Reproductive health intelligent nursing decision-making method and system based on deep learning

Abstract

The invention relates to the technical field of deep learning, and discloses a reproduction health intelligent nursing decision-making method and system based on deep learning, and the method comprises the following steps: obtaining the comprehensive information of an in-service pregnant woman and the historical data information of the in-service pregnant woman; first one-dimensional structure data is obtained based on in-service pregnant woman comprehensive information sorting, and second one-dimensional structure data is obtained based on in-service pregnant woman historical data information sorting; and inputting the first one-dimensional structure data and the second one-dimensional structure data into a nursing model, and outputting an in-service pregnant woman nursing strategy and probability distribution of occurrence risks of the in-service pregnant woman. According to the invention, a comprehensive risk prediction and nursing strategy can be provided for an in-service pregnant woman, and normal operation of work can be ensured while self health is considered.

Description

Technical Field

[0001] The present invention relates to the field of deep learning technology, and more specifically, to a reproductive health intelligent nursing decision-making method and system based on deep learning. Background Art

[0002] The present invention proposes a reproductive health intelligent nursing decision-making method and system based on deep learning, aiming to use artificial intelligence technology to achieve health care and risk warning for working pregnant women. With the progress of society and the improvement of women's status, more and more pregnant women choose to continue working during pregnancy. However, various risk factors in the working environment may pose a threat to the health of mothers and babies. In addition, due to the physiological and psychological particularities of pregnancy, working pregnant women face higher health risks. Traditional pregnancy care mainly relies on medical institutions and families, and it is difficult to carry out continuous and dynamic health management of working pregnant women. Therefore, it is urgent to develop an intelligent reproductive health care decision-making system to provide personalized and precise health services for working pregnant women. Summary of the invention

[0003] The present invention provides a method and system for intelligent reproductive health care decision-making based on deep learning. The method first obtains comprehensive information and historical data of working pregnant women and performs structured coding processing on them. Then, a nursing model composed of a multi-layer neural network is used to extract features and perform fusion analysis on the encoded data. Through training, the model can automatically evaluate the health risks and work risks of pregnant women, provide personalized nursing measures, and predict the probability of occurrence of various risks, solving technical problems such as incomplete information and inaccurate risk assessment in traditional reproductive health care.

[0004] The present invention provides a reproductive health intelligent nursing decision-making method based on deep learning, comprising the following steps:

[0005] Step 100, obtaining comprehensive information of working pregnant women, including: health data information of working pregnant women, monitoring video image information, and work information of working pregnant women; obtaining historical data information of working pregnant women, including: historical health data information of working pregnant women, medical history information of working pregnant women, and historical work information of working pregnant women;

[0006] Step 200, based on the comprehensive information of working pregnant women, the first one-dimensional structure data is obtained, and the first one-dimensional structure data includes data items sorted by time, The data item represents the The first two-dimensional structure data is generated by the comprehensive information of working pregnant women collected at each moment;

[0007] Step 300, based on the historical data information of working pregnant women, the second one-dimensional structure data is obtained. The second one-dimensional structure data includes data items sorted by time, The data item represents the Historical data information of working pregnant women collected at each moment;

[0008] Step 400, input the one-dimensional structure data No. 1 and the one-dimensional structure data No. 2 into the nursing model, the nursing model includes a first intermediate layer, a second intermediate layer, a third intermediate layer, a first data fusion layer, a first output layer, and a second output layer; the one-dimensional structure data No. 1 is input into the first intermediate layer, and the first intermediate representation data is output to the second intermediate layer; the second intermediate layer outputs the second intermediate representation data to the first intermediate fusion layer; the one-dimensional structure data No. 2 is input into the third intermediate layer, and the third intermediate representation data is output to the first intermediate fusion layer and the second output layer; the first intermediate fusion layer outputs the first fusion representation data to the first output layer; the first output layer outputs the nursing strategy for working pregnant women, and the second output layer outputs the probability distribution of risks occurring to working pregnant women.

[0009] In a preferred embodiment, the health data information of the working pregnant woman includes: a description text of the working pregnant woman's own health data consisting of symptoms, emotions, diet, activities, heart rate, blood pressure, blood sugar, fetal heart rate, and body fluid indicators; the monitoring video image information includes: video monitoring of the pregnant woman's office; the working pregnant woman's work information includes: a description text of the working pregnant woman's work consisting of environmental information and daily work information;

[0010] The historical health data information of working pregnant women includes: a description text of the historical health data of working pregnant women composed of historical symptoms, emotions, diet, activities, heart rate, blood pressure, blood sugar, fetal heart rate, and body fluid indicators; the medical history information of working pregnant women is represented by a description text of the medical history composed of information on diseases the working pregnant women suffered before pregnancy; the historical work information of working pregnant women includes: a description text of the historical work of working pregnant women composed of historical environmental information and daily work information.

