An intelligent early warning system for respiratory medicine based on data fusion

Abstract

The present invention relates to the field of data processing technology, and in particular to an intelligent early warning system for respiratory medicine based on data fusion. The contents include: monitoring and collecting multi-source physiological data of patients to obtain original physiological data; preprocessing the original physiological data to obtain preprocessed physiological data; intelligent fusion processing of the preprocessed physiological data to obtain comprehensive physiological characteristic data; abnormality detection and analysis of the comprehensive physiological characteristic data to obtain analysis results of the respiratory state; risk level classification based on the analysis results of the respiratory state; and formulating medical advice based on the risk level. The system solves the technical problems of inaccurate fusion of multi-source physiological data and low accuracy in judging the health status of patients in the current respiratory monitoring system in the process of realizing intelligent early warning in respiratory medicine.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to an intelligent early warning system for respiratory medicine based on data fusion. Background Art

[0002] With the development of medical technology and people's increasing attention to health, the application of intelligent health monitoring systems has gradually increased in the management of various chronic diseases, especially in the prevention and early intervention of respiratory diseases. Respiratory diseases, including chronic obstructive pulmonary disease (COPD), asthma and pneumonia, are diseases with high morbidity and mortality rates worldwide. Early symptoms are usually not obvious, but the disease deteriorates rapidly and the symptoms are acute. Therefore, an intelligent system that can monitor respiratory physiological data in real time, accurately identify abnormal changes, quickly warn and provide response suggestions is of great significance to improving patient prognosis and reducing medical burden. Most of the current respiratory monitoring systems rely on single or limited sensor data (such as blood oxygen monitoring, heart rate monitoring, etc.), and mainly focus on the setting of static thresholds, which are difficult to adapt to individual differences and dynamic physiological changes. The traditional single threshold setting method often leads to false alarms and missed alarms, which is not conducive to the long-term monitoring and management of chronic diseases.

[0003] In summary, in the process of implementing intelligent early warning in the respiratory department, the above-mentioned technologies have technical problems such as inaccurate fusion of multi-source physiological data and low accuracy in judging the patient's health status. Summary of the invention

[0004] The present invention provides a respiratory intelligent early warning system based on data fusion to solve the technical problems of inaccurate fusion of multi-source physiological data and low accuracy in judging the health status of patients in the current respiratory monitoring system during the implementation of respiratory intelligent early warning.

[0005] The present invention provides a respiratory intelligent early warning system based on data fusion, which specifically includes the following technical solutions:

[0006] A respiratory department intelligent early warning method based on data fusion, comprising the following steps:

[0007] S1. Monitor and collect multi-source physiological data of patients to obtain original physiological data; pre-process the original physiological data to obtain pre-processed physiological data; perform intelligent fusion processing on the pre-processed physiological data to obtain comprehensive physiological characteristic data;

[0008] S2. Perform anomaly detection and analysis on the comprehensive physiological characteristic data to obtain the analysis results of the respiratory status; classify the risk level based on the analysis results of the respiratory status; and formulate medical recommendations based on the risk level.

[0009] Preferably, the S1 specifically includes:

[0010] The pre-processed physiological data is intelligently fused and processed by a data fusion algorithm based on multi-level dynamic weighting and hierarchical feature extraction to obtain comprehensive physiological characteristic data; the pre-processed physiological data is a data set containing time series attributes.

[0011] Preferably, the S1 specifically includes:

[0012] In the process of implementing the data fusion algorithm based on multi-level dynamic weighting and hierarchical feature extraction, the preprocessed physiological data is subjected to feature extraction to obtain physiological feature data; the time series correlation coefficient is introduced, and the formula is as follows:

[0013] ,

[0014] in, is the temporal correlation coefficient; is the length of the entire time series; For the Physiological characteristics data The value of the moment; For the The mean of physiological characteristic data in time series; For the Physiological characteristics data The value of the moment; For the The mean of physiological characteristic data in time series;

[0015] Based on the time series correlation coefficient and the change rate of the physiological characteristic data, the initial weighted weights of the physiological characteristic data are obtained; based on the initial weighted weights of the physiological characteristic data, an initial weighted weight matrix is ​​obtained.

