A maternal prenatal status intelligent analysis system and method

Abstract

The present invention discloses a maternal prenatal state intelligent analysis system and method, which relates to the field of biomedical technology, wherein the method comprises: real-time collection of maternal multimodal data and preprocessing; fusing the preprocessed multimodal data to generate a multimodal feature vector; constructing a deep learning model to evaluate the prenatal state and obtain an evaluation result; generating a personalized intervention plan based on the evaluation result; collecting maternal real-time data and feedback information, monitoring the intervention effect, and evaluating the effectiveness of the intervention plan. By generating a personalized intervention plan based on the evaluation results, precise intervention on maternal is achieved, which not only improves the pertinence and effectiveness of the intervention, but also can timely discover and solve potential risk factors, thereby ensuring the safety of mother and baby.

Description

Technical Field

[0001] The present invention relates to the field of biomedical technology, and in particular to a system and method for intelligently analyzing the prenatal status of parturients. Background Art

[0002] In recent years, the integrated application of imaging data, physiological data, and psychological data has shown great potential in prenatal status assessment. Imaging data, such as ultrasound images, can provide detailed information on fetal development and fetal position; physiological data, such as heart rate, blood pressure, and blood oxygen saturation, reflect the physical condition of the mother; psychological data, such as emotional index, reveals the psychological state of the mother. The fusion of these multimodal data not only improves the accuracy of the assessment, but also provides a basis for personalized intervention. However, existing prenatal status assessment methods still have many shortcomings, especially in data processing and model construction.

[0003] Existing methods for assessing prenatal status mainly rely on a single data source, such as traditional ultrasound examinations or simple physiological parameter monitoring, which often cannot fully reflect the overall status of the mother. First, the limitations of a single data source lead to one-sidedness and inaccuracy in the assessment results. Second, existing data processing methods are usually relatively simple and lack the ability to deeply integrate multimodal data and extract advanced features. In addition, most of the existing assessment models are based on traditional statistical methods, such as logistic regression and support vector machines, which perform poorly when processing high-dimensional, complex multimodal data and have difficulty capturing deep features in the data. Finally, existing methods for generating intervention plans lack dynamism and personalization, and are usually one-time static recommendations that cannot be dynamically adjusted based on real-time data and feedback information. Summary of the invention

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

[0005] Therefore, the present invention provides a system and method for intelligent analysis of maternal prenatal status, which solves the problems of single data source limitation, rough data processing and lack of dynamic and personalization of intervention plans in existing prenatal status assessment methods.

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

[0007] In a first aspect, an embodiment of the present invention provides a method for intelligently analyzing a maternal prenatal state, which includes:

[0008] Collect multimodal data of parturients in real time and perform preprocessing;

[0009] The preprocessed multimodal data is fused to generate a multimodal feature vector;

[0010] Build a deep learning model to conduct prenatal status assessment and obtain the assessment results;

[0011] Generate personalized intervention plans based on the assessment results;

[0012] Collect real-time data and feedback from mothers, monitor intervention effects, and evaluate the effectiveness of intervention plans;

[0013] Wherein, the multimodal data includes imaging data, physiological data and psychological data;

[0014] The physiological data refers to the mother's heart rate, blood pressure and blood oxygen saturation;

[0015] The psychological data refers to the emotional index of the parturient;

[0016] The preprocessing includes noise removal, edge detection, data cleaning and quantization processing;

[0017] The steps of fusing the preprocessed multimodal data to generate a multimodal feature vector include:

[0018] Extracting feature maps from preprocessed ultrasonic images to form image feature vectors;

[0019] Extracting time series features from normalized physiological data to form physiological feature vectors;

[0020] Extract emotional change characteristics from the quantified emotional index to form a psychological feature vector;

[0021] Consider each parturient as a node in the graph;

[0022] According to the social support relationship between the mothers, the edges were established, and the weights of the edges were defined based on the frequency of interaction and the degree of emotional support;

[0023] The image feature vector, physiological feature vector and psychological feature vector are input into the graph neural network as node features;

[0024] The node features are aggregated through the graph convolution layer, and the features of neighboring nodes are considered to extract the influence of related pregnant women in the social network;

[0025] Through the propagation mechanism of the graph neural network, the information of each node is fused to obtain a new feature representation;

[0026] The feature representation processed by the graph neural network is concatenated with the feature vectors of each modality to generate the final multimodal feature vector.