[0011] In a preferred embodiment, corresponding features are obtained after feature engineering is performed on the comprehensive information of working pregnant women and the historical data information of working pregnant women;

[0012] The health data information of working pregnant women includes the description text of their own health data, and the WordEmbedding algorithm is used as a feature engineering method;

[0013] The surveillance video image information includes image information, which is used as a feature engineering method through the convolutional neural network (CNN) algorithm;

[0014] The work information of working pregnant women includes the job description text of working pregnant women, and the WordEmbedding algorithm is used as a feature engineering method;

[0015] The historical health data of working pregnant women includes the historical description text of the health data of working pregnant women, and the WordEmbedding algorithm is used as a feature engineering method;

[0016] The medical history information of working pregnant women includes the text describing the medical history, and the WordEmbedding algorithm is used as a feature engineering method;

[0017] The historical work information of employed pregnant women includes the historical work description text of employed pregnant women, and the WordEmbedding algorithm is used as a feature engineering method.

[0018] In a preferred embodiment, the constructed No. 1 two-dimensional structure data includes a data matrix and a relationship matrix, a unit of the data matrix represents comprehensive information of an independent object of a working pregnant woman, the independent object includes a working pregnant woman, a monitoring video image, and work, and a unit of the data matrix only contains comprehensive information of the working pregnant woman of the independent object represented by it;

[0019] The element in the i-th row and j-th column of the relationship matrix represents the association between the i-th cell of the data matrix and the independent object represented by the j-th cell. If there is an association, the value of this element of the relationship matrix is ​​1, otherwise it is 0.

[0020] In a preferred embodiment, the association relationship between independent objects in the relationship matrix is ​​expressed as follows:

[0021] There is a correlation between two working pregnant women, which means that the pregnancy period of the two working pregnant women is similar, and the pregnancy period is similar when the difference between the two working pregnant women is within one month;

[0022] There is a correlation between jobs if: the nature of the two jobs is similar;

[0023] There is a relationship between a working pregnant woman and her job if: the job belongs to the working pregnant woman;

[0024] The correlation between the surveillance video image and the working pregnant woman means that the working pregnant woman is within the surveillance range when working;

[0025] The existence of a correlation between surveillance video images and work means that the workplace is within the surveillance scope.

[0026] In a preferred embodiment, the nursing strategy for working pregnant women includes: health risks of working pregnant women, work risks of working pregnant women, nursing measures based on health risks of working pregnant women, nursing measures based on work risks of working pregnant women, and emergency nursing measures for working pregnant women.

[0027] In a preferred embodiment, the third intermediate layer is combined with the second output layer for separate pre-training, and the training nursing model has the ability to accurately judge the health and work risks of working pregnant women. The loss function calculation formula of the training is as follows:

[0028] ;

[0029] in is the total classification loss, Represents the classification weight parameter, the default value is 0.5, Indicates the major category of loss, Represents sub-category losses. The major category of losses is represented by the classified losses of health risks and work risks. The sub-category losses represent the classified losses of miscarriage or premature birth risks, pregnancy complication risks, physical risks, biological risks, and psychosocial risks.

[0030] ;

[0031] in Represents the loss value of the major category, represents the total number of working pregnant women, It is The actual risk classification label of a working pregnant woman is 0 or 1, where 1 means that the risk category of the working pregnant woman is her own health risk, and 0 means that the risk category of the working pregnant woman is work risk. The nursing model predicts Classification tags for working pregnant women, It represents the logarithmic function with the natural constant e as the base;

[0032] ;

[0033] in represents the loss value of the subcategory, where the subcategory represents the risk of miscarriage or premature birth, the risk of pregnancy complications, physical risk, biological risk, and psychosocial risk. M represents the total number of risk categories. represents the total number of working pregnant women, It is The symbol value of a working pregnant woman, if the The actual classification of the working pregnant women is If there is a subclass, its value is 1, otherwise it is 0. The nursing model predicts The number of working pregnant women belongs to The probability of a subclass, Represents the logarithmic function with the natural constant e as the base.

[0034] In a preferred embodiment, the nursing model is trained, and the steps of nursing model training include:

[0035] Step 101, initializing the parameters of the nursing model;

[0036] Step 102: Observe the comprehensive information of working pregnant women at time e , e-time implementation of the in-service maternity care strategy 、Comprehensive information of working pregnant women at the time of e+1 , implement strategies for on-the-job maternity care Rewards ;

[0037] Step 103, then calculate the error of the nursing strategy for pregnant women in service:

[0038] ;

[0039] in represents the nursing strategy error of working pregnant women at time e, represents the discount factor, , Represents the nursing model input The maximum probability value in the first output vector output when Represents the nursing model input The first output vector outputted at the time corresponds to the in-service maternity care strategy The probability value of

[0040] Step 104, updating the nursing model, the updated formula is as follows:

[0041] ;

[0042] , represents the step size of deep learning, Indicates delivery update;

[0043] Step 105, iterate steps 102-104 until the nursing model converges or the number of iterations reaches a set value; the default value of this value is 85.

[0044] In a preferred embodiment, the reward function calculation formula is as follows:

[0045] ;

[0046] in Indicates reward, represents the total number of visits to the hospital by all working pregnant women recorded in the intelligent nursing decision system, Indicates the number of times working pregnant women go to the hospital except for normal prenatal checkups. represents the total number of pregnant women in the intelligent nursing decision system, Indicates the number of people who actually implement the in-service maternity care strategy, represents the expected average nursing cost of working pregnant women in the intelligent nursing decision system, represents the actual cost of implementing the on-the-job maternity care strategy, It represents the total number of days of pregnancy of working pregnant women recorded in the intelligent nursing decision-making system. Indicates the actual number of working days for employed pregnant women.

