Breast cancer risk management and control system and medium

Abstract

The invention relates to a breast cancer risk management and control system and a medium, the system comprises at least one memory in which a computer program is stored, at least one processor is configured to implement a breast cancer risk management and control method when executing the computer program, and the method comprises the following specific steps: training a model by using a preset and trained random forest algorithm, calculating the contribution degree of each feature in decision tree construction, giving feature importance scores, and calculating the weight of each risk factor to the breast cancer risk value according to different feature importance scores; analyzing and determining a breast cancer detection period according to a corresponding relation between a weight range in which the weight of the specific risk factor to the breast cancer risk value falls, a risk value evaluation range in which the breast cancer risk value falls and the detection period; and forming a patient breast cancer related detection time planning table according to the analyzed and determined breast cancer detection period. The method has the advantages that accurate control over the breast cancer risk is achieved, and the prevention and treatment effects are improved.

Description

Technical Field

[0001] The present invention relates to the field of medical health, and in particular to a breast cancer risk management and control system and a medium. Background Art

[0002] The pathogenesis of breast cancer is complex and is affected by a combination of factors, including genetic factors, lifestyle, and environmental factors. Early detection, diagnosis, and intervention are crucial to improving the cure rate of breast cancer and the survival rate of patients. With the continuous development of science and technology, the application of digital and intelligent technologies in the medical field is becoming more and more extensive, providing new ideas and methods for breast cancer risk management.

[0003] Currently, traditional breast cancer detection mainly relies on regular clinical examinations, such as mammography, breast ultrasound, etc.

[0004] Although these examination methods can detect breast lesions to a certain extent, they have limitations. On the one hand, the examination cycle is usually arranged at fixed time intervals and lacks personalized basis. This may cause some high-risk patients to develop their diseases without being discovered in time during the interval between two examinations, or low-risk patients to be over-examined, resulting in a waste of medical resources. On the other hand, traditional detection methods are difficult to comprehensively consider the impact of various factors in patients' daily lives on the risk of breast cancer. They only focus on the local physiological characteristics of the breast and lack effective integration and utilization of multi-source data, making it difficult to meet the needs of precision medicine and efficient resource allocation. Summary of the invention

[0005] In order to achieve accurate control of breast cancer risk and improve prevention and treatment effects, the present application provides a breast cancer risk control system and medium.

[0006] In a first aspect, the present application provides a breast cancer risk management system, which adopts the following technical solution: A breast cancer risk management system, comprising: Breast monitoring equipment; at least one processor; At least one memory storing a computer program, and at least one processor configured to implement a breast cancer risk management method when executing the computer program, the specific steps are as follows: Obtain patient-related data collected by breast monitoring equipment and perform data preprocessing; Inputting the patient-related data that has completed data preprocessing into the trained breast cancer risk value assessment model that considers the patient's personalized data, and outputting the breast cancer risk value; Use the preset and trained random forest algorithm to train the model and calculate the weight of each risk factor for breast cancer risk value; Analyze and determine the breast cancer detection cycle based on the corresponding relationship between the weight range of the breast cancer risk value of the specific risk factor, the risk value assessment range of the breast cancer risk value, and the detection cycle; Form a time schedule for breast cancer related testing for the patient based on the breast cancer testing cycle determined by the analysis; According to the corresponding relationship between the weight interval of each risk factor for the breast cancer risk value and the quantitative suggestion, quantitative suggestion information is formed, and the formed quantitative suggestion information is sent to the terminal held by the patient; Collect the latest data of the patient within the preset period after the quantitative recommendation is sent, and re-analyze the breast cancer risk value and the weight of each risk factor to the breast cancer risk value based on the latest data; Analyze whether the reduction in breast cancer risk value produced by each risk factor alone reaches or exceeds the preset range; If yes, the original breast cancer detection cycle is maintained unchanged, and the original breast cancer detection cycle is determined according to the corresponding relationship between the weight range of the breast cancer risk value of the specific risk factor, the risk value assessment range of the breast cancer risk value, and the detection cycle; If not, then analyze and obtain the risk factors whose decline is less than the preset range as the target risk factors, and calculate the ratio of the decline of the target risk factor to the preset range. According to the ratio of the decline of the target risk factor to the preset range and the weight of the target risk factor to the breast cancer risk value, calculate the adjustment coefficient of all target factors with respect to the breast cancer detection cycle, and then calculate the adjusted breast cancer detection cycle according to the adjustment coefficient and the original breast cancer detection cycle; Update the patient's breast cancer related testing schedule based on the finalized breast cancer testing cycle.

[0007] By adopting the above technical solutions, precise management can be achieved through real-time monitoring, personalized risk assessment and dynamic adjustment of the detection cycle. Quantitative suggestions generated based on data-driven weight analysis can effectively intervene in high-risk factors and improve early detection rates. Intelligent optimization of detection frequency can reduce medical costs, while ensuring the timeliness and effectiveness of risk management through a continuous tracking and feedback mechanism.

[0008] Optionally, the method further includes steps after collecting patient-related data and performing data preprocessing, and before using a preset and trained random forest algorithm to train the model, as follows: The patients were divided into different age groups, including young, middle-aged and elderly groups; If the patient is in the young or middle-aged group, the original settings are maintained; If the patient was in the elderly group, variables associated with breast cancer risk were selected, and the stepwise regression method was used for variable screening to establish a preliminary multivariate regression model; The preprocessed data is divided into a training set and a validation set, and the multivariate regression model is trained on the training set. The performance of the multivariate regression model is adjusted by adjusting the parameters of the multivariate regression model until the accuracy of the multivariate regression model on the validation set reaches the preset accuracy within a preset number of consecutive times, and the training of the multivariate regression model is determined to be completed, and the regression coefficient corresponding to each risk factor is checked; According to the regression coefficient corresponding to each risk factor, the sum of the absolute values ​​of all regression coefficients is calculated, and the ratio of the absolute value of the regression coefficient of different risk factors to the sum of the absolute values ​​of all regression coefficients is calculated to determine the weight of each risk factor for breast cancer risk value; The breast cancer detection cycle is determined by analysis based on the correspondence between the weight range of the breast cancer risk value for the specific risk factors, the risk value assessment range within which the breast cancer risk value falls, and the detection cycle.

[0009] By adopting the above technical solutions and introducing age group classification, a more detailed risk assessment can be conducted for elderly patients. The stepwise regression method is used to screen variables and establish a multivariate regression model to more accurately analyze the weight of each risk factor on breast cancer risk. This innovation not only improves the accuracy of risk assessment, but also tailors a more reasonable detection cycle for elderly patients.

[0010] Optionally, the method further includes a step of analyzing and obtaining risk factors with a decrease range less than a preset range as target risk factors, which is as follows: Patients whose breast cancer risk values ​​resulting from risk factors alone do not decrease by the preset range are considered to be patients who need to be followed up; Start the follow-up process, contact the patient through intelligent voice methods, and obtain the patient's implementation results of quantitative advice information, including exercise duration and dietary adjustments; If the patient's execution result of the quantitative suggestion information does not match the quantitative content mentioned in the quantitative suggestion information, the patient is advised to execute it again according to the quantitative content mentioned in the quantitative suggestion information; If the patient's execution result on the quantitative advice information is consistent with the quantitative content mentioned in the quantitative advice information, the quantitative advice information is analyzed for health-related quantitative content not mentioned, and suggestions for other quantitative content are provided to be added to the quantitative advice information to form new quantitative advice information, which is then sent to the patient's terminal again.

[0011] By adopting the above technical solutions and refined follow-up management, we provide patients with more considerate health services. For patients whose risk reduction does not meet expectations, we initiate the follow-up process, communicate in time and adjust the quantitative recommendations. This not only improves patient compliance, but also adjusts the health plan according to actual conditions and realizes personalized health management.

[0012] Optionally, updating the patient's breast cancer related testing schedule according to the finalized breast cancer testing cycle includes: Obtain relevant information from the patient's electronic medical records and extract key factors that affect breast cancer test results, including ongoing treatment plans for other diseases, specific drugs used, recent surgical history, and the severity of the patient's underlying disease; According to the correspondence between the influence degree value range of the preset key factors affecting the breast cancer detection results and the delay time, the overall delay time caused by the key factors affecting the breast cancer detection results is analyzed and determined; According to the overall delay time caused by the key factors affecting the breast cancer detection results and the breast cancer detection cycle determined by analysis, the breast cancer detection cycle is re-analyzed and the patient's breast cancer-related detection time planning table is re-formed.

[0013] By adopting the above technical solutions and integrating key factors of electronic medical records (such as treatment plans, drugs, surgical history, etc.), the detection cycle can be dynamically adjusted to avoid interference of other medical interventions on the test results and improve the accuracy of the test. Real-time analysis of the impact of treatment and intelligent postponement can not only ensure the effectiveness of the test, but also optimize resource allocation, ensure that patients are screened at the best time, and enhance the comprehensiveness and adaptability of risk management.

[0014] In a second aspect, the present application provides a computer storage medium, which adopts the following technical solution: A computer storage medium comprises a memory, a processor and a program stored in the memory and executable on the processor, wherein the program can be loaded and executed by the processor to implement the breast cancer risk management method in the breast cancer risk management system as described in the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flow chart of a breast cancer risk management method in a breast cancer risk management system according to an embodiment of the present application.

[0016] Figure 2 It is a flowchart of another embodiment of the present application, which is located after collecting patient-related data and performing data preprocessing, and before using a preset and trained random forest algorithm to train a model.

[0017] Figure 3 It is a flowchart diagram in parallel with the step of "if not, analyzing and obtaining risk factors with a decline amplitude less than a preset amplitude as target risk factors" in another embodiment of the present application. DETAILED DESCRIPTION

[0018] The present application is further described in detail below in conjunction with the accompanying drawings.

[0019] Reference Figure 1 , is a breast cancer risk management system disclosed in the present application, comprising a breast monitoring device, at least one processor and at least one memory storing a computer program, wherein the at least one processor is configured to implement a breast cancer risk management method when executing the computer program, and the specific steps are as follows: Step S100, obtaining patient-related data collected by a breast monitoring device and performing data preprocessing.