[0027] As a preferred embodiment of the method for intelligent analysis of maternal prenatal status of the present invention, constructing a deep learning model, performing prenatal status assessment, and obtaining the assessment results include the following steps:

[0028] Randomly extract data at multiple time points from each parturient's multimodal data as positive sample pairs;

[0029] Randomly extract data from multimodal data of different parturients as negative sample pairs;

[0030] Use contrastive loss function to train the model, shorten the distance between positive sample pairs and increase the distance between negative sample pairs.

[0031] Design a shared encoder and add a projection head after the encoder to project the encoder's multimodal data output into a low-dimensional space;

[0032] Dynamically adjust feature weights based on the specific circumstances of the mother;

[0033] According to the dynamically adjusted feature weights, the multimodal data in the low-dimensional space is input into the model, and the prenatal assessment status results are output. The expression is: ;in, is the prenatal status assessment vector, is the image feature, For physiological characteristics, For psychological characteristics, is the weight of the image feature, is the weight of the physiological characteristics, is the weight of the psychological characteristics;

[0034] According to the specific components in the evaluation vector, the childbirth difficulty index and emotional stability index are calculated, and the expression is: ; ;in, The difficulty index of childbirth. is the emotional stability index, is the score for the difficulty of childbirth in the assessment results. For the emotional stability score in the evaluation results, The evaluation results are elements, The index of the element in the evaluation result.

[0035] As a preferred embodiment of the method for intelligent analysis of maternal prenatal status of the present invention, generating a personalized intervention plan according to the evaluation results includes the following steps:

[0036] The calculated childbirth difficulty index and emotional stability index are comprehensively scored to evaluate whether they are suitable for generating an intervention plan. The expression is: ;in, It is a comprehensive score of the childbirth difficulty index and emotional stability index. For time The difficulty index of childbirth at For time Emotional stability index, is the attenuation coefficient, is the time variable, The time span for the assessment;

[0037] A threshold is set based on the comprehensive score. When the comprehensive score is greater than the threshold, it is considered necessary to generate a personalized intervention plan.

[0038] The current health status, historical records, delivery difficulty index, and emotional stability index of the parturient are defined as the state space, and various possible intervention measures are defined as the action space;

[0039] Call the genomics analysis module to obtain the genetic information of the mother;

[0040] Design a reward function based on the mother’s current state and genetic information;

[0041] When an intervention measure can significantly improve the mood stability index of pregnant women and is consistent with their genetic characteristics, it will receive a higher positive reward;

[0042] Using a multi-layer perceptron in a deep Q-network, the state space and genetic information are input to obtain the value of each intervention measure;

[0043] The combined scores of the labor difficulty index and the emotional stability index are input into the trained model to generate a personalized intervention plan, which is expressed as: ;in, The intensity of the personalized intervention program, is the number of intervention measures, Index of intervention measures, Generate sensitivity to changes in composite scores for adjusting intervention options, For the Thresholds for intervention measures, For the The weight of the intervention measures, is the activation function.

[0044] As a preferred solution of the method for intelligent analysis of maternal prenatal status of the present invention, collecting real-time data and feedback information of the maternal, monitoring the intervention effect, and evaluating the effectiveness of the intervention plan include the following steps:

[0045] Gathering maternal feedback through a mobile app;

[0046] Combining real-time data and feedback information, the effectiveness of the intervention plan is evaluated, and the expression is: ;in, is the predictive value of the effectiveness of the intervention program. is the number of effectiveness indicators, is the sensitivity parameter for effectiveness prediction, For the The validation threshold of the effectiveness indicator is It is The weight of the effectiveness indicator, The index of effectiveness indicator.

