A multimodal physiological data monitoring system and method based on machine learning

Abstract

This invention discloses a multimodal physiological data monitoring system and method based on machine learning, belonging to the field of data analysis technology. The system includes: acquiring historical monitoring records from a hospital; identifying and extracting feature records; extracting monitoring information and patients from these feature records; determining the test results of each physiological indicator for the patient; extracting and analyzing keywords from the daily medical records and monitoring information of medical staff to identify target personnel; obtaining the correlation between patients and physiological indicators; establishing a target set based on the patient information and the monitoring interval of the physiological indicators; calculating a warning coefficient based on the set display duration of each physiological indicator; and providing warning prompts for the display duration. This invention, by acquiring historical monitoring records and daily medical records, obtains the correlation between patients and physiological indicators and determines whether the display duration is reasonable, which helps to capture abnormal indicators in a timely manner and reduce the risk of errors in clinical judgment.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, specifically a multimodal physiological data monitoring system and method based on machine learning. Background Technology

[0002] When hospital medical staff monitor patients' physiological functions, they typically use multimodal data collection and monitoring because there are close correlations between various physiological indicators. This allows medical staff to quickly grasp the real-time dynamics of various physiological indicators and make timely and accurate judgments on patients' health behaviors. Current physiological monitoring equipment is designed to display multimodal physiological data simultaneously or individually. When displayed individually, only one physiological indicator is shown at a time, requiring manual setting of the display duration of each physiological indicator in turn. Since manual setting is subjective and does not analyze historical records, the display duration of each physiological indicator may be unreasonable, causing delays in the timely detection of abnormal indicators and increasing the risk of errors in clinical judgment. Summary of the Invention

[0003] The purpose of this invention is to provide a multimodal physiological data monitoring system and method based on machine learning to solve the problems raised in the prior art.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A machine learning-based method for monitoring multimodal physiological data includes the following steps:

[0006] Step S1: Obtain the hospital's historical monitoring records of patients' physiological indicators, extract the total monitoring time and monitoring interval of each physiological indicator from the monitoring records, obtain the mean deviation of each physiological indicator based on the monitoring interval of each physiological indicator, and determine and extract the feature records in the monitoring records based on the mean deviation.

[0007] Step S2: Extract the monitoring information and patients corresponding to the feature records, and obtain the patient's disease information during monitoring. Based on the monitoring information, medical staff determine the test results of each physiological indicator of the patient. Extract the daily medical records of each medical staff member, extract and analyze the keywords in the daily medical records and monitoring information, and then identify the target personnel among the medical staff.

[0008] Step S3: Based on the monitoring information, the target personnel re-determine the test results of each physiological indicator of the patient. Based on the test results, the correlation degree of each patient with each physiological indicator is obtained. The patient information of the patients to be tested and the monitoring interval of each physiological indicator of the patients to be tested are obtained, and the target set corresponding to the patients to be tested is established.

[0009] Step S4: Extract the display duration of various physiological indicators of the patient to be tested, establish the corresponding display set, obtain the warning coefficient of the display duration based on the display set and the target set, and determine whether to issue a warning prompt for the display duration based on the warning coefficient.

[0010] Preferably, step S1 includes:

[0011] Step S1-1: Obtain the hospital's historical monitoring records of the patient's physiological indicators. The monitoring records are formed by medical staff monitoring the changes in the patient's physiological parameters according to the preset monitoring intervals for various physiological indicators.

[0012] Extract the total monitoring duration MD of a certain physiological indicator X from a certain monitoring record R. X and monitoring interval duration ID X Based on the total monitoring duration MD X The total number N of physiological indicator X recorded internally X The time monitoring bias of physiological index X in monitoring record R was obtained as follows: The average deviation of physiological indicator X over time is calculated in several monitoring records to obtain the mean deviation of physiological indicator X, and then the mean deviation of each physiological indicator is obtained.

