Method and management system for bcrl risk prediction based on patient 360 view

Abstract

This application discloses a BCRL risk prediction method and management system based on a patient 360-degree view, belonging to the field of medical information technology. This application obtains static medical data of patients from a locally pre-set patient 360-degree view system and dynamic postoperative data of patients before discharge; based on the static medical data and dynamic postoperative data, a pre-set risk scoring model is used to determine the real-time risk value of lymphedema in patients; and based on the real-time risk value, static medical data, and dynamic postoperative data, the postoperative recovery level of target patients is risk-stratified; based on the risk stratification results, an adaptive follow-up plan is generated, wherein the adaptive follow-up plan includes at least differentiated follow-up frequencies and follow-up content for different patients and / or the same patient; the adaptive follow-up plan is sent to relevant medical personnel for execution, and for regular monitoring of the postoperative lymphedema risk of each patient.

Description

Technical Field

[0001] This application relates to the field of medical information technology, and in particular to a BCRL risk prediction method and management system based on a 360-degree view of the patient. Background Technology

[0002] Breast cancer is one of the most common malignant tumors in women, and breast cancer-related lymphedema (BCRL) is a common chronic complication after surgery, severely impacting patients' quality of life. Early detection and intervention are key to managing lymphedema.

[0003] Currently, the following technical deficiencies exist in clinical practice: existing systems mainly integrate and visualize multi-source medical data, which are essentially "data dashboards" and lack in-depth analysis and prediction functions for specific diseases, thus failing to fully explore the value of the data. Furthermore, existing follow-up systems often adopt fixed-cycle homogeneous follow-ups, lacking specialist targeting and ignoring the risk of specialist complications, leading to patients with high-risk complications being easily overlooked or missed.

[0004] Therefore, there is an urgent need for a program that can tailor specific follow-up plans for different patients' conditions. Summary of the Invention

[0005] The main purpose of this application is to provide a BCRL risk prediction method and management system based on a 360-degree view of the patient, which aims to solve the technical problem of not being able to formulate specific follow-up plans for different patients' conditions.

[0006] To achieve the above objectives, this application provides a BCRL risk prediction method based on a 360-degree view of the patient, the BCRL risk prediction method based on a 360-degree view of the patient includes the following steps: The system retrieves static medical data of the patient from a locally pre-set 360° view system, and also retrieves dynamic postoperative data of the patient before discharge. Based on the static medical data and the dynamic postoperative data, the real-time risk value of the patient's lymphedema is determined through a preset risk scoring model. Based on the real-time risk value, the static medical data, and the dynamic postoperative data, the postoperative recovery level of the target patient is stratified by risk. Based on the results of risk stratification, an adaptive follow-up plan is generated, wherein the adaptive follow-up plan includes at least differentiated follow-up frequencies and follow-up content for different patients and / or the same patient; The adaptive follow-up plan will be sent to relevant medical personnel so that they can implement the adaptive follow-up plan and regularly monitor the risk of postoperative lymphedema in each patient with breast cancer.

[0007] In one embodiment, the step of generating an adaptive follow-up plan based on the risk stratification results, wherein the adaptive follow-up plan includes a high-risk-based follow-up plan and a low-risk-based follow-up plan, includes any one of the following: If the risk stratification result is high risk, a follow-up plan based on high risk is generated with the first follow-up frequency and the follow-up content is arm circumference measurement, bioelectrical impedance measurement and joint range of motion measurement. If the risk stratification result is low risk, a follow-up plan based on low risk is generated with the second follow-up frequency and the follow-up content being arm circumference measurement, bioelectrical impedance measurement and joint range of motion measurement. Wherein, the first follow-up frequency is higher than the second follow-up frequency, and the follow-up plan based on high risk and the follow-up plan based on low risk are respectively set with the expected phased results corresponding to the first follow-up frequency and the second follow-up frequency.

[0008] In one embodiment, after the step of sending the adaptive follow-up plan to relevant medical personnel for execution and monitoring the risk of postoperative lymphedema in each patient, the method further includes: Receive real-time follow-up data provided by the patient; Based on the real-time follow-up data, the real-time risk value is corrected using the risk scoring model. The real-time follow-up data was annotated with risk prediction and assessment using relevant expert annotation methods. The adaptive follow-up plan is dynamically adjusted based on the results of the correction and the evaluation labels.

[0009] In one embodiment, the step of dynamically adjusting the adaptive follow-up plan based on the correction results and the evaluation labeling results includes: If the value corresponding to the correction result is greater than or equal to the preset risk threshold, or if the evaluation and labeling results do not meet the safety range defined by the expected results of the stage, then the corresponding follow-up frequency will be increased. If the value corresponding to the correction result is less than the preset risk threshold, or if the evaluation label indicates that there is no risk of lymphedema, then the average score of each item corresponding to the follow-up content in the evaluation result is calculated, and the correction result and the average calculation result are weighted and calculated. Based on the deviation between the weighted calculation result and the real-time risk value, the corresponding follow-up frequency is reduced.

