Postoperative vital signs monitoring system based on medical information processing

Abstract

The present invention discloses a postoperative vital sign monitoring system based on medical information processing, including a data acquisition module, a vital sign feature extraction module, a postoperative vital sign evaluation model establishment module and a postoperative vital sign monitoring module. The present invention belongs to the field of medical information processing, and specifically refers to a postoperative vital sign monitoring system based on medical information processing. The present scheme identifies the fluctuation trend at different time points; quantifies the total change trend of heart rate fluctuation within the monitoring time; comprehensively analyzes the changes in blood pressure and heart rate to provide a quantitative basis for the postoperative recovery process; improves the vital sign monitoring effect; focuses on the key time period after surgery and dynamically adjusts it through the time weight function, the recovery stage weight and the environmental correction factor; introduces deviation weights for the fluctuations of heart rate and blood pressure, dynamically adjusts the loss coefficient, designs a segmented loss function according to the vital sign value, handles the situations of different risk intervals respectively, and adapts to the individual differences of postoperative patients and the uncertainty of vital sign fluctuations; thereby improving the accuracy of postoperative vital sign monitoring.

Description

Technical Field

[0001] The present invention relates to the field of medical information processing, and in particular to a postoperative vital sign monitoring system based on medical information processing. Background Art

[0002] The postoperative vital sign monitoring system is a medical auxiliary system that is mainly used to continuously monitor, analyze and evaluate the physiological vital sign data of postoperative patients, so as to help medical staff understand the patient's postoperative recovery, detect potential risks and abnormalities, and take timely intervention measures to ensure patient safety. However, the general postoperative vital sign monitoring system has low sensitivity to abnormal vital signs and cannot capture the dynamic characteristics of vital sign changes, which leads to poor postoperative vital sign monitoring effects; the general postoperative vital sign monitoring system has weak ability to improve the evaluation of abnormal vital signs and cannot balance the diversity of vital signs and individual characteristic differences, which leads to poor accuracy of postoperative vital sign monitoring. Summary of the invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a postoperative vital sign monitoring system based on medical information processing. In view of the problem that the general postoperative vital sign monitoring system has low sensitivity to abnormal vital signs and cannot capture the dynamic characteristics of vital sign changes, which leads to poor postoperative vital sign monitoring effect, this solution is based on calculating the time difference of blood pressure fluctuations to identify the fluctuation trend at different time points; quantifying the total change trend of heart rate fluctuations within the monitoring time to reflect the overall level of postoperative heart activity; comprehensively analyzing blood pressure and heart rate changes, calculating the degree of postoperative vital sign activity through energy changes, and providing a quantitative basis for the postoperative recovery process; extracting the slope peaks of the heart rate and blood pressure curves, identifying abnormal fluctuations, and improving the sensitivity to abnormal vital signs; thereby comprehensively optimizing the basic vital sign data and improving the vital sign monitoring effect. Results: In view of the problem that the general postoperative vital sign monitoring system has weak ability to improve the assessment of abnormal signs, cannot balance the diversity of signs and individual characteristic differences, and thus leads to poor accuracy of postoperative vital sign monitoring, this solution focuses on the key time period after surgery through dynamic adjustment of time weight function, recovery stage weight and environmental correction factor, so as to more accurately reflect the fluctuation characteristics of postoperative vital signs; introduces deviation weights for fluctuations in heart rate and blood pressure, dynamically adjusts the loss coefficient, designs a segmented loss function according to the distance between the vital sign value and the target range, handles the situations in different risk intervals respectively, and improves the ability to assess the comprehensive vital sign risk; calculates the risk expectation value based on the dynamic health score, refines and adapts to the individual differences of postoperative patients and the uncertainty of vital sign fluctuations; thereby improving the accuracy of postoperative vital sign monitoring.

