Mobile terminal-based postpartum hepatitis c referral management system

Abstract

The application discloses a postpartum hepatitis C referral management system based on a mobile terminal and relates to the medical management field.The system comprises a medical communication module, a data entry module and an index preset module.The medical communication module is used for integrating the communication permission of the mobile terminal and providing a plurality of communication channels.The data entry module is used for periodically collecting and entering the liver function indexes, the hepatitis C virus load data and the emotional evaluation scores of puerpera through the medical communication module.The index preset module is used for presetting the benchmark feeding indexes of future periods according to a newborn growth model.The risk prediction module is used for collecting the data of the current period data entry module and the preset feeding indexes of the next period based on the risk prediction model.The risk coefficient of breastfeeding behavior is dynamically predicted by using the risk prediction model, the potential risks under different feeding schemes are quantitatively evaluated, the high-risk puerpera are early identified, the disease deterioration or the risk of mother-to-child transmission caused by delayed intervention is reduced, the human error is reduced, and the accuracy and response speed of postpartum hepatitis C management are improved.

Description

Technical Field

[0001] This invention relates to the field of medical management technology, specifically a postpartum hepatitis C referral management system based on mobile terminals. Background Technology

[0002] Pregnant women infected with hepatitis C virus require long-term monitoring of liver function and viral load after delivery to assess disease progression and the risk of mother-to-child transmission. The development of mobile health technologies has made it possible to optimize postpartum hepatitis C management, but currently there is no dedicated intelligent referral and feeding decision support system for mothers with hepatitis C (HCV).

[0003] The current medical system has relatively weak follow-up management of postpartum HCV-infected patients, relying on in-person follow-up visits. Due to factors such as the burden of childcare and inconvenient transportation, mothers are easily lost to follow-up, resulting in the loss of key data such as viral load and liver function, which affects the judgment of treatment timing. At the same time, the safety assessment of breastfeeding relies on empirical judgment and does not take into account abnormal liver function and dynamic changes in feeding behavior, which may lead to misjudgment of risks. In addition, postpartum emotional fluctuations may affect treatment compliance and feeding method selection, but the current management model often neglects the comprehensive assessment of psychological factors. Summary of the Invention

[0004] (a) Technical problems to be solved

[0005] In view of the above-mentioned shortcomings of the existing technology, the present invention provides a postpartum hepatitis C referral management system based on mobile terminals, which can effectively solve the problems of the existing technology.

[0006] (II) Technical Solution

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] This invention discloses a postpartum hepatitis C referral management system based on a mobile terminal, comprising:

[0009] The medical communication module is used to integrate the communication permissions of the mobile terminal and provides several communication channels;

[0010] The data entry module is used to periodically collect and enter the mother's liver function indicators, hepatitis C virus load data, and emotional assessment scores through the medical communication module. Periodic collection is achieved through a combination of mobile application reminders, direct data connection with home portable testing devices via Bluetooth or Wi-Fi, or manual entry of test report images for OCR recognition.

[0011] The indicator preset module is used to preset the baseline feeding indicators for several future cycles based on the newborn growth model.

[0012] The risk prediction module is used to collect data from the current cycle data entry module and preset feeding indicators for the next cycle, and calculate the risk coefficient of breastfeeding behavior in the next cycle through a preset risk prediction model. The risk prediction model takes the mother's physiological indicators and emotional state as input variables and the estimated risk under specific feeding behaviors as output.

[0013] The anomaly detection module is used to compare the calculated risk coefficient with the preset standard risk threshold to determine whether there is any abnormal risk in the feeding plan for the next cycle.

[0014] The indicator adjustment module is used to assess and generate positive or negative adjustments to preset feeding indicators for several future cycles when an anomaly is detected, based on the degree to which the risk coefficient exceeds the threshold; the positive adjustment points to more conservative feeding recommendations, and the negative adjustment points to more aggressive feeding recommendations.

[0015] The indicator generation module is used to modify the preset baseline feeding indicators based on the adjustment range, and generate the final personalized feeding indicators for newborns in the future for several cycles.