[0010] In one embodiment, prior to the step of generating an adaptive follow-up plan based on the results of risk stratification, the method further includes: If any of the following conditions are met: the real-time risk value is greater than or equal to a preset risk threshold, the number of times the patient has undergone axillary lymph node dissection is greater than a preset number, the patient has received local radiotherapy, or the patient's body mass index is greater than a preset standard index, then the risk stratification result is high risk.

[0011] In one embodiment, before the step of determining the real-time risk value of the patient's lymphedema based on the static medical data and the dynamic postoperative data using a preset risk scoring model, the method further includes: Obtain sample data and corresponding sample labels for cases of lymphedema; Based on the sample data and the sample labels, the corresponding symptom characteristics of the lymphedema cases are determined, and the corresponding time periods for each symptom characteristic are determined. Based on the disease characteristics and the time period, add risk expectation value labels corresponding to the disease characteristics at different periods to the sample labels; Based on the sample data, the sample labels, and the risk expectation value labels, the neural network model to be trained is iteratively trained to obtain a risk scoring model.

[0012] In one embodiment, after the step of iteratively training the neural network model to be trained based on the sample data, the sample labels, and the risk expectation value labels to obtain a risk scoring model, the method further includes: A portion of the sample data is extracted as a test set; The risk scoring model is tested using the test set. If the test fails, the parameters of the risk scoring model will be adjusted and the network structure optimized based on the test results.

[0013] In one embodiment, after the step of iteratively training the neural network model to be trained based on the sample data, the sample labels, and the risk expectation value labels to obtain a risk scoring model, the method further includes: According to a preset update cycle, the latest case sample data and its corresponding sample labels are obtained, and the risk scoring model is updated and optimized based on the latest case sample data and its corresponding sample labels.

[0014] In this embodiment, a management system based on a BCRL risk prediction method using a 360-degree view of the patient is also proposed, comprising: The data acquisition module is used to acquire the patient's static medical data from the locally preset patient 360-degree view system, and to acquire the patient's dynamic postoperative data before discharge. The risk prediction module is used to determine the real-time risk value of the patient's lymphedema based on the static medical data and the dynamic postoperative data through a preset risk scoring model, and to perform risk stratification of the target patient's postoperative recovery level based on the real-time risk value, the static medical data and the dynamic postoperative data. The plan generation module is used to generate an adaptive follow-up plan based on the results of risk stratification, wherein the adaptive follow-up plan includes at least differentiated follow-up frequencies and follow-up content for different patients and / or the same patient; The information transmission module is used to send the adaptive follow-up plan to relevant medical personnel so that they can implement the adaptive follow-up plan and regularly monitor the risk of postoperative lymphedema in each patient with breast cancer.

[0015] In one embodiment, the locally preset patient 360-degree view system is a visualized dynamic health record that integrates the patient's static medical data throughout their entire life cycle, including at least the patient's surgical records, examination reports, pathology reports, patient medical history records, and outpatient visit records.

[0016] One or more technical solutions proposed in this application have at least the following technical effects: acquiring static medical data of a patient from a locally preset 360° patient view system, and acquiring dynamic postoperative data of the patient before discharge; determining the real-time risk value of lymphedema of the patient using a preset risk scoring model based on the static medical data and the dynamic postoperative data, and performing risk stratification of the postoperative recovery degree of the target patient based on the real-time risk value, the static medical data, and the dynamic postoperative data; generating an adaptive follow-up plan based on the risk stratification results, wherein the adaptive follow-up plan includes at least differentiated follow-up frequency and content for different patients and / or the same patient; and transferring the self- The adaptive follow-up plan is sent to relevant medical personnel for implementation, enabling regular monitoring of postoperative lymphedema risk in each patient. This avoids the limitations of the 360-degree view system, which only integrates data without analysis. By combining static medical data with corresponding dynamic postoperative data and transforming it into risk prediction features, the system achieves in-depth data value mining. Furthermore, through a risk scoring model, it enables real-time risk value prediction for different patient conditions and risk stratification. Based on this, an adaptive matching system for follow-up management tailored to different patient conditions is established, allowing medical resources to be precisely targeted at high-risk groups. This improves follow-up efficiency and management effectiveness, ensuring timely early intervention. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating an embodiment of the BCRL risk prediction method and management system based on a 360-degree view of the patient, as provided in this application.

[0020] Figure 2 This is a schematic diagram of the system architecture corresponding to the BCRL risk prediction method and management system based on the patient's 360-degree view in this application.