[0004] The technical solution adopted by the present invention is as follows: the postoperative vital signs monitoring system based on medical information processing provided by the present invention comprises a data acquisition module, a vital signs feature extraction module, a postoperative vital signs evaluation model establishment module and a postoperative vital signs monitoring module;

[0005] The data acquisition module collects historical postoperative medical data;

[0006] The physical sign feature extraction module extracts the time difference feature, cumulative integral feature, energy change feature and slope peak feature of the basic physical sign, and optimizes the postoperative physical sign data to construct a time series data set;

[0007] The postoperative physical sign evaluation model establishment module captures the long-term dependency of time series through residual bidirectional LSTM, combines the time-related weight function, the recovery stage weight and the environmental correction factor, focuses on the abnormal evaluation of the key time period after surgery, and thus realizes the establishment of the postoperative physical sign evaluation model;

[0008] The postoperative vital sign monitoring module performs vital sign monitoring on the postoperative medical data collected in real time based on the established postoperative vital sign evaluation model.

[0009] Furthermore, in the data acquisition module, the historical postoperative medical data includes basic vital signs data, postoperative biochemical indicators, behavioral data, time, surgical procedure category and abnormal vital signs risk assessment level; the abnormal vital signs risk assessment level includes normal, low risk, medium risk and high risk; the collected data is converted and standardized.

[0010] Furthermore, the physical sign feature extraction module optimizes the basic physical sign data, specifically including the following contents:

[0011] Time difference feature; calculate the time difference of blood pressure fluctuations at different time points after surgery , expressed as: ;in, and They are the timestamps of the blood pressure values ​​at the current time point and the reference time point respectively; It’s blood pressure;

[0012] Cumulative integral feature: Calculate the cumulative integral value of heart rate fluctuation, expressed as: ; Wherein, t1 and t2 are the start and end time of vital sign monitoring; HR(·) is the heart rate; t is the time;

[0013] Energy change characteristics: The energy change is calculated based on the total amount of changes in the postoperative physical sign data, expressed as: ; Among them, E2 is the energy change characteristic of physical signs, reflecting the degree of physical sign activity after surgery; BP(·) is blood pressure;

[0014] Slope peak feature: extract the peak value through the slope change of the heart rate and blood pressure curves, indicating: ;in, is the slope peak feature, which indicates the maximum rate of change of heart rate relative to time and is used to detect abnormal fluctuations; and They represent the heart rate difference between two adjacent moments and the time difference between two adjacent acquisition times respectively;

[0015] The basic sign data were replaced by the sign feature extraction results to construct a postoperative sign time series dataset, which was divided into a test set and a training set.

[0016] The postoperative physical sign assessment model establishment module specifically includes the following contents:

[0017] Dependency capture unit: introduce residual Bi-LSTM to capture the long-term dependency in the postoperative vital sign time series; specifically: input the vital sign time series data, and use the bidirectional LSTM to capture the dependency between the previous and next time steps, expressed as: ; Add residual connection to the LSTM output to retain the original input feature information, expressed as: ; Introduce time-dependent weight function f1(t), recovery phase weight and environmental correction factors , based on the attention mechanism, focusing on the key time period after surgery, expressed as: ; ; ; ; ; ;in, is the bidirectional LSTM output of the previous time step; and They are the forward LSTM output and the reverse LSTM output respectively; is the output after residual connection; and are the weight and bias of the residual connection respectively; is the input feature of the current time step; is the output after nonlinear transformation; , and They are the early postoperative stage, the postoperative stable stage and the critical window period. , and is the weight parameter of each stage; tanh(·) is the tanh activation function; is the attention weight; T is the total time step; is the final representation of the postoperative vital sign time series; Dropout(·) is a regularization technique; It is a critical time point; is the width of the time distribution; is the time step decay coefficient; and They are the average risk level of historical patients after surgery and the current surgical procedure category. and is the environmental coefficient; exp(·) is the exponential function;

[0018] Convolution and pooling unit: Perform convolution and pooling operations on the vital sign time series to extract high-level time series features and reduce the dimension; specifically: transpose the input tensor, exchange the time step and feature dimension, expressed as: ; Apply one-dimensional convolution, expressed as: ; Apply the average pooling operation, expressed as: ;in, is the transposed input tensor; Transpose(·) is the tensor dimension exchange operation; X is the time series of postoperative vital sign monitoring data; It is the time dimension; is the feature dimension; Y(·) is the output after the convolution operation; K(·) is the convolution kernel weight; X(·) is the sliding window data; b is the convolution pooling bias; f(·) is the activation function; a is the sliding window index; M is the sliding window size; k3 is the pooling window, is the pooling window index; is the time series data within the pooling window; is the average pooling output;