[0020] Among them, patient-related data include real-time physiological data, daily activity data, environmental data and personalized data. Data preprocessing includes the following steps, as follows: 1. Clean the collected raw data, remove abnormal values, duplicate values ​​and missing values, and ensure the accuracy and completeness of the data. 2. Standardize the data to eliminate the dimensional differences between different data types so that the adjusted risk assessment model can be analyzed more accurately. 3. Integrate the processed data into a format suitable for input into the breast cancer risk value assessment model. Among them, it should be noted that the patient information, real-time physiological data, daily activity data, environmental data, personalized data and other related data involved in this application have been fully authorized by the patient before acquisition. In the process of data collection, use and processing, this application strictly abides by the laws, regulations and standards of relevant countries and regions in terms of data privacy protection, medical data management, etc. to ensure the security, legality and compliance of the data and effectively protect the personal rights and interests of patients.

[0021] Among them, the patient's real-time physiological data refers to the physiological index data of the patient's breast and its surrounding areas collected in real time by breast monitoring equipment. These data can reflect the health status of the patient's breast, including but not limited to the temperature, humidity, blood flow velocity, electrophysiological signals, etc. of the breast tissue. Among them, the patient's real-time physiological data is obtained as follows: through wearable breast monitoring equipment, such as smart bras or patch sensors, the patient's breast physiological data is collected in real time. These devices can use wireless communication technology to transmit data to the analysis platform.

[0022] Daily activity data refers to the patient's behavioral habits and activity patterns in daily life, such as exercise frequency, exercise intensity, sleep duration, work stress, etc. These data help analyze the impact of the patient's lifestyle on breast cancer risk. Daily activity data is obtained as follows: Record the patient's daily activity data through smartphone applications or wearable devices (such as smart watches, health bracelets, etc.). Patients can also manually enter some activity information, such as diet records, work hours, etc.

[0023] Environmental data refers to information about the patient's environment, including air quality, UV intensity, eating habits, living environment, etc. Environmental data is obtained as follows: Use environmental monitoring equipment, such as air quality detectors, UV sensors, etc., to collect data about the patient's environment. At the same time, patients can record some environmental information through mobile phone applications, such as changes in residence and eating habits.

[0024] Patient-specific data includes the patient's unique biomarkers, genotype, lifestyle habits, and family medical history.

[0025] Assume that patient B wears a smart breast patch, a smart watch, and uses an environmental monitoring mobile phone application in daily life. The smart breast patch collects temperature and density data of breast tissue in real time, and the smart watch records the patient's daily activity and sleep quality. At the same time, the mobile phone application monitors and records the air quality index and ultraviolet exposure time of the patient's environment. These data are regularly uploaded to the cloud server for preprocessing. The preprocessing process includes steps such as data cleaning, standardization, and integration to ensure the accuracy and availability of the data.

[0026] Step S200: inputting the patient-related data that has undergone data preprocessing into a trained breast cancer risk value assessment model that takes into account the patient's personalized data, and outputting a breast cancer risk value.

[0027] The formation process of the breast cancer risk value assessment model that has been trained and considers the personalized data of patients is as follows: 1. Collect a large number of patients' real-time physiological data, daily activity data, environmental data, personalized data and corresponding diagnosis results (whether they have breast cancer) as training data sets. Preprocess the data, including missing value processing, outlier processing, data standardization, etc. 2. Select features related to breast cancer risk from the collected data, such as specific physiological indicators, daily activity patterns, environmental factors, etc.; 3. Randomly select a part of the data as a training set and another part as a test set, and determine the number of decision trees in the random forest (for example, 100 trees). For each decision tree: randomly select a feature subset and split the data recursively until the predetermined stopping condition is reached (such as the number of samples contained in the leaf node is less than the threshold, the depth of the tree reaches the upper limit, etc.). 4. Use the training set to train each decision tree so that each tree can classify or predict the training data as accurately as possible. 5. Use the test set to evaluate the performance of the random forest model. Common evaluation indicators include accuracy, recall, F1 value, etc. If the model performance does not meet the requirements, adjust the model parameters (such as the number of decision trees, feature subset size, split conditions, etc.) and retrain until the model accuracy reaches 95% for three consecutive times to determine the formation of the breast cancer risk value assessment model.

[0028] Suppose we have collected a dataset of 1,000 women, 500 of whom were diagnosed with breast cancer and the other 500 were healthy people. The dataset includes real-time physiological data (such as breast temperature, blood flow velocity), daily activity data (such as weekly exercise, eating habits), and environmental data (such as the air quality index of the place of residence).

[0029] After analysis, we selected the following characteristics associated with breast cancer risk: physiological indicators: breast temperature, blood flow velocity, breast tissue density; daily activities: weekly exercise volume, frequency of high-fat diet; environmental factors: air quality index of residence, long-term ultraviolet exposure.

[0030] We used 800 women as training data and the remaining 200 as test data. We decided to use a random forest model with 100 decision trees. During training, each decision tree randomly selected a subset of features and recursively split the data based on these features. We set a minimum of 5 samples per leaf node and a maximum tree depth of 10.

[0031] After training, we used the test set to evaluate the model. The initial model had an accuracy of 90%, a recall of 85%, and an F1 value of 87.5%. To improve model performance, we adjusted the following parameters: increasing the number of decision trees to 150, reducing the size of the feature subset, and adjusting the splitting conditions to make the tree deeper but with fewer leaf node samples. After several rounds of adjustment and optimization, the model's accuracy reached 95% for three consecutive times. We believe that the model is stable enough and performs well. Now, this trained breast cancer risk value assessment model can receive real-time physiological data, daily activity data, and environmental data of new patients, and output a breast cancer risk value.

[0032] Step S300, using a preset and trained random forest algorithm training model to calculate the weight of each risk factor to the breast cancer risk value.

[0033] Among them, the process of calculating the weight of each risk factor to the breast cancer risk value is as follows: Calculate the contribution of each feature in the decision tree construction and give the feature importance score, and calculate the weight of each risk factor to the breast cancer risk value according to different feature importance scores. The process of calculating the contribution of each feature in the decision tree construction and giving the feature importance score is as follows: In the random forest algorithm, the contribution of each feature is usually calculated using the following common methods, such as average impurity reduction or average precision reduction. Take average impurity reduction as an example: When building a decision tree, each time a feature is selected for node splitting to reduce the impurity of the data (usually measured by the Gini coefficient or information entropy). The impurity reduction of each feature in all decision trees is averaged to obtain the average impurity reduction value of the feature as a measure of its contribution. The feature importance score can be obtained by standardizing or normalizing the contribution so that the sum of the importance scores of all features is 1 or in the range of 0 to 1.

[0034] Suppose the random forest model constructs 100 decision trees. For the feature of age: In these 100 decision trees, the total impurity reduction when age is used to split nodes is 500. For the feature of hormone level: The total impurity reduction when used to split nodes is 800. For the feature of exercise duration: The total impurity reduction is 200. For the feature of family history: The total impurity reduction is 600. Then calculate the feature importance score: The importance score of age = 500 / (500+800+200+600) = 0.2. The importance score of hormone level = 800 / 2100≈0.38. The importance score of exercise duration = 200 / 2100≈0.095. The importance score of family history = 600 / 2100≈0.286.

[0035] The process of calculating the weight of each risk factor for breast cancer risk is as follows: Taking the importance score as the weight, assuming that the risk factors we define are: high hormone level risk, family history risk, and low exercise duration risk. The weight of high hormone level risk = 0.38; the weight of family history risk = 0.286; the weight of low exercise duration risk = 0.095.

[0036] Through the above calculations, we obtained the weight of each risk factor for the breast cancer risk value, which can be used for subsequent assessment of breast cancer risk and determination of detection cycles.

[0037] Furthermore, after the random forest model outputs the patient's breast cancer risk value according to the calculated weights of each risk factor, the SHAP analysis method can be introduced. The specific steps are as follows: Step S301, based on the SHAP analysis method, attribute and explain the breast cancer risk value output by the model. SHAP analysis is based on cooperative game theory, and generates the contribution of each risk factor to the individual risk score by calculating the marginal contribution of each feature in the model prediction. For example, if a patient's breast cancer risk score is 80%, the SHAP analysis may show: high hormone levels (+35%), family history (+25%), low exercise time (+15%), age factors (+5%) and other factors contribute specifically to the risk score.

[0038] Step S302, dynamically adjust the SHAP analysis display threshold according to the patient's real-time risk level. The system divides the patient into real-time risk levels according to the breast cancer risk value output by the model. For high-risk patients (risk value ≥ 0.7), the SHAP analysis display threshold is reduced from the default 10% to 5%. This is because there may be multiple potential factors that affect the risk of breast cancer in high-risk patients. Lowering the display threshold can fully present these potential influencing factors and avoid missing important information. For example, factors such as "long-term mental stress" and "bad emotional state" that were not displayed due to their small impact ratio can be presented under dynamic threshold adjustment. For low-risk patients (risk value < 0.3), the SHAP analysis display threshold is increased to 15%. The main goal of low-risk patients is to prevent the occurrence of breast cancer. Increasing the display threshold can help focus on the most critical factors that affect the risk, such as "high hormone levels" and "family history", so as to formulate prevention strategies more accurately.

[0039] Step S303, based on the local linear model, simulate the impact of multi-scenario intervention on the predicted risk value. For high-risk patients: Combined with the key influencing factors determined by the SHAP analysis in step S325, such as long-term mental stress, high hormone levels, etc., use the local linear model to simulate the comprehensive intervention scenario. For example, psychological counseling is set as professional psychological counseling twice a week, 50 minutes each time; work and rest adjustment requires patients to ensure 7-8 hours of high-quality sleep every day and fall asleep before 11 o'clock in the evening; drug intervention uses specific regulating drugs according to the patient's specific hormone levels and other conditions; exercise program is 3-4 times a week, 30-45 minutes each time aerobic exercise, such as jogging, yoga, etc. The model predicts the decrease in the patient's breast cancer risk value after these intervention measures are implemented for 3-6 months, such as specific quantitative results such as from 0.8 to 0.6. For low-risk patients: If the SHAP analysis shows that unreasonable diet structure and insufficient exercise are key factors, the local linear model simulates the corresponding prevention scenario. For example, patients reduce their intake of high-sugar foods by 30%, increase their intake of foods rich in dietary fiber, and increase outdoor exercise by 2 hours per week. The model predicts the change in the patient's risk value after 6-12 months, such as remaining in the low-risk range from 0.2 or further decreasing to 0.15.

