Intelligent follow-up visit assistant system for community health care of pregnant and lying-in women and method thereof

Abstract

The invention relates to the technical field of medical information, in particular to an intelligent follow-up visit assistant system for community health care of pregnant and lying-in women and a method thereof. The system comprises a follow-up visit strategy module, a health guidance module which is in communication connection with the follow-up visit strategy module, a medical resource cooperation module which is in communication connection with the health guidance module, and a decision support module which is in communication connection with the medical resource cooperation module and analyzes historical data and current health conditions. Intelligent decision suggestions are provided for community doctors; the system comprises a decision support module, a dynamic monitoring module in communication connection with the decision support module, a data management module in communication connection with the dynamic monitoring module, and a man-machine interaction module in communication connection with the data management module, so that the efficiency and quality of pregnant and lying-in woman community health care service are improved, and a new solution is provided for optimal configuration of medical resources.

Description

Technical Field

[0001] The present invention relates to the field of medical information technology, in particular to a pregnant and lying-in woman community health care intelligent follow-up assistant system and method thereof. Background Art

[0002] In recent years, with the continuous improvement of my country's medical and health system, maternal health care services have gradually become an important part of community medical care. However, the traditional maternal community health care model still has many shortcomings and cannot meet the growing health management needs.

[0006] There is an urgent need for an intelligent maternal community health care follow-up system that can comprehensively utilize advanced technologies such as artificial intelligence, big data analysis, and the Internet of Things. The system should be able to achieve comprehensive data collection, accurate risk prediction, personalized follow-up strategy formulation, intelligent medical resource allocation, and safe and reliable data management. At the same time, the system should also have the ability to self-learn and optimize, and be able to continuously improve service quality. Summary of the invention

[0007] The intelligent follow-up assistant system for maternal and child community health care of the present invention is designed to solve the above technical problems. The system realizes intelligent management of the whole process from data collection to decision support through the collaborative work of multiple modules, effectively improving the quality and efficiency of maternal and child community health care services.

[0008] The present invention proposes an intelligent follow-up assistant system for community health care of pregnant women, comprising:

[0009] Data acquisition module for:

[0010] Collect basic information, medical information and daily health care information of pregnant women;

[0011] Obtain postpartum physical data of pregnant women by connecting with the delivery hospital system;

[0012] A data processing module is connected to the data acquisition module for:

[0013] Receiving the maternal and child health data sent by the data acquisition module;

[0014] Based on the maternal health data, perform natural language processing and data analysis;

[0015] A risk prediction module is connected to the data processing module for:

[0016] Based on the analysis results of the data processing module, a risk prediction model with multiple input modes is used to perform health risk assessment;

[0017] Generate maternal health risk prediction reports;

[0018] A follow-up strategy module is connected in communication with the risk prediction module and is used to:

[0019] Develop personalized follow-up strategies based on the health risk prediction report;

[0020] Use reinforcement learning algorithms to dynamically optimize follow-up plans;

[0021] The health guidance module is connected to the follow-up strategy module for:

[0022] Generating targeted health advice and health care plans based on the personalized follow-up strategy;

[0023] Push health guidance information to pregnant women through mobile terminals;

[0024] The medical resource collaboration module is connected to the health guidance module for:

[0025] Integrate community medical, family doctor and hospital resources;

[0026] Intelligently allocate medical resources based on the health risk prediction report and personalized follow-up strategy;

[0027] The decision support module is communicatively connected with the medical resource collaboration module and is used to:

[0028] Analyze historical data and current health status;

[0029] Provide intelligent decision-making suggestions for community doctors;

[0030] A dynamic monitoring module is in communication with the decision support module and is used to:

[0031] Real-time monitoring of the health status of pregnant women;

[0032] Triggering a multi-level early warning mechanism based on preset thresholds;

[0033] The data management module is in communication with the dynamic monitoring module and is used to:

[0034] All data in the system is stored in a hierarchical manner and backed up in the cloud;

[0035] Ensure data security and accessibility;

[0036] The human-computer interaction module is connected to the data management module for:

[0037] Provide doctor operation interface and maternal user interface;

[0038] Realize the combination of artificial intelligence-assisted decision-making and doctors' experience and judgment.

[0039] The intelligent follow-up method based on the intelligent follow-up assistant system for pregnant women and mothers in community health care includes the following steps:

[0040] S1, collection of maternal health data:

[0041] The data collection module collects the basic information, medical information, daily health information and postpartum physical data of pregnant women;

[0042] S2, data processing and risk prediction:

[0043] The data processing module performs natural language processing and data analysis on the collected data, and the risk prediction module performs health risk assessment based on the processing results;

[0044] S3, develop a personalized follow-up strategy:

[0045] The follow-up strategy module uses reinforcement learning algorithms to develop and optimize personalized follow-up strategies based on risk prediction results;

[0046] S4, generate health guidance plan:

[0047] The health guidance module generates targeted health advice and health care plans based on personalized follow-up strategies;

[0048] S5, coordinated deployment of medical resources:

[0049] The medical resource coordination module intelligently allocates medical resources based on the needs and health status of pregnant women;

[0050] S6, provide decision support:

[0051] The decision support module analyzes historical data and current health conditions to provide community doctors with intelligent decision-making suggestions;

[0052] S7, implement dynamic health monitoring:

[0053] The dynamic monitoring module monitors the health status of pregnant women in real time and triggers the early warning mechanism;

[0054] S8, Data Security Management:

[0055] The data management module performs hierarchical storage, encryption and backup of system data to ensure data security;

[0056] S9, human-machine collaborative decision-making:

[0057] Through the human-computer interaction module, combined with AI-assisted decision-making and doctors' experience and judgment, the final diagnosis and treatment decision is made;

[0058] S10, Evaluation and Optimization:

[0059] Regularly evaluate follow-up results and use reinforcement learning to continuously optimize follow-up strategies and system performance.