[0021] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0022] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application. (See also...) Figure 1 In the first embodiment, the BCRL risk prediction method and management system based on a 360-degree view of the patient includes the following steps: S10, Obtain the patient's static medical data from the locally preset patient 360-degree view system, and obtain the patient's dynamic postoperative data before discharge. S20, based on the static medical data and the dynamic postoperative data, the real-time risk value of the patient's lymphedema is determined through a preset risk scoring model, and the postoperative recovery level of the target patient is risk-stratified according to the real-time risk value, the static medical data and the dynamic postoperative data; S30, Based on the results of risk stratification, generate an adaptive follow-up plan, wherein the adaptive follow-up plan includes at least differentiated follow-up frequencies and follow-up content for different patients and / or the same patient; S40, The adaptive follow-up plan is sent to the relevant medical personnel so that they can implement the adaptive follow-up plan and regularly monitor the risk of postoperative lymphedema in each patient after breast cancer surgery.

[0023] In this embodiment, a complete risk prediction and management architecture for postoperative lymphedema in breast cancer is constructed. Its core lies in integrating static medical data from the patient's 360-degree view system with dynamic postoperative monitoring data collected from the actual patient before discharge. Through an intelligent risk scoring model, the system achieves accurate prediction and hierarchical management of lymphedema risk and generates personalized follow-up plans based on the risk level.

[0024] The Patient 360 View System refers to a comprehensive medical data platform that integrates information from the entire patient care process. This platform gathers data from multiple sources, including electronic health records, laboratory information systems, and image archiving systems, to provide patients with a comprehensive view of their health information.

[0025] Static medical data refers to relatively stable and infrequently changing basic medical information of patients in the patient 360-degree view system, including demographic data, history of underlying diseases, surgical procedures, and treatment plans.

[0026] Dynamic postoperative data refers to monitoring indicators that change over time during the patient's postoperative recovery process, such as postoperative physiological status indicators, postoperative wound healing degree, and changes in postoperative symptoms.

[0027] Among them, the risk scoring model refers to a mathematical model for predicting the risk of lymphedema based on machine learning algorithms. It can use algorithms such as logistic regression, random forest or neural network, input multidimensional features, and output the probability of the patient developing lymphedema after surgery.

[0028] Specifically, the above solution involves first establishing a unified data collection and integration channel to ensure the integrity and timeliness of patient information. Static medical data of patients are obtained from the locally preset 360-degree view system, including basic information (age, BMI, and education level), medical history (chemotherapy history, radiotherapy history), and surgical records (number of axillary lymph node dissections, location of breast tumors, and surgical method). At the same time, dynamic postoperative data of patients are obtained before their discharge, including real-time monitoring indicators such as changes in limb circumference, bioelectrical impedance values, joint range of motion, and subjective symptoms.

[0029] Before using the data, it is necessary to perform corresponding standardization processing, including terminology standardization, format unification, and quality verification of heterogeneous medical data, to ensure data quality.

[0030] Furthermore, after obtaining static medical data and dynamic postoperative data, a quantitative assessment of lymphedema risk is performed based on multi-dimensional features. Specifically, based on the static medical data and the dynamic postoperative data, a preset risk scoring model is used to determine the real-time risk value of the patient's lymphedema. This value is any continuous value between 0 and 14. At the same time, based on the real-time risk value, the static medical data, and the dynamic postoperative data, the postoperative recovery level of the target patient is stratified by risk, and the patient is divided into two levels: high risk and low risk.

[0031] Furthermore, differentiated follow-up strategies can be developed based on risk levels. Specifically, an adaptive follow-up plan can be generated based on the results of risk stratification. This plan includes differentiated follow-up frequencies and content for different risk levels. At the same time, the feasibility and execution efficiency of the follow-up plan can be optimized based on the patient's specific condition and the availability of medical resources. In addition, to ensure the effective implementation of the follow-up plan and the continuity of risk monitoring, the adaptive follow-up plan needs to be sent to relevant medical personnel, including attending physicians, rehabilitation therapists, and community nurses, so that they can implement the adaptive follow-up plan, record the implementation status and patient feedback, and regularly monitor the risk of postoperative lymphedema in each patient after breast cancer surgery, forming a long-term tracking mechanism.

[0032] It should be noted that in current technologies, follow-up recommendations from follow-up systems are typically tailored to a large number of patients, with identical or similar long-term follow-up tasks. This reduces the complexity of task execution for staff and the complexity of patient information management. However, fixed long-term follow-up plans are not suitable for every patient. For high-risk patients with weaker constitutions, slower postoperative recovery, or a high risk of lymphedema, a long follow-up period may result in lymphedema going undetected even after the condition has developed. While the above-mentioned plan in this embodiment increases the workload for relevant medical personnel, the combination of risk scoring models and adaptive follow-up plan generation allows for the effective development of different follow-up plans for different risk groups. This enables more precise and flexible control of postoperative lymphedema risk in different patient groups, effectively improving the management of lymphedema risk.

[0033] In this embodiment, the step of generating an adaptive follow-up plan based on the risk stratification results includes an adaptive follow-up plan based on high risk and an adaptive follow-up plan based on low risk, and includes any one of the following: If the risk stratification result is high risk, a follow-up plan based on high risk is generated with the first follow-up frequency and the follow-up content is arm circumference measurement, bioelectrical impedance measurement and joint range of motion measurement. If the risk stratification result is low risk, a follow-up plan based on low risk is generated with the second follow-up frequency and the follow-up content being arm circumference measurement, bioelectrical impedance measurement and joint range of motion measurement. Wherein, the first follow-up frequency is higher than the second follow-up frequency, and the follow-up plan based on high risk and the follow-up plan based on low risk are respectively set with the expected phased results corresponding to the first follow-up frequency and the second follow-up frequency.