[0019] Loss function design unit; define the single feature loss function , introduce bias weight and , expressed as: ; ; ; The physical signs are weighted averagely as the final loss; The physical sign abnormality risk assessment level is used as the data label of the physical sign abnormality risk assessment unit; Whether there is a risk is used as the data label of other units; Among them, is the mean square error loss function; and is the loss increment factor; and They are the risk loss values ​​when the physical sign is below the lower limit and the upper limit respectively; and are the lower and upper limits of the target range, respectively; and are the baseline weighting factors for heart rate fluctuation and blood pressure, respectively; and are the heart rate variability and the baseline value of heart rate variability at the current time step, respectively; and are the blood pressure value and the baseline value of blood pressure at the current time step, respectively; m is the target value; e is the natural base number;

[0020] Abnormal physical sign risk assessment unit: conduct risk assessment of specific abnormal physical signs after surgery, including:

[0021] Calculate the dynamic health score; specifically: calculate the health score of each sign, expressed as: ; Get the comprehensive health score OS, expressed as: ;in, It is the physical health score; is the health score of the physical sign in the neighborhood of the target value m; and is the rate of change of the score on both sides of the target value; N is the number of signs; is the sign weight; is the i-th sign; y is the sign;

[0022] The fluctuation of postoperative patient's physical signs is uncertain. The distribution of physical sign values ​​is modeled with triangular distribution, the expected loss is calculated, and the optimal physical sign target value is determined. The physical signs of postoperative patients are regarded as random variables, and the target range is represented by triangular distribution. ; The probability density function F(·) of the triangular distribution of postoperative patients is expressed as: ; Where a, b and c are the minimum, target and maximum values ​​of the sign respectively; T is the tolerance range; x is the offset of the process mean;

[0023] For single-sign data, the expected gain loss value within the target range It is expressed as, ; For abnormal signs, the truncated normal distribution is used to model and calculate the expected risk loss. When the sign distribution is truncated by the upper and lower limits, the probability density of the truncated normal distribution is expressed as, ; The extreme sign loss is obtained as the total risk expectation value E, expressed as: ;in, is the basic gain value, which indicates the health gain when the physical sign value is within the target range; and is the loss coefficient, which describes the growth rate of the loss on both sides of the target value; and They are the lower and upper limits of the sign distribution, respectively; is the difference in the cumulative distribution function;

[0024] Risk assessment of abnormal physical signs; Defined as normal; Defined as low risk; Defined as medium risk; Defined as high risk; , and is the risk assessment threshold; based on the total risk expectation, the abnormal risk assessment result of the postoperative patient's vital sign data is obtained; when both the double-label data meet the standards based on the training set and the test set, the postoperative vital sign assessment model is established.

[0025] Furthermore, the postoperative vital sign monitoring module is based on the established postoperative vital sign assessment model, which collects the patient's postoperative medical data in real time, and inputs the data into the postoperative vital sign assessment model after being processed by the vital sign feature extraction module. The abnormal vital sign risk assessment level output by the model is used as the assessment result, and an early warning is issued when the assessment result is medium risk or high risk, so as to realize postoperative vital sign monitoring.

[0026] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0027] (1) In view of the problem that general postoperative vital sign monitoring systems have low sensitivity to abnormal vital signs and cannot capture the dynamic characteristics of vital sign changes, which leads to poor postoperative vital sign monitoring effects, this solution is based on calculating the time difference of blood pressure fluctuations to identify the fluctuation trend at different time points; quantifying the total change trend of heart rate fluctuations during the monitoring time to reflect the overall level of postoperative cardiac activity; comprehensively considering blood pressure and heart rate changes, calculating the degree of postoperative vital sign activity through energy changes, and providing a quantitative basis for the postoperative recovery process; extracting the slope peak of the heart rate and blood pressure curves to identify abnormal fluctuations and improve the sensitivity to abnormal vital signs; and thus comprehensively optimizing the basic vital sign data to improve the vital sign monitoring effect.