[0060] The beneficial effects of the present invention are mainly reflected in the following aspects:

[0061] First, in terms of data collection and processing, the system realizes the intelligent fusion of multi-source heterogeneous data. Through the collaborative work of the data collection module and the data processing module, the system can comprehensively collect the basic information, medical information, daily health information and postpartum physical data of pregnant women, and use advanced natural language processing technology for in-depth analysis. This all-round data collection and processing lays a solid foundation for subsequent risk prediction and decision support.

[0062] Secondly, in terms of risk prediction and follow-up strategy formulation, this system breaks through the limitations of traditional methods. The risk prediction module adopts a multi-input mode prediction model, combined with a machine learning algorithm, which can accurately identify potential health risks. The follow-up strategy module innovatively introduces a reinforcement learning algorithm to achieve dynamic optimization of the follow-up strategy, greatly improving the pertinence and effectiveness of follow-up.

[0063] In terms of medical resource coordination and decision support, this system realizes intelligent resource allocation and decision support. The medical resource coordination module realizes efficient allocation of medical resources through multi-objective optimization algorithm. The decision support module uses data mining, case reasoning and other technologies to provide community doctors with accurate decision-making suggestions, effectively improving the quality of diagnosis and treatment.

[0064] In terms of dynamic monitoring and early warning, the system has established a complete set of real-time monitoring and multi-level early warning mechanisms. Through the dynamic monitoring module, the system can timely capture changes in the health status of pregnant women and quickly trigger early warnings when abnormalities occur. This mechanism greatly reduces health risks and provides safer protection for pregnant women.

[0065] In terms of data security and privacy protection, this system uses advanced encryption technology and access control mechanisms. The data management module not only realizes the secure storage and transmission of sensitive data, but also establishes a fine-grained access control strategy to effectively protect the privacy information of pregnant women.

[0066] Finally, a major innovation of this system lies in its self-learning and optimization capabilities. By continuously collecting feedback data and applying reinforcement learning algorithms, the system can continuously optimize its follow-up strategies and decision-making models to achieve continuous improvement in service quality.

[0067] In summary, the intelligent follow-up assistant system for community health care for pregnant women of the present invention realizes the intelligence and personalization of multiple links such as data collection, risk prediction, follow-up management, resource allocation, and decision support through the collaborative work of multiple modules. The system not only improves the efficiency and quality of community health care services for pregnant women, but also provides a new solution for the optimal allocation of medical resources. At the same time, the self-learning ability of the system ensures the continuous improvement of service quality. These innovations not only solve many problems existing in the prior art, but also point out the direction for the future development of community health care services for pregnant women, and have important practical application value and far-reaching social significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is a logic block diagram of the entire system of the present invention.

[0069] Figure 2 It is a logic block diagram of the data acquisition module of the present invention.

[0070] Figure 3 It is a logic block diagram of the risk prediction module of the present invention.

[0071] Figure 4 It is a logic block diagram of the follow-up strategy module of the present invention.

[0072] Figure 5 It is a logic block diagram of the health guidance module of the present invention.

[0073] Figure 6 This is a logical block diagram of the medical resource collaboration module of the present invention.

[0074] Figure 7 It is a logic block diagram of the decision support module of the present invention.

[0075] Figure 8 It is a logic block diagram of the dynamic monitoring module of the present invention.

[0076] Fig. 9 It is a logic block diagram of the data management module of the present invention. DETAILED DESCRIPTION

[0077] See also Figure 1-9 The present invention provides a maternal community health care intelligent follow-up assistant system and method thereof, aiming to improve the maternal community health care service level through intelligent means. The specific implementation methods of the present invention will be described in detail below.

[0078] The intelligent follow-up assistant system for maternal and child community health care of the present invention comprises a data collection module 1, a data processing module 2, a risk prediction module 3, a follow-up strategy module 4, a health guidance module 5, a medical resource coordination module 6, a decision support module 7, a dynamic monitoring module 8, a data management module 9 and a human-computer interaction module 10. These modules form a closed-loop intelligent maternal and child health management system through close collaboration.

[0079] The data collection module 1 is used to collect the basic information, medical information and daily health care information of pregnant women, and obtain the postpartum physical data of pregnant women by connecting with the delivery hospital system. Preferably, the data collection module 1 of the present invention can collect information through multiple channels, such as questionnaires, mobile applications, wearable devices, etc. In a specific embodiment, the data collection module 1 can regularly push health questionnaires to pregnant women to collect information such as their daily diet, exercise and sleep conditions.

[0080] The data processing module 2 is connected to the data acquisition module 1 in communication, and is used to receive the maternal health data sent by the data acquisition module 1, and perform natural language processing and data analysis on the data. The data processing module 2 of the present invention adopts advanced natural language processing technology, which can extract valuable information from unstructured text data. For example, when processing the chief complaint information of pregnant women, the data processing module 2 can use named entity recognition (NER) technology to identify key information such as symptoms and drug names. Specifically, a NER model based on BiLSTM-CRF can be used, and its loss function can be expressed as:

[0081]

[0082] Among them, n is the sequence length, x is the input sequence, y i is the label of the ith position.

[0083] The risk prediction module 3 is in communication with the data processing module 2. Based on the analysis results of the data processing module 2, a risk prediction model with multiple input modes is used to perform health risk assessment and generate a maternal health risk prediction report. The risk prediction module 3 of the present invention integrates a variety of machine learning algorithms, including but not limited to logistic regression, random forest and deep neural network. In a preferred embodiment, the risk prediction module 3 adopts an integrated learning method, specifically using the XGBoost algorithm. The objective function of XGBoost can be expressed as:

[0084]

[0085] Among them, l is the loss function, Ω is the regularization term, and f kRepresents the kth tree model. By optimizing this objective function, the risk prediction module 3 can obtain a high-precision risk prediction model.

[0086] In one embodiment of the present invention, the data collection module 1 includes a basic information collection unit 11, a medical information collection unit 12, a health information collection unit 13, a postpartum data interface unit 14 and a data verification unit 15. These units work together to ensure the comprehensiveness and accuracy of the collected data.