[0034] It is understood that the focus of the above embodiments is on the prediction and adaptive follow-up plan generation for patients with different risks. Specifically, it is on the method of generating differentiated follow-up strategies. The core of this risk-based personalized follow-up strategy is to configure a follow-up plan with corresponding intensity and content according to the different risk levels of patients, so as to achieve precise allocation and efficient utilization of medical resources.

[0035] Therefore, when generating a corresponding follow-up plan, a risk identification approach is needed to determine the appropriate high- or low-frequency follow-up strategy for the current patient. Based on this, the corresponding decision-making plan logic can be generated. Specifically, in the risk level identification process, the risk stratification results output by the risk scoring model are used to identify the patient's risk category. The magnitude of the risk value and the degree of clinical characteristics are considered comprehensively. If the risk stratification result is high risk, a high-risk follow-up plan is generated with the first follow-up frequency and the follow-up content includes arm circumference measurement, bioelectrical impedance analysis, and joint range of motion measurement. If the risk stratification result is low risk, a low-risk follow-up plan is generated with the second follow-up frequency and the follow-up content includes arm circumference measurement, bioelectrical impedance analysis, and joint range of motion measurement.

[0036] Based on this, the expected phased results corresponding to the first follow-up frequency and the second follow-up frequency are set for the high-risk follow-up plan and the low-risk follow-up plan, respectively, as the benchmark for evaluating the follow-up effect. When generating the follow-up plan, the first follow-up frequency is set to once a week to continuously monitor changes in key indicators, and the second follow-up frequency is set to once a month to conduct routine monitoring and health education.

[0037] The first follow-up frequency refers to the enhanced follow-up interval set for high-risk patients, which can be set as follows: once a week within 1 month after surgery, once every two weeks from 1 to 3 months, and once a month from 3 to 6 months.

[0038] The second follow-up frequency refers to the routine follow-up interval set for low-risk patients. This second follow-up frequency must be less than the second follow-up frequency. Specifically, it can be set as follows: once a month for 1-3 months after surgery, once every two months for 3-6 months, and once a quarter after 6 months.

[0039] The follow-up content refers to the information that medical personnel need to inquire about when following up with patients, specifically including arm circumference measurement, bioelectrical impedance measurement, and joint range of motion measurement.

[0040] It should be noted that when generating this follow-up plan, due to the differences among patients, the follow-up indicators will also differ for different patients. The indicators in this follow-up plan specifically refer to the expected results at different stages. These expected results refer to the clinical target values ​​that the patient's wound healing degree, changes in the patient's physiological indicators, and other indicators are expected to reach at different time points after surgery. These can include the increase in arm circumference of the affected limb not exceeding 2cm from the baseline, bioelectrical impedance values ​​within the normal range, and joint range of motion meeting functional requirements, etc.

[0041] In this embodiment, after the step of sending the adaptive follow-up plan to relevant medical personnel for execution and monitoring the risk of postoperative lymphedema in each patient after breast cancer surgery, the method further includes: Receive real-time follow-up data provided by the patient; Based on the real-time follow-up data, the real-time risk value is corrected using the risk scoring model. The real-time follow-up data was annotated with risk prediction and assessment using relevant expert annotation methods. The adaptive follow-up plan is dynamically adjusted based on the results of the correction and the evaluation labels.

[0042] In this embodiment, during the implementation of the adaptive follow-up plan, there may be discrepancies between the actual postoperative recovery of some patients and the expected recovery, or the deterioration of the condition due to the patient's bad habits after discharge. Therefore, when medical personnel conduct follow-ups, they also need to dynamically optimize the established follow-up plan based on the actual situation and real-time data feedback from the patient. This establishes a continuous optimization mechanism for the follow-up plan, corrects risk predictions through real-time follow-up data and expert annotation information, and dynamically adjusts the follow-up plan based on the correction results, forming a closed-loop management.

[0043] Specifically, in this closed-loop management process, follow-up data collection and processing are first carried out from the patient's side. Real-time follow-up data provided by the patient is received, including measurements of the affected limb, changes in symptoms, and functional status. The data is then cleaned and standardized to ensure data quality and consistency. Based on this real-time data, the results output by the risk scoring model are corrected. Specifically, based on the real-time follow-up data, the real-time risk value is corrected through the risk scoring model to improve prediction accuracy. In addition, the real-time follow-up data is annotated with risk prediction assessments by relevant experts to obtain consistent evaluations from clinical experts.

[0044] Furthermore, the dynamic adjustment of the follow-up plan mainly relies on the corrected risk values ​​and expert annotation results to assess the applicability of the original follow-up plan, and then makes corresponding adjustments to the follow-up frequency, content and intensity based on the assessment results.

