Early monitoring and early warning system and method for obstetric patients

Abstract

The present invention discloses an early monitoring and early warning system and method for obstetric patients, which relates to the field of biomedical engineering technology, including: collecting and preprocessing data sets including physiological signals, genetic information, medical history records, and lifestyles through wearable devices and hospital information systems; extracting time domain, frequency domain and nonlinear features from the preprocessed data sets, and fusing them to form health perception features; constructing an anomaly detection model using a statistical baseline model and a machine learning model, identifying and filtering outliers in the health perception features, and outputting a purified feature set; constructing a health risk prediction model based on the purified feature set and the machine learning model, adapting to the physiological state of each patient, and outputting a predicted risk level value; inputting real-time monitoring data into the health risk prediction model, capturing the long-term dependency of the time series, and outputting a real-time risk assessment result; when the real-time risk assessment result exceeds a preset safety threshold, the system immediately triggers an early warning mechanism.

Description

Technical Field

[0001] The invention relates to the technical field of biomedical engineering, and in particular to an early monitoring and warning system and method for obstetric patients. Background Art

[0002] The obstetric field has long been committed to improving the accuracy and efficiency of patient monitoring, especially in early warning systems. With the advancement of science and technology, traditional monitoring methods have gradually been replaced by intelligent and personalized monitoring systems. Early obstetric monitoring systems mainly relied on regular observation and manual recording by medical staff. This method is not only time-consuming and labor-intensive, but also easily affected by subjective judgment and difficult to capture the ever-changing changes in vital signs. In recent years, with the rapid development of wearable technology, the Internet of Things (IoT) and big data analysis technology, obstetric patient monitoring systems have begun to shift towards automated and continuous monitoring. By integrating multiple sensors and algorithms, real-time monitoring of patients' physiological status has been achieved.

[0003] Although modern technology has made significant progress in the field of obstetric monitoring, there are still some key challenges that limit its application in clinical practice. First, traditional monitoring systems often focus on the monitoring of a single or a few physiological indicators, while ignoring the comprehensive assessment of individual differences and overall health status, resulting in limited sensitivity and specificity of the early warning system. Secondly, data preprocessing and feature extraction usually use relatively simple statistical methods, which fail to fully utilize the rich information of time domain, frequency domain and nonlinear characteristics, limiting the accuracy of anomaly detection and risk prediction. Furthermore, the existing anomaly detection models are based on a single statistical method and machine learning algorithm, lack sufficient consideration of data complexity and diversity, and are prone to false positive or false negative results. Finally, most early warning systems are relatively simple in the design of the trigger mechanism, and do not fully consider the refined management of the warning level and the effective allocation of resources, which may lead to overtime response and waste of resources when responding to emergencies. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides an early monitoring and early warning system and method for obstetric patients to solve the problem that traditional monitoring systems often focus on the monitoring of a single or a few physiological indicators, while ignoring the comprehensive assessment of individual differences and comprehensive health conditions, resulting in limited sensitivity and specificity of the early warning system. Secondly, data preprocessing and feature extraction links usually adopt relatively simple statistical methods, fail to fully utilize the rich information of time domain, frequency domain and nonlinear characteristics, and limit the accuracy of anomaly detection and risk prediction. Furthermore, the existing anomaly detection model is based on a single statistical method and machine learning algorithm, lacks sufficient consideration of data complexity and diversity, and is prone to false positive or false negative results. Finally, most early warning systems are relatively simple in the design of trigger mechanisms, and do not fully consider the refined management of warning levels and the effective allocation of resources, which may lead to problems such as overtime response and waste of resources when responding to emergencies.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides an early monitoring and early warning method for obstetric patients, comprising:

[0008] Through wearable devices and hospital information systems, data sets including physiological signals, genetic information, medical history records, and lifestyle are collected and preprocessed;

[0009] Extract time domain, frequency domain and nonlinear features from the preprocessed data set and fuse them to form health perception features;

[0010] Build an anomaly detection model using statistical baseline models and machine learning models to identify and filter outliers in health perception features and output a purified feature set;

[0011] Build a health risk prediction model based on the purified feature set and machine learning model, quickly adapt to the physiological state of each patient, and output the predicted risk level value;

[0012] Input real-time monitoring data into the health risk prediction model to capture the long-term dependencies of time series and output real-time risk assessment results;

[0013] When the real-time risk assessment results exceed the preset safety threshold, the system immediately triggers the early warning mechanism.