[0016] Preferably, the S1 specifically includes:

[0017] Based on the initial weighted weight matrix, the physiological characteristic data is subjected to multi-feature weighted fusion processing to obtain comprehensive physiological characteristic data; comprehensive physiological characteristic data Any element in The calculation formula is as follows:

[0018] ,

[0019] in, It is The initial weighting of each physiological characteristic data; For the Adaptive coefficient of physiological characteristic data; For the Physiological characteristic data; For the Periodic parameters of physiological characteristic data;

[0020] The characteristic function that changes dynamically over time is introduced to expand the comprehensive physiological characteristic data into time series dynamic data.

[0021] Preferably, the S1 specifically includes:

[0022] Based on the comprehensive physiological characteristic data, a correction coefficient of each physiological characteristic data is calculated; based on the correction coefficient of the physiological characteristic data, the initial weighted weight of the physiological characteristic data is dynamically adjusted to obtain the adjusted weight of the physiological characteristic data.

[0023] Preferably, the S1 specifically includes:

[0024] Based on the current comprehensive physiological characteristic data, the historical data memory matrix is ​​introduced to adaptively optimize the adjusted weights of each physiological characteristic data to obtain the final optimized weights; the adaptive optimization formula is as follows:

[0025] ,

[0026] in, After final optimization The weight of each physiological characteristic data represents the final optimized weight; For the The weight of each physiological characteristic data after adjustment; is the learning rate; It is the memory window period; It is Physiological characteristics data The value of the moment; It is the comprehensive physiological characteristic data in The value of the moment;

[0027] Based on the final optimized weights, the comprehensive physiological characteristic data are recalculated.

[0028] Preferably, the S2 specifically includes:

[0029] Through a combined model based on convolutional neural networks and long short-term memory networks, feature processing and anomaly detection analysis are performed on the comprehensive physiological characteristic data to obtain a health risk score; based on the health risk score, a personalized adaptive threshold generation algorithm is introduced to dynamically generate personalized dynamic thresholds.

[0030] Preferably, the S2 specifically includes:

[0031] In the process of implementing the personalized adaptive threshold generation algorithm, the patient's historical health risk score and individual characteristic data are introduced to calculate the basic threshold; based on the basic threshold, combined with the patient's current health risk score, the basic threshold is dynamically adjusted to generate a personalized dynamic threshold; the dynamic adjustment formula is as follows:

[0032] ,

[0033] in, is the personalized dynamic threshold; is the basic threshold; is the influence coefficient of controlling adaptation; is the total number of disease histories in the individual characteristic data; It is The weight coefficient of physiological characteristic data related to each disease; It is Physiological characteristic values ​​related to diseases; It is the current moment Health risk scores; is the smoothing factor.

[0034] Preferably, the S2 specifically includes:

[0035] The health risk score is compared with the personalized dynamic threshold to obtain the analysis result of the respiratory status; when the health risk score is less than the personalized dynamic threshold, it means that the patient is in a normal respiratory state; when the health risk score is greater than or equal to the personalized dynamic threshold, it means that the analysis result of the respiratory status is abnormal; when the analysis result of the respiratory status is abnormal, the risk level is divided by calculating the difference between the patient's current health risk score and the personalized dynamic threshold, and hierarchical medical advice is generated based on the risk level.

[0036] An intelligent early warning system for respiratory medicine based on data fusion, including the following parts:

[0037] Data acquisition module, data preprocessing module, multi-source data fusion module, anomaly detection and judgment module, early warning and suggestion module;

[0038] The data acquisition module monitors and collects the multi-source physiological data of the patient to obtain the original physiological data; and sends the original physiological data to the data preprocessing module;

[0039] The data preprocessing module preprocesses the original physiological data to obtain preprocessed physiological data; and sends the preprocessed physiological data to the multi-source data fusion module;

[0040] The multi-source data fusion module performs intelligent fusion processing on the pre-processed physiological data to obtain comprehensive physiological characteristic data; the comprehensive physiological characteristic data is sent to the abnormality detection and judgment module;

[0041] The abnormality detection and judgment module performs abnormality detection and analysis on the comprehensive physiological characteristic data to obtain the analysis results of the respiratory state; the analysis results of the respiratory state are sent to the early warning and suggestion module;

[0042] The early warning and advice module divides the risk level based on the analysis results of the respiratory status and formulates medical advice based on the risk level.