[0028] (2) In view of the fact that general postoperative vital sign monitoring systems have weak ability to improve the assessment of abnormal vital signs and cannot balance the diversity of vital signs and individual characteristic differences, which leads to poor accuracy of postoperative vital sign monitoring, this solution focuses on the key time period after surgery and dynamically adjusts it through time weight function, recovery stage weight and environmental correction factor to more accurately reflect the fluctuation characteristics of postoperative vital signs; introduces deviation weights for fluctuations in heart rate and blood pressure, dynamically adjusts the loss coefficient, and designs a segmented loss function based on the distance between the vital sign value and the target range to handle different risk intervals respectively, thereby improving the ability to assess comprehensive vital sign risks; calculates the expected risk value based on the dynamic health score, refines and adapts to the individual differences of postoperative patients and the uncertainty of vital sign fluctuations; thereby improving the accuracy of postoperative vital sign monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 A schematic diagram of the process of a postoperative vital sign monitoring system based on medical information processing provided by the present invention;

[0030] Figure 2Schematic diagram of the process of establishing modules for the postoperative vital signs assessment model.

[0031] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0032] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0033] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0034] Example 1, see Figure 1 The postoperative vital signs monitoring system based on medical information processing provided by the present invention includes a data acquisition module, a vital signs feature extraction module, a postoperative vital signs evaluation model establishment module and a postoperative vital signs monitoring module;

[0035] The data acquisition module collects historical postoperative medical data; and sends the data to the physical sign feature extraction module;

[0036] The physical sign feature extraction module extracts the time difference feature, cumulative integral feature, energy change feature and slope peak feature of the basic physical sign, optimizes the postoperative physical sign data to construct a time series data set; and sends the data to the postoperative physical sign evaluation model establishment module;

[0037] The postoperative physical sign evaluation model establishment module captures the long-term dependency of time series through residual bidirectional LSTM, combines the time-related weight function, the recovery stage weight and the environmental correction factor, focuses on the abnormal evaluation of the key time period after surgery, and then realizes the establishment of the postoperative physical sign evaluation model; and sends the data to the postoperative physical sign monitoring module;

[0038] The postoperative vital sign monitoring module performs vital sign monitoring on the postoperative medical data collected in real time based on the established postoperative vital sign evaluation model.

[0039] Example 2, see Figure 1This embodiment is based on the above embodiment. In the data acquisition module, the historical postoperative medical data includes basic vital signs data, postoperative biochemical indicators, behavioral data, time, surgical procedure category and abnormal physical signs risk assessment level; the abnormal physical signs risk assessment level includes normal, low risk, medium risk and high risk; the collected data is converted and standardized; the basic vital signs data includes blood pressure, heart rate, blood oxygen saturation, body temperature and respiratory rate; the postoperative biochemical indicators include C-reactive protein, white blood cell count, platelet count, blood sugar, urea nitrogen and prothrombin time; the behavioral data includes activity intensity, resting time, exercise steps and sleep duration.

[0040] Example 3, see Figure 1 This embodiment is based on the above embodiment. The physical sign feature extraction module optimizes the basic physical sign data, and specifically includes the following contents:

[0041] Time difference feature; calculate the time difference of blood pressure fluctuations at different time points after surgery , expressed as: ;in, and They are the timestamps of the blood pressure values ​​at the current time point and the reference time point respectively; It’s blood pressure;

[0042] Cumulative integral feature: Calculate the cumulative integral value of heart rate fluctuation, expressed as: ; Wherein, t1 and t2 are the start and end time of vital sign monitoring; HR(·) is the heart rate; t is the time;

[0043] Energy change characteristics: The energy change is calculated based on the total amount of changes in the postoperative physical sign data, expressed as: ; Among them, E2 is the energy change characteristic of physical signs, reflecting the degree of physical sign activity after surgery; BP(·) is blood pressure;

[0044] Slope peak feature: extract the peak value through the slope change of the heart rate and blood pressure curves, indicating: ;in, is the slope peak feature, which indicates the maximum rate of change of heart rate relative to time and is used to detect abnormal fluctuations; and They represent the heart rate difference between two adjacent moments and the time difference between two adjacent acquisition times respectively;

[0045] The basic sign data were replaced by the sign feature extraction results to construct a postoperative sign time series dataset, which was divided into a test set and a training set.