[0047] In a second aspect, the present invention provides a maternal prenatal status intelligent analysis system, comprising:

[0048] The acquisition and preprocessing module collects multimodal data of parturients in real time and performs preprocessing;

[0049] The feature fusion module fuses the preprocessed multimodal data to generate a multimodal feature vector;

[0050] The evaluation module builds a deep learning model to conduct prenatal status assessment and obtain the evaluation results;

[0051] The intervention plan generation module generates personalized intervention plans based on the evaluation results;

[0052] The effect monitoring and optimization module collects real-time data and feedback information from parturients, monitors the effect of intervention, and evaluates the effectiveness of intervention plans;

[0053] The multimodal data includes imaging data, physiological data and psychological data; the physiological data refers to the heart rate, blood pressure and blood oxygen saturation of the parturient; the psychological data refers to the emotional index of the parturient; the preprocessing includes noise removal, edge detection, data cleaning and quantification processing;

[0054] Wherein, the feature fusion module is specifically used for:

[0055] Extracting feature maps from preprocessed ultrasonic images to form image feature vectors;

[0056] Extracting time series features from normalized physiological data to form physiological feature vectors;

[0057] Extract emotional change characteristics from the quantified emotional index to form a psychological feature vector;

[0058] Consider each parturient as a node in the graph;

[0059] According to the social support relationship between the mothers, the edges were established, and the weights of the edges were defined based on the frequency of interaction and the degree of emotional support;

[0060] The image feature vector, physiological feature vector and psychological feature vector are input into the graph neural network as node features;

[0061] The node features are aggregated through the graph convolution layer, and the features of neighboring nodes are considered to extract the influence of related pregnant women in the social network;

[0062] Through the propagation mechanism of the graph neural network, the information of each node is fused to obtain a new feature representation;

[0063] The feature representation processed by the graph neural network is concatenated with the feature vectors of each modality to generate the final multimodal feature vector.

[0064] In a third aspect, an embodiment of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein the computer program, when executed by the processor, implements any step of the method for intelligent analysis of maternal prenatal status as described in the first aspect of the present invention.

[0065] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements any step of the method for intelligent analysis of maternal prenatal status as described in the first aspect of the present invention.

[0066] The beneficial effects of the present invention are as follows: by collecting multimodal data of parturients in real time, comprehensive monitoring of the parturient's status is achieved, the health status of the parturient can be reflected from different angles, and the comprehensiveness and accuracy of the assessment can be improved; by fusing the preprocessed multimodal data to generate multimodal feature vectors, comprehensive processing of multiple data sources is achieved; by constructing a deep learning model and conducting prenatal status assessment, comprehensive and accurate assessment results are obtained, capturing the complex changes in the parturient's status; by generating personalized intervention plans based on the assessment results, precise intervention for parturients is achieved, the pertinence and effectiveness of the intervention are improved, and the safety and health of mothers and babies are guaranteed. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0068] Figure 1 This is a flow chart of the intelligent analysis method for the prenatal status of a parturient in Example 1.

[0069] Figure 2This is a flow chart of the intelligent analysis system for the prenatal status of parturients in Example 1. DETAILED DESCRIPTION

[0070] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0071] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0072] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0073] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, and provides a method for intelligent analysis of maternal prenatal status, such as Figure 1 As shown, the method comprises the following steps:

[0074] S1, real-time collection of maternal multimodal data and preprocessing;

[0075] Among them, multimodal data includes imaging data, physiological data and psychological data; physiological data refers to the mother's heart rate, blood pressure and blood oxygen saturation; psychological data refers to the mother's emotional index; preprocessing includes noise removal, edge detection, data cleaning and quantification processing.

[0076] Furthermore, noise removal refers to the use of Gaussian filters to remove noise from images and improve image quality; edge detection refers to the use of the Canny edge detection algorithm to detect edges in images and highlight important features; data cleaning refers to the use of the Z-score method to detect and remove outliers and interpolate to fill missing values; quantitative processing refers to converting psychological data results into numerical form through questionnaires.

[0077] S2, fusing the preprocessed multimodal data to generate a multimodal feature vector includes the following steps:

[0078] Extract feature graphs from preprocessed ultrasound images to form image feature vectors; extract time series features from normalized physiological data to form physiological feature vectors; extract emotional change features from quantified emotional indexes to form psychological feature vectors; regard each parturient as a node in the graph;

[0079] According to the social support relationship between mothers, edges are established, and the weights of the edges are defined based on the frequency of interaction and the degree of emotional support. The image feature vector, physiological feature vector and psychological feature vector are input into the graph neural network as node features. The node features are aggregated through the graph convolution layer, and the features of neighboring nodes are considered to extract the influence of related mothers in the social network. The information of each node is fused through the propagation mechanism of the graph neural network to obtain a new feature representation. The feature representation processed by the graph neural network is spliced ​​with the feature vectors of each modality to generate the final multimodal feature vector.