[0013] Step S1-2: Extract the physiological indicators corresponding to a certain monitoring record T, as well as the total monitoring time and monitoring interval for each physiological indicator. Calculate the time monitoring deviation for each physiological indicator and obtain the variance between all time monitoring deviations. If each time monitoring deviation is not greater than the corresponding mean deviation and the variance is less than the preset variance threshold, then the monitoring record T is used as a feature record, and thus all feature records in the monitoring record are obtained.

[0014] In actual monitoring, the actual monitoring data may deviate from the preset values ​​due to the insensitivity of physiological monitoring equipment, resulting in unreliable detection records. Here, by analyzing the monitoring interval of various physiological indicators during the monitoring process, the corresponding time monitoring deviation and the average deviation are obtained, thereby extracting the feature records. The following analysis is performed on the feature records. For the present invention, this can make the calculation results more accurate and reliable.

[0015] Preferably, step S2 includes:

[0016] Step S2-1: Extract the monitoring information and patient corresponding to the feature record, and obtain the patient's disease information during monitoring. The monitoring information is the monitoring data generated during the monitoring of the patient's physiological parameters; extract the monitoring information M corresponding to a certain feature record F. F and patient P FInstruct a medical staff member P1 to monitor information M F Patient P was obtained F The test result of each physiological indicator is taken as the first test result, which indicates whether each physiological indicator is in a normal or abnormal state.

[0017] Step S2-2: Obtain the daily progress record (CR) of another medical staff member P2. The daily progress record (CR) is a record formed by medical staff member P2 based on daily observations of changes in the patient's condition, treatment measures, examination results, and doctor-patient communication.

[0018] Extract monitoring information M F For each physiological indicator, keywords are identified and a keyword set S1 is established. Q keywords are randomly extracted from daily medical records (CR) and a keyword set S2 is established. The number of keywords in keyword set S2 whose word similarity to a keyword K1 in keyword set S1 is greater than a preset similarity threshold is taken as the target number of keyword K1. Then, the target number corresponding to each keyword in keyword set S1 is obtained. The sum of the target numbers corresponding to each keyword for normal physiological indicators in the first detection result is taken as the first number N1, and the sum of the target numbers corresponding to each keyword for abnormal physiological indicators is taken as the second number N2.

[0019] The weights of the first quantity N1 and the second quantity N2 are set to W1 and W2 respectively. The feature quantity N0 of medical staff P2 is obtained as N0 = W1 × N1 + W2 × N2. Then, Q keywords are randomly extracted from the daily medical records of each medical staff member other than medical staff P1, and the feature quantity of each medical staff member is finally obtained. The medical staff member with the maximum value is taken as the target personnel.

[0020] Preferably, step S3 includes:

[0021] Step S3-1: Instruct the target personnel to use the monitoring information M F Patient P was obtained F The test result of each physiological indicator is used as the second test result. If the test results of each physiological indicator in the first and second test results are consistent, then the feature record F is marked; obtain all marked records when the patient is in a certain patient Z, the number of which is G. Z The number of marked records when a certain physiological indicator Y is abnormal is taken as G. Z Y Therefore, the degree of correlation between patient Z and physiological indicator Y is G. Z Y / G Z And obtain the correlation degree of each patient with each physiological indicator;

[0022] The correlation between various physiological indicators varies among different patients. For example, patients with hypertension generally have abnormal blood pressure but normal heart rate. However, for patients with coronary heart disease, the correlation between heart rate and condition is more prominent, so their heart rate is generally abnormal while their blood pressure is generally normal. Therefore, in this protocol, the following judgment should be made in close combination with the patient's condition to make the judgment more reliable, help to capture abnormal indicators in a timely manner, and reduce the risk of errors in clinical judgment.

[0023] Step S3-2: Obtain the patient information of the patient to be tested and the various physiological indicators currently monitored for the patient to be tested. Obtain the correlation degree of each patient in the patient information with a certain physiological indicator X1 monitored for the patient to be tested, and take the maximum correlation degree as the first target value corresponding to physiological indicator X1, and then obtain the first target value corresponding to each physiological indicator.