[0087] The basic information collection unit 11 is used to collect the basic information of the pregnant woman, such as name, age, ID number, etc. In practical applications, the basic information collection unit 11 can automatically obtain the basic information of the pregnant woman by connecting with the HIS system of the hospital, reducing repeated input and improving efficiency.

[0088] The medical information collection unit 12 is used to collect medical information such as the medical history, medication records, and examination results of pregnant women. The medical information collection unit 12 of the present invention can not only collect structured medical data, but also process unstructured medical records. For example, for medical record text, a medical text classification algorithm can be used for automatic classification and information extraction. A possible text classification model is a multi-label classification model based on BERT, and its loss function can be expressed as:

[0089]

[0090] Where N is the number of samples, C is the number of categories, and y ij is the true label, x ij is the logit of the model output, and σ is the sigmoid function. In this way, the model can continuously adapt to the new data distribution and maintain the prediction performance.

[0091] The health information collection unit 13 is used to collect health information such as daily diet, exercise, and sleep of pregnant women. The health information collection unit 13 of the present invention can automatically collect daily health data of pregnant women through IoT devices such as smart bracelets and smart weight scales. For example, indicators such as the number of steps per day, sleep duration, and heart rate changes of pregnant women can be recorded. These data can help the system to more comprehensively evaluate the health status of pregnant women.

[0092] The postpartum data interface unit 14 is used to connect data with the delivery hospital system to obtain postpartum body data. The postpartum data interface unit 14 of the present invention adopts a standardized data exchange protocol, such as HL7FHIR, to ensure compatibility with different hospital systems. In this way, key data such as delivery information, newborn information, and postpartum recovery status of pregnant women can be seamlessly obtained.

[0093] The data verification unit 15 is used to verify the integrity and validity of the collected data. The data verification unit 15 of the present invention adopts a multi-layer verification mechanism, including format verification, range verification and logic verification. For example, for blood pressure data, it is necessary not only to verify whether it is within the normal range (such as systolic pressure 90-140mmHg, diastolic pressure 60-90mmHg), but also to check whether the systolic pressure is greater than the diastolic pressure. This strict data verification ensures the accuracy of subsequent analysis.

[0094] The risk prediction module 3 includes a data preprocessing unit 31, a multi-model fusion unit 32, a risk scoring unit 33, a prediction result interpretation unit 34 and a model updating unit 35. The collaborative work of these units makes the risk prediction more accurate and interpretable.

[0095] The data preprocessing unit 31 is used to clean, normalize and extract features from the input data. In one embodiment of the present invention, the data preprocessing unit 31 standardizes the numerical features and converts them into a standard normal distribution with a mean of 0 and a variance of 1. For categorical features, One-hot encoding or Embedding technology is used for processing. In terms of feature extraction, principal component analysis (PCA) or autoencoder techniques can be used to reduce feature dimensions and improve the generalization ability of the model.

[0096] The multi-model fusion unit 32 is used to integrate the deep learning model and the traditional machine learning model. The multi-model fusion unit 32 of the present invention adopts the Stacking ensemble learning method, takes the outputs of multiple base learners as new features, and trains a meta-learner. Specifically, logistic regression, random forest, X GBoost, etc. can be selected as base learners, and logistic regression is used as a meta-learner. This method can make full use of the advantages of different models and improve the overall prediction performance.

[0097] The risk scoring unit 33 is used to calculate the health risk score based on the multi-model fusion results. The risk scoring unit 33 of the present invention adopts a scoring mechanism based on the SHAP (SHapley Additive exPlanations) value. The SHAP value can explain the contribution of each feature to the model prediction, and its calculation formula is:

[0098]

[0099] Among them, F is the set of all features, S is the subset that does not contain feature i, and f x (S) represents the predicted value of the model when only the feature subset S is used. By calculating the SHAP value for each feature, a more fair and interpretable risk score can be obtained.

[0100] The prediction result interpretation unit 34 is used to generate an interpretable risk prediction report. The prediction result interpretation unit 34 of the present invention not only gives the risk prediction result, but also provides a detailed explanation. For example, for high-risk pregnant women, the system will list the main factors leading to high risk and the importance of each factor. Such interpretable results help doctors and pregnant women better understand the source of risk and take targeted preventive measures.

[0101] The model updating unit 35 is used to regularly update the prediction model according to new data and feedback results. The model updating unit 35 of the present invention adopts online learning technology, which can receive new data samples in real time and update the model incrementally. Specifically, the FTRL (Follow-the-Regularized-Leader) algorithm can be used for online learning, and its update rule is:

[0102]

[0103] Among them, w is the model parameter, l i is the loss function of the ith sample, λ1 and λ2 are the L1 and L2 regularization coefficients respectively. In this way, the model can continuously adapt to the new data distribution and maintain the prediction performance.

[0104] Through the collaborative work of the above modules and units, the intelligent follow-up assistant system for community health care of pregnant women of the present invention can comprehensively collect maternal health data, conduct in-depth data analysis and risk prediction, and provide accurate health management services for pregnant women. The modular design of the system and the application of advanced algorithms make the entire follow-up process more intelligent and personalized, effectively improving the quality and efficiency of community health care services. In one embodiment of the present invention, the follow-up strategy module 4 includes a strategy generation unit 41, a reinforcement learning unit 42, a personalized adjustment unit 43, an execution monitoring unit 44 and an effect evaluation unit 45. The collaborative work of these units ensures the scientificity and effectiveness of the follow-up strategy.

[0105] The strategy generation unit 41 is used to generate an initial follow-up strategy based on the risk prediction results. The strategy generation unit 41 of the present invention adopts a rule-based expert system, combined with medical guidelines and clinical experience, to formulate corresponding follow-up strategies for pregnant women with different risk levels. For example, for pregnant women identified as high-risk, the system may recommend a follow-up once a week, while for low-risk pregnant women, it may recommend a follow-up once a month. This initial strategy provides a good starting point for subsequent optimization.