[0043] The beneficial effects of the technical solution of the present invention are:

[0044] 1. The present invention uses pre-processing technologies such as data cleaning, data denoising, data standardization, outlier monitoring and elimination, and a data fusion algorithm based on multi-level dynamic weighting and hierarchical feature extraction to achieve intelligent fusion of multi-source data. By introducing advanced methods such as time series correlation and frequency fluctuation analysis, comprehensive physiological characteristic data is generated to ensure the reliability and consistency of physiological characteristic data, providing a high-precision physiological data foundation for abnormality detection.

[0045] 2. Introduce a personalized adaptive threshold generation algorithm to generate personalized dynamic thresholds that adapt to individual differences based on the patient's historical health risk score and individual characteristic data, thereby enhancing the personalized adaptability of the respiratory department's intelligent early warning system, making the early warning results more targeted and effectively reducing false alarms and missed alarms. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a structural diagram of a respiratory department intelligent early warning system based on data fusion according to the present invention;

[0047] Figure 2 This is a flow chart of the intelligent early warning method for respiratory medicine based on data fusion described in the present invention. DETAILED DESCRIPTION

[0048] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] Unless defined otherwise, 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 belongs.

[0050] The specific scheme of the respiratory department intelligent early warning system based on data fusion provided by the present invention is described in detail below with reference to the accompanying drawings.

[0051] Refer to the attached Figure 1 , which shows a structure diagram of a respiratory department intelligent early warning system based on data fusion provided by an embodiment of the present invention, the system includes the following parts:

[0052] Data acquisition module, data preprocessing module, multi-source data fusion module, anomaly detection and judgment module, early warning and suggestion module;

[0053] The data acquisition module monitors and collects the multi-source physiological data of the patient to obtain the original physiological data; and sends the original physiological data to the data preprocessing module;

[0054] The original physiological data include parameters such as respiratory rate, blood oxygen saturation, heart rate, body temperature, etc.;

[0055] The data preprocessing module preprocesses the original physiological data to obtain preprocessed physiological data; and sends the preprocessed physiological data to the multi-source data fusion module;

[0056] The multi-source data fusion module performs intelligent fusion processing on the pre-processed physiological data to obtain comprehensive physiological characteristic data; the comprehensive physiological characteristic data is sent to the abnormality detection and judgment module;

[0057] The abnormality detection and judgment module performs abnormality detection and analysis on the comprehensive physiological characteristic data to obtain the analysis results of the respiratory state; the analysis results of the respiratory state are sent to the early warning and suggestion module;

[0058] The early warning and advice module divides the risk level based on the analysis results of the respiratory status; based on the risk level, medical advice is formulated to achieve intelligent early warning of the respiratory department based on data fusion.

[0059] Refer to the attached Figure 2 , which shows a flow chart of a respiratory department intelligent early warning method based on data fusion provided by an embodiment of the present invention, the method comprising the following steps:

[0060] S1. Monitor and collect multi-source physiological data of patients to obtain original physiological data; pre-process the original physiological data to obtain pre-processed physiological data; perform intelligent fusion processing on the pre-processed physiological data to obtain comprehensive physiological characteristic data;

[0061] Based on the patient's health needs, expert experience is used to select multiple sensors for real-time physiological parameter monitoring to obtain original physiological data; the sensors include respiratory sensors, blood oxygen sensors, heart rate sensors, body temperature sensors, etc.; the original physiological data includes parameters such as respiratory rate, blood oxygen saturation, heart rate, and body temperature.

[0062] Furthermore, the original physiological data is periodically preprocessed (such as data cleaning, data denoising, data standardization, outlier detection and elimination, and data format conversion) to obtain preprocessed physiological data; the techniques used in the preprocessing are all well-known technical means in the art and will not be described in detail here.

[0063] Among them, for a certain time (remember The original physiological data collected within (length of the time series) is preprocessed, and the preprocessed physiological data obtained is also a data set containing time series attributes.