[0024] Since the greater the correlation between a patient's physiological indicators and the indicators, the greater the likelihood of abnormalities in those indicators, the first target value here represents the degree of attention a physiological indicator should receive. The larger the first target value of a physiological indicator, the longer its display duration should be. Since the shorter the monitoring interval, the more frequently the physiological indicator data fluctuates, the greater the attention it should receive. Therefore, the larger the second target value of a physiological indicator, the longer its display duration should be. In conclusion, the larger the overall target value, the longer its display duration should be.

[0025] Obtain the monitoring interval t of the physiological indicator X1 of the patient to be tested, and obtain the second target value e corresponding to the physiological indicator X1. -k×t Where e is the natural constant and k is the target value correlation coefficient, the second target value corresponding to each physiological indicator is obtained; the weights of the first target value and the second target value are set respectively to obtain the total target value corresponding to each physiological indicator, and the target set corresponding to the patient to be tested is established.

[0026] Preferably, step S4 includes: extracting the display duration of various physiological indicators of the patient to be tested; establishing a corresponding display set according to the order of each physiological indicator in the target set based on the display duration; and calculating the cosine similarity between the display set and the target set as a warning coefficient for the display duration. If the warning coefficient is less than a preset numerical threshold, a warning prompt is given for the current display duration of each physiological indicator.

[0027] A multimodal physiological data monitoring system based on machine learning includes a feature recording and extraction module, a target personnel identification module, a target set establishment module, and an early warning module.

[0028] Feature record extraction module: used to obtain historical monitoring records of patients' physiological indicators in the hospital, extract the total monitoring time and monitoring interval time of each physiological indicator in the monitoring records, obtain the mean deviation of each physiological indicator based on the monitoring interval time of each physiological indicator, and judge and extract feature records in the monitoring records based on the mean deviation.

[0029] Target personnel identification module: This module is used to extract monitoring information and patients corresponding to feature records, and obtain patient information during monitoring. Based on the monitoring information, medical staff can determine the test results of each physiological indicator of the patient. It also extracts the daily medical records of each medical staff member, extracts and analyzes keywords in the daily medical records and monitoring information, and then identifies the target personnel among the medical staff.

[0030] Target set establishment module: It is used to enable target personnel to re-determine the test results of each physiological indicator of the patient based on the monitoring information, and obtain the correlation degree of each patient with each physiological indicator based on the test results; obtain the patient information of the current patients to be tested, as well as the monitoring interval of each physiological indicator of the patients to be tested, and establish the target set corresponding to the patients to be tested.

[0031] Early warning module: It is used to extract the display duration of various physiological indicators of the patients to be tested, establish the corresponding display set, obtain the early warning coefficient of the display duration based on the display set and the target set, and determine whether to issue an early warning for the display duration based on the early warning coefficient.

[0032] Preferably, the target personnel determination module includes a detection result determination unit and a target personnel determination unit;

[0033] The detection result determination unit is used to extract the monitoring information and patients corresponding to the feature records, and to obtain the patient's disease information during monitoring; it extracts the monitoring information and patients corresponding to the feature records, enabling medical staff to obtain the detection results of each physiological indicator of the patient based on the monitoring information;

[0034] Target personnel identification unit: used to obtain the daily medical records of medical staff, extract and analyze keywords in the daily medical records and monitoring information, obtain the feature quantity of each medical staff member, and identify the medical staff member with the maximum feature quantity as the target personnel.

[0035] Preferably, the target set establishment module includes a correlation degree calculation unit and a target set establishment unit;

[0036] The correlation degree calculation unit is used to enable the target personnel to obtain the detection results of each physiological indicator of the patient based on the monitoring information, and to determine whether to mark the feature records; to obtain all marked records when the patient is in a certain disease state, and to determine the correlation degree between a certain patient and a certain physiological indicator based on the number of marked records when a certain physiological indicator is abnormal.

