; The cardiopulmonary resuscitation module is configured to adjust the cardiopulmonary resuscitation compression depth according to the obtained compression parameters to complete the adaptive adjustment of cardiopulmonary resuscitation.

9. The cardiopulmonary resuscitation system based on non-invasive prediction of coronary perfusion pressure as described in claim 8, wherein in the control module, the current coronary perfusion pressure value obtained is subtracted from the set coronary perfusion pressure standard value to obtain the current error; if the output of the PID controller is positive, the compression depth of the cardiopulmonary resuscitation device is controlled to increase; if the output of the PID controller is negative, the compression depth is controlled to decrease, and the larger the output value, the faster the increase and decrease speed, thereby achieving adjustment of the compression parameters.

10. The cardiopulmonary resuscitation system based on non-invasive prediction of coronary artery perfusion pressure as claimed in claim 8, characterized in that: In the cardiopulmonary resuscitation module, based on the error between the real-time predicted coronary perfusion pressure value and the set coronary perfusion pressure reference value, combined with the PID controller, the compression depth of the cardiopulmonary resuscitation instrument is adaptively adjusted to achieve the gradual convergence of the predicted coronary perfusion pressure value and the set coronary perfusion pressure value, thereby completing the adjustment of adaptive cardiopulmonary resuscitation.

Images