[0047] In a preferred embodiment, a reproductive health intelligent nursing decision system based on deep learning includes the following modules:

[0048] A data acquisition module is used to obtain comprehensive information of working pregnant women and historical data information of working pregnant women;

[0049] A data encoding module is used to encode the comprehensive information of working pregnant women into a number one one-dimensional structure data, and to encode the historical data information of working pregnant women into a number two one-dimensional structure data;

[0050] The data processing module inputs the No. 1 one-dimensional structure data and the No. 2 one-dimensional structure data into the nursing model, and outputs the nursing strategy for working pregnant women and the probability distribution of the risks of working pregnant women;

[0051] In the working pregnant women care module, working pregnant women perform daily care according to the working pregnant women care strategy to ensure their own health level and smooth work. Compared with the traditional unified model, the nursing measures of the present invention are more precise, improve the nursing effect and pregnant women's satisfaction, and promote the health of mothers and babies.

[0052] The beneficial effects of the present invention are:

[0053] The present invention adopts deep learning technology to comprehensively analyze the multi-dimensional data of pregnant women, and can comprehensively evaluate the health status during pregnancy and the potential risks in the working environment. Compared with traditional methods, the risk identification of the present invention is more accurate and reduces the probability of critical situations.

[0054] The present invention provides personalized and dynamic nursing decisions based on the individual characteristics and risk factors of pregnant women, including lifestyle guidance, prenatal examination arrangements, medication recommendations, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1A flowchart of a reproductive health intelligent nursing decision-making method based on deep learning is disclosed in one embodiment of the present invention;

[0056] Figure 2 This is a set of desensitized data sample example 1 of the present invention;

[0057] Figure 3 This is Example 2 of a set of desensitized data samples of the present invention. DETAILED DESCRIPTION

[0058] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that the discussion of these implementations is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the present specification. Various examples may omit, replace, or add various processes or components as needed. In addition, the features described in some examples may also be combined in other examples.

[0059] At least one embodiment of the present invention discloses a reproductive health intelligent nursing decision-making method based on deep learning, such as Figure 1 As shown, including:

[0060] Step 100, obtaining comprehensive information of the working pregnant woman, the comprehensive information of the working pregnant woman including: health data information of the working pregnant woman herself, monitoring video image information, and work information of the working pregnant woman;

[0061] Obtaining historical data information of working pregnant women, including: historical health data information of working pregnant women, medical history information of working pregnant women, and historical work information of working pregnant women;

[0062] In one embodiment of the present invention, the health data information of working pregnant women includes: a description text of the health data of working pregnant women composed of symptoms, emotions, diet, activities, heart rate, blood pressure, blood sugar, fetal heart rate, and body fluid indicators; the monitoring video image information includes: video surveillance of the pregnant woman's office; the work information of working pregnant women includes: a work description text of working pregnant women composed of environmental information and daily work information; it should be noted that the video surveillance of the pregnant woman's office is obtained by an ordinary monitoring system used in normal office places, and is used to detect abnormal behaviors of pregnant women, such as falls, squatting for a long time, and other abnormalities.

[0063] The historical health data information of working pregnant women includes: a description text of the historical health data of working pregnant women composed of historical symptoms, emotions, diet, activities, heart rate, blood pressure, blood sugar, fetal heart rate, and body fluid indicators; the medical history information of working pregnant women is represented by a description text of the medical history composed of information on diseases the working pregnant women suffered before pregnancy; the historical work information of working pregnant women includes: a description text of the historical work of working pregnant women composed of historical environmental information and daily work information.

[0064] In one embodiment of the present invention, corresponding features are obtained after feature engineering is performed on the comprehensive information of working pregnant women and the historical data information of working pregnant women;

[0065] The health data information of working pregnant women includes the description text of their own health data, and the WordEmbedding algorithm is used as a feature engineering method;

[0066] The surveillance video image information includes image information, which is used as a feature engineering method through the convolutional neural network (CNN) algorithm;

[0067] The work information of working pregnant women includes the job description text of working pregnant women, and the WordEmbedding algorithm is used as a feature engineering method;

[0068] The historical health data of working pregnant women includes the historical description text of the health data of working pregnant women, and the WordEmbedding algorithm is used as a feature engineering method;

[0069] The medical history information of working pregnant women includes the text describing the medical history, and the WordEmbedding algorithm is used as a feature engineering method;

[0070] The historical work information of employed pregnant women includes the historical work description text of employed pregnant women, and the WordEmbedding algorithm is used as a feature engineering method;

[0071] In one embodiment of the present invention, symptoms include: nausea, vomiting, dizziness, abdominal pain, and the time, frequency, and severity of the symptoms are marked; the severity of the symptoms is expressed as mild, moderate, and severe; mild means that the symptom intensity is weak and the daily work can be completed normally; moderate means that the symptom intensity is obvious, and the working time needs to be reduced for recuperation; severe means that the symptom intensity is severe and the daily work cannot be carried out; emotions include: joy, calmness, irritability, anxiety, and the time and duration are marked; diet includes: the type, amount, and time of food consumed by working pregnant women; activities are expressed as daily activities of working pregnant women, including: work, rest, exercise, and the time and duration are marked; body fluid indicators include: protein, sugar, and ketone body content indicators in body fluids;

[0072] Facial expression images include: partial eyebrows, eyes, mouth corners images and overall facial expression images; body language includes: standing posture, sitting posture, gait;

[0073] Environmental information includes: air quality, noise, light, temperature and humidity; daily work information includes: working hours, work content, and number of work responses.