[0080] Furthermore, 3D convolutional neural network (3D CNN) is used to extract image features. 3D CNN can capture features in three-dimensional space and is suitable for processing ultrasound images.

[0081] Specifically, multiple 3D convolutional layers are used, each convolutional layer contains multiple 3D convolution kernels, such as (3, 3, 3) or (5, 5, 5); a 3D maximum pooling layer is added after each convolutional layer to reduce the dimension of the feature map and reduce the amount of computation; a batch normalization layer is added after each convolutional layer and pooling layer to accelerate the training process and improve the stability of the model; one or more fully connected layers are added after the last convolutional layer to map the extracted features to a higher-level abstract representation and output a fixed-length image feature vector.

[0082] The Transformer model is used to extract emotional change features. The Transformer can capture the subtle differences in emotional changes through the self-attention mechanism, which helps to assess the mental health status of pregnant women.

[0083] Specifically, a multi-head self-attention mechanism is used to capture the dependencies between different time steps. Each attention head can focus on different subspace features; a feed-forward network is added after each self-attention layer to further process the features; a layer normalization layer is added after each self-attention layer and feed-forward network to accelerate the training process and improve the stability of the model; one or more fully connected layers are added after the last Transformer layer to map the extracted features to a higher-level abstract representation and output a fixed-length psychological feature vector.

[0084] S3, building a deep learning model to conduct prenatal status assessment, and obtaining the assessment results includes the following steps:

[0085] Randomly extract data from multiple time points in the multimodal data of each parturient as positive sample pairs; randomly extract data from the multimodal data of different parturients as negative sample pairs; use the contrast loss function to train the model to shorten the distance between positive sample pairs and increase the distance between negative sample pairs; design a shared encoder, add a projection head after the encoder, and project the encoder's multimodal data output to a low-dimensional space; dynamically adjust the feature weights according to the specific situation of the parturient; input the multimodal data in the low-dimensional space into the model according to the dynamically adjusted feature weights, and output the prenatal assessment status results, expressed as: ;in, is the prenatal status assessment vector, is the image feature, For physiological characteristics, For psychological characteristics, is the weight of the image feature, is the weight of the physiological characteristics, is the weight of the psychological characteristics;

[0086] According to the specific components in the evaluation vector, the childbirth difficulty index and emotional stability index are calculated, and the expression is: ; ;in, The difficulty index of childbirth. is the emotional stability index, is the score for the difficulty of childbirth in the assessment results. is the score of emotional stability in the evaluation results. The evaluation results are elements, The index of the element in the evaluation result.

[0087] Furthermore, The value range is [0, 1], where 0 means that the evaluation result is very unsatisfactory and requires immediate intervention; 0.5 means that the evaluation result is average and requires attention and appropriate intervention; 1 means that the evaluation result is very ideal and no special intervention is required.

[0088] according to Generate a labor difficulty index score and recommendations if Greater than 0.7, further prenatal examination and consultation with a doctor are recommended; if Between 0.3 and 0.7, regular monitoring and moderate rest are recommended; if If it is less than 0.3, it means that the delivery is not difficult and you can just maintain a normal pace of life.

[0089] according to Generates a mood stability index score and recommendations if Greater than 0.7, psychological counseling and relaxation exercises are recommended; if Between 0.3 and 0.7, regular psychological assessment and moderate leisure activities are recommended; if If it is less than 0.3, it means that your emotions are relatively stable and you just need to maintain a positive attitude.

[0090] Furthermore, through positive sample pairs, the model can learn the feature changes of the same parturient at different time points, which helps to identify individual feature patterns; through negative sample pairs, the model can distinguish the feature differences between different parturients; through shared encoders, the model can extract common features of multimodal data; through projection heads, the model can compress high-dimensional feature vectors into low-dimensional space, reducing computational complexity and improving model efficiency.