[0106] The reinforcement learning unit 42 is used to optimize the follow-up strategy through the Q-learning algorithm. The reinforcement learning unit 42 of the present invention models the follow-up process as a Markov decision process, where the state is the health status of the pregnant woman, the action is the frequency and method of the follow-up, and the reward is the degree of improvement in the health status. The update formula of the Q-learning algorithm can be expressed as:

[0107]

[0108] Among them, Q(s t ,a t ) is the state s t Take action a t is the value estimate, α is the learning rate, r t is the immediate reward, and γ is the discount factor. By continuously learning and updating the Q value, the system can find the optimal follow-up strategy.

[0109] The personalized adjustment unit 43 is used to adjust the follow-up plan according to the pregnant woman's feedback and the doctor's advice. The personalized adjustment unit 43 of the present invention introduces an adaptive adjustment mechanism. For example, if a pregnant woman reports that the follow-up frequency is too high, the system will appropriately reduce the follow-up frequency while ensuring safety. At the same time, the system will also consider the doctor's professional advice and integrate the doctor's opinions into the follow-up plan through weighted fusion.

[0110] The execution monitoring unit 44 is used to track the execution of the follow-up plan. The execution monitoring unit 44 of the present invention uses multi-dimensional monitoring indicators, including follow-up completion rate, timeliness, effectiveness, etc. The system automatically records the details of each follow-up, including the time, method, content, etc. of the follow-up. When deviations are found in the execution of the follow-up plan, the system will issue a reminder in time to ensure the smooth implementation of the follow-up plan.

[0111] The effect evaluation unit 45 is used to evaluate the implementation effect of the follow-up strategy and provide optimization suggestions. The effect evaluation unit 45 of the present invention adopts a multi-index comprehensive evaluation method, including the degree of improvement of health indicators, maternal satisfaction, and medical resource utilization efficiency. The evaluation results are presented in a visual way, which is convenient for medical staff and managers to intuitively understand the follow-up effect. Based on the evaluation results, the system will automatically generate optimization suggestions to provide a basis for the continuous improvement of the follow-up strategy.

[0112] In a preferred embodiment of the present invention, the health guidance module 5 includes a knowledge base management unit 51, a guidance program generation unit 52, a content push unit 53, an interactive consultation unit 54 and an effect tracking unit 55. The coordinated work of these units ensures the pertinence and effectiveness of health guidance.

[0113] The knowledge base management unit 51 is used to maintain the maternal and infant health knowledge base. The knowledge base management unit 51 of the present invention adopts the semantic web technology to organize and store the maternal and infant health knowledge in the form of ontology. The construction of the knowledge graph uses a relationship extraction method based on deep learning, and its loss function can be expressed as:

[0114] L=L cls +λL reg ,

[0115] Among them, L cls is the cross entropy loss for relation classification, L reg is the regularization term and λ is the balancing factor. In this way, the system can better understand and utilize complex medical knowledge.

[0116] The guidance program generation unit 52 is used to generate a personalized guidance program based on the individual health status and risk prediction results. The guidance program generation unit 52 of the present invention adopts case-based reasoning (CBR) technology. The system retrieves the case that is most similar to the current maternal situation from the historical case library, and then performs weighted fusion based on the similarity to generate a preliminary guidance program. The similarity calculation adopts cosine similarity:

[0117]

[0118] Among them, A and B represent the feature vectors of the current case and the historical case respectively.

[0119] The content push unit 53 is used to push health guidance information to pregnant women through multiple channels. The content push unit 53 of the present invention adopts an intelligent recommendation algorithm to select the most appropriate content and push time according to the reading habits, interest preferences and current health status of pregnant women. The recommendation system uses a collaborative filtering algorithm based on deep learning, and its objective function can be expressed as:

[0120]

[0121] Among them, U and V are the latent vector matrices of users and items respectively, r i j is the rating of user i on item j, Ω is the observed rating set, and λ is the regularization coefficient.

[0122] The interactive consultation unit 54 is used to provide real-time online consultation services. The interactive consultation unit 54 of the present invention integrates an intelligent question-and-answer system and manual customer service. The intelligent question-and-answer system is fine-tuned based on the BERT model and can quickly and accurately answer common questions of pregnant women. For complex questions, the system will automatically transfer to manual customer service to ensure the quality and efficiency of the consultation service.

[0123] The effect tracking unit 55 is used to track the execution and effect of health guidance. The effect tracking unit 55 of the present invention adopts multi-source data fusion technology to comprehensively analyze data from multiple channels such as wearable devices, APP usage records, questionnaire feedback, etc. The system uses time series analysis methods, such as ARIMA model, to evaluate the changing trend of health indicators. The ARIMA model can be expressed as:

[0124] φ(B)(1-B) d X t =θ(B)∈ t ,

[0125] Where B is the lag operator, d is the difference order, φ(B) and θ(B) are the autoregressive and moving average polynomials, respectively.

[0126] Preferably, the medical resource collaboration module 6 of the present invention includes a resource information collection unit 61, a demand analysis unit 62, a resource matching unit 63, a reservation management unit 64 and a referral management unit 65. The collaborative work of these units realizes efficient utilization and accurate allocation of medical resources.

[0127] The resource information collection unit 61 is used to collect and update the resource information of medical institutions at all levels. The resource information collection unit 61 of the present invention adopts distributed crawler technology to obtain the resource information of equipment, personnel, beds, etc. of each medical institution in real time. The system uses the Scrapy framework to build a crawler and uses the Bloom filter to remove URL duplication, thereby improving the efficiency and accuracy of information collection.

[0128] The demand analysis unit 62 is used to analyze medical needs according to the health status of pregnant women. The demand analysis unit 62 of the present invention adopts a demand prediction model based on machine learning. The model uses a random forest algorithm and takes into account multiple factors, such as gestational age, past medical history, current symptoms, etc. The Gini impurity calculation formula of the random forest is:

[0129]

[0130] Where D is the data set, p i is the proportion of samples in the i-th category. By minimizing the Gini impurity, the system can predict medical needs more accurately.