[0040] Step S304: Output the contribution and quantitative health advice to the doctor's terminal simultaneously. The system outputs the contribution of each risk factor to the individual risk score obtained by SHAP analysis, as well as the quantitative health advice generated after the local linear model simulates multi-scenario intervention, such as specific psychological counseling programs, drug dosages and exercise plans for high-risk patients, and detailed dietary adjustments and exercise increase advice for low-risk patients, to the doctor's terminal in the form of a clear and intuitive report.

[0041] Step S400, analyzing and determining the breast cancer detection cycle according to the corresponding relationship between the weight range of the breast cancer risk value of the specific risk factor, the risk value assessment range of the breast cancer risk value, and the detection cycle.

[0042] Among them, specific risk factors usually refer to those factors that have a significant impact on the risk of breast cancer, including but not limited to the following: 1. Physiological factors, such as abnormal hormone levels (such as estrogen and progesterone), specific physiological characteristics of breast tissue, etc. 2. Genetic factors, such as specific gene mutations (such as BRCA1, BRCA2 gene mutations), multiple relatives with breast cancer in the family history, etc. 3. Lifestyle-related factors, such as long-term smoking habits, excessive drinking, lack of exercise, unhealthy diet (high-fat, high-calorie diet), etc. 4. Environmental factors, such as long-term exposure to radiation environment, contact with certain chemicals, etc. 5. Fertility-related factors, such as older age of first birth, no childbirth, short breastfeeding time, etc.

[0043] Determine the weight range for specific risk factors: First, according to the previous calculations and analysis, set different weight ranges. For example, divide the weight range into low (0-0.2), medium (0.2-0.5) and high (0.5-1).

[0044] The breast cancer risk value assessment range is determined as follows: Different risk value assessment ranges are also set, such as low risk (0-0.3), medium risk (0.3-0.7) and high risk (0.7-1).

[0045] The corresponding relationship between the weight range, risk value assessment range and detection cycle is established as follows: low weight and low risk value, the detection cycle is once every 2 years. Medium weight and medium risk value, the detection cycle is once a year. High weight and high risk value, the detection cycle is once every six months.

[0046] Step S500: forming a patient's breast cancer related detection time planning table according to the analyzed and determined breast cancer detection cycle.

[0047] The detection cycle is determined mainly based on the breast cancer detection cycle analyzed in step S400 to clarify the frequency of breast detection required by the patient, such as weekly, monthly or quarterly.

[0048] The timetable is developed as follows: Based on the determined testing cycle, specific testing time points are planned for the patient. These time points can be a list of specific dates or a time pattern that repeats according to the cycle. The planned testing time points are then organized into a clear timetable that includes the test date, test time (optional), test location (such as hospital, clinic or home self-test) and any additional precautions or reminders.

[0049] Step S600, forming quantitative suggestion information according to the corresponding relationship between the weight interval of each risk factor for the breast cancer risk value and the quantitative suggestion association, and sending the formed quantitative suggestion information to the terminal held by the patient.

[0050] In this step, multiple weight intervals are pre-set, and corresponding quantitative recommendations are made for each weight interval. Assume that the following three weight intervals are set: low weight interval (0-0.2), medium weight interval (0.2-0.5), and high weight interval (0.5-1).

[0051] Here are some examples of specific risk factors and their corresponding quantitative recommendations: Risk factor "hormone imbalance": If the weight falls into the low weight range: It is recommended that patients maintain a regular work and rest schedule, avoid staying up late, and monitor their hormone levels once a month. If the weight falls into the medium weight range: It is recommended that patients maintain a regular work and rest schedule, conduct professional hormone level testing every three months, and adjust their diet according to the doctor's advice to reduce the intake of hormone-rich foods. If the weight falls into the high weight range: It is recommended that patients seek medical attention immediately, follow the doctor's advice for drug treatment to regulate hormone levels, conduct hormone level testing once a week, and strictly control their diet.

[0052] Risk factor "excessive drinking": If the weight falls into the low weight range: It is recommended that the patient control the amount of drinking, and drink no more than twice a month, and no more than one glass each time. If the weight falls into the medium weight range: It is recommended that the patient gradually reduce the amount of drinking in the next two months until the number of drinks per month does not exceed once. If the weight falls into the high weight range: It is recommended that the patient stop drinking immediately, join a mutual aid group for alcohol abstinence, and receive professional supervision.

[0053] Step S700, collecting the latest data of the patient within a preset period after the quantitative recommendation is sent, and reanalyzing the breast cancer risk value and the weight of each risk factor to the breast cancer risk value based on the latest data.

[0054] The preset period can be set according to the actual situation, such as three months or six months. During this period, the latest data of the patient will be collected through various channels.

[0055] The collected data may include but are not limited to the following aspects: 1. The patient's daily physiological indicators, such as body temperature, blood pressure, heart rate, etc. 2. Changes in the patient's living habits, such as exercise frequency, sleep duration, diet structure, etc. 3. The patient's emotional state and psychological stress assessment results.

[0056] Based on the latest data collected, the previously established models and algorithms were reapplied to analyze the risk value of breast cancer and the weight of each risk factor to that risk value.

[0057] Step S800, analyzing whether the decrease range of the breast cancer risk value generated by each risk factor individually reaches or exceeds the preset range. If yes, execute step S900; if no, execute step Sa00.

[0058] First, you need to clearly define the preset reduction for each risk factor. For example, for the risk factor of "chronic stress", the preset reduction may be 20%.

[0059] Then, the actual decrease in the breast cancer risk value caused by each risk factor after a period of intervention (such as three months) is calculated. For example, after three months of psychological adjustment and relaxation training, the impact of the risk factor of "long-term excessive stress" on the breast cancer risk value is re-evaluated. The original risk value caused by this factor was 0.3, which became 0.21 after intervention, and the actual decrease was 30%, exceeding the preset 20%. For another example, for the risk factor of "staying up late for a long time", the preset decrease was 15%. After re-evaluation, it was found that the original risk value of 0.25 changed to 0.22, and the actual decrease was 12%, which did not reach the preset 15%.

[0060] Step S900, maintaining the original breast cancer detection cycle unchanged. The original breast cancer detection cycle is determined according to the corresponding relationship between the weight range of the breast cancer risk value of the specific risk factor, the risk value assessment range of the breast cancer risk value, and the detection cycle.

[0061] Step Sa00, analyze and obtain the risk factors whose decline is less than the preset range, as the target risk factors, and calculate the ratio of the decline of the target risk factor to the preset range, according to the ratio of the decline of the target risk factor to the preset range and the weight of the target risk factor to the breast cancer risk value, calculate the adjustment coefficient of all target factors with respect to the breast cancer detection cycle, and then calculate the adjusted breast cancer detection cycle according to the adjustment coefficient and the original breast cancer detection cycle.

[0062] By analyzing the decrease in breast cancer risk caused by each risk factor, we identify those risk factors whose decrease is less than the preset value as target risk factors. For each target risk factor, we calculate the ratio of its actual decrease to the preset value. This ratio reflects the gap between the risk reduction effect of the risk factor and the expected effect. Then, we will use this ratio and the weight of the risk factor to the breast cancer risk value to calculate an adjustment coefficient for the breast cancer detection cycle.

[0063] For example: Suppose we have a risk factor called "unhealthy eating habits". After a period of quantitative recommendations, we find that the decrease in breast cancer risk caused by this risk factor does not reach the preset range.

[0064] We first determine "unhealthy eating habits" as the target risk factor. Assuming that the preset range is a 20% decrease in the risk value, but the actual decrease is only 10%, the ratio of the decrease to the preset range is 0.5. If the weight of "unhealthy eating habits" on the breast cancer risk value is 0.3, then we can calculate an adjustment factor. This adjustment factor may be the product of the ratio of the decrease to the preset range and the risk factor weight, that is, 0.5*0.3=0.15. This adjustment factor means that due to the poor improvement effect of the risk factor of "unhealthy eating habits", we may need to shorten the adjusted breast cancer detection cycle by 15% (or make other corresponding adjustments according to the specific situation) in order to monitor the patient's health status more frequently.

[0065] Step Sb00, updating the patient's breast cancer related detection time planning table according to the finally determined breast cancer detection cycle.

[0066] A breast cancer risk management system further includes steps after step S300, which are as follows: Step S3a0, analyzing whether complete data of patients reaching or exceeding the preset number of cases have been collected. These data should comprehensively cover patient personalized data (such as family cancer history, gene test results, etc.), real-time physiological data (such as elastic imaging data of breast tissue, tumor marker levels in blood, etc.), daily activity data (such as daily dietary calorie intake, weekly physical labor hours, etc.) and environmental data (such as electromagnetic radiation intensity of long-term living environment, air pollutant concentration, etc.).

[0067] Among them, the preset number of cases can be 1,200 cases.

[0068] Step S3b0: if no, continue with the subsequent steps.

[0069] Step S3c0, if yes, then select variables related to breast cancer risk, use stepwise regression method to screen variables, establish a preliminary multiple regression model, divide the preprocessed data into a training set and a validation set, train the multiple regression model on the training set, adjust the performance of the multiple regression model by adjusting the parameters of the multiple regression model, until the accuracy of the multiple regression model on the validation set reaches the preset accuracy within a preset number of consecutive times, determine that the multiple regression model training is completed, and check the regression coefficient corresponding to each risk factor, calculate the sum of the absolute values ​​of all regression coefficients, and calculate the ratio of the absolute value of the regression coefficient of different risk factors to the sum of the absolute values ​​of all regression coefficients, and determine the weight of each risk factor for the breast cancer risk value.