[0091] S4, based on the assessment results, generating a personalized intervention plan includes the following steps:

[0092] The calculated childbirth difficulty index and emotional stability index are comprehensively scored to evaluate whether they are suitable for generating an intervention plan. The expression is: ;in, It is a comprehensive score of the childbirth difficulty index and emotional stability index. For time The difficulty index of childbirth at For time Emotional stability index, is the attenuation coefficient, is the time variable, The time span for the assessment;

[0093] A threshold is set based on the comprehensive score. When the comprehensive score is greater than the threshold, it is considered that a personalized intervention plan needs to be generated. The current health status, historical records, delivery difficulty index and emotional stability index of the parturient are defined as the state space, and various possible intervention measures are defined as the action space. The genomics analysis module is called to obtain the genetic information of the parturient. The reward function is designed based on the current state and genetic information of the parturient. When an intervention measure can significantly improve the emotional stability index of the parturient and is consistent with her genetic characteristics, a higher positive reward will be obtained. The multi-layer perceptron in the deep Q network is used to input the state space and genetic information to obtain the value of each intervention measure. The comprehensive score of the delivery difficulty index and the emotional stability index is input into the trained model to generate a personalized intervention plan, which is expressed as: ;in, is the intensity of the personalized intervention plan, with a value range of [0, 1]. A value close to 1 indicates that the intervention plan is strong and needs more attention. is the number of intervention measures, Index of intervention measures, Generate sensitivity to changes in composite scores for adjusting intervention options, >0, For the The threshold of an intervention measure indicates at what level of the composite score the intervention measure becomes necessary. For the The weight of an intervention measure indicates the relative importance of the intervention measure among all intervention measures. is the activation function.

[0094] Furthermore, when 0≤ When the risk is less than 0.3, low-intensity intervention is used, including providing basic prenatal education, regular prenatal examinations, psychological support and consultation. It is suitable for pregnant women with relatively low delivery difficulty and stable emotions. Basic intervention measures are used to ensure the health and safety of pregnant women. ... When the prevalence is less than 0.7, moderate-intensity intervention is used to provide more frequent prenatal examinations, increase the frequency of psychological support and consultation, and provide nutritional guidance and exercise advice. This is suitable for women with moderate delivery difficulties or slight mood swings. Through moderate-intensity intervention measures, the health status and emotional stability of women can be further improved. When 0.7≤ When the BP / mdl ratio is ≤1, close medical monitoring should be provided, the frequency of prenatal examinations should be increased, professional psychological treatment should be provided, hospitalization for observation should be arranged, and detailed nutrition and exercise plans should be provided. This is suitable for women who have a difficult delivery or are emotionally unstable. Through high-intensity intervention measures, the safety of the mother and fetus can be ensured and the risk of delivery can be reduced.

[0095] Furthermore, by setting thresholds, the model can automatically determine whether an intervention plan needs to be generated, thereby improving the automation and accuracy of decision-making; by defining a detailed state space, the model can comprehensively consider the various characteristics of the mother and improve the accuracy of the assessment; by defining a rich action space, the model can provide a variety of intervention measures to meet the needs of different mothers; by designing a reasonable reward function, the model can optimize according to the effects of the intervention measures to improve the effectiveness of the intervention; by calculating the intensity of the intervention plan, the model can provide personalized intervention measures based on the specific situation of the mother, thereby improving the effectiveness and pertinence of the intervention; by recommending intervention measures of different intensities in segments, the model can provide appropriate intervention plans based on the needs of different mothers and improve their health and safety.

[0096] S5, collecting real-time data and feedback information from parturients, monitoring intervention effects, and evaluating the effectiveness of intervention programs include the following steps:

[0097] Feedback from mothers was collected through mobile applications. The effectiveness of the intervention program was evaluated by combining real-time data and feedback information, expressed as: ;in, is the predicted value of the effectiveness of the intervention plan, with a value range of [0, 1]. A value close to 1 indicates that the predicted intervention plan is very effective. is the number of effectiveness indicators, indicating how many different effectiveness indicators are considered to evaluate the intervention program. It is the sensitivity parameter of effectiveness prediction, which is used to adjust the response speed of the prediction model to changes in the intensity of the intervention program. >0, For the The verification threshold of the effectiveness indicator indicates the intensity of the intervention program at which the effectiveness indicator can be met. It is The weight of the effectiveness indicator indicates the relative importance of the effectiveness indicator in the entire effectiveness evaluation. The index of effectiveness indicator.