[0014] As a preferred solution of the method for early monitoring and early warning of obstetric patients of the present invention, the data sets including physiological signals, genetic information, medical history records, and lifestyle are collected and preprocessed through wearable devices and hospital information systems, including:

[0015] Wearable devices can be used to monitor the patient's heart rate, blood pressure, fetal heart rate, uterine contraction frequency, and blood oxygen saturation in real time.

[0016] Collect patients’ genetic information, medical history records, and lifestyle data through hospital information systems;

[0017] Preprocess the collected data by cleaning and removing noise;

[0018] Let the preprocessed dataset be X.

[0019] As a preferred solution of the method for early monitoring and early warning of obstetric patients of the present invention, the extraction of time domain, frequency domain and nonlinear features from the preprocessed data set and fusion to form health perception features include:

[0020] The time domain features, frequency domain features and nonlinear features are extracted from the preprocessed data set X respectively, which are expressed as:

[0021] ;

[0022] ;

[0023] ;

[0024] Among them, F time represents the extracted time domain feature set, μ X represents the mean value of the data set X, σ X represents the standard deviation of the data set X, max(X) and min(X) represent the maximum and minimum values ​​in the data set X, respectively, γ1(X) represents the skewness of the data set X, and β(X) represents the kurtosis of the data set X; F freq represents the extracted frequency domain feature set, f p represents the peak frequency in the spectrum, E f Indicates the energy distribution of the spectrum; F n represents the extracted nonlinear feature set, S S (X) represents sample entropy, D f (X) represents fractal dimension;

[0025] Based on the nonlinear transformation of the sigmoid function, the final health perception feature is obtained, which is expressed as:

[0026] ;

[0027] ;

[0028] Among them, H(X) represents the health perception characteristics, Represents the comprehensive feature vector after integrating time domain features, frequency domain features and nonlinear features.

[0029] As a preferred solution of the early monitoring and early warning method for obstetric patients described in the present invention, the method of using a statistical baseline model and a machine learning model to build an abnormality detection model, identifying and filtering abnormal values ​​in health perception features, and outputting a purified feature set includes:

[0030] Anomaly detection function is constructed based on the statistical baseline model and the processed data set X. The expression is:

[0031] ;

[0032] Among them, A std represents the anomaly detection function, is the threshold parameter, otherwise means when Default case when conditions are not met;

[0033] Based on the isolation forest algorithm in the machine learning model, an isolation forest anomaly detection function is constructed to identify outliers. The expression is:

[0034] ;

[0035] Among them, A iso represents the isolation forest anomaly detection function;

[0036] Only when both models mark an observation as abnormal, it is considered as an outlier, the outliers in the dataset X are filtered out, and the purified dataset is retained.

[0037] As a preferred solution of the method for early monitoring and early warning of obstetric patients of the present invention, the health risk prediction model is constructed based on the purified feature set and the machine learning model, which quickly adapts to the physiological state of each patient and outputs the predicted risk level value, including:

[0038] Use the gradient boosting decision tree model GBDT as the basis of the health risk prediction model.

[0039] Based on the purified feature set Construct health risk prediction function;

[0040] The LSTM layer in the health risk prediction model receives a feature vector sequence as input and outputs a hidden state vector containing the context information of the time series. The LSTM expression is:

[0041] ;

[0042] Among them, H t represents the hidden state vector at time step t, represents the purified feature vector at time step t, H t-1represents the hidden state vector at time step t-1;

[0043] The health risk prediction function is constructed based on the GBDT model, and the expression is:

[0044] ;

[0045] Among them, R represents the health risk prediction function;

[0046] Experts label the patient's health data and obtain a risk level label Y;

[0047] The cross entropy loss function is used to optimize the model parameters. At the same time, a regularization term is introduced to penalize the complexity of the model. The expression is:

[0048] ;

[0049] in, represents the predicted risk level value of the health risk prediction model, Y represents the actual risk level label, θ represents the model parameter, λ represents the regularization parameter, Ω(θ) represents the regularization term, It means to find the optimal value of parameter θ. represents the cross entropy loss function.