[0070] First, we need to select those variables that are considered to be closely related to the risk of breast cancer from the collected and pre-processed data. For example, they may include the patient's age, hormone levels (such as estrogen, progesterone), age of menarche, number of births, duration of breastfeeding, BMI (body mass index), eating habits (whether to prefer high-fat foods), exercise habits (number of hours of exercise per week), family history (number of first-degree relatives with breast cancer), etc.

[0071] Next, we use the stepwise regression method to screen variables. Stepwise regression will gradually introduce or eliminate variables according to certain criteria to find the optimal combination of variables.

[0072] Then, a preliminary multivariate regression model is established based on the selected variables. Assuming that we have selected three variables: age, hormone level, and family history, the model may be expressed as: breast cancer risk value = β0 + β1*age + β2*hormone level + β3*family history.

[0073] After that, the preprocessed data is randomly divided into a training set and a validation set. For example, if we have data on 100 patients, we may use 70 of them as training sets and 30 as validation sets. When training a multivariate regression model on the training set, we will constantly adjust the model's parameters, such as learning rate, regularization parameters, etc., just like adjusting the engine and parts of a car, to optimize the model's performance. When evaluating the model, we will use the validation set to check the model's accuracy. If the model's accuracy on the validation set reaches the preset 80% for, say, three consecutive times, we consider the multivariate regression model training complete.

[0074] After the training is completed, we look at the regression coefficients corresponding to each risk factor. Suppose the regression coefficient of age is 0.3, the regression coefficient of hormone level is 0.5, and the regression coefficient of family history is 0.2.

[0075] Then calculate the sum of the absolute values ​​of all regression coefficients, that is, |0.3|+|0.5|+|0.2|=1.

[0076] Finally, the ratio of the absolute value of the regression coefficient of different risk factors to the total is calculated to determine the weight of each risk factor for breast cancer risk. The weight of age is |0.3| / 1=0.3, the weight of hormone level is |0.5| / 1=0.5, and the weight of family history is |0.2| / 1=0.2.

[0077] Step S3d0, input the risk value weight output by the random forest model and the risk value weight output by the multivariate regression model as new features into the trained decision tree fusion model, and output the comprehensive risk weight assessment result.

[0078] Among them, the decision tree fusion model is a model that combines multiple decision trees to improve prediction performance and stability.

[0079] The construction of decision trees is as follows: a certain number of decision trees are constructed, for example, 8 decision trees are constructed. Each decision tree is grown based on a randomly selected feature subset and a random split point when it is constructed.

[0080] The training is as follows: 1. Each decision tree is trained with the existing training data. 2. During the training process, by adjusting the split features and split points, each decision tree can reasonably classify or predict the input risk value weight features.

[0081] The fusion mechanism is as follows: When the risk value weights output by the random forest model are RF_weight1, RF_weight2, ..., RF_weightn, and the risk value weights output by the multivariate regression model are MR_weight1, MR_weight2, ..., MR_weightn, their average value is calculated as the fused weight. The fused weight is as follows: Fusion_weighti=(RF_weighti+MR_weighti) / 2.

[0082] The output of the comprehensive risk weight assessment result is as follows: The fusion weights of all risk factors are calculated comprehensively. Finally, a comprehensive risk weight assessment result is output, which reflects the average performance of the two models.

[0083] First, we have obtained a set of risk value weights through the random forest model, and also calculated another set of risk value weights through the multivariate regression model. Next, these two sets of weights are input as new features into the previously trained decision tree fusion model. This decision tree fusion model will comprehensively consider the information of these two sets of weights, and analyze and process the input weight features through its internal decision rules and calculation logic.

[0084] .Assume that the risk value weights output by the random forest model are: Factor A: 0.4; Factor B: 0.3; Factor C: 0.2.

[0085] The risk value weights output by the multivariate regression model are: Factor A: 0.3; Factor B: 0.4; Factor C: 0.3.

[0086] These weights are input as new features into the decision tree fusion model. After calculation and analysis, the decision tree fusion model may output the following comprehensive risk weights: Factor A: 0.35; Factor B: 0.38; Factor C: 0.27.

[0087] Step S3e0, analyzing and determining the breast cancer detection cycle according to the corresponding relationship between the comprehensive risk weight output by the decision tree fusion model, the risk value assessment range into which the breast cancer risk value falls, and the detection cycle.

[0088] The comprehensive risk weight output by the decision tree fusion model will be divided into different intervals, each of which corresponds to a specific breast cancer risk value assessment range.

[0089] Assume that we set the following intervals and corresponding risk assessment ranges and detection cycles: Low-risk range: comprehensive risk weight <0.3; risk value assessment range: 0-0.2; detection cycle: once every 3 years; Medium risk range: 0.3<=comprehensive risk weight<0.7; risk value assessment range: 0.2-0.6; Inspection cycle: once every 2 years; High-risk range: comprehensive risk weight>=0.7; risk value assessment range: 0.6-1; detection cycle: once a year.

[0090] For example, if the comprehensive risk weight output by the decision tree fusion model is 0.25, then it is in the low risk range. The corresponding breast cancer risk value assessment range is 0-0.2, and the recommended detection cycle is once every 3 years. For another example, if the comprehensive risk weight is 0.8, it is in the high risk range. The corresponding breast cancer risk value assessment range is 0.6-1, and the detection cycle should be set to once a year.

[0091] Reference Figure 2 A breast cancer risk management system further includes steps after collecting patient-related data and preprocessing the data, and before using a preset and trained random forest algorithm to train a model, specifically as follows: Step SA00, dividing the patients into different age groups, including a youth group, a middle-aged group and an elderly group.

[0092] First, it is necessary to clarify the specific criteria for dividing age groups. Usually, the youth group can be defined as those aged between 18 and 40, the middle-aged group as those aged between 41 and 60, and the elderly group as those aged 61 and above. Then, obtain the patient's age information. This can be obtained through the patient's medical records, questionnaires, or direct inquiries. According to the age information obtained, the patient is classified into the corresponding age group.

[0093] Step SB00, if the patient is in the young group or the middle-aged group, maintain the original setting.

[0094] Among them, the original setting is to continue to execute step S300, that is, to use the preset and trained random forest algorithm to train the model.

[0095] Step SC00: If the patient is in the elderly group, variables related to breast cancer risk are selected, and the stepwise regression method is used to screen the variables and establish a preliminary multivariate regression model.

[0096] Variables associated with breast cancer risk include, but are not limited to, the following: 1. Physiological indicators, such as changes in hormone levels (estrogen, progesterone, etc.) and breast tissue density. 2. Health status, such as whether you have chronic diseases such as diabetes and hypertension. 3. Lifestyle factors, such as diet (whether it is high in oil and fat), exercise habits (whether you exercise regularly and the intensity of exercise). 4. Genetic factors, specific gene mutations or details of family medical history.

[0097] The process of variable screening using the stepwise regression method is as follows: First, all possible variables are included in the model. Then, according to certain statistical criteria (such as AIC, BIC, etc.), the variables that do not contribute significantly to the model are gradually eliminated. Finally, the variables that have a significant impact on the prediction of breast cancer risk are left. Based on the screened variables, a preliminary multivariate regression model is established. The general form of the multivariate regression model is: Breast cancer risk value = β0 + β1X1 + β2X2 + … + βn*Xn, where X1, X2, etc. are screened variables, β0 is the intercept, and β1, β2, etc. are regression coefficients.

[0098] Suppose we have the following data of an elderly patient: Patient D: Age: 75 years old; Estrogen level: 25pg / mL; Progesterone level: 10pg / mL; Suffering from diabetes (yes / no): yes; Diet structure (healthy / unhealthy): unhealthy; Weekly exercise duration: 2 hours; There is 1 relative in the family with breast cancer.

[0099] First, we selected variables that may be associated with breast cancer risk, such as estrogen levels, progesterone levels, diabetes, diet, weekly exercise minutes, and family history.

[0100] Using the stepwise regression method, it may be found that the contribution of diet structure and weekly exercise duration to the model is not significant and is eliminated.

[0101] The final multivariate regression model may be: breast cancer risk value = β0 + β1 estrogen level + β2 progesterone level + β3 whether or not there is diabetes + β4 family history.

[0102] Through this process, a preliminary multivariate regression model was established for elderly patients to facilitate further analysis and determination of risk weights and detection cycles.

[0103] Step SD00, divide the preprocessed data into a training set and a validation set, train the multivariate regression model on the training set, adjust the performance of the multivariate regression model by adjusting the parameters of the multivariate regression model, until the accuracy of the multivariate regression model on the validation set reaches the preset accuracy for a consecutive preset number of times, determine that the training of the multivariate regression model is completed, and check the regression coefficient corresponding to each risk factor.

[0104] First, the preprocessed data is randomly divided. Usually, a certain ratio can be used, such as 70% of the data as a training set and 30% of the data as a validation set. Then, the training set is used to train the multivariate regression model. During the training process, the performance of the model is optimized by adjusting the parameters of the model, such as the learning rate, regularization parameters, etc. After each round of training, the validation set is used to evaluate the accuracy of the model. If the accuracy of the model on the validation set reaches the preset accuracy (for example, 80%) for a preset number of consecutive times (for example, 5 times), the multivariate regression model training is considered complete.

[0105] Next, check the regression coefficients corresponding to each risk factor in the trained model. These regression coefficients reflect the degree of influence of each risk factor on the breast cancer risk value.

[0106] Step SE00, according to the regression coefficient corresponding to each risk factor, calculate the sum of the absolute values ​​of all regression coefficients, and calculate the ratio of the absolute value of the regression coefficient of different risk factors to the sum of the absolute values ​​of all regression coefficients to determine the weight of each risk factor for the breast cancer risk value.

[0107] After obtaining the regression coefficients corresponding to each risk factor in the multivariate regression model, first calculate the sum of the absolute values ​​of all regression coefficients. Suppose we have three risk factors, and their regression coefficients are: β1=0.2, β2=-0.3, β3=0.4. Then calculate their absolute values ​​first: |β1|=0.2, |β2|=0.3, |β3|=0.4.