[0098] Furthermore, the types of data collected include but are not limited to physiological data, emotional data, behavioral data and feedback information; physiological data include heart rate, blood pressure, blood oxygen saturation, etc., emotional data include emotional index, anxiety level, sleep quality, etc., behavioral data include exercise volume, diet records, prenatal examination records, etc., and feedback information includes the satisfaction, feelings, and suggestions of the mother on the intervention measures. By collecting data in real time, the health status and intervention effects of the mother can be monitored in a timely manner, and the timeliness and effectiveness of the intervention can be improved; by collecting feedback information from the mother, the mother's real feelings and needs can be understood, and the personalization and satisfaction of the intervention plan can be improved.

[0099] Common effectiveness indicators include physiological indicators, emotional indicators, behavioral indicators and satisfaction indicators.

[0100] Physiological indicators include changes in heart rate, blood pressure, blood oxygen saturation, etc.; emotional indicators include changes in emotional index, anxiety level, sleep quality, etc.; behavioral indicators include changes in exercise volume, eating habits, prenatal check-up frequency, etc.; satisfaction indicators include the mother's satisfaction, feelings, and suggestions on the intervention measures.

[0101] This embodiment also provides a maternal prenatal status intelligent analysis system, such as Figure 2 As shown, the system includes:

[0102] The collection and preprocessing module collects the multimodal data of the parturient in real time and performs preprocessing; the feature fusion module fuses the preprocessed multimodal data to generate a multimodal feature vector; the evaluation module builds a deep learning model, performs prenatal status assessment, and obtains the evaluation results; the intervention plan generation module generates a personalized intervention plan based on the evaluation results; the effect monitoring and optimization module collects the real-time data and feedback information of the parturient, monitors the intervention effect, and evaluates the effectiveness of the intervention plan.

[0103] This embodiment also provides a computer device, which is suitable for the case of an intelligent analysis method for the prenatal status of a parturient, and includes: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the intelligent analysis method for the prenatal status of a parturient as proposed in the above embodiment.

[0104] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0105] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for intelligent analysis of the prenatal status of a parturient is implemented as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0106] In summary, the present invention realizes comprehensive monitoring of the maternal status by real-time collection of the maternal multimodal data, can reflect the maternal health status from different angles, and improves the comprehensiveness and accuracy of the evaluation; by fusing the preprocessed multimodal data to generate a multimodal feature vector, it realizes the comprehensive processing of multiple data sources; by constructing a deep learning model, a prenatal status assessment is performed, and a comprehensive and accurate assessment result is obtained, capturing the complex changes in the maternal status; by generating a personalized intervention plan based on the evaluation results, precise intervention for the maternal is achieved, the pertinence and effectiveness of the intervention are improved, and the safety and health of mother and baby are guaranteed.

[0107] Example 2, referring to Table 1, is the second example of the present invention. To further verify the technical solution of the present invention, experimental simulation data of the intelligent analysis method of the prenatal status of parturients are provided.

[0108] Collect maternal imaging data (ultrasound images), physiological data (heart rate, blood pressure, blood oxygen saturation) and psychological data (emotion index) in real time. Perform noise removal, edge detection, data cleaning and quantification. Extract imaging feature vectors, physiological feature vectors and psychological feature vectors from the preprocessed data and fuse them into multimodal feature vectors. Design a deep learning model combining convolutional neural network (CNN) and long short-term memory network (LSTM) for prenatal status assessment. Use composite loss function (cross entropy loss function and mean square error loss function) to train the model. Generate personalized intervention plan based on the evaluation results. Collect maternal real-time data and feedback information to monitor the intervention effect.