[0050] As a preferred solution of the early monitoring and early warning method for obstetric patients of the present invention, the real-time monitoring data is input into the health risk prediction model to capture the long-term dependency of the time series and output the real-time risk assessment results, including:

[0051] Collect real-time data to form real-time data sets and perform preprocessing, including data cleaning and noise removal;

[0052] Feature extraction and fusion are performed on multiple types of data in the processed data set to obtain real-time health perception features;

[0053] Input the real-time health perception features into the anomaly detection model, identify and filter outliers in the real-time health perception features, and output a purified feature set;

[0054] The purified feature set is input into the health risk prediction model, and the real-time predicted risk level value is output.

[0055] As a preferred solution of the early monitoring and early warning method for obstetric patients described in the present invention, when the real-time risk assessment result exceeds the preset safety threshold, the system immediately triggers the early warning mechanism, including:

[0056] Set the real-time prediction risk level value to R t ;

[0057] Based on real-time prediction of risk level value Rt Construct a comparison function, the expression is:

[0058] ;

[0059] Among them, C (R t , Λ) represents the comparison function, Λ represents the set safety threshold, and otherwise represents the default operation when the condition is not met;

[0060] C(R t , Λ) takes values ​​of 0 and 1, indicating that the warning mechanism is not triggered or triggered. t , Λ) is 0, the warning mechanism is not triggered. t , when the value of Λ) is 1, the early warning mechanism is triggered.

[0061] In a second aspect, the present invention provides an early monitoring and warning system for obstetric patients, comprising:

[0062] The acquisition module is responsible for collecting patient physiological signals, genetic information, medical history records, and lifestyle data through wearable devices and hospital information systems. This module also includes preprocessing processes for data cleaning and noise removal;

[0063] The feature processing module is responsible for extracting time domain, frequency domain and nonlinear features from the preprocessed data, and then fusing these features into health perception features through weights;

[0064] The anomaly detection module is responsible for using the statistical baseline model and machine learning model to identify and filter outliers in health perception features, purify the feature set, and build a health risk prediction model based on the purified feature set using the gradient boosting decision tree GBDT model and LSTM network to predict the patient's real-time risk level;

[0065] The early warning module is responsible for inputting real-time monitoring data into the health risk prediction model, capturing the long-term dependencies of time series, and outputting real-time risk assessment results. When the assessment results exceed the preset safety threshold, the early warning mechanism is triggered and corresponding response measures are taken.

[0066] In a third aspect, an embodiment of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, any step of the method for early monitoring and early warning of obstetric patients as described in the first aspect of the present invention is implemented.

[0067] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, any step of the method for early monitoring and early warning of obstetric patients as described in the first aspect of the present invention is implemented.

[0068] The beneficial effects of the present invention are as follows: the present invention realizes the all-round collection and preprocessing of the patient's physiological signals, genetic information, medical history records and lifestyle data through deep integration of wearable devices and hospital information systems; by constructing multi-dimensional health perception features and combining advanced abnormality detection and risk prediction models, it can more accurately identify the patient's health risks, especially those subtle but important changes in physiological state; in addition, the health risk prediction model adopted by the system can effectively capture the long-term dependencies of time series, thereby outputting more accurate real-time risk assessment results; when the assessment results exceed the preset safety threshold, the system can immediately trigger the early warning mechanism, provide timely intervention guidance for medical staff, and significantly improve the efficiency and safety of obstetric patient monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0070] Figure 1 This is a flow chart of the early monitoring and early warning method for obstetric patients in Example 1.

[0071] Figure 2 This is an interval diagram of the warning levels in Example 1. DETAILED DESCRIPTION

[0072] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0073] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0074] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0075] Example 1, reference Figure 1 and Figure 2, which is the first embodiment of the present invention, provides an early monitoring and early warning method for obstetric patients, comprising the following steps:

[0076] S1: Through wearable devices and hospital information systems, data sets including physiological signals, genetic information, medical history records, and lifestyle are collected and preprocessed;

[0077] Wearable devices can be used to monitor the patient's heart rate, blood pressure, fetal heart rate, uterine contraction frequency, and blood oxygen saturation in real time.

[0078] Collect patients’ genetic information, medical history records, and lifestyle data through hospital information systems;

[0079] Preprocess the collected data by cleaning and removing noise;

[0080] Set the preprocessed data set to X;

[0081] The interquartile range (IQR) is used to identify and exclude outliers in the data. The expression is:

[0082] ;

[0083] Among them, Xi represents the i-th observation in the data set, ∈ represents belonging to, Os represents the outlier in the data set, Q1 and Q3 represent the first quartile and the third quartile respectively, and k represents a constant factor used to determine the boundary of the outlier.