[0074] Step 200, based on the comprehensive information of working pregnant women, the first one-dimensional structure data is obtained, and the first one-dimensional structure data includes data items sorted by time, The data item represents the The first two-dimensional structure data is generated by the comprehensive information of working pregnant women collected at each moment;

[0075] The first two-dimensional structure data includes a data matrix and a relationship matrix. A unit of the data matrix represents comprehensive information of an independent object of pregnant women at work. The independent objects include pregnant women at work, surveillance video images, and work. A unit of the data matrix only contains comprehensive information of pregnant women at work of the independent object it represents.

[0076] The element in the i-th row and j-th column of the relationship matrix represents the association between the i-th unit of the data matrix and the independent object represented by the j-th unit. If there is an association, the value of this element of the relationship matrix is ​​1, otherwise it is 0;

[0077] There is a correlation between two working pregnant women, which means that the pregnancy period of the two working pregnant women is similar, and the pregnancy period is similar when the difference between the two working pregnant women is within one month;

[0078] There is a correlation between jobs if: the nature of the two jobs is similar;

[0079] There is a relationship between a working pregnant woman and her job if: the job belongs to the working pregnant woman;

[0080] The correlation between the surveillance video image and the working pregnant woman means that the working pregnant woman is within the surveillance range when working;

[0081] The existence of a correlation between surveillance video images and work means that the workplace is within the surveillance scope.

[0082] Step 300, based on the historical data information of working pregnant women, the second one-dimensional structure data is obtained. The second one-dimensional structure data includes data items sorted by time, The data item represents the Historical data information of working pregnant women collected at each moment;

[0083] In one embodiment of the present invention, , .

[0084] Step 400, input the one-dimensional structure data No. 1 and the one-dimensional structure data No. 2 into the nursing model, the nursing model includes a first intermediate layer, a second intermediate layer, a third intermediate layer, a first data fusion layer, a first output layer, and a second output layer; the one-dimensional structure data No. 1 is input into the first intermediate layer, and the first intermediate representation data is output to the second intermediate layer; the second intermediate layer outputs the second intermediate representation data to the first intermediate fusion layer; the one-dimensional structure data No. 2 is input into the third intermediate layer, and the third intermediate representation data is output to the first intermediate fusion layer and the second output layer; the first intermediate fusion layer outputs the first fusion representation data to the first output layer; the first output layer outputs the nursing strategy for working pregnant women, and the second output layer outputs the probability distribution of risks occurring to working pregnant women, and the risks are expressed as health risks for working pregnant women and work risks for working pregnant women.

[0085] In one embodiment of the present invention, the nursing strategy for working pregnant women includes: health risks of working pregnant women, work risks of working pregnant women, nursing measures based on the health risks of working pregnant women, nursing measures based on the work risks of working pregnant women, and emergency nursing measures for working pregnant women.

[0086] In one embodiment of the present invention, the calculation formula of the first intermediate layer is as follows:

[0087]

[0088]

[0089]

[0090] in Indicates the first one-dimensional structure data The first The first intermediate representation data of the unit, and Respectively represent the first one-dimensional structure data The first and The comprehensive information of working pregnant women represented by independent objects, Indicates the first one-dimensional structure data The first and The attention coefficient of each unit, Indicates the first one-dimensional structure data The first and The attention coefficient of each unit, Indicates the first one-dimensional structure data The first and The normalized attention coefficient of each unit, Indicates the first one-dimensional structure data The first The set of cells of the data matrix where there are associations between cells, represents an exponential function with a natural constant as base, represents the trainable attention vector parameter, CONCAT represents the concatenation operation, represents transpose, represents the two-dimensional recognition weight parameter, and LeakyReLU is the LeakyReLU activation function.

[0091] In one embodiment of the present invention, the calculation formula of the second intermediate layer is as follows:

[0092]

[0093] in Indicates A second intermediate representation data, Indicates A second intermediate representation data, , The total number of data items representing the one-dimensional structure data. , Represents the first one-dimensional structure data The set of cells of all data matrices with data items, and are the first and second weight parameters, is the first bias parameter and tanh is the hyperbolic tangent function.

[0094] In one embodiment of the present invention, the calculation formula of the third intermediate layer is as follows:

[0095]

[0096] in Indicates A third intermediate representation data, Indicates A third intermediate representation data, , The total number of data items representing the second-dimensional structure data, The first one-dimensional structure data data items, and are the third and fourth weight parameters, is the second bias parameter and tanh is the hyperbolic tangent function.