[0131] The resource matching unit 63 is used to perform intelligent matching based on demand and resource information. The resource matching unit 63 of the present invention adopts a multi-objective optimization algorithm, taking into account multiple factors such as the urgency of medical needs, the availability of resources, and the geographical location. The system uses a genetic algorithm to solve the multi-objective optimization problem, and its fitness function can be expressed as:

[0132] F=w1f1+w2f2+...+w nf n ,

[0133] Among them, f i is the i-th objective function, w i is the corresponding weight. By adjusting the weight, the system can flexibly balance the relationship between different goals.

[0134] The reservation management unit 64 is used to coordinate the reservation and use of medical resources. The reservation management unit 64 of the present invention realizes intelligent scheduling and dynamic scheduling functions. The system uses heuristic algorithms to solve resource scheduling problems, such as simulated annealing algorithms. The temperature drop function can be expressed as:

[0135] T(k)=T0α k ,

[0136] Among them, T0 is the initial temperature, α is the cooling coefficient, and k is the number of iterations. By gradually lowering the "temperature", the system can find a near-optimal reservation solution within a reasonable time.

[0137] The referral management unit 65 is used to assist in completing referrals between medical institutions of different levels when necessary. The referral management unit 65 of the present invention establishes an intelligent referral decision support system. The system uses a decision tree algorithm to assist doctors in making referral decisions. The decision tree is constructed using the ID3 algorithm, and its information gain calculation formula is:

[0138]

[0139] Where D is the data set, a is the feature, and Ent(D) is the information entropy of the data set D. By maximizing the information gain, the system can construct the optimal referral decision tree

[0140] Through the close collaboration of the above modules and units, the intelligent follow-up assistant system for maternal and child community health care of the present invention realizes the intelligent management of the whole process from follow-up strategy formulation, health guidance to medical resource coordination. The design of the system fully considers the special needs of maternal and child health care, and provides accurate, efficient and personalized services through advanced algorithms and technical means, which significantly improves the quality and efficiency of community health care.

[0141] In a preferred embodiment of the present invention, the decision support module 7 includes a data mining unit 71, a case reasoning unit 72, a rule reasoning unit 73, a decision tree generation unit 74 and a suggestion generation unit 75. The collaborative work of these units provides powerful decision support for community doctors and effectively improves the quality of diagnosis and treatment.

[0142] The data mining unit 71 is used to extract valuable decision support information from historical data. The data mining unit 71 of the present invention uses a variety of data mining techniques, including association rule mining, cluster analysis, and anomaly detection. For example, in association rule mining, the system uses the Apriori algorithm to discover the association between maternal symptoms and potential risks. The calculation formulas for support and confidence are as follows:

[0143]

[0144] Among them, A and B represent symptoms and risks respectively. By setting appropriate support and confidence thresholds, the system can discover meaningful association rules and provide reference for doctors' decision-making.

[0145] The case reasoning unit 72 is used to provide decision references based on similar cases. The case reasoning unit 72 of the present invention implements a case retrieval system based on K nearest neighbor (KNN). The system first uses the TF-ID F algorithm to extract the feature vector of the case, and then uses the cosine similarity to calculate the similarity between the cases. The selection of K nearest neighbors adopts an adaptive K value method, and the formula for determining the K value is:

[0146]

[0147] Where N is the total number of cases in the case library. This method can automatically adjust the K value according to the size of the case library and improve the accuracy of case reasoning.

[0148] The rule reasoning unit 73 is used to make inferences based on medical rules and guidelines. The rule reasoning unit 73 of the present invention adopts a forward chain reasoning engine based on the Rete algorithm. The system converts medical guidelines and clinical pathways into if-then rules and stores them in the rule base. The Rete algorithm greatly improves the speed of rule matching by constructing an efficient pattern matching network. Preferably, the system also introduces fuzzy rule reasoning to deal with the uncertainty in medical knowledge. The confidence calculation formula of the fuzzy rule is:

[0149] CF(H,E)=CF(H)*CF(E)*CF(Rule),

[0150] Among them, CF(H) is the confidence of the hypothesis, CF(E) is the confidence of the evidence, and CF(Rule) is the confidence of the rule. The decision tree generation unit 74 is used to construct a visual decision tree to assist doctors in making decisions. The decision tree generation unit 74 of the present invention uses the C4.5 algorithm to construct a decision tree. Compared with the D3 algorithm, the C4.5 algorithm introduces the concept of information gain rate, which effectively solves the preference problem for multi-valued attributes. The calculation formula of the information gain rate is:

[0151]

[0152] Among them, SplitInfo(D,a) is the split information of attribute a, which is used to penalize attributes with more values. By maximizing the information gain rate, the system can build a more reasonable and efficient decision tree.

[0153] The suggestion generation unit 75 is used to generate the final decision suggestion by integrating the results of each unit. The suggestion generation unit 75 of the present invention adopts a multi-source information fusion method based on the Dempster-Shafer evidence theory. This method can effectively handle uncertain and incomplete information from different units. The calculation formula of information fusion is:

[0154]

[0155] Among them, m(A) is the basic probability distribution of A after fusion, m1 and m2 are the basic probability distributions of the two evidence sources, and K is the conflict factor. In this way, the system can generate more comprehensive and reliable decision recommendations.

[0156] In another embodiment of the present invention, the dynamic monitoring module 8 includes a real-time data acquisition unit 81, a data analysis unit 82, an early warning rule setting unit 83, an early warning triggering unit 84 and an early warning notification unit 85. The coordinated work of these units realizes real-time monitoring and timely early warning of the health status of pregnant women.