[0108] Then calculate the sum of the absolute values: |β1|+|β2|+|β3|=0.2+0.3+0.4=0.9.

[0109] Next, the ratio of the absolute value of each risk factor regression coefficient to the sum of the absolute values ​​of all regression coefficients was calculated.

[0110] For risk factor 1: ratio = |β1| / (|β1|+|β2|+|β3|) = 0.2 / 0.9≈0.22.

[0111] For risk factor 2: ratio = |β2| / (|β1|+|β2|+|β3|) = 0.3 / 0.9≈0.33.

[0112] For risk factor 3: ratio = |β3| / (|β1|+|β2|+|β3|) = 0.4 / 0.9≈0.44.

[0113] The ratio calculated in this way determines the weight of each risk factor for breast cancer risk value.

[0114] Step SF00, analyzing and determining the breast cancer detection cycle according to the corresponding relationship between the weight range of the breast cancer risk value of the specific risk factor, the risk value assessment range of the breast cancer risk value, and the detection cycle.

[0115] First, it is necessary to clarify the different ranges of weights of specific risk factors for breast cancer risk values. For example, the low weight range can be set to 0-0.2, the medium weight range to 0.2-0.5, and the high weight range to 0.5-1.

[0116] At the same time, the assessment range of breast cancer risk value should also be determined, such as the low risk range is 0-0.3, the medium risk range is 0.3-0.7, and the high risk range is 0.7-1.

[0117] Then, the weight of the specific risk factor to the breast cancer risk value and the breast cancer risk value calculated in the previous steps are obtained.

[0118] According to the range of weights and the range of risk values, a preset corresponding relationship is searched to determine the breast cancer detection cycle.

[0119] For example, if the weight is in the low weight range and the risk value is in the low risk range, the corresponding detection cycle may be once every 3 years; if the weight is in the medium weight range and the risk value is in the medium risk range, the corresponding detection cycle may be once every 2 years; if the weight is in the high weight range and the risk value is in the high risk range, the corresponding detection cycle may be once a year.

[0120] Reference Figure 3 , a breast cancer risk management system also includes other steps in parallel with the step of "if not, analyzing and obtaining risk factors with a decrease range less than a preset range as target risk factors", which are as follows: Step Sh00, patients whose breast cancer risk values ​​generated by risk factors alone have not decreased by a preset amount are considered as patients who need to be followed up; First, the system analyzes the reduction in breast cancer risk due to each risk factor. This is determined by comparing it to the risk value of the last assessment. If a risk factor does not reduce the risk by the preset amount, then the risk factor is considered a target for focus. For those patients with risk factors that do not achieve the preset reduction, the system will mark them as those who need to be followed up.

[0121] Suppose we have a patient whose BMI (body mass index) is too high, which is a significant breast cancer risk factor. After a period of quantitative recommendations, we find that her BMI does not reduce her breast cancer risk by as much as we expected. We first identify BMI as the target risk factor for this patient. Since the reduction in BMI did not reach the expected level, we mark this patient as someone who needs to be followed up.

[0122] For this patient, we can give specific quantitative suggestions to help her reduce her BMI and breast cancer risk more effectively. For example, we recommend that she do at least 150 minutes of moderate-intensity aerobic exercise per week, such as brisk walking, swimming, or cycling. Adjust her diet structure, reduce the intake of high-calorie, high-fat foods, and increase the intake of vegetables, fruits, and whole grains. Monitor BMI changes regularly and record them to observe progress.

[0123] Step Si00, start the follow-up process, contact the patient through intelligent voice method, and obtain the patient's execution results on quantitative advice information, including exercise duration and diet adjustment.

[0124] After identifying the patient who needs follow-up, the system automatically starts the intelligent voice follow-up function. This function uses advanced speech recognition and speech synthesis technology to communicate with the patient naturally and fluently. The system will first send an intelligent voice greeting to the patient, such as "Hello, this is the Breast Cancer Risk Management Follow-up System. Are you [patient name]?" After the patient confirms his identity, the follow-up process will continue.

[0125] Next, the system will ask the patient about the implementation of the previously sent quantitative advice information, especially the specific data on exercise duration and diet adjustment. For example, the system may ask: "How long did you exercise on average per day in the past week?" or "Have you adjusted your diet structure as recommended to reduce the intake of high-fat foods?" The patient's answers will be automatically recorded by the system and converted into an analyzable data format. These data will be used to evaluate the patient's performance, as well as the adjusted risk assessment and recommended adjustments.

[0126] Step Sj00: If the patient's execution result of the quantitative suggestion information does not match the quantitative content mentioned in the quantitative suggestion information, the patient is advised to execute again according to the quantitative content mentioned in the quantitative suggestion information.

[0127] Compare the implementation results provided by the patient with the original quantitative recommendations to check whether the patient has made corresponding health behavior adjustments according to the recommendations. If the patient's implementation results do not match the quantitative recommendations, it means that the patient has failed to meet the recommended standards in some aspects. For patients who do not meet the quantitative recommendations, we will re-emphasize the importance of quantitative recommendations and recommend that the patient strictly follow the quantitative recommendations again.

[0128] Step Sk00, if the patient's execution result on the quantitative advice information is consistent with the quantitative content mentioned in the quantitative advice information, then analyze the quantitative content related to health that is not mentioned in the quantitative advice information, and provide suggestions for other quantitative content to be added to the quantitative advice information to form new quantitative advice information, and send it to the terminal held by the patient again.

[0129] If the patient has already implemented the previous quantitative recommendations well, the system will further analyze other health-related risk factors that may not have been covered by the quantitative recommendations before. Based on the above analysis, the system will generate new quantitative recommendations to further optimize the patient's health behavior and reduce the risk of breast cancer.

[0130] Taking the new risk factor of "lack of sleep" as an example, the corresponding quantitative recommendations can be: ensure 7-9 hours of high-quality sleep every night. Avoid using electronic devices before going to bed to reduce the impact of blue light on sleep quality. Create a comfortable sleeping environment and make sure that the mattress and pillow are suitable for your body curve and sleeping habits. Record the sleep time and quality every night for self-monitoring and continuous improvement. By implementing these new quantitative recommendations, patients can further reduce the risk of breast cancer and improve their overall health.

[0131] Update the patient's breast cancer related testing schedule based on the finalized breast cancer testing cycle, including: Step S410, obtaining the patient's electronic medical record related information and extracting the key factors affecting the breast cancer detection results, wherein the key factors affecting the breast cancer detection results include ongoing treatment plans for other diseases, specific drugs used, recent surgical history, and the severity of the patient's underlying disease.

[0132] From the collected electronic medical record information, the key factors that affect the breast cancer detection results are extracted. These key factors include but are not limited to: 1. Ongoing treatment for other diseases: Certain treatment plans may affect the detection results of breast cancer, especially those involving hormone therapy or immunotherapy. 2. Specific medications used: Certain medications, such as hormone drugs, may interfere with the detection of breast cancer. 3. Recent surgical history: Surgery may cause changes in the physiological environment inside the body, thereby affecting the test results. 4. The severity of the patient's underlying disease: The severity of the underlying disease may affect the patient's physiological state, which in turn affects the test results.

[0133] Step S420, analyzing and determining the total delay time caused by the key factors affecting the breast cancer detection result according to the corresponding relationship between the preset influence degree value range of the key factors affecting the breast cancer detection result and the delay time.

[0134] First, the ranges of different impact values ​​and the corresponding delay times are pre-set. For example, the impact values ​​are divided into three ranges: mild (0-0.3), moderate (0.3-0.6), and severe (0.6-1).

[0135] For the key factor of "ongoing treatment for other diseases", if the impact of the treatment on the body is assessed to be mild, the corresponding delay time may be set at 1 month; if it is a moderate impact, the delay time may be 2 months; if it is a severe impact, it may be 3 months.

[0136] Assuming that a patient is undergoing a treatment for another disease that has a moderate impact on the body, in this step, based on the preset correspondence, it will be determined that the delay caused by this treatment is 2 months.

[0137] For example, for the key factor of "specific medications used", if the impact of the medication is assessed to be mild, the delay time may be half a month; moderate impact is 1 month; severe impact is 1.5 months.

[0138] If a patient was taking a specific medication that was assessed as having severe effects, the resulting delay was determined to be 1.5 months.

[0139] Step S430, based on the overall time of delay caused by the key factors affecting the breast cancer detection result and the analyzed and determined breast cancer detection cycle, the breast cancer detection cycle is re-analyzed and the patient's breast cancer related detection time planning table is re-formed.

[0140] In this step, the overall time of the delay caused by the key factors determined in the previous step (S420) and the original breast cancer detection cycle that has been analyzed and determined before are first obtained. For example, the original breast cancer detection cycle is to conduct a test every 8 months, but the overall delay time caused by the key factors calculated by step S420 is 2 months. Next, the delay time will be added to the original detection cycle for re-analysis. Taking the above example, the recalculated detection cycle becomes a test every 10 months. Then, based on this re-determined detection cycle, a detailed breast cancer-related detection time plan is formulated for the patient.

[0141] A breast cancer risk management system further includes steps after updating the patient's breast cancer related detection time planning table according to the finally determined breast cancer detection cycle, which are as follows: Step Sd00, collecting specific information of the patient's breast cancer related tests and performing data preprocessing.

[0142] The specific information of breast cancer related tests includes but is not limited to breast cancer pathology data, physical condition information, basic disease information and breast cancer risk assessment results. Data preprocessing. This may include the following operations: 1. Data cleaning, removing duplicate, erroneous or incomplete data. 2. Data standardization: converting data from different sources and formats into a standard format that can be compared and analyzed. 3. Data coding: coding some non-numerical data for the purpose of adjusted analysis and calculation.

[0143] Step Se00, selecting features related to auxiliary treatment decision-making from the specific information collected from the patient's breast cancer related tests.

[0144] Characteristics relevant to adjuvant therapy decisions include tumor size, grade, patient age, and severity of underlying disease.

[0145] In this step, the detailed information collected from a large amount of patient breast cancer-related tests is carefully screened and analyzed to extract the key features that are most relevant to adjuvant treatment decisions.