[0097] In one embodiment of the present invention, the calculation formula of the first data fusion layer is as follows:

[0098]

[0099] in, represents the fusion representation data, represents the fusion function, Indicates A second intermediate representation data, Indicates A third intermediate representation data, represents the sum weight matrix, Represents the summed bias parameter.

[0100] In one embodiment of the present invention, the calculation formula of the first output layer is as follows:

[0101]

[0102] in represents the first output vector, and its cth component value represents the probability value of the cth nursing strategy for working pregnant women. The nursing strategy for working pregnant women with the largest probability value is selected as the output. The nursing strategy group for working pregnant women contains all the nursing strategies for working pregnant women that can be executed. is the first output weight parameter, is the first output bias parameter, Represents the sigmoid function.

[0103] In one embodiment of the present invention, a nursing strategy for working pregnant women is represented by a matrix, in which the a-th column of the first row of the matrix represents the health risk of the a-th working pregnant woman; the b-th column of the second row of the matrix represents the work risk of the b-th working pregnant woman; the c-th column of the third row of the matrix represents the nursing measures for the c-th working pregnant woman based on the health risks of the working pregnant woman; the d-th column of the fourth row of the matrix represents the nursing measures for the d-th working pregnant woman based on the work risks of the working pregnant woman; and the e-th column of the fifth row of the matrix represents the emergency nursing measures for the e-th working pregnant woman.

[0104] In one embodiment of the present invention, the health risks of working pregnant women include: risk of miscarriage or premature birth, risk of pregnancy complications, and pregnancy complications include: gestational hypertension, gestational diabetes;

[0105] Nursing measures based on the health risks of working pregnant women include: uterine contraction monitoring, blood pressure management, and blood sugar management; uterine contraction monitoring means recording the frequency and duration of uterine contractions; blood pressure management and blood sugar management mean controlling blood pressure and blood sugar to normal levels through diet, exercise, and medication;

[0106] The risks of working pregnant women include: physical risks, biological risks, and psychosocial risks;

[0107] Physical risks include: varicose veins, lower limb edema, and back pain caused by long working hours; biological risks include: office radiation and office environment pollution; psychosocial risks include: high work pressure, difficulty in work-life balance, and workplace discrimination or lack of support;

[0108] Nursing measures based on the work risks of working pregnant women include: rest arrangements, radiation protection, environmental air improvement, and stress management; rest arrangements refer to arranging regular rest time, such as taking a 15-minute break every hour; radiation protection refers to controlling the time working pregnant women use electronic devices, such as using computers or other devices for no more than 40 minutes or wearing radiation-proof clothing to work; environmental air improvement includes: controlling the dryness and humidity in the office environment, and removing harmful substances such as PM2.5 and formaldehyde in the air; stress management includes: participating in relaxation activities, such as yoga and meditation;

[0109] Emergency care measures include: first aid support and emergency care; first aid support means requesting rescue support, such as ambulances and emergency personnel; first aid care includes: oxygen inhalation, cardiopulmonary resuscitation, and left side lying position.

[0110] In one embodiment of the present invention, the calculation formula of the second output layer is as follows:

[0111]

[0112] in represents the second output vector, and its fth component represents the probability distribution of the risk of the fth working pregnant woman. The risks include: health risk of working pregnant women, work risk of working pregnant women, Indicates A third intermediate representation data, is the second output weight parameter, is the second output bias parameter, Represents the sigmoid function.

[0113] In one embodiment of the present invention, the third intermediate layer is combined with the second output layer for separate pre-training, and the training nursing model has the ability to accurately judge the health and work risks of working pregnant women. The loss function calculation formula of the training is as follows:

[0114]

[0115] in is the total classification loss, Represents the classification weight parameter, the default value is 0.5, Indicates the major category of loss, Represents sub-category losses. The major category of losses is represented by classified losses of health risks and work risks. The sub-category losses represent classified losses of miscarriage or premature birth risks, pregnancy complications risks, physical risks, biological risks, and psychosocial risks.

[0116]

[0117] in Represents the loss value of the major category, represents the total number of working pregnant women, It is The actual risk classification label of a working pregnant woman is 0 or 1, where 1 means that the risk category of the working pregnant woman is her own health risk, and 0 means that the risk category of the working pregnant woman is work risk. The nursing model predicts Classification tags for working pregnant women, Represents the logarithmic function with the natural constant e as the base.

[0118]

[0119] in represents the loss value of the sub-category, where the sub-category represents the risk of miscarriage or premature birth, risk of pregnancy complications, physical risk, biological risk, and psychosocial risk. M represents the total number of classification categories. represents the total number of working pregnant women, It is The symbol value of a working pregnant woman, if the The actual classification of the working pregnant women is If there is a subclass, its value is 1, otherwise it is 0. The nursing model predicts The number of working pregnant women belongs to The probability of a subclass, Represents the logarithmic function with the natural constant e as the base.