[0046] By performing the above operations, the general postoperative vital sign monitoring system has a low sensitivity to abnormal vital signs and cannot capture the dynamic characteristics of vital sign changes, which leads to poor postoperative vital sign monitoring effect. This solution is based on calculating the time difference of blood pressure fluctuations to identify the fluctuation trend at different time points; quantifying the total change trend of heart rate fluctuations during the monitoring time to reflect the overall level of postoperative cardiac activity; comprehensively considering blood pressure and heart rate changes, calculating the degree of postoperative vital sign activity through energy changes, and providing a quantitative basis for the postoperative recovery process; extracting the slope peak of the heart rate and blood pressure curves, identifying abnormal fluctuations, and improving the sensitivity to abnormal vital signs; and then comprehensively optimizing the basic vital sign data to improve the vital sign monitoring effect.

[0047] Example 4, see Figure 1 and Figure 2 Based on the above embodiment, this embodiment includes the following contents:

[0048] Dependency capture unit: introduce residual Bi-LSTM to capture the long-term dependency in the postoperative vital sign time series, and combine the attention mechanism to focus on the period of abnormal heart rate fluctuation after surgery; specifically: input vital sign time series data, use bidirectional LSTM to capture the dependency between the previous and next time steps, expressed as: ; Add residual connection to the LSTM output to retain the original input feature information, expressed as: ; Introduce time-dependent weight function f1(t), recovery phase weight and environmental correction factors , based on the attention mechanism, focusing on the key time period after surgery, expressed as: ; ; ; ; ; ;in, is the bidirectional LSTM output of the previous time step; and They are the forward LSTM output and the reverse LSTM output respectively; is the output after residual connection; and are the weight and bias of the residual connection respectively; is the input feature of the current time step; is the output after nonlinear transformation; , and They are the early postoperative stage, the postoperative stable stage and the critical window period. , and is the weight parameter of each stage, and its value range is [0.5, 0.7], [0.2, 0.4] and [0.1, 0.3] respectively; tanh(·) is the tanh activation function; is the attention weight; T is the total time step; is the final representation of the postoperative vital sign time series; Dropout(·) is a regularization technique; is the key time point, and its value range is [7,72]; is the width of the time distribution, ranging from [1,5]; is the time step attenuation coefficient, ranging from [0.01, 0.5]; and They are the average risk level of historical patients after surgery and the current surgical procedure category. and is the environmental coefficient, with a value range of [0.1,2]; exp(·) is the exponential function;

[0049] Convolution and pooling unit: Perform convolution and pooling operations on the vital sign time series to extract high-level time series features and reduce the dimension; specifically: transpose the input tensor, exchange the time step and feature dimension, expressed as: ; Apply one-dimensional convolution, expressed as: ; Apply the average pooling operation, expressed as: ;in, is the transposed input tensor; Transpose(·) is the tensor dimension exchange operation; X is the time series of postoperative vital sign monitoring data; It is the time dimension; is the feature dimension; Y(·) is the output after the convolution operation; K(·) is the convolution kernel weight, ranging from [-0.1, 0.1]; X(·) is the sliding window data; b is the convolution pooling bias; f(·) is the activation function; a is the sliding window index; M is the sliding window size, ranging from [3, 10]; k3 is the pooling window, ranging from [2, 5], is the pooling window index; is the time series data within the pooling window; is the average pooling output;