[0064] Furthermore, the pre-processed physiological data is intelligently fused using a data fusion algorithm based on multi-level dynamic weighting and hierarchical feature extraction to obtain comprehensive physiological characteristic data; the data fusion algorithm based on multi-level dynamic weighting and hierarchical feature extraction gradually forms comprehensive and high-precision comprehensive physiological characteristic data by introducing multi-level processing of adaptive weights, time series correlation and frequency fluctuation analysis. The specific implementation process is as follows:

[0065] Firstly, the preprocessed physiological data is subjected to feature extraction using existing feature engineering techniques to obtain physiological feature data;

[0066] Furthermore, the timing correlation coefficient is introduced It represents the temporal correlation between the physiological characteristic data. The calculation formula is as follows:

[0067] ,

[0068] in, is the time series correlation coefficient, indicating the The physiological characteristics data and The degree of temporal correlation between physiological characteristic data (e.g., respiratory rate and blood oxygen saturation); is the number of time points, i.e. the length of the entire time series; For the Physiological characteristics data The value of the moment; For the The mean of physiological characteristic data in time series; For the Physiological characteristics data The value of the moment; For the The mean of physiological characteristic data in the time series. The calculation of the time series correlation coefficient can ensure that the time series correlation between physiological characteristic data can be fully considered in the subsequent fusion process;

[0069] Furthermore, the time series correlation coefficient and the change rate of the physiological characteristic data are used to form the initial weighted weight of the physiological characteristic data; the calculation formula of the initial weighted weight of the physiological characteristic data is:

[0070] ,

[0071] in, It is The initial weighting of each physiological characteristic data; It is The sum of all time series correlation coefficients of physiological characteristic data, that is, ; For the The change rate of each physiological characteristic data reflects the degree of change of each physiological characteristic data within a specific time; For the Physiological characteristics data The standard deviation of The volatility of physiological characteristic data; is the total number of physiological characteristic data; It is The sum of all time series correlation coefficients of physiological characteristic data, that is, ; For the The rate of change of physiological characteristic data; For the Physiological characteristics data Further, the initial weighted weight matrix is ​​obtained based on the initial weighted weights of the physiological characteristic data. ; It is The initial weighting of each physiological feature data.

[0072] Furthermore, the physiological characteristic data is subjected to multi-feature weighted fusion processing to form comprehensive physiological characteristic data; based on the initial weighted weight matrix , the physiological characteristic data are superimposed and nonlinearly combined to generate comprehensive physiological characteristic data Any element in The calculation formula is as follows:

[0073] ,

[0074] in, For the The adaptive coefficient of each physiological characteristic data is used to control the influence of each physiological characteristic data in the nonlinear transformation; For the Physiological characteristic data; For the The mean of physiological characteristic data in time series; For the The periodic parameters of physiological characteristic data are used to control the time period influence of physiological characteristic data; through the above processing, it is ensured that different physiological characteristic data can be complexly integrated on a unified scale, so as to obtain more comprehensive comprehensive physiological characteristic data. .

[0075] Obtaining comprehensive physiological characteristics data Finally, in order to enhance the temporal expressivity of the comprehensive physiological characteristic data, a characteristic function that changes dynamically over time is further generated to give the comprehensive physiological characteristic data a higher time sensitivity, that is, to expand the comprehensive physiological characteristic data into temporal dynamic data. The specific formula is as follows:

[0076] ,

[0077] in, Represents the adaptive frequency, which is used to control the speed of dynamic changes; It is a time variable, which is used to reflect the time sensitivity of comprehensive physiological characteristic data; Indicates Physiological characteristic data; It is the sum of the absolute values ​​of the physiological characteristic data and is used to control the amplitude of the dynamic gain. The characteristic function assigns the comprehensive physiological characteristic data With time-varying characteristics, comprehensive physiological characteristic data It can reflect the patient's physiological changes in real time.

[0078] In order to further improve the adaptive ability of physiological characteristic data during long-term monitoring, the initial weights of different physiological characteristic data are dynamically adjusted. By introducing the historical data memory matrix, the influence of each physiological characteristic data is dynamically corrected. The specific implementation process is as follows:

[0079] First, calculate the correction coefficient for each physiological characteristic data , to reflect the difference between the physiological characteristic data and the comprehensive physiological characteristic data, the formula is as follows:

[0080] ,

[0081] in, It is The correction coefficient of each physiological characteristic data reflects the deviation between the physiological characteristic data and the comprehensive physiological characteristic data; The total duration of monitoring; Indicates that comprehensive physiological characteristics data is The value of the moment; Indicates Physiological characteristics data The value of the moment; Indicates comprehensive physiological characteristics data The correction coefficient of the physiological characteristic data forms adaptive feedback information according to the degree of deviation between the physiological characteristic data and the comprehensive physiological characteristic data, so as to adjust the initial weight matrix in the subsequent steps;