[0084] S2: Extract time domain, frequency domain and nonlinear features from the preprocessed data set and fuse them to form health perception features;

[0085] The time domain features, frequency domain features and nonlinear features are extracted from the preprocessed data set X respectively, which are expressed as:

[0086] ;

[0087] ;

[0088] ;

[0089] Among them, F time represents the extracted time domain feature set, μ X represents the mean value of the data set X, σ X represents the standard deviation of the data set X, max(X) and min(X) represent the maximum and minimum values ​​in the data set X, respectively, γ1(X) represents the skewness of the data set X, and β(X) represents the kurtosis of the data set X; F freq represents the extracted frequency domain feature set, f p represents the peak frequency in the spectrum, Ef Indicates the energy distribution of the spectrum; F n represents the extracted nonlinear feature set, S S (X) represents sample entropy, D f (X) represents fractal dimension;

[0090] Use the SoftMax function to normalize the three features to ensure that all features are on the same scale. The expression is:

[0091] ;

[0092] ;

[0093] ;

[0094] in, , and Respectively represent the normalized time domain features, frequency domain features and nonlinear features;

[0095] Taking the sigmoid function as the final nonlinear transformation, the expression is:

[0096] ;

[0097] Among them, the sigmoid function is an activation function, which is an S-shaped curve function, and e represents the base of the natural logarithm;

[0098] Based on the nonlinear transformation of the sigmoid function, the final health perception feature is obtained, which is expressed as:

[0099] ;

[0100] ;

[0101] Among them, H(X) represents the health perception characteristics, Represents the comprehensive feature vector after integrating time domain features, frequency domain features and nonlinear features;

[0102] The value range of the health perception feature H(X) is from 0 to 1. When the value of H(X) is close to 0, it means that the physiological state is close to the lower end of the normal range. When the value of H(X) is close to 1, it means that the physiological state is close to the upper end of the normal range.

[0103] S3: Use the statistical baseline model and machine learning model to build an anomaly detection model, identify and filter outliers in health perception features, and output a purified feature set;

[0104] Anomaly detection function is constructed based on the statistical baseline model and the processed data set X. The expression is:

[0105] ;

[0106] Among them, A std represents the anomaly detection function, is the threshold parameter, which is used to define the boundary of abnormal values. It can be set according to needs. Otherwise means when Default case when conditions are not met;

[0107] Based on the isolation forest algorithm in the machine learning model, an isolation forest anomaly detection function is constructed to identify outliers. The expression is:

[0108] ;

[0109] Among them, A iso Represents the isolation forest anomaly detection function. In the isolation forest algorithm, each observation value X i are randomly placed into a series of decision trees. i At the shallower levels of the tree, it will be isolated (i.e., separated from other data points), which means that X i are outliers, because outliers require fewer splits to be singled out;

[0110] An observation is considered an outlier only if both models flag it as an outlier;

[0111] The outputs of the anomaly detection function and the isolation forest anomaly detection function are combined using the logical AND operation, and the comprehensive anomaly detection function is introduced at the same time. The expression is:

[0112] ;

[0113] Among them, A to represents the comprehensive anomaly detection function, ∧ represents the logical AND operation;

[0114] According to the comprehensive anomaly detection function A to (X), filter out the outliers in the data set X and keep the purified data set. The expression is:

[0115] ;

[0116] in, represents the purified feature set;

[0117] Anomaly detection function A std (X) and Isolation Forest Anomaly Detection Function A isoThe value range of (X) is 0 and 1, where 0 means the observation is considered normal and 1 means the observation is considered abnormal;

[0118] The value range of the comprehensive anomaly detection function Ato(X) is 0 and 1. Its value is 1 only when both anomaly detection functions mark an observation as abnormal.