[0050] Loss function design unit; define the single feature loss function , introduce bias weight and , expressed as: ; ; ; The physical signs are weighted averagely as the final loss; The physical sign abnormality risk assessment level is used as the data label of the physical sign abnormality risk assessment unit; Whether there is a risk is used as the data label of other units; Among them, is the mean square error loss function; and is the loss increment factor, ranging from [0.01,5]; and are the risk loss values ​​of physical signs below the lower and upper limits, respectively, with a value range of [10,50]; and are the lower and upper limits of the target range, respectively; and are the baseline weighting factors for heart rate fluctuation and blood pressure, respectively; and are the heart rate variability and the baseline value of heart rate variability at the current time step, respectively; and are the blood pressure value and the baseline value of blood pressure at the current time step, respectively; m is the target value; e is the natural base number;

[0051] Abnormal physical sign risk assessment unit: conduct risk assessment of specific abnormal physical signs after surgery, including:

[0052] Calculate the dynamic health score; specifically: calculate the health score of each sign, expressed as: ; Get the comprehensive health score OS, expressed as: ;in, It is the physical health score; is the health score of the physical sign in the neighborhood of the target value m; and is the rate of change of the score on both sides of the target value, ranging from [0.01,5], depending on the sensitivity of fluctuations; N is the number of signs; is the sign weight; is the i-th sign; y is the sign;

[0053] The fluctuation of postoperative patient's physical signs is uncertain. The distribution of physical sign values ​​is modeled with triangular distribution, the expected loss is calculated, and the optimal physical sign target value is determined. The physical signs of postoperative patients are regarded as random variables, and the target range is represented by triangular distribution. ; The probability density function F(·) of the triangular distribution of postoperative patients is expressed as: ; Where a, b and c are the minimum, target and maximum values ​​of the sign respectively; T is the tolerance range, which is 5%~20%; x is the offset of the process mean, which is 0%~10%;

[0054] For single-sign data, the expected gain loss value within the target range It is expressed as, ; For abnormal signs, the truncated normal distribution is used to model and calculate the expected risk loss. When the sign distribution is truncated by the upper and lower limits, the probability density of the truncated normal distribution is expressed as, ; The extreme sign loss is obtained as the total risk expectation value E, expressed as: ;in, and are the lower and upper limits of the target range, respectively; is the basic gain value, which indicates the health gain when the physical sign value is within the target range; and is the loss coefficient, which describes the growth rate of the loss on both sides of the target value; and They are the lower and upper limits of the sign distribution, respectively; is the difference in the cumulative distribution function;

[0055] Risk assessment of abnormal physical signs; Defined as normal; Defined as low risk; Defined as medium risk; Defined as high risk; , and are risk assessment thresholds, which are set to 10, 30, and 70 respectively; the abnormal risk assessment results of postoperative patient vital signs data are obtained based on the total risk expectation value; when both double-label data meet the standards based on the training set and the test set, the postoperative vital signs assessment model is established.

[0056] By performing the above operations, the general postoperative vital sign monitoring system has the problem of weak ability to improve the assessment of abnormal signs, unable to balance the diversity of signs and individual characteristic differences, which leads to poor accuracy of postoperative vital sign monitoring. This solution focuses on the key time period after surgery through dynamic adjustment of time weight function, recovery stage weight and environmental correction factor to more accurately reflect the postoperative vital sign fluctuation characteristics; introduces deviation weights for fluctuations in heart rate and blood pressure, dynamically adjusts the loss coefficient, designs a segmented loss function according to the distance between the vital sign value and the target range, handles different risk intervals respectively, and improves the ability to assess comprehensive vital sign risks; calculates the expected risk value based on the dynamic health score, refines and adapts to the individual differences of postoperative patients and the uncertainty of vital sign fluctuations; thereby improving the accuracy of postoperative vital sign monitoring.

[0057] Example 5, see Figure 1This embodiment is based on the above embodiment. Postoperative vital signs monitoring is based on the established postoperative vital signs assessment model. The patient's postoperative medical data is collected in real time. After being processed by the vital signs feature extraction module, it is input into the postoperative vital signs assessment model. The vital signs abnormality risk assessment level output by the model is used as the assessment result. When the assessment result is medium risk or high risk, an early warning is issued to realize postoperative vital signs monitoring.