[0146] First, we look at the characteristics of the tumor itself, such as its size. If the tumor is large, it may mean that more aggressive treatment is necessary; if the tumor is small, different treatment considerations may be necessary. The grade of the tumor is also an important characteristic.

[0147] High-grade tumors are generally more malignant and aggressive, which affects the choice of treatment options.

[0148] The patient's age is another key factor. Younger patients may tolerate treatment better, while older patients may need a gentler approach or one that takes into account their overall body condition.

[0149] The severity of the underlying disease cannot be ignored. If the patient also suffers from serious underlying diseases such as heart disease and diabetes, this will limit the application of certain treatment methods, or require special attention to the prevention and management of complications during treatment.

[0150] Step Sf00, input the features related to the auxiliary treatment decision into the pre-trained auxiliary treatment decision model, and output the result of whether to use auxiliary treatment and the treatment plan if auxiliary treatment is needed. The treatment plan if auxiliary treatment is needed includes surgery and drug provision plan.

[0151] In this step, the screened features related to adjuvant therapy decisions, such as tumor size, grade, patient age, severity of underlying disease, etc., will be input into an adjuvant therapy decision model that has been pre-trained with a large amount of data.

[0152] This model uses complex algorithms and machine learning techniques to comprehensively analyze and calculate the input features. For example, the model may assess the patient's tolerance to surgery based on his or her age and the severity of the underlying disease, and determine the malignancy and risk of progression of the tumor based on the size and grade of the tumor.

[0153] Then, the model will output two key results. One is the judgment of whether adjuvant treatment is needed. If the model believes that the patient's condition is more serious and simple initial treatment is not enough to control the condition, it will output the result that adjuvant treatment is needed; conversely, if the patient's condition is relatively good, it may output the conclusion that adjuvant treatment is not needed. Second, if adjuvant treatment is needed, the model will give a specific treatment plan, including a surgical plan and a drug delivery plan. The surgical plan may specify the type of surgery (such as breast-conserving surgery or total resection), the scope and difficulty of the surgery, etc. The drug delivery plan will list the types of drugs suitable for the patient (such as chemotherapy drugs, targeted drugs, etc.) as well as the corresponding dosages and frequencies of use.

[0154] Step Sg00, if the output is the result of needing auxiliary treatment and the treatment plan under the need for auxiliary treatment is surgery, the patient's medical imaging data is obtained, and the contours of the tumor and surrounding tissues are extracted using an image segmentation algorithm. Based on the extracted contours of the tumor and surrounding tissues, a path planning algorithm is used to simulate the entry path and operating range of the surgical instrument, and the surgical path and resection range are determined, thereby forming a surgical planning plan.

[0155] First, obtain the patient's medical imaging data, which may include the results of various imaging examinations such as magnetic resonance imaging (MRI), computed tomography (CT), and ultrasound.

[0156] Then, advanced image segmentation algorithms, such as convolutional neural networks (CNNs) in deep learning, are used to process medical images. Through the trained model, the tumor and its surrounding key tissue structures, such as blood vessels, nerves and other important organs, are automatically identified and segmented.

[0157] Based on the results of image segmentation, the contours of the tumor and surrounding tissues are accurately extracted. Using these contour data, a three-dimensional (3D) model is constructed to more intuitively display the location of the tumor and its relative relationship with the surrounding tissues.

[0158] Next, based on the extracted contour information, a path planning algorithm such as an ant colony algorithm or an A* algorithm is used to simulate the possible entry paths of surgical instruments with the goal of minimizing surgical trauma and maximizing tumor resection effects. Based on tumor characteristics and path planning results, combined with the size and operation requirements of the surgical instruments, the operating range of the surgical instruments in the tumor area is determined. The final surgical path and resection range are determined by comprehensively considering the path planning and operating range to ensure that the surrounding normal tissues are protected to the maximum extent while the tumor is completely removed.

[0159] Based on the above simulation results, the specific surgical path and resection range are finally determined. For example, determining which part to cut from can minimize damage to surrounding normal tissues while ensuring complete removal of the tumor; determining how far the resection boundary should be from the edge of the tumor to ensure that cancer cells are completely removed without excessive removal of healthy tissue.

[0160] For example, if the patient's tumor is located deep in the breast and close to important blood vessels and nerves, image segmentation and path planning may determine a surgical path that enters from the side and bypasses important structures, and determine a relatively conservative resection range that can ensure tumor removal.

[0161] For example, for a tumor with clear boundaries and a shallow location, the surgical approach may be to cut directly from the top, and the resection range can also be relatively small to preserve normal breast tissue and function to the greatest extent.

[0162] Step Sh00, if the output is the result of needing auxiliary treatment and the treatment plan under the need for auxiliary treatment is medication, the patient's genetic test results, drug sensitivity data and surgical plan are obtained, and based on the patient's genetic test results, drug sensitivity data and surgical plan, a preset drug matching algorithm is used to screen out suitable drug types and dosages as a drug provision plan.

[0163] Description of drug matching algorithm: Based on big data analysis and machine learning technology, by integrating the patient's genetic test results, drug sensitivity data and surgical plan, the most suitable drug type and dosage are intelligently matched. The algorithm will first analyze the genetic test results, identify genetic variants related to drug response, and predict the patient's metabolic rate and efficiency of different drugs. Then, combined with drug sensitivity data, evaluate the patient's tolerance and response to different drugs to avoid choosing drugs that may cause serious side effects or be ineffective. At the same time, consider the impact of the surgical plan on medication, such as medication adjustments during the postoperative recovery period. Finally, the algorithm will combine all the above information, perform intelligent matching through preset rules and models, and recommend the most suitable drug type and dosage for the patient.

[0164] According to the output of the drug matching algorithm, a personalized drug provision plan is formulated. The plan will list in detail key information such as the drug name, dosage, frequency of use, and time of use. At the same time, the plan will also take into account the actual situation of the patient, such as economic conditions and convenience of medication, to ensure the feasibility of the plan and patient compliance.

[0165] Assume that the genetic test results of patient B show that he has a high responsiveness to a new targeted drug, and the drug sensitivity test also supports the use of this drug. Considering that patient B has just undergone surgery and needs to adjust the drug dosage to reduce the burden on the body, the drug matching algorithm will recommend this targeted drug and give specific medication instructions.

[0166] The breast cancer risk management system also includes steps after providing a surgical planning solution, as follows: Step Sg01, classify and organize the collected physiological data, rehabilitation data, lifestyle data and psychological state data respectively.

[0167] Specifically, the collected physiological data (such as heart rate, blood pressure, body temperature, blood test results, etc.), rehabilitation data (wound healing status, activity recovery progress, etc.), lifestyle data (diet structure, exercise duration and type, etc.) and psychological state data (anxiety, depression level assessment results, etc.) will be classified separately for subsequent analysis.

[0168] Step Sg02, according to each category of data, analyze the trend of each category of data within a future preset time range, and analyze and determine the score proportion of different categories of data according to the data range that different categories of data fall into, the trend range and score proportion that different categories of data fall into within the future preset time range.

[0169] Specifically, for each type of data, use existing data and appropriate analysis methods (such as time series analysis, etc.) to predict its trend within a preset time range in the future. For example, whether the heart rate in physiological data is likely to remain stable, increase or decrease; whether the wound healing speed in rehabilitation data will accelerate, remain the same or slow down, etc.

[0170] Step Sg03, according to the data range and future trend range of different categories of data, combined with the pre-set corresponding relationship, the score proportion of each category of data is determined. For example, if the physiological data is currently in the normal range and the future trend is positive, a higher score proportion may be given, such as 80% of the total score of this category; if the rehabilitation data is currently recovering slowly and the future trend is uncertain, a lower score proportion may be given, such as 60% of the total score of this category.

[0171] Step Sg04, analyzing and determining the overall score based on the preset scores corresponding to the different categories of data and the score proportions of the different categories of data.

[0172] The scores corresponding to different categories of data are obtained as follows: the surgical level and the overall surgical success rate are obtained; based on the surgical level, the overall surgical success rate and the scores corresponding to different categories of data, the scores corresponding to different categories of data are analyzed and determined.

[0173] Specifically, the scores corresponding to different categories of data are initially determined based on the surgical level and the overall success rate of the surgery. For example, when the surgical level is high and the success rate is low, the initial scores of various types of data may be low; when the surgical level is low and the success rate is high, the initial scores of various types of data may be high.

[0174] Combine the score proportion of different categories of data and the preset scores of each category of data to calculate the actual score of each category of data. For example, if the preset score of physiological data is 40 points and the score proportion is 80%, the actual score is 32 points. Add up the actual scores of all categories to get the overall score.

[0175] Step Sg05, analyzing and determining the postoperative breast cancer detection cycle according to the corresponding relationship between the score range that the overall score falls into and the postoperative breast cancer detection cycle.

[0176] Suppose a patient has undergone breast cancer surgery and is recovering from surgery.

[0177] The data is classified and organized as follows: Physiological data: The current heart rate, blood pressure, and body temperature are basically normal, and some indicators of blood tests fluctuate slightly. Rehabilitation data: The wound heals slowly, and the recovery of mobility is average. Lifestyle data: The diet structure is relatively reasonable, and there is some exercise but it is irregular.

[0178] Psychological state data: Mild anxiety.

[0179] The future trends of each type of data are analyzed and the score ratios are determined as follows: Physiological data: Through analysis, it is expected that heart rate, blood pressure, etc. may remain stable in the future, and blood test indicators are also expected to gradually return to normal. The score for this type of data is 75%. Rehabilitation data: Due to the slow healing of wounds and the uncertainty of the recovery of mobility, the score is 60%. Lifestyle data: Considering that patients have room for improvement and may have a positive impact on rehabilitation, the score is 70%. Psychological state data: Given that mild anxiety may last for a period of time, the score is 65%.

[0180] The scores and overall scores of different categories of data are determined as follows: the surgical level is medium, and the overall success rate of the surgery is 75%. The preset score of physiological data is 40 points, the preset score of rehabilitation data is 30 points, the preset score of lifestyle data is 20 points, and the preset score of psychological state data is 10 points.