[0119] S4: Build a health risk prediction model based on the purified feature set and machine learning model, quickly adapt to the physiological state of each patient, and output the predicted risk level value;

[0120] Use the gradient boosted decision tree model GBDT as the basis of the health risk prediction model;

[0121] Based on the purified feature set Construct a health risk prediction function and predict the patient's risk level through the GBDT model;

[0122] The LSTM layer in the health risk prediction model receives a feature vector sequence as input and outputs a hidden state vector containing the context information of the time series. The LSTM expression is:

[0123] ;

[0124] Among them, H t represents the hidden state vector at time step t, LSTM() represents the function of the LSTM unit, which accepts two main inputs: the input vector of the current time step and the hidden state of the previous time step, represents the purified feature vector at time step t, H t-1 represents the hidden state vector at time step t-1;

[0125] The output H of LSTM t Connected to the GBDT model as an additional input feature for the health risk prediction model;

[0126] The health risk prediction function is constructed based on the GBDT model, and the expression is:

[0127] ;

[0128] Among them, R represents the health risk prediction function;

[0129] Experts label the patient's health data and obtain a risk level label Y;

[0130] The cross entropy loss function is used to optimize the model parameters. At the same time, the regularization term is introduced to penalize the complexity of the model and prevent overfitting. The expression is:

[0131] ;

[0132] in, represents the predicted risk level value of the health risk prediction model, Y represents the actual risk level label, θ represents the model parameter, λ represents the regularization parameter, Ω(θ) represents the regularization term, It means to find the optimal value of parameter θ so that the expression in the brackets is minimized. represents the cross entropy loss function;

[0133] Predicted risk level value The value range is from 0 to 1. When the value is close to 0, it means that the patient is in a lower risk state. When the value is close to 1, it means that the patient is in a higher risk state. When the value is close to 0.5, it means that the patient is at moderate risk.

[0134] S5: Input the real-time monitoring data into the health risk prediction model, capture the long-term dependencies of the time series, and output the real-time risk assessment results;

[0135] Collect real-time data to form real-time data sets and perform preprocessing, including data cleaning and noise removal;

[0136] Feature extraction and fusion are performed on multiple types of data in the processed data set to obtain real-time health perception features;

[0137] Input the real-time health perception features into the anomaly detection model, identify and filter outliers in the real-time health perception features, and output a purified feature set;

[0138] The purified feature set is input into the health risk prediction model, and the real-time predicted risk level value is output.

[0139] S6: When the real-time risk assessment results exceed the preset safety threshold, the system immediately triggers the early warning mechanism;

[0140] Set the real-time prediction risk level value to R t ;

[0141] Based on real-time prediction of risk level value R t Construct a comparison function, the expression is:

[0142] ;

[0143] Among them, C (R t , Λ) represents the comparison function, Λ represents the set safety threshold, and otherwise represents the default operation when the condition is not met;

[0144] C(R t, Λ) takes values ​​of 0 and 1, indicating that the warning mechanism is not triggered or triggered. t , Λ) is 0, the warning mechanism is not triggered. t , Λ) is 1, the early warning mechanism is triggered;

[0145] The patient's historical health status and physiological parameters are collected to form a patient parameter set P. Based on the comparison function C (R t , Λ) constructs the early warning trigger function, and the patient parameter set P is used as an additional input to assist in judgment. The expression is:

[0146] ;

[0147] Among them, W (R t ,Λ,P) represents the warning trigger function, Alert represents a string constant, when the comparison function C(R t , Λ) is 1, the output of the warning trigger function means that the system will trigger a warning, NoAction represents a string constant, when the comparison function C (R t , Λ) is not 1, that is, 0, the output of the warning trigger function means that the system will not take any warning action;

[0148] Use the support vector machine (SVM) model to build the warning level function, and set the warning level function to E(R t ,Λ,P), the expression is:

[0149] ;

[0150] Among them, L represents the warning level, which includes low-level warning L1, medium-level warning L2 and high-level warning L3;

[0151] According to the warning level L, send the corresponding warning notification;

[0152] When the warning level is low-level warning L1, the system automatically records the warning event and sends an electronic reminder to the responsible nurse to remind her to pay attention to the patient. The system recommends mild intervention measures to medical staff, including adjusting the patient's position, light activities and dietary suggestions to improve the patient's comfort and physiological state.

[0153] When the warning level is intermediate warning L2, the system immediately sends a telephone notification to the responsible doctor and displays the warning information on the doctor's workstation, prompting the doctor to check the patient's condition immediately. The system notifies the medical equipment management department to check and prepare necessary emergency equipment, including oxygen inhalation devices and ECG monitors;

[0154] When the warning level is high warning L3, the system automatically summons a rapid response team, including obstetricians, anesthesiologists and neonatologists, to quickly arrive at the patient's location. The system instructs medical staff to prepare for advanced life support measures, including endotracheal intubation and cardiopulmonary resuscitation. At the same time, the system prompts medical staff to communicate with the patient's family, explain the severity of the current situation, and inform them of the ongoing emergency measures.