[0058] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0059] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0060] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A postoperative vital signs monitoring system based on medical information processing, characterized by: The system includes a data acquisition module, a physical sign feature extraction module, a postoperative physical sign evaluation model establishment module and a postoperative physical sign monitoring module; The data acquisition module collects historical postoperative medical data; The physical sign feature extraction module extracts the time difference feature, cumulative integral feature, energy change feature and slope peak feature of the basic physical sign, and optimizes the postoperative physical sign data to construct a time series data set; The postoperative physical sign evaluation model establishment module captures the long-term dependency of time series through residual bidirectional LSTM, combines the time-related weight function, the recovery stage weight and the environmental correction factor, focuses on the abnormal evaluation of the key time period after surgery, and thus realizes the establishment of the postoperative physical sign evaluation model; The postoperative physical sign monitoring module performs physical sign monitoring on the postoperative medical data collected in real time based on the established postoperative physical sign evaluation model; The postoperative physical sign evaluation model establishment module includes: dependency capture unit; introducing residual Bi-LSTM to capture the long-term dependency relationship in the postoperative physical sign time series; specifically: input physical sign time series data, and use bidirectional LSTM to capture the dependency relationship between the previous and next time steps, expressed as: ; Add residual connection to the LSTM output to retain the original input feature information, expressed as: ; Introduce time-dependent weight function f1(t), recovery phase weight and environmental correction factors , based on the attention mechanism, focusing on the key time period after surgery, expressed as: ; ; ; ； ； ;in, is the bidirectional LSTM output of the previous time step; and They are the forward LSTM output and the reverse LSTM output respectively; is the output after residual connection; and are the weight and bias of the residual connection respectively; is the input feature of the current time step; is the output after nonlinear transformation; , and They are the early postoperative stage, the postoperative stable stage and the critical window period. , and is the weight parameter of each stage; tanh(·) is the tanh activation function; is the attention weight; T is the total time step; is the final representation of the postoperative vital sign time series; Dropout(·) is a regularization technique; It is a critical time point; is the width of the time distribution; is the time step decay coefficient; and They are the average risk level of historical patients after surgery and the current surgical procedure category. and is the environmental coefficient; t is the time; exp(·) is the exponential function.

2. The postoperative vital sign monitoring system based on medical information processing according to claim 1, characterized in that: The physical sign feature extraction module specifically includes the following contents: Time difference characteristics; Calculate the time difference of blood pressure fluctuations at different time points after surgery , expressed as: ;in, and They are the timestamps of the blood pressure values ​​at the current time point and the reference time point respectively; It’s blood pressure; Cumulative integral feature: Calculate the cumulative integral value of heart rate fluctuation, expressed as: ; Wherein, t1 and t2 are the start and end time of vital sign monitoring; HR(·) is the heart rate; Energy change characteristics: The energy change is calculated based on the total amount of changes in the postoperative physical sign data, expressed as: ; Among them, E2 is the energy change characteristic of physical signs, reflecting the degree of physical sign activity after surgery; BP(·) is blood pressure; Slope peak feature: extract the peak value through the slope change of the heart rate and blood pressure curves, indicating: ;in, is the slope peak feature, which indicates the maximum rate of change of heart rate relative to time and is used to detect abnormal fluctuations; and They represent the heart rate difference between two adjacent moments and the time difference between two adjacent acquisition times respectively; The basic sign data were replaced by the sign feature extraction results to construct a postoperative sign time series dataset, which was divided into a test set and a training set.