[0045] Real-time follow-up data refers to the clinical indicators and symptom data of patients collected in real time during the follow-up process, including objective measurement data (arm circumference, bioelectrical impedance) and subjective reporting data (symptom score, functional status).

[0046] Among them, risk prediction assessment and labeling refers to the manual assessment and labeling of the degree of lymphedema risk of patients by clinical experts. A unified lymphedema assessment standard can be adopted to ensure the consistency and reliability of the labeling.

[0047] It should be noted that the main purpose of the above-mentioned follow-up plan optimization process is to avoid the deviation between the risk scoring model and the patient's actual recuperation process. The prediction direction of the model is closer to the idealized situation of the data, while there may be many hidden dangers in the patient's life that are not conducive to the condition. For example, the patient's work and rest are not proper, the diet is not proper, or the patient is infected with other germs during the recuperation process, which may induce premature lymphedema. Therefore, optimizing the follow-up plan based on actual real-time data can further improve the management system and improve the monitoring effect of the patient's condition.

[0048] In this embodiment, the step of dynamically adjusting the adaptive follow-up plan based on the correction results and the evaluation labeling results includes: If the value corresponding to the correction result is greater than or equal to the preset risk threshold, or if the evaluation and labeling results do not meet the safety range defined by the expected results of the stage, then the corresponding follow-up frequency will be increased. If the value corresponding to the correction result is less than the preset risk threshold, or if the evaluation label indicates that there is no risk of lymphedema, then the average score of each item corresponding to the follow-up content in the evaluation result is calculated, and the correction result and the average calculation result are weighted and calculated. Based on the deviation between the weighted calculation result and the real-time risk value, the corresponding follow-up frequency is reduced.

[0049] In this embodiment, the adaptive adjustment process of follow-up frequency involves establishing follow-up frequency adjustment rules based on risk thresholds and phased goals to achieve precise control of follow-up intensity. If the value corresponding to the correction result is greater than or equal to the preset risk threshold, the patient is determined to be in a high-risk state. Or, if the evaluation label does not meet the safety range defined by the phased expected result, the patient's recovery progress is determined to be unsatisfactory. When any of the above conditions are met, the corresponding follow-up frequency is increased, such as from once a month to once a week. If the value corresponding to the correction result is less than the preset risk threshold, or the evaluation label indicates that there is no risk of lymphedema, then the frequency reduction assessment process is initiated.

[0050] In the frequency reduction assessment process, it is necessary to calculate the mean of the scores of each item corresponding to the follow-up content in the assessment results to obtain a comprehensive score. The correction results and the mean calculation results are weighted and calculated, with the weights allocated according to clinical importance. Based on the deviation between the weighted calculation results and the real-time risk value, the frequency adjustment range is determined, and the corresponding follow-up frequency is reduced.

[0051] It should be noted that the follow-up process involves multiple aspects, each with a assigned score. The average of the scores for each aspect is calculated, and this average is then weighted and calculated with the corrected result to obtain the actual result corresponding to the combined actual follow-up data and the corrected data. The follow-up frequency is adjusted based on the deviation between this weighted calculation result and the real-time risk value initially used to set the adaptive follow-up plan. Specifically, if the weighted calculation result is greater than the real-time risk value, the frequency is increased, and vice versa.

[0052] The preset risk threshold refers to the critical value set in advance to determine the risk level. This value can be set to 13 (the real-time risk value corresponds to 0-14, based on the Chinese Nursing Association Group Standard T / CNAS05─2020; Prevention and Nursing of Lymphedema after Breast Cancer Surgery). That is, when the risk value is ≥13, it is considered a high-risk state.

[0053] In the process of assigning weights to different indicators according to their importance and calculating the comprehensive score, the weight of clinical indicators is 0.6, the weight of patient-reported results is 0.3, and the weight of expert evaluation is 0.1.

[0054] In this embodiment, before the step of generating an adaptive follow-up plan based on the risk stratification results, the method further includes: If any of the following conditions are met: the real-time risk value is greater than or equal to a preset risk threshold, the number of times the patient has undergone axillary lymph node dissection is greater than a preset number, the patient has received local radiotherapy, or the patient's body mass index is greater than a preset standard index, then the risk stratification result is high risk.

[0055] Understandably, it is necessary to clarify the criteria for identifying high-risk patients and to establish rules for high-risk factors based on clinical evidence in order to achieve accurate identification and early intervention for high-risk groups.

[0056] Among them, the indicators for detecting high-risk factors include: Real-time risk value: If the real-time risk value is greater than or equal to the preset risk value (e.g., ≥13).

[0057] Surgical scope: The number of axillary lymph node dissections performed on the patient is greater than the preset number (e.g., ≥10).

[0058] Medical history: The patient had received local radiotherapy.

[0059] Physical condition: The patient's body mass index is greater than the preset standard index (e.g., BMI ≥ 24).