[0155] This embodiment also provides an early monitoring and warning system for obstetric patients, including:

[0156] The acquisition module is responsible for collecting patient physiological signals, genetic information, medical history records, and lifestyle data through wearable devices and hospital information systems. This module also includes preprocessing processes for data cleaning and noise removal;

[0157] The feature processing module is responsible for extracting time domain, frequency domain and nonlinear features from the preprocessed data, and then fusing these features into health perception features through weights;

[0158] The anomaly detection module is responsible for using the statistical baseline model and machine learning model to identify and filter outliers in health perception features, purify the feature set, and build a health risk prediction model based on the purified feature set using the gradient boosting decision tree GBDT model and LSTM network to predict the patient's real-time risk level;

[0159] The early warning module is responsible for inputting real-time monitoring data into the health risk prediction model, capturing the long-term dependencies of time series, and outputting real-time risk assessment results. When the assessment results exceed the preset safety threshold, the early warning mechanism is triggered and corresponding response measures are taken.

[0160] This embodiment also provides a computer device, which is suitable for the early monitoring and early warning method for obstetric patients, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the early monitoring and early warning method for obstetric patients proposed in the above embodiment.

[0161] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0162] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for early monitoring and early warning of obstetric patients proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0163] In summary, the present invention realizes the comprehensive collection and preprocessing of patients' physiological signals, genetic information, medical history records and lifestyle data through deep integration of wearable devices and hospital information systems. By constructing multi-dimensional health perception features and combining advanced abnormality detection and risk prediction models, it can more accurately identify patients' health risks, especially those subtle but important changes in physiological state. In addition, the health risk prediction model adopted by the system can effectively capture the long-term dependencies of time series, thereby outputting more accurate real-time risk assessment results. When the assessment results exceed the preset safety threshold, the system can immediately trigger the early warning mechanism to provide timely intervention guidance for medical staff, significantly improving the efficiency and safety of obstetric patient monitoring.

[0164] Example 2, referring to Table 1, is the second example of the present invention. To further verify the advancement of the present invention, experimental simulation data of an early monitoring and early warning method for obstetric patients are provided.

[0165] Before the trial began, we selected 50 patients of different ages, weights, and medical histories as test subjects. The test environment was set up in the hospital's maternity ward to ensure that the patients were monitored under comfortable and safe conditions;

[0166] First, we use wearable devices to monitor the patient's heart rate, blood pressure, fetal heart rate, uterine contraction frequency, and blood oxygen saturation in real time. At the same time, we collect the patient's genetic information, medical history records, and lifestyle data through the hospital information system;

[0167] The collected data is preprocessed to clean and remove noise, the preprocessed data set is set as X, and the interquartile range (IQR) is used to identify and exclude outliers in the data to ensure the accuracy of the data;

[0168] Then, we extract time domain, frequency domain and nonlinear features from the preprocessed data set and fuse them to form health perception features. By constructing time domain feature extraction function, frequency domain feature extraction function and nonlinear feature extraction function, we obtain rich health perception features.

[0169] Next, we used the statistical baseline model and the machine learning model to build an anomaly detection model to identify and filter outliers in the health perception features and output a purified feature set. We used the isolation forest algorithm and threshold parameters to combine the two models to improve the accuracy of anomaly detection.

[0170] Based on the purified feature set and machine learning model, a health risk prediction model is built to quickly adapt to the physiological state of each patient and output the predicted risk level value. We use the gradient boosting decision tree model GBDT and LSTM layer, combined with time series context information, to improve the accuracy of risk prediction;

[0171] Finally, the real-time monitoring data is input into the health risk prediction model to capture the long-term dependencies of the time series and output the real-time risk assessment results. When the real-time risk assessment results exceed the preset safety threshold, the system immediately triggers the early warning mechanism. We set the real-time prediction risk level value Rt and build a comparison function based on it to determine whether to trigger an early warning.