3. The postoperative vital signs monitoring system based on medical information processing according to claim 2 is characterized in that The postoperative physical sign assessment model establishment module specifically includes the following contents: Dependency capture unit; Convolution and pooling unit: performs convolution and pooling operations on the vital sign time series to extract high-level time series features and reduce the dimension; Specifically: transpose the input tensor and swap the time step and feature dimensions, expressed as: ; Apply one-dimensional convolution, expressed as: ; Apply the average pooling operation, expressed as: ;in, is the transposed input tensor; Transpose(·) is the tensor dimension exchange operation; X is the time series of postoperative vital sign monitoring data; It is the time dimension; is the feature dimension; Y(·) is the output after the convolution operation; K(·) is the convolution kernel weight; X(·) is the sliding window data; b is the convolution pooling bias; f(·) is the activation function; a is the sliding window index; M is the sliding window size; k3 is the pooling window, is the pooling window index; is the time series data within the pooling window; is the average pooling output; Loss function design unit; define the single feature loss function , introduce bias weight and , expressed as: ; ; ; The physical signs are weighted averagely as the final loss; The physical sign abnormality risk assessment level is used as the data label of the physical sign abnormality risk assessment unit; Whether there is a risk is used as the data label of other units; Among them, is the mean square error loss function; and is the loss increment factor; and They are the risk loss values ​​when the physical sign is below the lower limit and the upper limit respectively; and are the lower and upper limits of the target range, respectively; and are the baseline weighting factors for heart rate fluctuation and blood pressure, respectively; and are the heart rate variability and the baseline value of heart rate variability at the current time step, respectively; and are the blood pressure value and the baseline value of blood pressure at the current time step respectively; m is the target value; e is the natural base number; y is the physical sign; Abnormal physical sign risk assessment unit: conduct risk assessment of specific abnormal physical signs after surgery, including: Calculate the dynamic health score; specifically: calculate the health score of each sign, expressed as: ; Get the comprehensive health score OS, expressed as: ;in, It is the physical health score; is the health score of the physical sign in the neighborhood of the target value m; and is the rate of change of the score on both sides of the target value; N is the number of signs; is the sign weight; is the i-th sign; The fluctuation of postoperative patient's physical signs is uncertain. The distribution of physical sign values ​​is modeled with triangular distribution, the expected loss is calculated, and the optimal physical sign target value is determined. The physical signs of postoperative patients are regarded as random variables, and the target range is represented by triangular distribution. ; The probability density function F(·) of the triangular distribution of postoperative patients is expressed as: ; Where a, b and c are the minimum, target and maximum values ​​of the sign respectively; T is the tolerance range; x is the offset of the process mean; For single-sign data, the expected gain loss value within the target range It is expressed as, ; For abnormal signs, the truncated normal distribution is used to model and calculate the expected risk loss. When the sign distribution is truncated by the upper and lower limits, the probability density of the truncated normal distribution is expressed as, ; The extreme sign loss is obtained as the total risk expectation value E, expressed as: ;in, and are the lower and upper limits of the target range, respectively; is the basic gain value, which indicates the health gain when the physical sign value is within the target range; and is the loss coefficient, which describes the growth rate of the loss on both sides of the target value; and They are the lower and upper limits of the sign distribution, respectively; is the difference in the cumulative distribution function; Risk assessment of abnormal physical signs; Defined as normal; Defined as low risk; Defined as medium risk; Defined as high risk; , and is the risk assessment threshold; based on the total risk expectation, the abnormal risk assessment result of the postoperative patient's vital sign data is obtained; when both the double-label data meet the standards based on the training set and the test set, the postoperative vital sign assessment model is established.

4. The postoperative vital sign monitoring system based on medical information processing according to claim 3 is characterized in that: In the data acquisition module, the historical postoperative medical data includes basic vital signs data, postoperative biochemical indicators, behavioral data, time, surgical procedure type and abnormal vital signs risk assessment level; the abnormal vital signs risk assessment level includes normal, low risk, medium risk and high risk; the collected data is converted and standardized.

5. The postoperative vital sign monitoring system based on medical information processing according to claim 4 is characterized in that: The postoperative vital sign monitoring module is based on the established postoperative vital sign assessment model. It collects the patient's postoperative medical data in real time, and inputs it into the postoperative vital sign assessment model after being processed by the vital sign feature extraction module. The abnormal vital sign risk assessment level output by the model is used as the assessment result. When the assessment result is medium risk or high risk, an early warning is issued to achieve postoperative vital sign monitoring.

Images