[0181] Actual score of physiological data: 40×75%=30 points.

[0182] Actual score of rehabilitation data: 30×60%=18 points.

[0183] Actual score of lifestyle data: 20×70%=14 points.

[0184] Actual score of psychological state data: 10×65%=6.5 points.

[0185] The overall score is 30+18+14+6.5=68.5 points.

[0186] The postoperative breast cancer detection cycle is determined as follows: According to the preset correspondence between the scoring range and the detection cycle, 68.5 points is between 60-80 points, which can determine that the postoperative detection cycle is once every three months.

[0187] Furthermore, the breast cancer risk management system also includes steps after providing a surgical planning solution, which are as follows: Step Sg06, using the SHAP analysis method, analyze the contribution of various types of postoperative data in risk prediction. The Holt-Winters triple exponential smoothing time series analysis method is used to analyze the change trend of postoperative data and obtain the trend stability score of the data. According to the surgical level and the patient's physiological indicators, the matching degree of postoperative data with the clinical standard rehabilitation pathway is evaluated to obtain the clinical pathway relevance score.

[0188] Among them, the Holt-Winters triple exponential smoothing method is used to analyze the trend, seasonality and irregular fluctuations of time series data and predict future data changes. Output: trend stability score (0-1), the higher the score, the smaller the data fluctuation.

[0189] The trend stability score is obtained as follows: 1. Collect multiple observations of a certain data (such as the patient's blood pressure value) over a period of time after surgery to form time series data. 2. Use the Holt-Winters three-exponential smoothing method to analyze the time series, which takes into account factors such as the level, trend, and seasonality of the data. 3. Based on the analysis results, calculate the trend stability score of the data. For example, the score is determined by calculating indicators such as the degree of deviation between the predicted value and the actual value.

[0190] Clinical pathway relevance: How well the postoperative data matches a standardized rehabilitation pathway (e.g., NCCN guidelines). Range: 0 (completely unmatched) to 1 (completely matched).

[0191] The clinical pathway relevance score is obtained as follows: 1. Establish a clinical standard rehabilitation pathway database, which contains the standard rehabilitation data range and change trend under different surgical levels and patient physiological index combinations. 2. Compare the patient's actual postoperative data with the clinical standard rehabilitation pathway corresponding to the corresponding surgical level and physiological index. 3. According to the comparison results, the clinical pathway relevance score is calculated by the preset clinical pathway relevance score algorithm. For example, if a patient's indicator is within the standard pathway range, a higher score is given; if the deviation range is large, a lower score is given.

[0192] The steps involved in the pre-defined clinical pathway relevance scoring algorithm are as follows: 1. Data standardization: Perform Z-score standardization on postoperative data (such as body temperature, heart rate, wound healing, etc.), and perform the same standardization on the corresponding indicators of the clinical standard rehabilitation pathway.

[0193] 2. Dynamic Time Warping (DTW) alignment: Use the DTW algorithm to solve the time axis misalignment problem between postoperative data and the standard path, and calculate the optimal matching path between time series. Example: The postoperative temperature data time points are [Day 1, Day 3, Day 5], and the standard path is [Day 2, Day 4, Day 6]. DTW will find the time warp paths of the two to match.

[0194] 3. Cosine similarity calculation: Calculate the cosine similarity between the aligned postoperative data sequence and the standard path.

[0195] 4. Weighted adjustment of surgical level: Set the weight coefficient α∈[0.7,1.3] according to the type of surgery (such as breast-conserving surgery / radical resection). Example: The weight coefficient of radical resection is set to 1.2, and that of breast-conserving surgery is set to 0.9.

[0196] 5. Physiological index correction: Introduce physiological indexes such as BMI and age as correction factor β.

[0197] The specific formula is as follows: .

[0198] 6. Calculation of comprehensive score: Clinical pathway relevance score = sim × α × β.

[0199] Example: Similarity 0.85, surgical weight 1.2, physiological correction factor 0.95, the final score is 0.85×1.2×0.95=0.969.

[0200] Step Sg07, according to the preset weight coefficient, combined with SHAP contribution, trend stability score and clinical pathway relevance score, a dynamic weight is assigned to each postoperative data. The formula is as follows: Weight = α × (SHAP contribution / total SHAP contribution) + β × trend stability score + γ × clinical pathway relevance score.

[0201] The preset weight coefficients are defined as follows: Three coefficients pre-set in the breast cancer risk management system, denoted by α, β, and γ, are used to balance the relative importance of SHAP contribution, trend stability score, and clinical pathway relevance score in determining the weight of postoperative data.

[0202] The preset weight coefficient is obtained as follows: It is usually determined by medical experts and data scientists based on clinical experience, research results, and a large amount of historical data statistical analysis. For example, in some studies, through the analysis of postoperative rehabilitation data of multiple groups of breast cancer patients, it was found that the SHAP contribution had a greater impact on risk prediction, so the value of α can be appropriately increased.

[0203] Step Sg08, standardize the data of various indicators after surgery, and determine the expected recovery curve of each indicator at different stages after surgery. According to the comparison between the standardized value of the indicator and the expected recovery curve, calculate the score of each indicator. According to the importance of different indicators, assign corresponding weights to the scores of each indicator to calculate the comprehensive rehabilitation score.

[0204] Various indicator data after surgery: including physiological data (such as heart rate, blood pressure, body temperature, etc.), rehabilitation data (such as wound healing status, activity recovery progress, etc.), lifestyle data (such as diet structure, exercise duration and type, etc.) and psychological state data (such as anxiety, depression level assessment results, etc.).

[0205] Standardization processing: Z-score standardization was performed on various index data after surgery to make their mean value 0 and standard deviation 1, so as to eliminate the influence of dimensions and dimensions between different index data.

[0206] Expected recovery curve: Based on a large amount of clinical data and medical research, an ideal recovery trajectory is set for each postoperative indicator at different stages. It reflects the numerical change trend of the indicator over time under normal recovery conditions.

[0207] Indicator score calculation: The standardized indicator data is compared with the expected recovery curve, and the score of each indicator is calculated based on the comparison. The scoring rules can be set according to the clinical significance and recovery progress of the indicator. For example, if the actual value of an indicator deviates less from the expected value, a higher score is given; otherwise, a lower score is given.

[0208] Weight allocation: According to the importance of different indicators in the rehabilitation process, the corresponding weight is assigned to the score of each indicator. For example, heart rate and blood pressure in physiological indicators may have higher weights than dietary structure in lifestyle indicators. Comprehensive rehabilitation score calculation: Multiply the score of each indicator by its corresponding weight, and then add all weighted scores to obtain a comprehensive rehabilitation score. This score can fully reflect the patient's postoperative rehabilitation status.

[0209] For example, suppose we have the following three indicators and their weights: Physiological indicator A (weight 0.5): the standardized value is 0.8, the expected recovery curve value is 0.7, and the scoring rule is 80 points for a deviation within ±0.2, and 10 points will be deducted for every 0.1 exceeding the limit.

[0210] The score S1 of physiological index A is: S1=80-10×(0.8-0.7) / 0.1=80-10×1=70. Rehabilitation index B (weight 0.3): the standardized value is 0.6, the expected recovery curve value is 0.5, and the scoring rule is 100 points for deviation within ±0.2, and 10 points are deducted for every 0.1 over deviation. The score S2 of rehabilitation index B is: S2=100-10×(0.6-0.5) / 0.1=100-10×1=90.

[0211] Lifestyle index C (weight 0.2): The standardized value is 0.9, the expected recovery curve value is 0.8, and the scoring rule is 90 points for deviation within ±0.2, and 10 points are deducted for every 0.1 over deviation. The score S3 of lifestyle index C is: S3=90-10×(0.9-0.8) / 0.1=90-10×1=80. The calculation of comprehensive rehabilitation score S is as follows: S=S1×0.5+S2×0.3+S3×0.2=70×0.5+90×0.3+80×0.2=35+27+16=78.

[0212] Step Sg09, according to the standardized postoperative index data and the weight value of the dynamic weight allocation, the current risk value is obtained by weighted calculation, and the rate of change of the current risk value relative to the baseline risk value is calculated to reflect the dynamic change of the risk level. Formula: Risk change rate = (current risk value - baseline risk value) / baseline risk value. Compare the patient's actual rehabilitation score with the expected rehabilitation score to calculate the rehabilitation deviation. The rehabilitation score can be constructed based on an adaptive scoring matrix, taking into account the current status, rate of change and acceleration of the indicator. Formula: Rehabilitation deviation = (actual rehabilitation score - expected rehabilitation score) / (full score rehabilitation score - minimum rehabilitation score).

[0213] Current risk value: obtained by multiplying the standardized postoperative index data by the corresponding dynamic weights and then adding up all weighted data. It is a quantitative indicator that comprehensively reflects the patient's current risk status.

[0214] Baseline risk value: The initial risk value of the patient at the first postoperative assessment, which serves as a reference for subsequent risk changes. It is usually output by the preoperative risk prediction model and reflects the patient's basic risk level immediately after surgery.

[0215] Risk change rate: calculated by the formula (current risk value - baseline risk value) / baseline risk value, used to reflect the dynamic changes in risk levels. Negative values ​​indicate a decrease in risk, and positive values ​​indicate an increase in risk.

[0216] Actual rehabilitation score: It is constructed based on the adaptive scoring matrix, taking into account the current status, rate of change and acceleration of indicators. It reflects the actual rehabilitation status of patients.

[0217] Expected recovery score: Based on a large amount of clinical data and medical research, an ideal recovery score is set for each postoperative indicator at different stages. It reflects the level of recovery that the patient should achieve under normal recovery conditions.

[0218] In step Sg0a, the risk change rate and the recovery deviation are used as input variables of the fuzzy inference system, and different fuzzy sets and membership functions are defined for each input variable. For example, the risk change rate is divided into three fuzzy sets of "low", "medium" and "high", and the membership function of each set is determined.