[0172] The details are shown in Table 1:

[0173] Table 1 Experimental record table:

[0174] Test subject name Heart rate (beats / minute) Blood pressure (mmHg) Fetal heart rate (beats / minute) Uterine contraction frequency (times / hour) Blood oxygen saturation (%) Genetic information Medical history lifestyle Real-time risk assessment results Warning level Patient 1 80 120 / 80 160 4 95 normal normal normal 0.5 L2 Patient 2 70 110 / 70 150 3 96 normal normal normal 0.6 L1 Patient 3 90 130 / 90 170 5 94 normal normal normal 0.7 L2 Patient 4 60 100 / 60 140 2 97 normal normal normal 0.4 L1 Patient 5 85 125 / 85 165 4 95 normal normal normal 0.6 L2 Patient 6 75 115 / 75 155 3 96 normal normal normal 0.5 L1

[0175] By comparing the data in the table, we can see that the present invention has the following creative and novel effects compared with the shortcomings of the prior art in the embodiments:

[0176] By real-time monitoring and collecting the patient's physiological signals, genetic information, medical history records and lifestyle data, the present invention can fully understand the patient's health status and provide more basis for early warning. The existing technology often relies on only a single physiological indicator and cannot fully reflect the patient's overall condition.

[0177] By extracting time domain, frequency domain and nonlinear features and fusing them to form health perception features, the present invention can more accurately assess the patient's physiological state and improve the accuracy of early warning. The existing technology often only considers a single or a few indicators and cannot fully reflect the patient's physiological state.

[0178] By using the statistical baseline model and the machine learning model to build an anomaly detection model, the present invention can automatically identify outliers and improve the accuracy of early warning. The existing technology often relies on manual judgment and is easily affected by subjective factors.

[0179] The health risk prediction model is constructed based on the purified feature set and machine learning model. The present invention can quickly adapt to the physiological state of each patient and output the predicted risk level value. The existing technology often takes a long time to adjust and optimize the model and cannot quickly adapt to the physiological changes of the patient.

[0180] Input real-time monitoring data into the health risk prediction model, capture the long-term dependencies of the time series, and output real-time risk assessment results. Existing technologies often cannot update the model in real time and cannot reflect the patient's latest status in a timely manner;

[0181] Through these innovative effects, the present invention can monitor and warn obstetric patients' health conditions more accurately and timely, thereby improving patients' safety and satisfaction.

[0182] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should be included in the scope of the claims of the present invention.

Claims

1. An early monitoring and early warning method for obstetric patients, characterized in that: include: Through wearable devices and hospital information systems, data sets including physiological signals, genetic information, medical history records, and lifestyle are collected and preprocessed; Extract time domain, frequency domain and nonlinear features from the preprocessed data set and fuse them to form health perception features; Build an anomaly detection model using statistical baseline models and machine learning models to identify and filter outliers in health perception features and output a purified feature set; Build a health risk prediction model based on the purified feature set and machine learning model, quickly adapt to the physiological state of each patient, and output the predicted risk level value; Input real-time monitoring data into the health risk prediction model to capture the long-term dependencies of time series and output real-time risk assessment results; When the real-time risk assessment results exceed the preset safety threshold, the system immediately triggers the early warning mechanism; The extracting of time domain, frequency domain and nonlinear features from the preprocessed data set and fusing them to form health perception features include: The time domain features, frequency domain features and nonlinear features are extracted from the preprocessed data set X respectively, which are expressed as: ； ； ； Among them, F time represents the extracted time domain feature set, μ X represents the mean value of the data set X, σ X represents the standard deviation of the data set X, max(X) and min(X) represent the maximum and minimum values ​​in the data set X, respectively, γ1(X) represents the skewness of the data set X, and β(X) represents the kurtosis of the data set X; F freq represents the extracted frequency domain feature set, f p represents the peak frequency in the spectrum, E f Indicates the energy distribution of the spectrum; F n represents the extracted nonlinear feature set, S S (X) represents sample entropy, D f (X) represents fractal dimension; Based on the nonlinear transformation of the sigmoid function, the final health perception feature is obtained, which is expressed as: ； ； Among them, H(X) represents the health perception characteristics, Represents the comprehensive feature vector after integrating time domain features, frequency domain features and nonlinear features; The statistical baseline model and the machine learning model are used to build an anomaly detection model, identify and filter outliers in health perception features, and output a purified feature set, including: Anomaly detection function is constructed based on the statistical baseline model and the processed data set X. The expression is: Among them, A std represents the anomaly detection function, is the threshold parameter, otherwise means when Default case when conditions are not met; Based on the isolation forest algorithm in the machine learning model, an isolation forest anomaly detection function is constructed to identify outliers. The expression is: ； Among them, A iso represents the isolation forest anomaly detection function; Only when both models mark an observation as abnormal, it is considered as an outlier, the outliers in the data set X are filtered out, and the purified data set is retained; The health risk prediction model is constructed based on the purified feature set and the machine learning model, which quickly adapts to the physiological state of each patient and outputs the predicted risk level value, including: Use the gradient boosted decision tree model GBDT as the basis of the health risk prediction model; Based on the purified feature set Construct health risk prediction function; The LSTM layer in the health risk prediction model receives a feature vector sequence as input and outputs a hidden state vector containing the context information of the time series. The LSTM expression is: ； Among them, H t represents the hidden state vector at time step t, represents the purified feature vector at time step t, H t-1 represents the hidden state vector at time step t-1; The health risk prediction function is constructed based on the GBDT model, and the expression is: ； Among them, R represents the health risk prediction function; Experts label the patient's health data and obtain a risk level label Y; The cross entropy loss function is used to optimize the model parameters. At the same time, a regularization term is introduced to penalize the complexity of the model. The expression is: ； in, represents the predicted risk level value of the health risk prediction model, Y represents the actual risk level label, θ represents the model parameter, λ represents the regularization parameter, Ω(θ) represents the regularization term, It means to find the optimal value of parameter θ. represents the cross entropy loss function.