[0016] The referral decision module is used to comprehensively assess the impact of the mother's hepatitis C status on feeding safety by combining the current risk coefficient prediction trend with the final feeding indicators, and to formulate a recommended medical referral intervention cycle based on the preset referral trigger rules.

[0017] Furthermore, the working logic of the indicator preset module is as follows:

[0018] Based on the newborn's gestational age, birth weight, current age and sex, a standard newborn growth and nutrition demand model is constructed, dividing the postpartum timeline into continuous cycles in weeks, and setting a baseline time point for each future cycle.

[0019] Based on the growth nutrition requirement model, the average weight and daily total energy requirement of newborns at this time point are obtained. Combined with the data collected from mothers in the data entry module, the theoretical safe upper limit of breast milk intake during this period is defined.

[0020] The theoretical safe intake limit is converted into a baseline feeding index that includes the recommended total number of breastfeedings per day, the recommended maximum duration of each breastfeeding session, and the recommended milk volume range per bottle feeding.

[0021] The calculated baseline feeding index sequence for all future periods is stored and output as an unadjusted initial feeding plan.

[0022] Furthermore, the steps for constructing the risk prediction model in the risk prediction module include:

[0023] Step 41: Training set construction: Collect liver function indicators, hepatitis C virus RNA load, standardized mood scores and corresponding feeding behavior records of several postpartum women during the postpartum follow-up period as training datasets, and clinical outcome datasets including whether there were suspected mother-to-child transmission events, acute deterioration of postpartum liver function events or severe mood disorder events as labels.

[0024] Step 42, Data Processing: Preprocess the training dataset, including handling missing values, normalization, and feature construction; the constructed features include: the rate of change of physiological indicators within the period, the interaction term between emotion scores and physiological indicators, and compliance quantification features based on feeding behavior records.

[0025] Step 43, Model Training: Divide the processed dataset into a training set and a validation set, train it using a neural network algorithm, evaluate the predictive performance of each model through the validation set, and select the model with the best overall performance in terms of discrimination and calibration as the initial risk prediction model.

[0026] Step 44, Data Correction: Time series analysis is introduced to correct the trend of the output of the initial risk prediction model, and Bayesian optimization method is used to fine-tune the model hyperparameters, ultimately forming an integrated and dynamic risk prediction model. The trained risk prediction model is deployed in the risk prediction module, and a regular update mechanism is established. New data is collected to perform incremental learning or periodic full retraining of the model to maintain the prediction accuracy and timeliness of the model.

[0027] Furthermore, the working logic of the indicator adjustment module includes:

[0028] The risk coefficient output by the anomaly detection module is mapped to discrete risk levels based on multiple preset risk threshold ranges, with each risk level corresponding to a qualitative description of the severity of the risk.

[0029] Based on the current risk level obtained from the mapping, a predefined adjustment strategy matrix is ​​queried. The matrix is ​​arranged with the risk level as the row and each of the future several cycles as the column. Each element in the matrix is ​​an adjustment instruction tuple, which includes: the adjustment direction and the basic adjustment magnitude for three dimensions: feeding method, maximum duration of a single feeding, and total number of feedings per day. The adjustment direction is determined by the risk level, and the basic adjustment magnitude is determined by the risk level and the distance between the target cycle and the current cycle. The closer the cycle is, the greater the adjustment magnitude.

[0030] The system obtains the mother's historical adjustment records, feeding behavior compliance score, and current emotional assessment score to generate a personalized adjustment factor. This adjustment factor is then weighted and calculated with the baseline adjustment range to obtain the final personalized adjustment range. If the historical compliance is poor or the emotional state is particularly vulnerable, the positive adjustment range is amplified and the negative adjustment range is reduced.

[0031] The calculated final personalized adjustment range for each feeding dimension over several future cycles is output to the indicator generation module for revising the baseline feeding indicators.

[0032] Furthermore, the adjustment strategy matrix defines several adjustment ranges and directions for feeding methods, maximum duration of a single feeding, and total number of feedings per day under different risk levels.