[0157] The real-time data acquisition unit 81 is used to collect real-time health data through wearable devices and other channels. The real-time data acquisition unit 81 of the present invention supports a variety of wearable devices and smart home devices, such as smart bracelets, smart weight scales, blood pressure monitors, etc. The system uses the MQTT (Message Queuing Telemetry Transport) protocol for data transmission, ensuring the real-time and reliability of data transmission. Preferably, the system also implements edge computing functions, which can perform preliminary data processing and anomaly detection on the device side, reducing the amount of data transmission and improving the system response speed.

[0158] The data analysis unit 82 is used to analyze and detect anomalies in real-time data. The data analysis unit 82 of the present invention uses a time series anomaly detection algorithm based on a long short-term memory network (LSTM). In this way, the system can effectively capture the long-term dependencies of health data and accurately identify abnormal situations.

[0159] The warning rule setting unit 83 is used to set multi-level warning rules according to medical standards. The warning rule setting unit 83 of the present invention adopts a multi-level warning rule system based on fuzzy logic. The system defines multiple fuzzy sets, such as "normal", "mild abnormal", "moderate abnormal" and "severe abnormal". The membership function of the fuzzy set adopts a Gaussian function:

[0160]

[0161] Among them, c is the center of the fuzzy set, and σ controls the width of the function. By adjusting these parameters, the system can flexibly set early warning rules to adapt to the individual differences of different pregnant women.

[0162] The early warning trigger unit 84 is used to trigger an early warning of a corresponding level when an abnormality is detected. The early warning trigger unit 84 of the present invention implements an early warning trigger mechanism based on a state machine. The system defines multiple health states, such as "normal", "attention", "warning" and "danger". The conditions for state transition are determined by the early warning rules. The calculation formula of the state transition probability matrix P is:

[0163]

[0164] Among them, n ij is the number of transitions from state i to state j. By analyzing the state transition probability, the system can more accurately determine the changing trend of health status and reduce false positives and false negatives.

[0165] The early warning notification unit 85 is used to send early warning information to relevant personnel. The early warning notification unit 85 of the present invention adopts a multi-channel and intelligent notification strategy. The system will select the most appropriate notification method, such as SMS, App push, phone call, etc., according to the warning level and the characteristics of the recipient (such as role, time, location, etc.). The notification priority calculation formula is:

[0166] Priority=w1*Severity+w2*Urgency+w3*Trend,

[0167] Among them, Severity is the severity of the warning, Urgency is the urgency, Trend is the trend of change, and w1, w2, and w3 are the corresponding weights. In this way, the system can ensure that important warning information can be delivered to relevant personnel in a timely manner.

[0168] In another embodiment of the present invention, the data management module 9 includes a data classification unit 91, a storage strategy unit 92, an encryption processing unit 93, an access control unit 94 and a data backup unit 95. The coordinated work of these units ensures the security, integrity and availability of system data.

[0169] The data classification unit 91 is used to classify the sensitivity of the system data. The data classification unit 91 of the present invention adopts an automatic classification method based on machine learning. The system uses a support vector machine (SVM) algorithm to classify the data, and its decision function is:

[0170]

[0171] Among them, α i is the Lagrange multiplier, y i is the class label, K(x i ,x) is the kernel function. By selecting an appropriate kernel function (such as the RBF kernel), the system can effectively classify data with different sensitivities.

[0172] The storage strategy unit 92 is used to formulate storage strategies for data of different sensitivities. The storage strategy unit 92 of the present invention implements a multi-layer storage architecture, including multiple storage layers such as memory database, solid-state hard disk and mechanical hard disk. The system uses the LRU-K algorithm to manage the migration of data between different storage layers. The core idea of ​​the LRU-K algorithm is to record the time of the most recent K accesses to the data, and give priority to eliminating data with fewer recent accesses. This method can effectively balance storage efficiency and access speed.

[0173] The encryption processing unit 93 is used to encrypt sensitive data. The encryption processing unit 93 of the present invention uses homomorphic encryption technology, allowing calculations to be performed directly on encrypted data without decryption. The system uses the Paillier encryption algorithm, and its encryption function is:

[0174] E(m)=g m ·r m mod n 2 ,

[0175] Among them, m is plaintext, g and r are random numbers, and n is the product of two large prime numbers. Through homomorphic encryption, the system can perform necessary data analysis while protecting data privacy.

[0176] The access control unit 94 is used to manage the access rights of data. The access control unit 94 of the present invention implements an attribute-based access control (ABAC) mechanism. The system defines a series of attributes (such as user role, time, location, etc.) and access policies. The calculation formula for access decision is:

[0177] Decision = f(Subject a Attributes,Object a Attributes,Action,Environment a ttributes),

[0178] Among them, f is a Boolean function that determines whether access is allowed based on various attributes. This fine-grained access control mechanism can better protect sensitive data. The data backup unit 95 is used to regularly back up data to cloud storage. The data backup unit 95 of the present invention adopts a strategy that combines incremental backup and differential backup. The system uses snapshot technology to achieve efficient incremental backup, and at the same time determines whether the data has changed by calculating the hash value of the data block. The backup frequency is determined by an adaptive algorithm, and the backup frequency is dynamically adjusted according to the data change rate. The formula for frequency adjustment is:

[0179] Frequency=Base F requency1+k*Change R ate),

[0180] Among them, Base_Frequency is the basic backup frequency, Change_Rate is the data change rate, and k is the adjustment coefficient. In this way, the system can minimize the impact of backup on system performance while ensuring data security.

[0181] Finally, the present invention also provides an intelligent follow-up method based on the above-mentioned intelligent follow-up assistant system for pregnant women and mothers in community health care. The method comprises the following steps:

[0182] S1, collecting maternal health data: collecting maternal basic information, medical information, daily health information and postpartum physical data through data collection module 1. In this step, the system will collect data through a variety of channels such as questionnaires, wearable devices, and hospital HIS systems to ensure the comprehensiveness and accuracy of the data.