[0037] The target set establishment unit is used to obtain the patient information and various physiological indicators monitored by the current patients to be tested, and to obtain the first target value corresponding to the physiological indicators based on the degree of correlation; to obtain the monitoring interval of the physiological indicators monitored by the patients to be tested, to obtain the second target value corresponding to the physiological indicators, and then to establish the target set corresponding to the patients to be tested.

[0038] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention provides a multimodal physiological data monitoring system and method based on machine learning, including: acquiring historical monitoring records from a hospital, identifying and extracting feature records; extracting monitoring information and patients from the feature records, determining the detection results of each physiological indicator for the patient, extracting and analyzing keywords from the daily medical records and monitoring information of medical staff, and identifying target personnel; obtaining the degree of correlation between patients and physiological indicators, establishing a target set based on the patient information to be tested and the monitoring interval of physiological indicators, calculating a warning coefficient based on the set display duration of each physiological indicator, and providing a warning prompt for the display duration. This invention, by acquiring historical monitoring records and daily medical records, obtains the degree of correlation between patients and physiological indicators, and determines whether the display duration is reasonable, which helps to capture abnormal indicators in a timely manner and reduce the risk of errors in clinical judgment. Attached Figure Description

[0039] Figure 1 This is a flowchart illustrating a multimodal physiological data monitoring method based on machine learning according to the present invention. Detailed Implementation

[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] This embodiment provides a multimodal physiological data monitoring method based on machine learning, and its corresponding flowchart is shown below. Figure 1 As shown, the specific steps include:

[0042] Step S1: Obtain the hospital's historical monitoring records of patients' physiological indicators, extract the total monitoring time and monitoring interval of each physiological indicator from the monitoring records, obtain the mean deviation of each physiological indicator based on the monitoring interval of each physiological indicator, and determine and extract the feature records in the monitoring records based on the mean deviation.

[0043] Step S1-1: Obtain the hospital's historical monitoring records of the patient's physiological indicators. The monitoring records are formed by medical staff monitoring the changes in the patient's physiological parameters according to the preset monitoring intervals for various physiological indicators.

[0044] Extract the total monitoring duration MD of a certain physiological indicator X from a certain monitoring record R. X and monitoring interval duration ID X Based on the total monitoring duration MD X The total number N of physiological indicator X recorded internally X The time monitoring bias of physiological index X in monitoring record R was obtained as follows: The average deviation of physiological indicator X over time is calculated in several monitoring records to obtain the mean deviation of physiological indicator X, and then the mean deviation of each physiological indicator is obtained.

[0045] Total monitoring duration MD X N refers to the total duration of monitoring physiological indicator X. X This refers to the total monitoring duration MD X The total number of recorded physiological index X data within the MD is then... X With N X The ratio represents the time required to record a single data point, and the monitoring interval duration ID. X The time monitoring deviation B can be obtained by subtracting the time required to record a single data point from the time required to record the data point and taking the absolute value. X For example: if the total monitoring time MD X The monitoring interval is 120 seconds. (ID) X The duration is 10 seconds, and the total number of data points is N. X When the value is 12, the time monitoring deviation B can be obtained. X It is 0 seconds, but when the total monitoring duration MD X The monitoring interval is 120 seconds. (ID) X The duration is 10 seconds, and the total number of data points is N. X When the value is 10, the time monitoring deviation B is obtained. X The interval is 2 seconds. Then, based on the deviations monitored over numerous time periods, the average deviation is obtained and used to determine the following characteristic records, as follows:

[0046] Step S1-2: Extract the physiological indicators corresponding to a certain monitoring record T, as well as the total monitoring time and monitoring interval for each physiological indicator. Calculate the time monitoring deviation for each physiological indicator and obtain the variance between all time monitoring deviations. If each time monitoring deviation is not greater than the corresponding mean deviation and the variance is less than the preset variance threshold, then the monitoring record T is used as a feature record, and thus all feature records in the monitoring record are obtained.