[0033] Furthermore, the formula for calculating the risk coefficient in the risk prediction module is as follows:

[0034] ;

[0035] In the formula, This represents the risk coefficient for predicted breastfeeding behavior in the next cycle, with a value range of [0,1]. A higher value indicates a higher risk. Represents the activation function. The total number of representative features Representing the The weight coefficients of each feature are learned during the model training process and reflect the importance of that feature in risk prediction. The representative feature function represents the feature extracted from the input data. The value of each feature, This represents a data vector of postpartum women collected during the current cycle, including but not limited to: liver function indicators, hepatitis C viral load, and standardized mood assessment scale scores. The vector representing the pre-defined newborn feeding indicators for the next cycle includes, but is not limited to: planned feeding method, planned duration of a single breastfeeding session, planned total number of breastfeeding sessions per day, and the percentage of planned breastfeeding sessions in total feeding. This represents the model bias term.

[0036] Furthermore, the referral triggering rule of the referral decision module is a multi-factor weighted decision model. The input factors include: the number of times the risk coefficient exceeds the threshold for M consecutive cycles, the upward slope of the risk coefficient, the frequency of continuous and significant negative adjustments to the feeding index, and the deterioration trend of the maternal emotional assessment score. The output is the urgency and time window for recommending referral.

[0037] Furthermore, the medical communication module is interconnected with a data management module via a wireless network. The data management module is used to encrypt and store all maternal data, feeding indicators, risk coefficient records, adjustment logs, and referral decision records, and provides a structured query interface.

[0038] Furthermore, the referral decision module is connected to a push module via a wireless network. The push module is used to push the final personalized feeding indicators, health guidance suggestions, and referral intervention reminders to the medical communication module, and to display the risk coefficient change curve, feeding indicator adjustment trajectory, and referral priority queue of the managed mothers.

[0039] Furthermore, the medical communication module is interconnected with the data entry module and the indicator preset module via a wireless network; the risk prediction module is interconnected with the data entry module, the indicator preset module, and the anomaly judgment module via a wireless network; the indicator adjustment module is interconnected with the anomaly judgment module and the indicator generation module via a wireless network; and the referral decision module is interconnected with the anomaly judgment module and the indicator generation module via a wireless network.

[0040] (III) Beneficial Effects

[0041] Compared with the known prior art, the technical solution provided by this invention has the following beneficial effects:

[0042] 1. The system collects the mother's liver function indicators, viral load, and emotional state in real time through a mobile terminal integrated with the data entry module. Combined with preset newborn feeding indicators, the system uses a risk prediction model to dynamically predict the risk coefficient of breastfeeding behavior, quantitatively assess the potential risks under different feeding plans, and automatically trigger warnings or adjustment suggestions based on risk thresholds. Through periodic data monitoring and model iteration optimization, the system can identify high-risk mothers at an early stage, reduce the risk of disease deterioration or mother-to-child transmission caused by delayed intervention, reduce human error, ensure the objectivity and consistency of medical decisions, and improve the accuracy and response speed of postpartum hepatitis C management.

[0043] 2. By dynamically adjusting newborn feeding indicators, the system provides personalized feeding advice to each mother, balancing breastfeeding needs with disease prevention. When the risk coefficient exceeds the threshold, the system automatically generates a conservative feeding plan, such as reducing the duration or frequency of breastfeeding to reduce the risk of virus exposure. Conversely, it gradually relaxes restrictions to support breastfeeding, which not only ensures the nutritional safety of newborns but also reduces the psychological stress on mothers. Referral decisions are based on multi-dimensional risk assessment and intelligent recommendation of referral priorities, helping medical institutions to allocate resources efficiently, prioritize high-risk cases, avoid over-referrals or missed diagnoses, and improve overall medical efficiency.

[0044] 3. Receive personalized feeding guidance, health reminders and psychological support through the medical communication module, reducing anxiety or erroneous behavior caused by information asymmetry. Medical staff can also use the management platform to have a global grasp of the patient's risk level, feeding adjustment records and referral needs, and quickly develop intervention plans. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0046] Figure 1 This is a schematic diagram of the framework of the present invention;

[0047] Figure 2 This is a flowchart illustrating the steps involved in constructing the risk prediction model in this invention.