[0120] In one embodiment of the present invention, the nursing model is trained, and the steps of nursing model training include:

[0121] Step 101, initializing the parameters of the nursing model;

[0122] Step 102: Observe the comprehensive information of working pregnant women at time e , e-time implementation of the in-service maternity care strategy 、Comprehensive information of working pregnant women at the time of e+1 , implement strategies for on-the-job maternity care Rewards ;

[0123] Step 103, then calculate the error of the nursing strategy for pregnant women in service:

[0124]

[0125] in represents the nursing strategy error of working pregnant women at time e, represents the discount factor, , Represents the nursing model input The maximum probability value in the first output vector output when Represents the nursing model input The first output vector outputted at the time corresponds to the in-service maternity care strategy The probability value of

[0126] Step 104, updating the nursing model, the updated formula is as follows:

[0127]

[0128] , represents the step size of deep learning, Indicates delivery update;

[0129] Step 105, iterate steps 102-104 until the nursing model converges or the number of iterations reaches a set value. The default value of this value is 85.

[0130] In one embodiment of the present invention, a reproductive health intelligent nursing decision-making system based on deep learning includes the following modules:

[0131] A data acquisition module is used to obtain comprehensive information of working pregnant women and historical data information of working pregnant women;

[0132] A data encoding module is used to encode the comprehensive information of working pregnant women into a number one one-dimensional structure data, and to encode the historical data information of working pregnant women into a number two one-dimensional structure data;

[0133] The data processing module inputs the No. 1 one-dimensional structure data and the No. 2 one-dimensional structure data into the nursing model, and outputs the nursing strategy for working pregnant women and the probability distribution of the risks of working pregnant women;

[0134] In the working pregnant women care module, working pregnant women perform daily care according to the working pregnant women care strategy to ensure their own health level and smooth work.

[0135] In one embodiment of the present invention, an example of a specific application scenario of the aforementioned reproductive health intelligent nursing decision-making method based on deep learning is provided:

[0136] This example focuses on the health care of working pregnant women, especially urban working pregnant women. Working pregnant women face unique health risks and pressures, such as pregnancy complications, occupational risks, work-life balance and other issues. Therefore, it is necessary to provide working pregnant women with full-process, personalized health management services to maximize the safety of mothers and babies and promote equal employment for pregnant women.

[0137] Data collection

[0138] Comprehensive information for working pregnant women

[0139] Working pregnant women’s own health data: descriptive text of physiological and psychological indicators such as symptoms, emotions, diet, and activities.

[0140] Surveillance video images: visual information such as facial expressions and body language.

[0141] Work information for working pregnant women: descriptive text of the work environment and daily work content.

[0142] Historical data of pregnant women in work

[0143] Historical personal health data: text describing the health status at each stage of pregnancy.

[0144] Past medical history: A text description of any medical conditions you had before pregnancy.

[0145] Historical work information: text describing past work environment and content.

[0146] Data Annotation

[0147] Health risk labeling:

[0148] Positive samples: pregnant women who have experienced miscarriage, premature birth or pregnancy complications.

[0149] Negative samples: pregnant women with no obvious health problems during pregnancy.

[0150] Labeling method: Based on the medical record information of the pregnant woman, determine whether it is a positive sample and form a binary label.

[0151] Job risk labeling:

[0152] Positive examples of physical risk: pregnant women who engage in high-intensity physical labor such as standing for long periods of time and carrying heavy objects for a long time.

[0153] Physical risk negative sample: pregnant women doing easy jobs or office work.

[0154] Positive samples of biological risks: pregnant women who are exposed to dangerous factors such as radiation and toxic chemicals in their working environment.

[0155] Negative biological risk sample: pregnant women working in a safe and harmless environment.

[0156] Positive sample of psychosocial risk: pregnant women with high work pressure and severe discrimination in the workplace.

[0157] Negative sample of psychosocial risk: pregnant women working in a friendly atmosphere and with low stress.

[0158] Labeling method: Based on the work information of pregnant women, determine whether they are positive samples of various risks and form multiple labels.

[0159] Data Source

[0160] Physiological indicators of pregnant women collected by health testing equipment.

[0161] Mobile App: records pregnant women’s subjective feelings, diet, activities and other information.

[0162] Video surveillance system: obtains video data such as pregnant women’s facial expressions, body language, etc.

[0163] Electronic medical record system: provides medical history data of pregnant women.

[0164] Human Resources Management System: Provides information on pregnant women’s job positions, work content, work environment, etc. Data preprocessing

[0165] Feature Engineering

[0166] Text information: WordEmbedding is used to map the description text into a semantic vector.

[0167] Image information: Convolutional neural network (CNN) is used to extract visual features.

[0168] Model Pre-training

[0169] The third intermediate layer and the second output layer are trained separately to enable them to have risk assessment capabilities.

[0170] The cross entropy loss function is used to calculate the classification losses for health risks, work risks and their subclasses respectively to improve the accuracy of risk identification.

[0171] Model training (reinforcement learning)

[0172] Environment: Health and work status of working pregnant women.

[0173] Action: Care strategies for working pregnant women (health interventions, work adjustments, emergency plans, etc.).

[0174] Rewards: Considers factors such as number of medical visits, working days, and care costs.

[0175] The Q-learning algorithm is used for training, and rewards are used for iterative updates to gradually improve the quality of nursing strategies.