[0082] Furthermore, the initial weight of the physiological characteristic data is dynamically adjusted based on the correction coefficient of the physiological characteristic data to obtain the adjusted weight of the physiological characteristic data; the specific formula is:

[0083] ,

[0084] in, For the The weight of each physiological characteristic data after adjustment; It is Correction coefficients for physiological characteristic data; It is Correction coefficients for physiological characteristic data; An adaptive adjustment index is used to control the update rate; For the The initial weights of the physiological characteristic data. Based on the adjusted weights of the physiological characteristic data, the adjusted weight matrix is ​​constructed It is used to recalculate the comprehensive physiological characteristics data, so that the stability and accuracy of the comprehensive physiological characteristics data are further improved. Indicates The weight of each physiological characteristic data is adjusted.

[0085] Furthermore, in order to improve the accuracy of the adjusted weights of physiological characteristic data, the historical data memory matrix is ​​introduced , used to record a certain period of time in the past The comprehensive physiological characteristic data and weight adjustment in the memory matrix provide a reference for adaptive learning optimization, where historical data comes from the existing database. Recording time period Comprehensive physiological characteristics data within And the weight change at each time step is as follows:

[0086] ,

[0087] in, To integrate physiological characteristics data The value of the moment; To integrate physiological characteristics data The value of the moment; for The adjusted weight matrix at time; for The adjusted weight matrix at time; The memory window period is used to refer to historical physiological characteristic data in subsequent optimization to optimize the intelligent fusion processing of current physiological characteristic data. Through the historical data memory matrix, historical data can be called to further optimize the accuracy of the intelligent fusion of current physiological characteristic data.

[0088] Based on historical data memory matrix and current comprehensive physiological characteristics data , the weights of each physiological characteristic data after adjustment are adaptively optimized to obtain the final optimized weights; the adaptive optimization formula is as follows:

[0089] ,

[0090] in, After final optimization The weight of the physiological characteristic data, that is, the final optimized weight, is calculated by The weights after adjusting the physiological characteristics data After adaptive optimization, we get: is the learning rate, which is used to control the adjustment rate; It is The periodic parameters of physiological characteristic data are used to control the sinusoidal fluctuations; It is Physiological characteristics data The value of the moment; It is the comprehensive physiological characteristic data in The adaptive optimization formula uses the difference between the current comprehensive physiological characteristic data and the historical data in the historical data memory matrix as feedback information, and automatically corrects the weight of the physiological characteristic data, so that each physiological characteristic data can better adapt to the individual physiological changes of the patient. Within a certain period, the final optimized weight is used to recalculate the comprehensive physiological characteristic data, so that the stability and accuracy of the comprehensive physiological characteristic data are further improved.

[0091] S2. Perform anomaly detection and analysis on the comprehensive physiological characteristic data to obtain the analysis results of the respiratory status; classify the risk level based on the analysis results of the respiratory status; and formulate medical recommendations based on the risk level.

[0092] The existing combined model based on convolutional neural network and long short-term memory network is used for the comprehensive physiological characteristic data to realize feature processing and anomaly detection analysis. The convolutional neural network is used to optimize the features of the comprehensive physiological characteristic data to obtain representative physiological characteristic data. The representative physiological characteristic data is used as the output of the convolutional neural network and input into the long short-term memory network. The time series characteristics of the representative physiological characteristic data are processed by the long short-term memory network to capture the long-term dependence of the physiological characteristic data and obtain a health risk score. The health risk score is compared with the personalized dynamic threshold to obtain the analysis result of the respiratory state.