[0048] The numbers in the diagram represent: 1. Medical Communication Module; 2. Data Entry Module; 3. Indicator Preset Module; 4. Risk Prediction Module; 5. Anomaly Detection Module; 6. Indicator Adjustment Module; 7. Indicator Generation Module; 8. Referral Decision Module; 9. Push Module; 10. Data Management Module. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0050] The present invention will be further described below with reference to embodiments.

[0051] This embodiment features a mobile terminal-based postpartum hepatitis C referral management system, such as... Figure 1 As shown, it includes:

[0052] The medical communication module 1 is used to integrate the communication permissions of the mobile terminal and provides several communication channels. The medical communication module 1 is connected to the data management module 10 through a wireless network. The data management module 10 is used to encrypt and store all maternal data, feeding indicators, risk coefficient records, adjustment logs and referral decision records, and provides a structured query interface.

[0053] Data entry module 2 is used to periodically collect and enter the mother's liver function indicators, hepatitis C virus load data, and emotional assessment scores through medical communication module 1. Periodic collection is achieved through a combination of mobile application reminders, direct data connection with home portable testing devices via Bluetooth or Wi-Fi, or manual entry of test report images for OCR recognition.

[0054] Indicator preset module 3 is used to preset baseline feeding indicators for several future cycles based on the newborn growth model; the working logic of indicator preset module 3 is as follows:

[0055] Based on the newborn's gestational age, birth weight, current age and sex, a standard newborn growth and nutrition demand model is constructed, dividing the postpartum timeline into continuous cycles in weeks, and setting a baseline time point for each future cycle.

[0056] Based on the growth nutrition requirement model, the average weight and daily total energy requirement of newborns at this time point are obtained. Combined with the maternal data collected in data entry module 2, the theoretical safe upper limit of breast milk intake during this period is defined.

[0057] The theoretical safe intake limit is converted into a baseline feeding index that includes the recommended total number of breastfeedings per day, the recommended maximum duration of each breastfeeding session, and the recommended milk volume range per bottle feeding.

[0058] The calculated baseline feeding index sequence for all future periods is stored and output as an unadjusted initial feeding plan.

[0059] The risk prediction module 4 is used to calculate the risk coefficient of breastfeeding behavior in the next cycle based on the data collected by the current cycle data entry module 2 and the preset feeding indicators for the next cycle through the preset risk prediction model. The risk prediction model takes the mother's physiological indicators and emotional state as input variables and the estimated risk under specific feeding behaviors as output.

[0060] The anomaly detection module 5 is used to compare the calculated risk coefficient with the preset standard risk threshold to determine whether there is an abnormal risk in the feeding plan for the next cycle.

[0061] The indicator adjustment module 6 is used to assess and generate positive or negative adjustments to preset feeding indicators for several future cycles when an anomaly is detected, based on the degree to which the risk coefficient exceeds the threshold. Positive adjustments indicate more conservative feeding recommendations, while negative adjustments indicate more aggressive feeding recommendations. The working logic of indicator adjustment module 6 includes:

[0062] The risk coefficient output by the anomaly judgment module 5 is mapped to discrete risk levels based on multiple preset risk threshold ranges, with each risk level corresponding to a qualitative description of the severity of the risk.

[0063] Based on the current risk level obtained from the mapping, a predefined adjustment strategy matrix is ​​queried. The adjustment strategy matrix defines several adjustment magnitudes and directions for feeding methods, maximum duration of a single feeding, and total number of feedings per day under different risk levels. The matrix is ​​organized with risk levels as rows and each of the next few cycles as columns. Each element in the matrix is ​​an adjustment instruction tuple, which includes the adjustment direction and basic adjustment magnitude value for the three dimensions of feeding methods, maximum duration of a single feeding, and total number of feedings per day. The adjustment direction is determined by the risk level, and the basic adjustment magnitude value is jointly determined by the risk level and the distance between the target cycle and the current cycle. The closer the cycle is, the larger the adjustment magnitude.