[0183] S2, data processing and risk prediction: Data processing module 2 performs natural language processing and data analysis on the collected data, and risk prediction module 3 performs health risk assessment based on the processing results. In this step, the system will use the various machine learning algorithms described above, such as the BERT model for text processing and the XGB oost algorithm for risk prediction.

[0184] S3, formulate personalized follow-up strategy: Follow-up strategy module 4 uses reinforcement learning algorithm to formulate and optimize personalized follow-up strategy based on risk prediction results. In this step, the system will consider the individual differences of pregnant women and the professional advice of doctors to dynamically adjust the follow-up plan.

[0185] S4, Generate health guidance plan: The health guidance module 5 generates targeted health advice and health care plans based on the personalized follow-up strategy. In this step, the system will extract relevant information from the knowledge base based on the specific situation of the pregnant woman and generate a health guidance plan that is easy to understand and implement.

[0186] S5, coordinated allocation of medical resources: The medical resource coordination module 6 intelligently allocates medical resources according to the needs and health status of pregnant women. In this step, the system will comprehensively consider multiple factors such as resource availability, geographical location, and professional matching to achieve the optimal allocation of medical resources.

[0187] S6, provide decision support: Decision support module 7 analyzes historical data and current health status, and provides intelligent decision suggestions for community doctors in the preferred embodiment of the present invention. In this step, the system will comprehensively use multiple technologies such as data mining, case reasoning, rule reasoning, etc. to generate comprehensive and accurate decision suggestions. For example, the system may provide personalized treatment suggestions for doctors based on the treatment plans of similar cases and the specific conditions of the current pregnant women.

[0188] S7, implement dynamic health monitoring: The dynamic monitoring module 8 monitors the health status of pregnant women in real time and triggers the early warning mechanism. In this step, the system will continuously collect the physiological indicators of pregnant women, such as heart rate, blood pressure, blood sugar, etc., through wearable devices and other channels. When an indicator exceeds the preset threshold, the system will immediately trigger the corresponding level of early warning. For example, if it is detected that the blood pressure of a pregnant woman continues to rise, the system may issue a "high blood pressure risk" warning and recommend further examination.

[0189] S8, Data security management: The data management module 9 performs hierarchical storage, encryption and backup of system data to ensure data security. In this step, the system will take different levels of security measures according to the sensitivity of the data. For example, for highly sensitive personal medical information, the system may use homomorphic encryption technology for encrypted storage and implement strict access control.

[0190] S9, human-machine collaborative decision-making: Through the human-machine interaction module 10, the final diagnosis and treatment decision is made by combining AI-assisted decision-making and the doctor's experience and judgment. In this step, the system will present the decision suggestions generated by AI to the doctor in an intuitive way, and allow the doctor to adjust the suggestions based on his own experience and judgment. This human-machine collaborative approach can give full play to the efficiency of AI and the professionalism of doctors to achieve the best diagnosis and treatment results.

[0191] S10, Evaluation and Optimization: Regularly evaluate the follow-up effect, and use reinforcement learning to continuously optimize the follow-up strategy and system performance. In this step, the system will collect various data during the follow-up process, such as the improvement of maternal health indicators, the implementation of the follow-up plan, and the feedback from doctors and pregnant women. Based on this data, the system will use reinforcement learning algorithms to continuously optimize the follow-up strategy and improve the overall performance of the system.

[0192] The intelligent follow-up method of the present invention realizes the full process of intelligent maternal health management through the above steps. From data collection to decision support, to follow-up execution and effect evaluation, each link is integrated with advanced artificial intelligence technology. This method not only improves the efficiency and accuracy of follow-up, but also continuously optimizes itself according to the actual effect, providing continuously improved health management services for pregnant women.

[0193] Preferably, in step S2, the multi-input mode risk prediction model adopted by the risk prediction module 3 can be expressed as:

[0194] Risk=f(X1,X2,...,X n ),

[0195] Among them, X i Represents different types of input features, such as basic information, medical records, daily health data, etc. Function f can be a complex nonlinear function, which is approximated by a deep learning model. Specifically, a multi-layer perceptron (MLP) can be used to implement this function, and its forward propagation process can be expressed as:

[0196] h (l) =σ(W (l) h (l-1) +b (l) ),

[0197] Among them, h (l) is the output of layer l, W (l) and b (l) are the weight and bias of the layer respectively, and σ is the activation function, such as the ReLU function.

[0198] In step S3, the goal of the reinforcement learning algorithm is to maximize the cumulative reward, which can be expressed as:

[0199]

[0200] Among them, π θ is the parameterized policy function, γ is the discount factor, r t is the reward obtained at time step t. The system optimizes this objective function through the gradient ascent method to obtain the optimal follow-up strategy.

[0201] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. The intelligent follow-up assistant system for maternal and child community health care is characterized by: include: Data acquisition module for: Collect basic information, medical information and daily health care information of pregnant women; Obtain postpartum physical data of pregnant women by connecting with the delivery hospital system; A data processing module is connected to the data acquisition module for: Receiving maternal health data sent by the data acquisition module; Based on the maternal health data, perform natural language processing and data analysis; A risk prediction module is connected to the data processing module for: Based on the analysis results of the data processing module, a risk prediction model with multiple input modes is used to perform health risk assessment; Generate maternal health risk prediction reports; A follow-up strategy module is connected in communication with the risk prediction module and is used to: Develop personalized follow-up strategies based on the health risk prediction report; Use reinforcement learning algorithms to dynamically optimize follow-up plans; The health guidance module is connected to the follow-up strategy module for: Generating targeted health advice and health care plans based on the personalized follow-up strategy; Push health guidance information to pregnant women through mobile terminals; The medical resource collaboration module is connected to the health guidance module for: Integrate community medical, family doctor and hospital resources; Intelligently allocate medical resources based on the health risk prediction report and personalized follow-up strategy; The decision support module is communicatively connected with the medical resource collaboration module and is used to: Analyze historical data and current health status; Provide intelligent decision-making suggestions for community doctors; A dynamic monitoring module is connected in communication with the decision support module and is used to: Real-time monitoring of the health status of pregnant women; Triggering a multi-level early warning mechanism based on preset thresholds; The data management module is in communication with the dynamic monitoring module and is used to: All data in the system is stored in a hierarchical manner and backed up in the cloud; Ensure data security and accessibility; The human-computer interaction module is connected to the data management module for: Provide doctor operation interface and maternal user interface; Realize the combination of artificial intelligence-assisted decision-making and doctors' experience and judgment.