[0047] In actual monitoring, the actual monitoring data may deviate from the preset values ​​due to the insensitivity of physiological monitoring equipment, resulting in unreliable detection records. Here, by analyzing the monitoring interval of various physiological indicators during the monitoring process, the corresponding time monitoring deviation and the average deviation are obtained, thereby extracting the feature records. The following analysis is performed on the feature records. For the present invention, this can make the calculation results more accurate and reliable.

[0048] Step S2: Extract the monitoring information and patients corresponding to the feature records, and obtain the patient's disease information during monitoring. Based on the monitoring information, medical staff determine the test results of each physiological indicator of the patient. Extract the daily medical records of each medical staff member, extract and analyze the keywords in the daily medical records and monitoring information, and then identify the target personnel among the medical staff.

[0049] Step S2-1: Extract the monitoring information and patient corresponding to the feature record, and obtain the patient's disease information during monitoring. The monitoring information is the monitoring data generated during the monitoring of the patient's physiological parameters; extract the monitoring information M corresponding to a certain feature record F. F and patient P F Instruct a medical staff member P1 to monitor information M F Patient P was obtained F The test result of each physiological indicator is taken as the first test result, which indicates whether each physiological indicator is in a normal or abnormal state.

[0050] Step S2-2: Obtain the daily progress record (CR) of another medical staff member P2. The daily progress record (CR) is a record formed by medical staff member P2 based on daily observations of changes in the patient's condition, treatment measures, examination results, and doctor-patient communication.

[0051] Extract monitoring information M FFor each physiological indicator, keywords are identified and a keyword set S1 is established. Q keywords are randomly extracted from daily medical records (CR) and a keyword set S2 is established. The number of keywords in keyword set S2 whose word similarity to a keyword K1 in keyword set S1 is greater than a preset similarity threshold is taken as the target number of keyword K1. Then, the target number corresponding to each keyword in keyword set S1 is obtained. The sum of the target numbers corresponding to each keyword for normal physiological indicators in the first detection result is taken as the first number N1, and the sum of the target numbers corresponding to each keyword for abnormal physiological indicators is taken as the second number N2.

[0052] The first quantity here is the number of test results with all normal physiological indicators, and the second quantity is the number of test results with all abnormal physiological indicators. Here, W1=0.4 and W2=0.6 are set.

[0053] The weights of the first quantity N1 and the second quantity N2 are set to W1 and W2 respectively. The feature quantity N0 of medical staff P2 is obtained as N0 = W1 × N1 + W2 × N2. Then, Q keywords are randomly extracted from the daily medical records of each medical staff member other than medical staff P1, and the feature quantity of each medical staff member is finally obtained. The medical staff member with the maximum value is taken as the target personnel.

[0054] Step S3: Based on the monitoring information, the target personnel re-determine the test results of each physiological indicator of the patient. Based on the test results, the correlation degree of each patient with each physiological indicator is obtained. The patient information of the patients to be tested and the monitoring interval of each physiological indicator of the patients to be tested are obtained, and the target set corresponding to the patients to be tested is established.

[0055] Step S3-1: Instruct the target personnel to use the monitoring information M F Patient P was obtained F The test result of each physiological indicator is used as the second test result. If the test results of each physiological indicator in the first and second test results are consistent, then the feature record F is marked; obtain all marked records when the patient is in a certain patient Z, the number of which is G. Z The number of marked records when a certain physiological indicator Y is abnormal is taken as G. Z Y Therefore, the degree of correlation between patient Z and physiological indicator Y is G. Z Y / G Z And obtain the correlation degree of each patient with each physiological indicator;

[0056] Step S3-2: Obtain the patient information of the patient to be tested and the various physiological indicators currently monitored for the patient to be tested. Obtain the correlation degree of each patient in the patient information with a certain physiological indicator X1 monitored for the patient to be tested, and take the maximum correlation degree as the first target value corresponding to physiological indicator X1, and then obtain the first target value corresponding to each physiological indicator.