[0064] The system obtains the mother's historical adjustment records, feeding behavior compliance score, and current emotional assessment score to generate a personalized adjustment factor. This adjustment factor is then weighted and calculated with the baseline adjustment range to obtain the final personalized adjustment range. If the historical compliance is poor or the emotional state is particularly vulnerable, the positive adjustment range is amplified and the negative adjustment range is reduced.

[0065] The calculated final personalized adjustment range for each feeding dimension over several future cycles is output to the indicator generation module 7 for revising the baseline feeding indicators.

[0066] The indicator generation module 7 is used to modify the preset baseline feeding indicators based on the adjustment range, and generate the final personalized feeding indicators for newborns in the future for several cycles.

[0067] Referral decision module 8 is used to comprehensively assess the impact of the mother's hepatitis C status on feeding safety by combining the current risk coefficient prediction trend and the final feeding indicators, and to formulate a recommended medical referral intervention cycle based on the preset referral trigger rules. The referral trigger rules are a multi-factor weighted decision model. The input factors include: the number of times the risk coefficient exceeds the threshold for M consecutive cycles, the slope of the risk coefficient increase, the frequency of continuous and significant negative adjustments to the feeding indicators, and the deterioration trend of the mother's emotional assessment score. The output is the urgency and time window of the recommended referral.

[0068] The referral decision module 8 is connected to the push module 9 via a wireless network. The push module 9 is used to push the final personalized feeding indicators, health guidance suggestions and referral intervention reminders to the medical communication module 1, and to display the risk coefficient change curve, feeding indicator adjustment trajectory and referral priority queue of the managed mothers.

[0069] Medical communication module 1 is connected to data entry module 2 and indicator preset module 3 via a wireless network. Risk prediction module 4 is connected to data entry module 2, indicator preset module 3 and anomaly judgment module 5 via a wireless network. Indicator adjustment module 6 is connected to anomaly judgment module 5 and indicator generation module 7 via a wireless network. Referral decision module 8 is connected to anomaly judgment module 5 and indicator generation module 7 via a wireless network.

[0070] Compared with existing technologies, this technology breaks through the traditional static assessment that relies solely on viral load, integrates liver function, emotional state, and feeding behavior data, improves the comprehensiveness of risk prediction, calculates risk coefficients in real time based on risk prediction, and automatically generates quantifiable feeding indicator adjustment suggestions, thus solving the problem of lack of evidence-based basis for feeding decisions in existing technologies.

[0071] By using risk trend analysis and threshold triggering mechanisms, premature or delayed referrals can be avoided, reducing the risk of adverse maternal and infant outcomes. Professional medical judgment can be brought down to the patient end, and the fragmented nature of traditional follow-up can be compensated for through periodic data collection and intelligent push, thereby improving management compliance.

[0072] At other levels, in this embodiment, such as Figure 2 As shown, a risk prediction model is constructed using the following steps:

[0073] Step 41: Training set construction: Collect liver function indicators, hepatitis C virus RNA load, standardized mood scores and corresponding feeding behavior records of several postpartum women during the postpartum follow-up period as training datasets, and clinical outcome datasets including whether there were suspected mother-to-child transmission events, acute deterioration of postpartum liver function events or severe mood disorder events as labels.

[0074] Step 42, Data Processing: Preprocess the training dataset, including handling missing values, normalization, and feature construction; the constructed features include: the rate of change of physiological indicators within the period, the interaction term between emotion scores and physiological indicators, and compliance quantification features based on feeding behavior records.

[0075] Step 43, Model Training: Divide the processed dataset into a training set and a validation set, train it using a neural network algorithm, evaluate the predictive performance of each model through the validation set, and select the model with the best overall performance in terms of discrimination and calibration as the initial risk prediction model.

[0076] Step 44, Data Correction: Time series analysis is introduced to correct the trend of the output of the initial risk prediction model, and Bayesian optimization method is used to fine-tune the model hyperparameters, finally forming an integrated and dynamic risk prediction model. The trained risk prediction model is deployed in risk prediction module 4, and a regular update mechanism is established. New data is collected to perform incremental learning or periodic full retraining of the model to maintain the prediction accuracy and timeliness of the model.