2. The method for early monitoring and early warning of obstetric patients according to claim 1, characterized in that: The data sets including physiological signals, genetic information, medical history records, and lifestyle are collected and preprocessed through wearable devices and hospital information systems, including: Wearable devices can be used to monitor the patient's heart rate, blood pressure, fetal heart rate, uterine contraction frequency, and blood oxygen saturation in real time. Collect patients’ genetic information, medical history records, and lifestyle data through hospital information systems; Preprocess the collected data by cleaning and removing noise; Let the preprocessed dataset be X.

3. The method for early monitoring and early warning of obstetric patients according to claim 2, characterized in that: The real-time monitoring data is input into the health risk prediction model to capture the long-term dependencies of the time series and output real-time risk assessment results, including: Collect real-time data to form real-time data sets and perform preprocessing, including data cleaning and noise removal; Feature extraction and fusion are performed on multiple types of data in the processed data set to obtain real-time health perception features; Input the real-time health perception features into the anomaly detection model, identify and filter outliers in the real-time health perception features, and output a purified feature set; The purified feature set is input into the health risk prediction model, and the real-time predicted risk level value is output.

4. The method for early monitoring and early warning of obstetric patients according to claim 3, characterized in that: When the real-time risk assessment result exceeds the preset safety threshold, the system immediately triggers the early warning mechanism, including: Set the real-time prediction risk level value to R t ; Based on real-time prediction of risk level value R t Construct a comparison function, the expression is: ； Among them, C (R t , Λ) represents the comparison function, Λ represents the set safety threshold, and otherwise represents the default operation when the condition is not met; C(R t , Λ) takes the values ​​of 0 and 1, indicating that the warning mechanism is not triggered or triggered. t , Λ) is 0, the warning mechanism is not triggered. t , when the value of Λ) is 1, the early warning mechanism is triggered.

5. An early monitoring and early warning system for obstetric patients, based on the early monitoring and early warning method for obstetric patients according to any one of claims 1 to 4, characterized in that: include: The acquisition module is responsible for collecting patients’ physiological signals, genetic information, medical history records, and lifestyle data through wearable devices and hospital information systems. This module also includes preprocessing processes for data cleaning and noise removal; The feature processing module is responsible for extracting time domain, frequency domain and nonlinear features from the preprocessed data, and then fusing these features into health perception features through weights; The anomaly detection module is responsible for using the statistical baseline model and machine learning model to identify and filter outliers in health perception features, purify the feature set, and build a health risk prediction model based on the purified feature set using the gradient boosting decision tree GBDT model and LSTM network to predict the patient's real-time risk level; The early warning module is responsible for inputting real-time monitoring data into the health risk prediction model, capturing the long-term dependencies of time series, and outputting real-time risk assessment results. When the assessment results exceed the preset safety threshold, the early warning mechanism is triggered and corresponding response measures are taken.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method for early monitoring and early warning of obstetric patients described in any one of claims 1 to 4 are implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for early monitoring and early warning of obstetric patients described in any one of claims 1 to 4 are implemented.

Images