[0176] The reward function calculation formula is as follows:

[0177]

[0178] in Indicates reward, represents the total number of visits to the hospital by all working pregnant women recorded in the intelligent nursing decision system, Indicates the number of times working pregnant women go to the hospital except for normal prenatal checkups. represents the total number of pregnant women in the intelligent nursing decision system, Indicates the number of people who actually implement the in-service maternity care strategy, represents the expected average nursing cost of working pregnant women in the intelligent nursing decision system, represents the actual cost of implementing the on-the-job maternity care strategy, It represents the total number of days of pregnancy of working pregnant women recorded in the intelligent nursing decision-making system. Indicates the actual number of working days for employed pregnant women.

[0179] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation mode. The above-mentioned specific implementation mode is merely illustrative and not restrictive. Under the guidance of this embodiment, ordinary technicians in this field can also make more forms of equivalent embodiments, all of which are within the protection of this embodiment.

Claims

1. A reproductive health intelligent nursing decision-making method based on deep learning, characterized in that: The following steps are involved: Step 100, obtaining comprehensive information of working pregnant women, including: health data information of working pregnant women, monitoring video image information, and work information of working pregnant women; obtaining historical data information of working pregnant women, including: historical health data information of working pregnant women, medical history information of working pregnant women, and historical work information of working pregnant women; Step 200, based on the comprehensive information of working pregnant women, the first one-dimensional structure data is obtained, and the first one-dimensional structure data includes data items sorted by time, The data item represents the The first two-dimensional structure data is generated by the comprehensive information of working pregnant women collected at each moment; Step 300, based on the historical data information of working pregnant women, the second one-dimensional structure data is obtained. The second one-dimensional structure data includes data items sorted by time, The data item represents the Historical data information of working pregnant women collected at each moment; Step 400, input the one-dimensional structure data No. 1 and the one-dimensional structure data No. 2 into the nursing model, the nursing model includes a first intermediate layer, a second intermediate layer, a third intermediate layer, a first data fusion layer, a first output layer, and a second output layer; the one-dimensional structure data No. 1 is input into the first intermediate layer, and the first intermediate representation data is output to the second intermediate layer; the second intermediate layer outputs the second intermediate representation data to the first intermediate fusion layer; the one-dimensional structure data No. 2 is input into the third intermediate layer, and the third intermediate representation data is output to the first intermediate fusion layer and the second output layer; the first intermediate fusion layer outputs the first fusion representation data to the first output layer; the first output layer outputs the nursing strategy for working pregnant women, and the second output layer outputs the probability distribution of risks occurring to working pregnant women.

2. According to a deep learning-based reproductive health intelligent nursing decision-making method according to claim 1, it is characterized in that: The working pregnant woman’s own health data information includes: symptoms, emotions, diet, activities, heart rate, blood pressure, blood sugar, fetal heart rate, and body fluid indicators, which are composed of the working pregnant woman’s own health data description text; monitoring video image information includes: video monitoring of the pregnant woman’s office; working pregnant woman’s work information includes: working pregnant woman’s work description text, which is composed of environmental information and daily work information; The historical health data information of working pregnant women includes: a description text of the historical health data of working pregnant women composed of historical symptoms, emotions, diet, activities, heart rate, blood pressure, blood sugar, fetal heart rate, and body fluid indicators; the medical history information of working pregnant women is represented by a description text of the medical history composed of information on diseases the working pregnant women suffered before pregnancy; the historical work information of working pregnant women includes: a description text of the historical work of working pregnant women composed of historical environmental information and daily work information.

3. According to a deep learning-based reproductive health intelligent nursing decision-making method according to claim 1, it is characterized in that: After performing feature engineering on the comprehensive information of working pregnant women and the historical data information of working pregnant women, the corresponding features are obtained; The health data information of working pregnant women includes the description text of their own health data, and the WordEmbedding algorithm is used as a feature engineering method; The surveillance video image information includes image information, which is used as a feature engineering method through the convolutional neural network (CNN) algorithm; The work information of working pregnant women includes the job description text of working pregnant women, and the WordEmbedding algorithm is used as a feature engineering method; The historical health data of working pregnant women includes the historical health data description text of working pregnant women, and the WordEmbedding algorithm is used as a feature engineering method; The medical history information of working pregnant women includes the text describing the medical history, and the WordEmbedding algorithm is used as a feature engineering method; The historical work information of employed pregnant women includes the historical work description text of employed pregnant women, and the WordEmbedding algorithm is used as a feature engineering method.

4. According to a deep learning-based reproductive health intelligent nursing decision-making method according to claim 1, it is characterized in that: The constructed No. 1 two-dimensional structure data includes a data matrix and a relationship matrix. A unit of the data matrix represents the comprehensive information of an independent object of a working pregnant woman. The independent object includes a working pregnant woman, a monitoring video image, and work. A unit of the data matrix only contains the comprehensive information of the working pregnant woman of the independent object it represents. The element in the i-th row and j-th column of the relationship matrix represents the association between the i-th cell of the data matrix and the independent object represented by the j-th cell. If there is an association, the value of this element of the relationship matrix is ​​1, otherwise it is 0.