[0077] The formula for calculating the risk coefficient is:

[0078] ;

[0079] In the formula, This represents the risk coefficient for predicted breastfeeding behavior in the next cycle, with a value range of [0,1]. A higher value indicates a higher risk. Represents the activation function. The total number of representative features Representing the The weight coefficients of each feature are learned during the model training process and reflect the importance of that feature in risk prediction. The representative feature function represents the feature extracted from the input data. The value of each feature, This represents a data vector of postpartum women collected during the current cycle, including but not limited to: liver function indicators, hepatitis C viral load, and standardized mood assessment scale scores. The vector representing the pre-defined newborn feeding indicators for the next cycle includes, but is not limited to: planned feeding method, planned duration of a single breastfeeding session, planned total number of breastfeeding sessions per day, and the percentage of planned breastfeeding sessions in total feeding. This represents the model bias term.

[0080] In summary, this invention achieves accurate risk assessment by integrating multiple categories of data, dynamically linking the mother's liver function indicators, viral load, emotional state, and newborn feeding plan, and using a risk prediction model to quantify and predict risk coefficients, providing an objective basis for clinical decision-making.

[0081] Employing a dynamic adjustment mechanism, the system automatically optimizes feeding plans and intelligently recommends referral times when predicted risks exceed thresholds. This ensures the safety of breastfeeding while avoiding excessive medical intervention, enabling personalized management. Through mobile terminal integration, the system simplifies the data collection process and pushes personalized suggestions in real time, improving the efficiency of doctor-patient communication. It is especially suitable for postpartum women with limited mobility.

[0082] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A postpartum hepatitis C referral management system based on mobile terminals, characterized in that, include: The medical communication module is used to integrate the communication permissions of the mobile terminal and provides several communication channels; The data entry module is used to periodically collect and enter the mother's liver function indicators, hepatitis C virus load data, and emotional assessment scores through the medical communication module. The indicator preset module is used to preset the baseline feeding indicators for several future cycles based on the newborn growth model. The risk prediction module is used to calculate the risk coefficient of breastfeeding behavior in the next cycle based on the data collected by the current cycle data entry module and the preset feeding indicators for the next cycle through the preset risk prediction model. The anomaly detection module is used to compare the calculated risk coefficient with the preset standard risk threshold to determine whether there is any abnormal risk in the feeding plan for the next cycle. The indicator adjustment module is used to assess and generate positive or negative adjustments to preset feeding indicators for several future cycles based on the degree to which the risk coefficient exceeds the threshold when an anomaly is detected. The indicator generation module is used to modify the preset baseline feeding indicators based on the adjustment range, and generate the final personalized feeding indicators for newborns in the future for several cycles. The referral decision module is used to comprehensively assess the impact of the mother's hepatitis C status on feeding safety by combining the current risk coefficient prediction trend with the final feeding indicators, and to formulate a recommended medical referral intervention cycle based on the preset referral trigger rules.

2. The postpartum hepatitis C referral management system based on a mobile terminal according to claim 1, characterized in that, The working logic of the indicator preset module is as follows: Based on the newborn's gestational age, birth weight, current age and sex, a standard newborn growth and nutrition demand model is constructed, dividing the postpartum timeline into continuous cycles in weeks, and setting a baseline time point for each future cycle. Based on the growth nutrition requirement model, the average weight and daily total energy requirement of newborns at this time point are obtained. Combined with the data collected from mothers in the data entry module, the theoretical safe upper limit of breast milk intake during this period is defined. The theoretical safe intake limit is converted into a baseline feeding index that includes the recommended total number of breastfeedings per day, the recommended maximum duration of each breastfeeding session, and the recommended milk volume range per bottle feeding. The calculated baseline feeding index sequence for all future periods is stored and output as an unadjusted initial feeding plan.