[0093] A personalized adaptive threshold generation algorithm is introduced to dynamically generate personalized dynamic thresholds based on health risk scores, patient historical health risk scores, and individual characteristic data (such as age, gender, and disease history). The specific implementation process is as follows:

[0094] Calculate the basic threshold based on the patient's historical health risk score and individual characteristic data (such as age, gender, medical history, etc.) :

[0095] ,

[0096] in, is the basic threshold, which is calculated based on the patient's historical health risk score and individual characteristic data and is used as the basis for generating personalized dynamic thresholds; is the total number of times the patient's historical health risk score is scored; The patient times historical health risk score; is the mean of the patient’s historical health risk scores, used to measure the average health risk level; is the standard deviation of the patient's historical health risk score, which is used to measure the dispersion of the patient's historical health risk score; is the age factor; It is the adjustment factor of individual characteristic data, which is used to control the impact of individual characteristic data on the basic threshold; It is an adjustment factor used to avoid excessive influence of age factor on the basic threshold; is the gender impact factor, which is used to adjust the impact of gender on the basic threshold. is the gender, 1 for male and 0 for female.

[0097] Furthermore, based on the basic threshold, the basic threshold is dynamically adjusted in combination with the patient's current health risk score and the disease history in the individual characteristic data to generate a threshold that adapts to the patient's current state, that is, a personalized dynamic threshold; the dynamic adjustment formula is as follows:

[0098] ,

[0099] in, is the personalized dynamic threshold; It is the influence coefficient of controlling the adaptation, which is used to adaptively adjust the influence of the current health risk score on the personalized dynamic threshold; is the total number of disease histories; It is The weight coefficient of the physiological characteristic data related to each disease is used to characterize the weight of a specific disease in the health risk score; It is The physiological characteristic values ​​related to each disease are derived from the individual characteristic data of the patient (e.g. ,otherwise ); It is the current moment The health risk score represents the health status indicator measured in real time; is a smoothing factor used to prevent the health risk score Mutations cause dramatic fluctuations in personalized dynamic thresholds.

[0100] Furthermore, the health risk score is compared with the personalized dynamic threshold to obtain the analysis result of the respiratory state. When the health risk score is greater than or equal to the personalized dynamic threshold, it means that the analysis result of the respiratory state is abnormal, indicating that the patient's respiratory state is abnormal; when the health risk score is less than the personalized dynamic threshold, it means that the patient is in a normal respiratory state. When the analysis result of the respiratory state is abnormal, a hierarchical medical recommendation is generated based on the patient's current health risk score and personalized dynamic threshold. To classify the risk level and determine the severity of the abnormality, the risk level classification is calculated as follows:

[0101] ,

[0102] like , then it is low risk; if , then it is medium risk; if , it is high risk. , It is a grading threshold set based on expert experience and is used to control the risk level classification.

[0103] Medical recommendations are made based on the level of risk:

[0104] Low risk: Patients are advised to monitor themselves and adjust their breathing status, improve rest, and avoid excessive fatigue. It is also recommended to increase daily breathing training to enhance the endurance of the respiratory system.

[0105] Medium risk: Patients are advised to see a doctor or contact medical staff to check for respiratory infections or other discomforts; consider using oxygen to assist breathing and avoid contact with possible allergens or respiratory irritants;

[0106] High risk: It is recommended to seek medical attention immediately and receive professional medical intervention. Oxygen support or ventilator assistance may be required. Further blood oxygen testing and lung function tests are recommended, and causes such as pneumonia and acute asthma attacks may be considered based on the patient's medical history.

[0107] In summary, an intelligent early warning system for respiratory diseases based on data fusion was completed.