2. The intelligent follow-up assistant system for community health care of pregnant women according to claim 1 is characterized in that: The data acquisition module comprises: Basic information collection unit, used to collect basic information of pregnant women, such as name, age, ID number, etc.; Medical information collection unit, used to collect medical information such as the medical history, medication records, and examination results of pregnant women; Health information collection unit, used to collect health information such as daily diet, exercise, sleep, etc. of pregnant women; Postpartum data interface unit, used to connect with the delivery hospital system to obtain postpartum body data; The data verification unit is used to verify the integrity and validity of the collected data.

3. The intelligent follow-up assistant system for community health care of pregnant women according to claim 1 is characterized in that: The risk prediction module includes: Data preprocessing unit, used to clean, normalize and extract features from input data; Multi-model fusion unit for integrating deep learning models and traditional machine learning models; A risk scoring unit, used to calculate health risk scores based on multi-model fusion results; Prediction result interpretation unit, used to generate interpretable risk prediction reports; The model updating unit is used to regularly update the prediction model based on new data and feedback results.

4. The intelligent follow-up assistant system for community health care of pregnant women according to claim 1 is characterized in that: The follow-up strategy module includes: A strategy generation unit, used for generating an initial follow-up strategy based on the risk prediction result; A reinforcement learning unit for optimizing follow-up strategies using a Q-learning algorithm; A personalized adjustment unit to adjust the follow-up plan based on maternal feedback and physician recommendations; An execution monitoring unit is used to track the execution of the follow-up plan; The effectiveness evaluation unit is used to evaluate the implementation effect of the follow-up strategy and provide optimization suggestions.

5. The intelligent follow-up assistant system for community health care of pregnant women according to claim 1 is characterized in that: The health guidance module includes: A knowledge base management unit, used to maintain a maternal and infant health knowledge base; A guidance program generation unit, used to generate a personalized guidance program based on individual health status and risk prediction results; Content push unit, used to push health guidance information to pregnant women through multiple channels; Interactive consultation unit, used to provide real-time online consultation services; The effect tracking unit is used to track the implementation and effect of health guidance.

6. The intelligent follow-up assistant system for community health care of pregnant women according to claim 1 is characterized in that: The medical resource collaboration module includes: Resource information collection unit, used to collect and update resource information of medical institutions at all levels; The needs analysis unit is used to analyze medical needs based on the health status of pregnant women; Resource matching unit, used for intelligent matching based on demand and resource information; An appointment management unit, used to coordinate the appointment and use of medical resources; The referral management unit is used to assist in completing referrals between different levels of medical institutions when necessary.

7. The intelligent follow-up assistant system for community health care of pregnant women according to claim 1 is characterized in that: The decision support module includes: Data mining unit, used to extract valuable decision support information from historical data; Case-based reasoning unit, used to provide decision references based on similar cases; a rule reasoning unit for reasoning based on medical rules and guidelines; Decision tree generation unit, used to build a visual decision tree to assist doctors in making decisions; The suggestion generation unit is used to integrate the results of each unit to generate the final decision suggestion.

8. The intelligent follow-up assistant system for community health care of pregnant women according to claim 1 is characterized in that: The dynamic monitoring module comprises: A real-time data collection unit, used to collect real-time health data through wearable devices and other channels; Data analysis unit, used for analysis and anomaly detection of real-time data; An early warning rule setting unit, used to set multi-level early warning rules according to medical standards; An early warning trigger unit is used to trigger an early warning of a corresponding level when an abnormality is detected; The early warning notification unit is used to send early warning information to relevant personnel.

9. The intelligent follow-up assistant system for community health care of pregnant women according to claim 1, characterized in that: The data management module comprises: A data classification unit, used to classify the sensitivity of system data; Storage policy unit, used to formulate storage policies for data of different sensitivities; An encryption processing unit, used for encrypting sensitive data; Access control unit, used to manage access rights to data; A data backup unit is used to regularly back up data to cloud storage.

10. An intelligent follow-up method based on the intelligent follow-up assistant system for community health care of pregnant women according to any one of claims 1 to 9, characterized in that: The following steps are involved: S1, collection of maternal health data: The data collection module collects the basic information, medical information, daily health information and postpartum physical data of pregnant women; S2, data processing and risk prediction: The data processing module performs natural language processing and data analysis on the collected data, and the risk prediction module performs health risk assessment based on the processing results; S3, develop a personalized follow-up strategy: The follow-up strategy module uses reinforcement learning algorithms to develop and optimize personalized follow-up strategies based on risk prediction results; S4, generate health guidance plan: The health guidance module generates targeted health advice and health care plans based on personalized follow-up strategies; S5, coordinated deployment of medical resources: The medical resource coordination module intelligently allocates medical resources based on the needs and health status of pregnant women; S6, provide decision support: The decision support module analyzes historical data and current health conditions to provide community doctors with intelligent decision-making suggestions; S7, implement dynamic health monitoring: The dynamic monitoring module monitors the health status of pregnant women in real time and triggers the early warning mechanism; S8, Data Security Management: The data management module performs hierarchical storage, encryption and backup of system data to ensure data security; S9, human-machine collaborative decision-making: Through the human-computer interaction module, combined with AI-assisted decision-making and doctors' experience and judgment, the final diagnosis and treatment decision is made; S10, Evaluation and Optimization: Regularly evaluate follow-up results and use reinforcement learning to continuously optimize follow-up strategies and system performance.

Images