3. The postpartum hepatitis C referral management system based on a mobile terminal according to claim 1, characterized in that, The steps for constructing the risk prediction model in the risk prediction module include: Step 41: Training set construction: Collect liver function indicators, hepatitis C virus RNA load, standardized mood scores and corresponding feeding behavior records of several postpartum women during the postpartum follow-up period as training datasets, and clinical outcome datasets including whether there were suspected mother-to-child transmission events, acute deterioration of postpartum liver function events or severe mood disorder events as labels. Step 42, Data Processing: Preprocess the training dataset; Step 43, Model Training: Divide the processed dataset into a training set and a validation set, train it using a neural network algorithm, evaluate the predictive performance of each model through the validation set, and select the model with the best overall performance in terms of discrimination and calibration as the initial risk prediction model. Step 44, Data Correction: Time series analysis is introduced to correct the trend of the output of the initial risk prediction model, and finally an integrated and dynamic risk prediction model is formed. The trained risk prediction model is then deployed in the risk prediction module.

4. The postpartum hepatitis C referral management system based on a mobile terminal according to claim 1, characterized in that, The working logic of the indicator adjustment module includes: The risk coefficient output by the anomaly detection module is mapped to discrete risk levels based on multiple preset risk threshold ranges; Based on the current risk level obtained from the mapping, a predefined adjustment strategy matrix is ​​queried; the matrix is ​​arranged with risk level as the row and each period in the future as the column, and each element in the matrix is ​​an adjustment instruction tuple, which includes: the adjustment direction and basic adjustment range value for three dimensions: feeding method, maximum duration of a single feeding, and total number of feedings per day. The system obtains the mother's historical adjustment records, feeding behavior compliance score, and current emotional assessment score to generate a personalized adjustment factor. This adjustment factor is then weighted and calculated with the baseline adjustment range to obtain the final personalized adjustment range. If the historical compliance is poor or the emotional state is particularly vulnerable, the positive adjustment range is amplified and the negative adjustment range is reduced. The calculated final personalized adjustment range for each feeding dimension over several future cycles is output to the indicator generation module.

5. The postpartum hepatitis C referral management system based on a mobile terminal according to claim 4, characterized in that, The adjustment strategy matrix defines several adjustment ranges and directions for feeding methods, maximum duration of a single feeding, and total number of feedings per day under different risk levels.

6. The postpartum hepatitis C referral management system based on a mobile terminal according to claim 1, characterized in that, The formula for calculating the risk coefficient in the risk prediction module is as follows: ； In the formula, This represents the risk factor for predicted breastfeeding behavior in the next cycle. Represents the activation function. The total number of representative features Representing the The weight coefficients of each feature The representative feature function represents the feature extracted from the input data. The value of each feature, This represents a vector of maternal data collected during the current cycle. This represents the vector of pre-defined newborn feeding indicators for the next cycle. This represents the model bias term.

7. The postpartum hepatitis C referral management system based on a mobile terminal according to claim 1, characterized in that, The referral decision module uses a multi-factor weighted decision model as its referral trigger rule. The input factors include: the number of times the risk coefficient exceeds the threshold for M consecutive cycles, the slope of the risk coefficient increase, the frequency of continuous and significant negative adjustments to the feeding indicators, and the deterioration trend of the maternal emotional assessment score. The output is the urgency and time window for recommending referral.

8. The postpartum hepatitis C referral management system based on a mobile terminal according to claim 1, characterized in that, The medical communication module is connected to the data management module via a wireless network. The data management module is used to encrypt and store all maternal data, feeding indicators, risk coefficient records, adjustment logs and referral decision records, and provides a structured query interface.

9. The postpartum hepatitis C referral management system based on a mobile terminal according to claim 1, characterized in that, The referral decision module is connected to a push module via a wireless network. The push module is used to push the final personalized feeding indicators, health guidance suggestions and referral intervention reminders to the medical communication module, and to display the risk coefficient change curve, feeding indicator adjustment trajectory and referral priority queue of the managed mothers.

10. The postpartum hepatitis C referral management system based on a mobile terminal according to claim 1, characterized in that, The medical communication module, data entry module, and indicator preset module are interconnected via a wireless network. The risk prediction module, data entry module, indicator preset module, and anomaly judgment module are interconnected via a wireless network. The indicator adjustment module, anomaly judgment module, and indicator generation module are interconnected via a wireless network. The referral decision module, anomaly judgment module, and indicator generation module are interconnected via a wireless network.

Images