[0057] Obtain the monitoring interval t of the physiological indicator X1 of the patient to be tested, and obtain the second target value e corresponding to the physiological indicator X1. -k×t Where e is the natural constant and k is the target value correlation coefficient, the second target value corresponding to each physiological indicator is obtained; the weights of the first target value and the second target value are set respectively to obtain the total target value corresponding to each physiological indicator, and the target set corresponding to the patient to be tested is established.

[0058] Since the greater the correlation between a patient's physiological indicators and the indicators, the greater the likelihood of abnormalities in those indicators, the first target value here represents the degree of attention a physiological indicator should receive. The larger the first target value of a physiological indicator, the longer its display duration should be. Since the shorter the monitoring interval, the more frequently the physiological indicator data fluctuates, the greater the attention it should receive. Therefore, the larger the second target value of a physiological indicator, the longer its display duration should be. In conclusion, the larger the overall target value, the longer its display duration should be.

[0059] Among them, due to G Z Y Not greater than G Z Therefore, the correlation degree ranges from 0 to 1, and the first target value also ranges from 0 to 1, because the formula y=e -x In this context, when x > 0, y takes the value from 0 to 1, and y decreases as x increases. Therefore, the value range of the second target value is also from 0 to 1. In this application, the target value correlation coefficient k plays the role of data scaling. k > 0. If the weights of the first target value and the second target value are set to be added together to 1, then the total target value corresponding to each physiological indicator also takes the value range from 0 to 1.

[0060] Step S4: Extract the display duration of various physiological indicators of the patient to be tested, establish the corresponding display set, obtain the warning coefficient of the display duration based on the display set and the target set, and determine whether to issue a warning prompt for the display duration based on the warning coefficient.

[0061] Step S4 includes: extracting the display duration of various physiological indicators of the patient to be tested; establishing a corresponding display set according to the order of each physiological indicator in the target set based on the display duration; and calculating the cosine similarity between the display set and the target set as the warning coefficient for the display duration. If the warning coefficient is less than the preset numerical threshold, a warning prompt is given for the current display duration of each physiological indicator.

[0062] The greater the cosine similarity, the more reasonable the display duration setting is. Therefore, if the warning coefficient is too small, a warning should be issued regarding the current display duration of each physiological indicator.

[0063] This embodiment also provides a machine learning-based multimodal physiological data monitoring system, including a feature recording extraction module, a target personnel identification module, a target set establishment module, and an early warning module. The target personnel identification module includes a detection result determination unit and a target personnel identification unit, and the target set establishment module includes a correlation degree calculation unit and a target set establishment unit. When the system executes the computer program, it implements the above-mentioned machine learning-based multimodal physiological data monitoring method. Since this machine learning-based multimodal physiological data monitoring method has been described in detail above, it will not be repeated here.