[0060] Specifically, when a patient meets any one of the high-risk factors, the risk stratification result is high risk. Patients who meet multiple high-risk factors are marked as extremely high risk, and an intensive management program is initiated.

[0061] The preset number of times refers to the critical number of axillary lymph node dissections. The evidence-based basis is the positive correlation between the number of lymph node dissections and the risk of lymphedema shown in clinical studies. The specific number of times can be set according to actual needs.

[0062] In this embodiment, before the step of determining the real-time risk value of the patient's lymphedema based on the static medical data and the dynamic postoperative data using a preset risk scoring model, the method further includes: Obtain sample data and corresponding sample labels for cases of lymphedema; Based on the sample data and the sample labels, the corresponding symptom characteristics of the lymphedema cases are determined, and the corresponding time periods for each symptom characteristic are determined. Based on the disease characteristics and the time period, add risk expectation value labels corresponding to the disease characteristics at different periods to the sample labels; Based on the sample data, the sample labels, and the risk expectation value labels, the neural network model to be trained is iteratively trained to obtain a risk scoring model.

[0063] In this embodiment, the process of constructing a risk scoring model is provided. In this process, the main focus is on training a neural network model based on lymphedema case data through a supervised learning algorithm to establish a mapping relationship between multiple features and lymphedema risk.

[0064] It should be noted that the lymphedema case data is data that has been authorized by the corresponding patient users and has undergone corresponding user information desensitization processing. In addition, this type of data has a high degree of temporal characteristics.

[0065] During the construction process, training data preparation and data collection are first performed. Sample data of lymphedema cases and their corresponding sample labels are obtained, including basic patient information, treatment data, follow-up results, etc. (data after desensitization). The data is then cleaned, de-identified, and standardized. Based on the sample data and sample labels, the corresponding symptom characteristics of the lymphedema cases are determined, such as the difference in limb circumference, tissue stiffness, and subjective symptoms. This determines the time period corresponding to each symptom characteristic and clarifies the clinical significance of each characteristic at different time points.

[0066] Meanwhile, unlike the conventional model building process, risk expectation value labels corresponding to the disease characteristics of different periods are added to the sample labels to construct a time series labeling system. Based on the sample data, the sample labels, and the risk expectation value labels, the neural network model to be trained is iteratively trained. The cross-entropy loss function and the corresponding optimizer are used, and the early stopping method is used to prevent overfitting, thus obtaining the risk scoring model.

[0067] It should be noted that the main purpose of adding risk expectation value labels corresponding to the disease characteristics at different stages in the sample labels is to distinguish the different performance of different patients in different recovery cycles. The risk expectation value label represents the expected value of lymphedema risk at different time points. For example, patient A is a low-risk patient whose data changes significantly in a short period of time and is cured after three months. Or, for example, patient B is a high-risk patient whose disease recovers slowly within one month and experiences multiple relapses within three months. Therefore, it is necessary to set a standard for the disease situation of different patients at different stages, and on the basis of this standard, additional risk expectation values ​​are added to the sample labels for different stages. Specifically, for example, patient A should recover well within one month and be nearly cured within three months, corresponding to one-month and three-month time periods, respectively, and set low-risk characterization values. Similarly, if the actual physiological data performance of the patient in the corresponding one-month and three-month time periods is different from the expectation, then a high-risk condition is met at this time, and a high-risk characterization value is set. That is, two sets of expectation values ​​are set for whether the expected condition is met in the same cycle.

[0068] The sample labels involved in the above model training process are all real lymphedema occurrences labeled in the training data, and the labeling standard is a unified labeling based on the international lymphedema standard.

[0069] In this embodiment, after the step of iteratively training the neural network model to be trained based on the sample data, the sample labels, and the risk expectation value labels to obtain the risk scoring model, the method further includes: A portion of the sample data is extracted as a test set; The risk scoring model is tested using the test set. If the test fails, the parameters of the risk scoring model will be adjusted and the network structure optimized based on the test results.

[0070] In this embodiment, the model performance also needs to be verified and optimized. A strict verification mechanism for model performance is established, and the model performance is evaluated through an independent test set. Based on the test results, parameters and structure are optimized to ensure the accuracy and generalization ability of the model. A portion of the sample data is extracted as a test set, usually accounting for 20%-30% of the total sample size. The risk scoring model is tested using the test set, and performance indicators such as accuracy and recall are calculated. If the test fails, the parameters of the risk scoring model are adjusted and the network structure is optimized based on the test results. The parameter adjustment includes adjusting the learning rate and optimizing the regularization strength, and the network structure optimization includes adjusting the number of hidden layer nodes and optimizing the activation function.

[0071] The test set is an independent dataset used to evaluate the model's generalization ability. It is independent and identically distributed with the training set to ensure the objectivity of the evaluation results.

[0072] Among them, parameter tuning refers to the process of optimizing the model's hyperparameters, which can be achieved using automated hyperparameter tuning methods such as grid search, random search, or Bayesian optimization.