[0109] The details are shown in Table 1 below:

[0110] Table 1 Data table of prenatal intelligent analysis experiment for pregnant women

[0111] Name Heart rate (BPM) Blood pressure (mmHg) Blood oxygen saturation (%) Sentiment Index Delivery difficulty index Mood stability index Intervention effectiveness Experimental group: maternal 1 75 120 / 80 98 0.8 0.65 0.85 0.78 Experimental group: Maternal 2 80 130 / 85 97 0.7 0.60 0.90 0.82 Experimental group: maternal 3 70 115 / 75 96 0.9 0.70 0.75 0.74 Control group: maternal 1 72 118 / 78 95 0.6 0.75 0.80 0.65 Control group: maternal 2 78 132 / 86 94 0.8 0.68 0.82 0.68 Control group: 3 68 114 / 74 93 0.5 0.72 0.78 0.62

[0112] The average delivery difficulty index of the experimental group was 0.65, and that of the control group was 0.72, indicating that delivery difficulty was lower in the experimental group.

[0113] The average emotional stability index of the experimental group was 0.83, and that of the control group was 0.80, indicating that the emotions of the mothers in the experimental group were more stable.

[0114] The average effectiveness of the intervention program in the experimental group was 0.78, and that in the control group was 0.65, indicating that the intervention program in the experimental group was more effective.

[0115] In summary, the intelligent analysis method for maternal prenatal status described in the present invention has significant innovations and advantages in data collection and preprocessing, multimodal data fusion, deep learning model evaluation, generation of personalized intervention plans, and evaluation of intervention effects. Through these innovative technologies and methods, comprehensive, accurate, and personalized intelligent analysis of maternal prenatal status can be achieved, providing doctors and maternal scientific and reliable prenatal care plans.

[0116] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for intelligent analysis of maternal prenatal status, characterized in that: include: Collect multimodal data of parturients in real time and perform preprocessing; The preprocessed multimodal data is fused to generate a multimodal feature vector; Build a deep learning model to conduct prenatal status assessment and obtain the assessment results; Generate personalized intervention plans based on the assessment results; Collect real-time data and feedback from mothers, monitor intervention effects, and evaluate the effectiveness of intervention plans; The multimodal data includes imaging data, physiological data and psychological data; the physiological data refers to the heart rate, blood pressure and blood oxygen saturation of the parturient; the psychological data refers to the emotional index of the parturient; the preprocessing includes noise removal, edge detection, data cleaning and quantification processing; The step of fusing the preprocessed multimodal data to generate a multimodal feature vector includes the following steps: Extracting feature maps from preprocessed ultrasonic images to form image feature vectors; Extracting time series features from normalized physiological data to form physiological feature vectors; Extract emotional change characteristics from the quantified emotional index to form a psychological feature vector; Consider each parturient as a node in the graph; According to the social support relationship between the mothers, the edges were established, and the weights of the edges were defined based on the frequency of interaction and the degree of emotional support; The image feature vector, physiological feature vector and psychological feature vector are input into the graph neural network as node features; The node features are aggregated through the graph convolution layer, and the features of neighboring nodes are considered to extract the influence of related pregnant women in the social network; Through the propagation mechanism of the graph neural network, the information of each node is fused to obtain a new feature representation; The feature representation processed by the graph neural network is concatenated with the feature vectors of each modality to generate the final multimodal feature vector; Dynamically adjust feature weights based on the specific circumstances of the mother; According to the dynamically adjusted feature weights, the multimodal data in the low-dimensional space is input into the model, and the prenatal assessment status results are output. The expression is: ;in, is the prenatal status assessment vector, is the image feature, For physiological characteristics, For psychological characteristics, is the weight of the image feature, is the weight of the physiological characteristics, is the weight of the psychological characteristics; According to the specific components in the evaluation vector, the childbirth difficulty index and emotional stability index are calculated, and the expressions are: ; ;in, The difficulty index of childbirth. is the emotional stability index, is the score for the difficulty of childbirth in the assessment results. For the emotional stability score in the evaluation results, The evaluation results are elements, The index of the element in the evaluation result.

2. The method for intelligent analysis of maternal prenatal status according to claim 1, characterized in that: Constructing a deep learning model to conduct prenatal status assessment and obtain the assessment results includes the following steps: Randomly extract data at multiple time points from each parturient's multimodal data as positive sample pairs; Randomly extract data from multimodal data of different parturients as negative sample pairs; Use contrastive loss function to train the model, shorten the distance between positive sample pairs and increase the distance between negative sample pairs. We design a shared encoder and add a projection head after the encoder to project the encoder's multimodal data output into a low-dimensional space.