5. A reproductive health intelligent nursing decision-making method based on deep learning according to claim 4, characterized in that: The association relationship between independent objects in the relationship matrix is ​​expressed as follows: There is a correlation between two working pregnant women, which means that the pregnancy period of the two working pregnant women is similar, and the pregnancy period is similar when the difference between the two working pregnant women is within one month; There is a correlation between jobs if: the nature of the two jobs is similar; There is a relationship between a working pregnant woman and her job if: the job belongs to the working pregnant woman; The correlation between the surveillance video image and the working pregnant woman means that the working pregnant woman is within the surveillance range when working; The existence of a correlation between surveillance video images and work means that the workplace is within the surveillance scope.

6. The method for reproductive health intelligent nursing decision-making based on deep learning according to claim 1 is characterized in that: The nursing strategies for working pregnant women include: health risks of working pregnant women, work risks of working pregnant women, nursing measures based on health risks of working pregnant women, nursing measures based on work risks of working pregnant women, and emergency nursing measures for working pregnant women.

7. The method for reproductive health intelligent nursing decision-making based on deep learning according to claim 1 is characterized in that: The third intermediate layer is combined with the second output layer for separate pre-training. The training nursing model has the ability to accurately judge the health and work risks of working pregnant women. The loss function calculation formula of the training is as follows: ； in is the total classification loss, Represents the classification weight parameter, the default value is 0.5, Indicates the major category of loss, Represents sub-category losses. The major category of losses is represented by the classified losses of health risks and work risks. The sub-category losses represent the classified losses of miscarriage or premature birth risks, pregnancy complication risks, physical risks, biological risks, and psychosocial risks. ； in Represents the loss value of the major category, represents the total number of working pregnant women, It is The actual risk classification label of a working pregnant woman is 0 or 1, where 1 means that the risk category of the working pregnant woman is her own health risk, and 0 means that the risk category of the working pregnant woman is work risk. The nursing model predicts Classification tags for working pregnant women, It represents the logarithmic function with the natural constant e as the base; ； in represents the loss value of the subcategory, where the subcategory represents the risk of miscarriage or premature birth, the risk of pregnancy complications, physical risk, biological risk, and psychosocial risk. M represents the total number of risk categories. represents the total number of working pregnant women, It is The symbol value of a working pregnant woman, if the The actual classification of the working pregnant women is If there is a subclass, its value is 1, otherwise it is 0. The nursing model predicts The number of working pregnant women belongs to The probability of a subclass, Represents the logarithmic function with the natural constant e as the base.

8. The method for reproductive health intelligent nursing decision-making based on deep learning according to claim 1 is characterized in that: The nursing model is trained. The steps of nursing model training include: Step 101, initializing the parameters of the nursing model; Step 102: Observe the comprehensive information of working pregnant women at time e , e-time implementation of the in-service maternity care strategy 、Comprehensive information of working pregnant women at the time of e+1 , implement strategies for on-the-job maternity care Rewards ; Step 103, then calculate the error of the nursing strategy for pregnant women in service: ； in represents the nursing strategy error of working pregnant women at time e, represents the discount factor, , Represents the nursing model input The maximum probability value in the first output vector output when Represents the nursing model input The first output vector outputted at the time corresponds to the in-service maternity care strategy The probability value of Step 104, updating the nursing model, the updated formula is as follows: ； , represents the step size of deep learning, Indicates delivery update; Step 105, iterate steps 102-104 until the nursing model converges or the number of iterations reaches a set value; the default value of this value is 85.

9. The method for reproductive health intelligent nursing decision-making based on deep learning according to claim 8, characterized in that: The reward function calculation formula is as follows: ； in Indicates reward, represents the total number of visits to the hospital by all working pregnant women recorded in the intelligent nursing decision system, Indicates the number of times working pregnant women go to the hospital except for normal prenatal checkups. represents the total number of pregnant women in the intelligent nursing decision system, Indicates the number of people who actually implement the in-service maternity care strategy, represents the expected average nursing cost of working pregnant women in the intelligent nursing decision system, represents the actual cost of implementing the on-the-job maternity care strategy, It represents the total number of days of pregnancy of working pregnant women recorded in the intelligent nursing decision-making system. Indicates the actual number of working days for employed pregnant women.

10. A reproductive health intelligent nursing decision-making system based on deep learning, characterized in that: A method for making decisions on reproductive health intelligent nursing care based on deep learning is used to implement any one of claims 1 to 9, and a reproductive health intelligent nursing care decision system based on deep learning comprises the following modules: A data acquisition module is used to obtain comprehensive information of working pregnant women and historical data information of working pregnant women; A data encoding module is used to encode the comprehensive information of working pregnant women into a number one one-dimensional structure data, and to encode the historical data information of working pregnant women into a number two one-dimensional structure data; The data processing module inputs the No. 1 one-dimensional structure data and the No. 2 one-dimensional structure data into the nursing model, and outputs the nursing strategy for working pregnant women and the probability distribution of the risks of working pregnant women; In the working pregnant women care module, working pregnant women perform daily care according to the working pregnant women care strategy to ensure their own health level and smooth work.

Images