[0073] In this embodiment, after the step of iteratively training the neural network model to be trained based on the sample data, the sample labels, and the risk expectation value labels to obtain the risk scoring model, the method further includes: According to a preset update cycle, the latest case sample data and its corresponding sample labels are obtained, and the risk scoring model is updated and optimized based on the latest case sample data and its corresponding sample labels.

[0074] In this embodiment, a continuous evolution mechanism for the model is established. This is mainly achieved by regularly collecting new clinical case data and continuously updating and optimizing the model through incremental learning to maintain the timeliness and accuracy of the model's predictive ability. According to a preset update cycle (such as quarterly), the latest case sample data and its corresponding sample labels are obtained, and the new data undergoes the same cleaning and standardization processing as the original training data. At the same time, the risk scoring model is updated and optimized based on the latest case sample data and its corresponding sample labels.

[0075] Incremental learning algorithms can be used to incorporate new features and patterns while retaining existing knowledge, and to monitor performance changes before and after model updates to ensure that model performance does not decline after updates.

[0076] The preset update cycle refers to the fixed time interval for model updates. This cycle can be set to 3-6 months to balance the timeliness and stability requirements of the model.

[0077] Incremental learning refers to the technique of continuing to train on the basis of the original model using new data. Algorithms such as elastic weight consolidation can be used to prevent catastrophic forgetting caused by new knowledge overwriting old knowledge.

[0078] This embodiment acquires static medical data of the patient from a locally pre-set patient 360-degree view system, and also acquires dynamic postoperative data of the patient before discharge. Based on the static medical data and the dynamic postoperative data, a pre-set risk scoring model is used to determine the real-time risk value of the patient's lymphedema. Based on the real-time risk value, the static medical data, and the dynamic postoperative data, the postoperative recovery level of the target patient is risk-stratified. Based on the risk stratification results, an adaptive follow-up plan is generated, wherein the adaptive follow-up plan includes at least differentiated follow-up frequencies and content for different patients and / or the same patient. The adaptive follow-up plan is then sent to relevant [relevant departments / organizations]. The system is designed for medical personnel to implement the adaptive follow-up plan and regularly monitor the risk of postoperative lymphedema in each patient after breast cancer surgery. This avoids the limitations of the 360-degree view system, which only integrates data without analysis. It combines static medical data with corresponding dynamic postoperative data and transforms them into risk prediction features, achieving in-depth data value mining. At the same time, through a risk scoring model, it can predict the real-time risk value of different patients and stratify different patients by risk. Based on this, an adaptive matching for follow-up management is established for different patient conditions, enabling medical resources to be accurately directed to high-risk groups, improving follow-up efficiency and management effectiveness, and ensuring the timeliness of early intervention.

[0079] According to the above embodiments, refer to Figure 2 The BCRL risk prediction method based on a 360-degree view of the patient also proposes a corresponding management system, which includes: The data acquisition module is used to acquire the patient's static medical data from the locally preset patient 360-degree view system, and to acquire the patient's dynamic postoperative data before discharge. The risk prediction module is used to determine the real-time risk value of the patient's lymphedema based on the static medical data and the dynamic postoperative data through a preset risk scoring model, and to perform risk stratification of the target patient's postoperative recovery level based on the real-time risk value, the static medical data and the dynamic postoperative data. The plan generation module is used to generate an adaptive follow-up plan based on the results of risk stratification, wherein the adaptive follow-up plan includes at least differentiated follow-up frequencies and follow-up content for different patients and / or the same patient; The information transmission module is used to send the adaptive follow-up plan to relevant medical personnel so that they can implement the adaptive follow-up plan and regularly monitor the risk of postoperative lymphedema in each patient with breast cancer.

[0080] In this embodiment, the locally preset patient 360-degree view system is a visualized dynamic health record that integrates the patient's static medical data throughout the entire life cycle, including at least the patient's surgical records, examination reports, pathology reports, patient medical history records, and outpatient visit records.

[0081] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A BCRL risk prediction method based on a 360-degree view of the patient, characterized in that, The BCRL risk prediction method based on a 360-degree view of the patient includes the following steps: The system retrieves static medical data of the patient from a locally pre-set 360° view system, and also retrieves dynamic postoperative data of the patient before discharge. Based on the static medical data and the dynamic postoperative data, the real-time risk value of the patient's lymphedema is determined through a preset risk scoring model. Based on the real-time risk value, the static medical data, and the dynamic postoperative data, the postoperative recovery level of the target patient is stratified by risk. Based on the results of risk stratification, an adaptive follow-up plan is generated, wherein the adaptive follow-up plan includes at least differentiated follow-up frequencies and follow-up content for different patients and / or the same patient; The adaptive follow-up plan will be sent to relevant medical personnel so that they can implement the adaptive follow-up plan and regularly monitor the risk of postoperative lymphedema in each patient with breast cancer.