3. The method for intelligent analysis of maternal prenatal status according to claim 2, characterized in that: Based on the assessment results, generating a personalized intervention plan includes the following steps: The calculated childbirth difficulty index and emotional stability index are comprehensively scored to evaluate whether they are suitable for generating an intervention plan. The expression is: ;in, It is a comprehensive score of the childbirth difficulty index and emotional stability index. For time The difficulty index of childbirth at For time Emotional stability index, is the attenuation coefficient, is the time variable, The time span for the assessment; A threshold is set based on the comprehensive score. When the comprehensive score is greater than the threshold, it is considered necessary to generate a personalized intervention plan. The current health status, historical records, delivery difficulty index, and emotional stability index of the parturient are defined as the state space, and various possible intervention measures are defined as the action space; Call the genomics analysis module to obtain the genetic information of the mother; Design a reward function based on the mother’s current state and genetic information; When an intervention measure can significantly improve the mood stability index of pregnant women and is consistent with their genetic characteristics, it will receive a higher positive reward; Using a multi-layer perceptron in a deep Q-network, the state space and genetic information are input to obtain the value of each intervention measure; The comprehensive scores of the labor difficulty index and the emotional stability index are input into the trained model to generate a personalized intervention plan, which is expressed as: ;in, The intensity of the personalized intervention program, is the number of intervention measures, Index of intervention measures, To generate sensitivity to changes in composite scores for adjusting interventions, For the Thresholds for intervention measures, For the The weight of the intervention measures, is the activation function.

4. The method for intelligent analysis of maternal prenatal status according to claim 3, characterized in that: Collecting real-time data and feedback from mothers, monitoring intervention effects, and evaluating the effectiveness of intervention programs include the following steps: Gathering maternal feedback through a mobile app; Combining real-time data and feedback information, the effectiveness of the intervention plan is evaluated, and the expression is: ;in, is the predictive value of the effectiveness of the intervention program. is the number of effectiveness indicators, is the sensitivity parameter for effectiveness prediction, For the The validation threshold of the effectiveness indicator is It is The weight of the effectiveness indicator, The index of effectiveness indicator.

5. A maternal antenatal status intelligent analysis system, based on the maternal antenatal status intelligent analysis method according to any one of claims 1 to 4, characterized in that: include: The acquisition and preprocessing module collects multimodal data of parturients in real time and performs preprocessing; The feature fusion module fuses the preprocessed multimodal data to generate a multimodal feature vector; The evaluation module builds a deep learning model to conduct prenatal status assessment and obtain the evaluation results; The intervention plan generation module generates personalized intervention plans based on the evaluation results; The effect monitoring and optimization module collects real-time data and feedback information from parturients, monitors the effect of intervention, and evaluates the effectiveness of intervention plans; The multimodal data includes imaging data, physiological data and psychological data; the physiological data refers to the heart rate, blood pressure and blood oxygen saturation of the parturient; the psychological data refers to the emotional index of the parturient; the preprocessing includes noise removal, edge detection, data cleaning and quantification processing; Wherein, the feature fusion module is specifically used for: Extracting feature maps from preprocessed ultrasonic images to form image feature vectors; Extracting time series features from normalized physiological data to form physiological feature vectors; Extract emotional change characteristics from the quantified emotional index to form a psychological feature vector; Consider each parturient as a node in the graph; According to the social support relationship between the mothers, the edges were established, and the weights of the edges were defined based on the frequency of interaction and the degree of emotional support; The image feature vector, physiological feature vector and psychological feature vector are input into the graph neural network as node features; The node features are aggregated through the graph convolution layer, and the features of neighboring nodes are considered to extract the influence of related pregnant women in the social network; Through the propagation mechanism of the graph neural network, the information of each node is fused to obtain a new feature representation; The feature representation processed by the graph neural network is concatenated with the feature vectors of each modality to generate the final multimodal feature vector.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method for intelligent analysis of the prenatal status of a parturient described in any one of claims 1 to 4 are implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the intelligent analysis method for the prenatal status of a parturient described in any one of claims 1 to 4 are implemented.

Images