[0064] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, specific embodiments have been described above. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims can be performed in a different order than that shown in the embodiments and still achieve the desired result. Additionally, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0065] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0066] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multimodal physiological data monitoring method based on machine learning, characterized in that, Includes the following steps: Step S1: Obtain the hospital's historical monitoring records of patients' physiological indicators, extract the total monitoring time and monitoring interval of each physiological indicator from the monitoring records, obtain the mean deviation of each physiological indicator based on the monitoring interval of each physiological indicator, and determine and extract the feature records in the monitoring records based on the mean deviation. Step S2: Extract the monitoring information and patients corresponding to the feature records, and obtain the patient's disease information during monitoring. Based on the monitoring information, medical staff determine the test results of each physiological indicator of the patient. Extract the daily medical records of each medical staff member, extract and analyze the keywords in the daily medical records and monitoring information, and then identify the target personnel among the medical staff. Step S3: Based on the monitoring information, the target personnel re-determine the test results of each physiological indicator of the patient, and based on the test results, obtain the correlation degree of each patient with each physiological indicator; Obtain the current patient information of the patients to be tested, as well as the monitoring interval of various physiological indicators of the patients to be tested, and establish the target set corresponding to the patients to be tested; Step S4: Extract the display duration of various physiological indicators of the patient to be tested, establish the corresponding display set, obtain the warning coefficient of the display duration based on the display set and the target set, and determine whether to issue a warning prompt for the display duration based on the warning coefficient; Step S2 includes: Step S2-1: Extract the monitoring information and patient corresponding to the feature record, and obtain the patient's disease information during monitoring. The monitoring information is the monitoring data generated during the monitoring of the patient's physiological parameters; extract the monitoring information M corresponding to a certain feature record F. F and patient P F Instruct a medical staff member P1 to monitor information M F Patient P was obtained F The test result of each physiological indicator is taken as the first test result, which indicates whether each physiological indicator is in a normal or abnormal state. Step S2-2: Obtain the daily medical record (CR) of another medical staff member P2. The daily medical record (CR) is a record formed by medical staff member P2 based on daily observations of changes in the patient's condition, treatment measures, examination results, and doctor-patient communication. Extract monitoring information M F For each physiological indicator, keywords are identified and a keyword set S1 is established. Q keywords are randomly extracted from daily medical records (CR) and a keyword set S2 is established. The number of keywords in keyword set S2 whose word similarity to a keyword K1 in keyword set S1 is greater than a preset similarity threshold is taken as the target number of keyword K1. Thus, the target number corresponding to each keyword in keyword set S1 is obtained. The sum of the target numbers corresponding to each keyword for normal physiological indicators in the first detection result is taken as the first number N1, and the sum of the target numbers corresponding to each keyword for abnormal physiological indicators is taken as the second number N2. The weights of the first quantity N1 and the second quantity N2 are set to W1 and W2 respectively. The feature quantity N0 of the medical staff P2 is obtained as W1×N1+W2×N2. Then, Q keywords are randomly extracted from the daily medical records of each medical staff member other than medical staff P1, and the feature quantity of each medical staff member is finally obtained. The medical staff member with the maximum value is taken as the target personnel. Step S3 includes: Step S3-2: Obtain the patient information of the patient to be tested and the various physiological indicators currently monitored for the patient to be tested. Obtain the correlation degree of each patient in the patient information with a certain physiological indicator X1 monitored for the patient to be tested, and take the maximum correlation degree as the first target value corresponding to the physiological indicator X1, and then obtain the first target value corresponding to each physiological indicator. The monitoring interval t of the physiological indicator X1 of the patient to be tested is obtained, and the second target value corresponding to the physiological indicator X1 is obtained as e. -k×t Where e is the natural constant and k is the target value correlation coefficient, the second target value corresponding to each physiological indicator is obtained; the weights of the first target value and the second target value are set respectively to obtain the total target value corresponding to each physiological indicator, and the target set corresponding to the patient to be tested is established.

2. The multimodal physiological data monitoring method based on machine learning according to claim 1, characterized in that, Step S1 includes: Step S1-1: Obtain the hospital's historical monitoring records of the patient's physiological indicators. The monitoring records are formed by medical staff monitoring the changes in the patient's physiological parameters according to the preset monitoring intervals for various physiological indicators. Extract the total monitoring duration MD of a certain physiological indicator X from a certain monitoring record R. X and monitoring interval duration ID X Based on the total monitoring duration MD X The total number N of physiological indicator X recorded internally X The time monitoring bias of physiological index X in monitoring record R was obtained as follows: The average deviation of physiological indicator X over time is calculated in several monitoring records to obtain the mean deviation of physiological indicator X, and then the mean deviation of each physiological indicator is obtained. Step S1-2: Extract the physiological indicators corresponding to a certain monitoring record T, as well as the total monitoring time and monitoring interval for each physiological indicator. Calculate the time monitoring deviation for each physiological indicator and obtain the variance between all time monitoring deviations. If each time monitoring deviation is not greater than the corresponding mean deviation and the variance is less than the preset variance threshold, then the monitoring record T is used as a feature record, thereby obtaining all feature records in the monitoring record.