2. The method as described in claim 1, characterized in that, The step of generating an adaptive follow-up plan based on the risk stratification results, wherein the adaptive follow-up plan includes a high-risk-based follow-up plan and a low-risk-based follow-up plan, and includes any one of the following: If the risk stratification result is high risk, a follow-up plan based on high risk is generated with the first follow-up frequency and the follow-up content is arm circumference measurement, bioelectrical impedance measurement and joint range of motion measurement. If the risk stratification result is low risk, a follow-up plan based on low risk is generated with the second follow-up frequency and the follow-up content being arm circumference measurement, bioelectrical impedance measurement and joint range of motion measurement. Wherein, the first follow-up frequency is higher than the second follow-up frequency, and the follow-up plan based on high risk and the follow-up plan based on low risk are respectively set with the expected phased results corresponding to the first follow-up frequency and the second follow-up frequency.

3. The method as described in claim 2, characterized in that, After the step of sending the adaptive follow-up plan to relevant medical personnel for execution and monitoring the risk of postoperative lymphedema in each patient, the method further includes: Receive real-time follow-up data provided by the patient; Based on the real-time follow-up data, the real-time risk value is corrected using the risk scoring model. The real-time follow-up data was annotated with risk prediction and assessment using relevant expert annotation methods. The adaptive follow-up plan is dynamically adjusted based on the results of the correction and the evaluation labels.

4. The method as described in claim 3, characterized in that, The step of dynamically adjusting the adaptive follow-up plan based on the correction results and evaluation labeling results includes: If the value corresponding to the correction result is greater than or equal to the preset risk threshold, or if the evaluation and labeling results do not meet the safety range defined by the expected results of the stage, then the corresponding follow-up frequency will be increased. If the value corresponding to the correction result is less than the preset risk threshold, or if the evaluation label indicates that there is no risk of lymphedema, then the average score of each item corresponding to the follow-up content in the evaluation result is calculated, and the correction result and the average calculation result are weighted and calculated. Based on the deviation between the weighted calculation result and the real-time risk value, the corresponding follow-up frequency is reduced.

5. The method as described in claim 2, characterized in that, Before the step of generating an adaptive follow-up plan based on the risk stratification results, the method further includes: If any of the following conditions are met: the real-time risk value is greater than or equal to a preset risk threshold, the number of times the patient has undergone axillary lymph node dissection is greater than a preset number, the patient has received local radiotherapy, or the patient's body mass index is greater than a preset standard index, then the risk stratification result is high risk.

6. The method as described in claim 1, characterized in that, Before the step of determining the real-time risk value of the patient's lymphedema based on the static medical data and the dynamic postoperative data using a preset risk scoring model, the method further includes: Obtain sample data and corresponding sample labels for cases of lymphedema; Based on the sample data and the sample labels, the corresponding symptom characteristics of the lymphedema cases are determined, and the corresponding time periods for each symptom characteristic are determined. Based on the disease characteristics and the time period, add risk expectation value labels corresponding to the disease characteristics at different periods to the sample labels; Based on the sample data, the sample labels, and the risk expectation value labels, the neural network model to be trained is iteratively trained to obtain a risk scoring model.

7. The method as described in claim 6, characterized in that, After the step of iteratively training the neural network model to be trained based on the sample data, the sample labels, and the risk expectation value labels to obtain the risk scoring model, the method further includes: A portion of the sample data is extracted as a test set; The risk scoring model is tested using the test set. If the test fails, the parameters of the risk scoring model will be adjusted and the network structure optimized based on the test results.

8. The method as described in claim 7, characterized in that, After the step of iteratively training the neural network model to be trained based on the sample data, the sample labels, and the risk expectation value labels to obtain the risk scoring model, the method further includes: According to a preset update cycle, the latest case sample data and its corresponding sample labels are obtained, and the risk scoring model is updated and optimized based on the latest case sample data and its corresponding sample labels.

9. A management system for implementing the BCRL risk prediction method based on a 360-degree view of a patient as described in any one of claims 1-8, characterized in that, include: The data acquisition module is used to acquire the patient's static medical data from the locally preset patient 360-degree view system, and to acquire the patient's dynamic postoperative data before discharge. The risk prediction module is used to determine the real-time risk value of the patient's lymphedema based on the static medical data and the dynamic postoperative data through a preset risk scoring model, and to perform risk stratification of the target patient's postoperative recovery level based on the real-time risk value, the static medical data and the dynamic postoperative data. The plan generation module is used to generate an adaptive follow-up plan based on the results of risk stratification, wherein the adaptive follow-up plan includes at least differentiated follow-up frequencies and follow-up content for different patients and / or the same patient; The information transmission module is used to send the adaptive follow-up plan to relevant medical personnel so that they can implement the adaptive follow-up plan and regularly monitor the risk of postoperative lymphedema in each patient with breast cancer.

10. The system as described in claim 9, characterized in that, include: The locally preset 360-degree view system for patients is a visualized dynamic health record that integrates static medical data from the patient's entire life cycle. It includes at least the patient's surgical records, laboratory reports, pathology reports, patient medical history records, and outpatient visit records.

Images