[0108] The order of the embodiments of the invention is for description only and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0109] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0110] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A respiratory intelligent early warning method based on data fusion, characterized in that: The following steps are involved: S1. Monitor and collect multi-source physiological data of patients to obtain original physiological data; Preprocessing the original physiological data to obtain preprocessed physiological data; The preprocessed physiological data is a data set containing time series attributes; the preprocessed physiological data is subjected to feature extraction through a data fusion algorithm based on multi-level dynamic weighting and hierarchical feature extraction to obtain physiological feature data; the time series correlation coefficient is introduced, and the formula is as follows: ， in, is the temporal correlation coefficient; is the length of the entire time series; For the Physiological characteristics data The value of the moment; For the The mean of physiological characteristic data in time series; For the Physiological characteristics data The value of the moment; For the The mean of physiological characteristic data in time series; Based on the time series correlation coefficient and the change rate of the physiological characteristic data, the initial weighted weight of the physiological characteristic data is obtained; based on the initial weighted weight of the physiological characteristic data, an initial weighted weight matrix is ​​obtained; based on the initial weighted weight matrix, multi-feature weighted fusion processing is performed on the physiological characteristic data to obtain comprehensive physiological characteristic data; comprehensive physiological characteristic data Any element in The calculation formula is as follows: ， in, It is The initial weighting weight of each physiological characteristic data; For the Adaptive coefficient of physiological characteristic data; For the Physiological characteristic data; For the Periodic parameters of physiological characteristic data; S2. Through the combined model of convolutional neural network and long short-term memory network, the comprehensive physiological characteristic data is processed and anomaly detection is analyzed to obtain the health risk score; based on the health risk score, the patient's historical health risk score and individual characteristic data are introduced through the personalized adaptive threshold generation algorithm to calculate the basic threshold; based on the basic threshold, combined with the patient's current health risk score, the basic threshold is dynamically adjusted to generate a personalized dynamic threshold; the dynamic adjustment formula is as follows: ， in, is the personalized dynamic threshold; is the basic threshold; is the influence coefficient of controlling adaptation; is the total number of disease histories in the individual characteristic data; It is The weight coefficient of physiological characteristic data related to each disease; It is Physiological characteristic values ​​related to diseases; It is the current moment Health risk scores; is the smoothing factor; Compare the health risk score with the personalized dynamic threshold to obtain the analysis results of the respiratory status; divide the risk level based on the analysis results of the respiratory status; and formulate medical recommendations based on the risk level.

2. The intelligent early warning method for respiratory department based on data fusion according to claim 1, characterized in that: The S1 specifically includes: The characteristic function that changes dynamically over time is introduced to expand the comprehensive physiological characteristic data into time series dynamic data.

3. The intelligent early warning method for respiratory department based on data fusion according to claim 2 is characterized in that: The S1 specifically includes: Based on the comprehensive physiological characteristic data, a correction coefficient of each physiological characteristic data is calculated; based on the correction coefficient of the physiological characteristic data, the initial weighted weight of the physiological characteristic data is dynamically adjusted to obtain the adjusted weight of the physiological characteristic data.

4. The intelligent early warning method for respiratory department based on data fusion according to claim 3 is characterized in that: The S1 specifically includes: Based on the current comprehensive physiological characteristic data, the historical data memory matrix is ​​introduced to adaptively optimize the adjusted weights of each physiological characteristic data to obtain the final optimized weights; the adaptive optimization formula is as follows: ， in, After final optimization The weight of each physiological characteristic data represents the final optimized weight; For the The weight of each physiological characteristic data after adjustment; is the learning rate; It is the memory window period; It is Physiological characteristics data The value of the moment; It is the comprehensive physiological characteristic data in The value of the moment; Based on the final optimized weights, the comprehensive physiological characteristic data are recalculated.

5. The intelligent early warning method for respiratory department based on data fusion according to claim 1 is characterized in that: The S2 specifically includes: When the health risk score is less than the personalized dynamic threshold, it means that the patient is in a normal breathing state; when the health risk score is greater than or equal to the personalized dynamic threshold, it means that the analysis result of the breathing state is abnormal; when the analysis result of the breathing state is abnormal, the risk level is divided by calculating the difference between the patient's current health risk score and the personalized dynamic threshold, and hierarchical medical advice is generated based on the risk level.

6. A respiratory intelligent early warning system based on data fusion, applied to the respiratory intelligent early warning method based on data fusion as claimed in claim 1, characterized in that: Includes the following sections: Data acquisition module, data preprocessing module, multi-source data fusion module, anomaly detection and judgment module, early warning and suggestion module; The data acquisition module monitors and collects multi-source physiological data of the patient to obtain the original physiological data; Sending raw physiological data to a data preprocessing module; A data preprocessing module preprocesses the original physiological data to obtain preprocessed physiological data; Send the preprocessed physiological data to the multi-source data fusion module; The multi-source data fusion module performs intelligent fusion processing on the pre-processed physiological data to obtain comprehensive physiological characteristic data; Sending the comprehensive physiological characteristic data to the abnormality detection and determination module; The abnormality detection and judgment module performs abnormality detection and analysis on the comprehensive physiological characteristic data to obtain the analysis results of the respiratory state; Send the analysis results of the respiratory status to the early warning and suggestion module; The warning and suggestion module divides the risk level based on the analysis results of the respiratory status; Based on the level of risk, medical recommendations are made.

Images