3. The multimodal physiological data monitoring method based on machine learning according to claim 1, characterized in that, Step S3 also includes: Step S3-1: Instruct the target personnel to use the monitoring information M F Patient P was obtained F The test result of each physiological indicator is used as the second test result. If the test results of each physiological indicator in the first and second test results are consistent, then the feature record F is marked; obtain all marked records when the patient is in a certain patient Z, the number of which is G. Z The number of marked records when a certain physiological indicator Y is abnormal is taken as G. Z Y Therefore, the degree of correlation between the patient Z and the physiological indicator Y is G. Z Y / G Z And obtain the correlation degree of each patient with each physiological indicator.

4. The multimodal physiological data monitoring method based on machine learning according to claim 1, characterized in that, Step S4 includes: extracting the display duration of various physiological indicators of the patient to be tested; establishing a corresponding display set according to the order of each physiological indicator in the target set based on the display duration; and calculating the cosine similarity between the display set and the target set as a warning coefficient for the display duration. If the warning coefficient is less than a preset numerical threshold, a warning prompt is given for the current display duration of each physiological indicator.

5. A multimodal physiological data monitoring system, used to execute the machine learning-based multimodal physiological data monitoring method according to any one of claims 1-4, characterized in that, The system includes a feature record extraction module, a target personnel identification module, a target set establishment module, and an early warning module. Feature record extraction module: used to obtain historical monitoring records of patients' physiological indicators in the hospital, extract the total monitoring time and monitoring interval time of each physiological indicator in the monitoring records, obtain the mean deviation of each physiological indicator based on the monitoring interval time of each physiological indicator, and judge and extract feature records in the monitoring records based on the mean deviation. Target personnel identification module: This module is used to extract monitoring information and patients corresponding to feature records, and obtain patient information during monitoring. Based on the monitoring information, medical staff can determine the test results of each physiological indicator of the patient. It also extracts the daily medical records of each medical staff member, extracts and analyzes keywords in the daily medical records and monitoring information, and then identifies the target personnel among the medical staff. Target set establishment module: This module enables target personnel to re-determine the test results of each physiological indicator of the patient based on the monitoring information, and obtain the correlation degree of each patient with each physiological indicator based on the test results. Obtain the current patient information of the patients to be tested, as well as the monitoring interval of various physiological indicators of the patients to be tested, and establish the target set corresponding to the patients to be tested; Early warning module: used to extract the display duration of various physiological indicators of the patient to be tested, establish a corresponding display set, obtain the early warning coefficient of the display duration based on the display set and the target set, and determine whether to issue an early warning for the display duration based on the early warning coefficient.

6. The multimodal physiological data monitoring system according to claim 5, characterized in that, The target personnel determination module includes a detection result determination unit and a target personnel determination unit; Detection result determination unit: used to extract the monitoring information and patients corresponding to the feature records, and obtain the patient's disease information at the time of monitoring; Extracting feature records corresponding to monitoring information and patients allows medical staff to obtain the test results of each physiological indicator of the patient based on the monitoring information; Target personnel identification unit: used to obtain the daily medical records of medical staff, extract and analyze keywords in the daily medical records and monitoring information, obtain the feature quantity of each medical staff member, and identify the medical staff member with the maximum feature quantity as the target personnel.

7. A multimodal physiological data monitoring system according to claim 5, characterized in that, The target set establishment module includes an association degree calculation unit and a target set establishment unit; Correlation calculation unit: used to enable target personnel to obtain the test results of each physiological indicator of the patient based on the monitoring information, and to determine whether to mark the feature records; Obtain all the labeled records of a patient when the patient is in a certain disease state, and determine the degree of correlation between a certain patient and a certain physiological indicator based on the number of labeled records when a certain physiological indicator is abnormal. Target set establishment unit: used to acquire the patient information and monitored physiological indicators of the patients to be tested, and obtain the first target value corresponding to the physiological indicators based on the degree of correlation; The monitoring interval of physiological indicators of the patients to be tested is obtained, the second target value corresponding to the physiological indicators is obtained, and then the target set corresponding to the patients to be tested is established.

Images