[0219] Definition of fuzzy reasoning system: Based on fuzzy logic theory, it maps input variables (such as risk change rate, recovery deviation) to fuzzy sets and makes inferences through fuzzy rules. Function: It processes uncertain data and simulates the decision-making process of human experts.

[0220] Input variable definition: The input parameters of the fuzzy inference system, which are the risk change rate and recovery deviation in this step.

[0221] Value range: Risk change rate: [-100%, +∞) (negative value means risk decrease, positive value means risk increase); Rehabilitation deviation: [-100%, +100%) (negative value means rehabilitation lag, positive value means rehabilitation advance) Fuzzy set definition: Discretization of continuous input variables, such as classifying risk change rate into "low", "medium" and "high". Features: Fuzzy boundaries, allowing partial membership (e.g., a risk change rate of 25% may belong to both "medium" and "high").

[0222] Membership function definition: describes the degree to which the input value belongs to a fuzzy set, with a value range of [0,1]. Common types: triangular, trapezoidal, Gaussian, etc.

[0223] Step Sg0b, according to the membership degree of the input variable, the fuzzy rules are used for reasoning to obtain the fuzzy output result, and the fuzzy output result is converted into a specific detection period adjustment value by using a defuzzification method (such as the centroid method).

[0224] Fuzzy rule definition: Based on clinical experience and medical knowledge, a conditional statement is created to map fuzzy input variables (such as “high” risk change rate, “large” recovery deviation) to output results (such as “shortened” detection cycle).

[0225] Fuzzy reasoning definition: The process of deriving the fuzzy set of output results based on fuzzy rules and the membership of input variables. Method: Mamdani reasoning method (based on maximum-minimum synthesis).

[0226] Definition of defuzzification method: The process of converting fuzzy output results into specific values ​​(such as shortening 2 weeks). Common methods: Centroid method: Calculate the centroid position of the fuzzy set. Maximum membership method: Take the point with the largest membership.

[0227] Detection cycle adjustment value definition: The detection cycle adjustment range (such as shortening by 15%) determined based on the fuzzy reasoning results. Range: Usually between -50% (shortening) and +50% (extending).

[0228] Step Sg0c, dynamically adjusting the current detection cycle according to the detection cycle adjustment value output by the fuzzy inference system.

[0229] Based on the same inventive concept, an embodiment of the present invention provides a computer storage medium, including a memory and a processor, wherein the memory stores a program that can be executed on the processor to implement the following Figures 1 to 3 Procedure for either method.

[0230] The embodiments of this specific implementation method are all preferred embodiments of the present application, and are not intended to limit the protection scope of the present application. Therefore, all equivalent changes made based on the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. A breast cancer risk management system, characterized in that: include: Breast monitoring equipment; at least one processor; At least one memory storing a computer program, and at least one processor configured to implement a breast cancer risk management method when executing the computer program, the specific steps are as follows: Obtain patient-related data collected by breast monitoring equipment and perform data preprocessing; Inputting the patient-related data that has completed data preprocessing into the trained breast cancer risk value assessment model that considers the patient's personalized data, and outputting the breast cancer risk value; Use the preset and trained random forest algorithm to train the model and calculate the weight of each risk factor for breast cancer risk value; Analyze and determine the breast cancer detection cycle based on the corresponding relationship between the weight range of the breast cancer risk value of the specific risk factor, the risk value assessment range of the breast cancer risk value, and the detection cycle; Form a time schedule for breast cancer related testing for patients based on the breast cancer testing cycle determined by analysis; According to the corresponding relationship between the weight interval into which the weight of each risk factor for the breast cancer risk value falls and the quantitative suggestion association, quantitative suggestion information is formed, and the formed quantitative suggestion information is sent to the terminal held by the patient.

2. A breast cancer risk management and control system according to claim 1, characterized in that: The step after sending the generated quantitative suggestion information to the terminal held by the patient is further included, which is as follows: Collect the latest data of the patient within the preset period after the quantitative recommendation is sent, and re-analyze the breast cancer risk value and the weight of each risk factor to the breast cancer risk value based on the latest data; Analyze whether the reduction in breast cancer risk value generated by each risk factor alone reaches or exceeds the preset range; If yes, the original breast cancer detection cycle is maintained unchanged, and the original breast cancer detection cycle is determined according to the corresponding relationship between the weight range of the breast cancer risk value of the specific risk factor, the risk value assessment range of the breast cancer risk value, and the detection cycle; If not, the risk factors with a decrease less than the preset range are analyzed and obtained as the target risk factors, and the ratio of the decrease of the target risk factor to the preset range is calculated. According to the ratio of the decrease of the target risk factor to the preset range and the weight of the target risk factor to the breast cancer risk value, the adjustment coefficients of all target factors for the breast cancer detection cycle are calculated, and then The adjusted breast cancer detection cycle is calculated based on the adjustment coefficient and the original breast cancer detection cycle; Update the patient's breast cancer related testing schedule based on the finalized breast cancer testing cycle.

3. A breast cancer risk management and control system according to claim 2, characterized in that: It also includes the following steps after collecting patient-related data and preprocessing the data, and before using the preset and trained random forest algorithm to train the model: The patients were divided into different age groups, including young, middle-aged and elderly groups; If the patient is in the young or middle-aged group, the original settings are maintained; If the patient was in the elderly group, variables associated with breast cancer risk were selected, and the stepwise regression method was used to screen the variables and establish a preliminary multivariate regression model; The preprocessed data is divided into a training set and a validation set, and the multivariate regression model is trained on the training set. The performance of the multivariate regression model is adjusted by adjusting the parameters of the multivariate regression model until the accuracy of the multivariate regression model on the validation set reaches the preset accuracy within a preset number of consecutive times, and the training of the multivariate regression model is determined to be completed, and the regression coefficient corresponding to each risk factor is checked; According to the regression coefficient corresponding to each risk factor, the sum of the absolute values ​​of all regression coefficients is calculated, and the ratio of the absolute value of the regression coefficient of different risk factors to the sum of the absolute values ​​of all regression coefficients is calculated to determine the weight of each risk factor for breast cancer risk value; The breast cancer detection cycle is determined by analysis based on the correspondence between the weight range of the breast cancer risk value for the specific risk factors, the risk value assessment range within which the breast cancer risk value falls, and the detection cycle.

4. A breast cancer risk management and control system according to claim 2, characterized in that: The steps of analyzing and obtaining the risk factors with a decrease range less than a preset range as target risk factors are also included, which are as follows: Patients whose breast cancer risk values ​​resulting from risk factors alone do not decrease by the preset range are considered to be patients who need to be followed up; Start the follow-up process, contact the patient through intelligent voice methods, and obtain the patient's implementation results of quantitative advice information, including exercise duration and dietary adjustments; If the patient's execution result of the quantitative suggestion information does not match the quantitative content mentioned in the quantitative suggestion information, the patient is advised to execute it again according to the quantitative content mentioned in the quantitative suggestion information; If the patient's execution result on the quantitative advice information is consistent with the quantitative content mentioned in the quantitative advice information, the quantitative advice information is analyzed for health-related quantitative content not mentioned, and suggestions for other quantitative content are provided to be added to the quantitative advice information to form new quantitative advice information, which is then sent to the patient's terminal again.

5. A breast cancer risk management and control system according to claim 2, characterized in that: Update the patient's breast cancer related testing schedule based on the finalized breast cancer testing cycle, including: Obtain relevant information from the patient's electronic medical records and extract key factors that affect breast cancer test results, including ongoing treatment plans for other diseases, specific drugs used, recent surgical history, and the severity of the patient's underlying disease; According to the correspondence between the influence degree value range of the preset key factors affecting the breast cancer detection results and the delay time, the overall delay time caused by the key factors affecting the breast cancer detection results is analyzed and determined; According to the overall delay time caused by the key factors affecting the breast cancer detection results and the breast cancer detection cycle determined by analysis, the breast cancer detection cycle is re-analyzed and the patient's breast cancer-related detection time planning table is re-formed.

6. A breast cancer risk management and control system according to claim 5, characterized in that: The method also includes the following steps after updating the patient's breast cancer related testing schedule according to the finalized breast cancer testing cycle: Collect specific information about patients' breast cancer-related tests and perform data preprocessing; Select features relevant to adjuvant treatment decisions from specific information collected on patients' breast cancer-related tests; Input the features related to adjuvant therapy decision-making into a pre-trained adjuvant therapy decision-making model, and output the result of whether adjuvant therapy is required and the treatment plan if adjuvant therapy is required, and the treatment plan if adjuvant therapy is required includes surgery and drug provision plan; If the result of adjuvant therapy is required and the treatment plan under adjuvant therapy is surgery, the patient's medical imaging data is obtained, and the contours of the tumor and surrounding tissues are extracted using an image segmentation algorithm; Based on the extracted contours of the tumor and surrounding tissues, a path planning algorithm is used to simulate the entry path and operating range of the surgical instrument, determine the surgical path and resection range, and thus form a surgical planning plan; If the output is a result that adjuvant therapy is needed and the treatment plan under adjuvant therapy is medication, the patient's genetic test results, drug sensitivity data and surgical plan are obtained, and based on the patient's genetic test results, drug sensitivity data and surgical plan, a preset drug matching algorithm is used to screen out suitable drug types and dosages as a drug provision plan.

7. A breast cancer risk management and control system according to claim 1, characterized in that: Also included are steps after calculating the weight of each risk factor for breast cancer risk, as follows: Analyze whether complete data of patients reaching or exceeding the preset number of cases have been collected; If no, continue with the next steps; If yes, then select variables related to breast cancer risk, use stepwise regression method to screen variables, establish a preliminary multiple regression model, divide the preprocessed data into a training set and a validation set, train the multiple regression model on the training set, adjust the performance of the multiple regression model by adjusting the parameters of the multiple regression model, until the accuracy of the multiple regression model on the validation set reaches the preset accuracy within a preset number of consecutive times, determine that the multiple regression model training is completed, check the regression coefficient corresponding to each risk factor, calculate the sum of the absolute values ​​of all regression coefficients, and calculate the ratio of the absolute value of the regression coefficient of different risk factors to the sum of the absolute values ​​of all regression coefficients, and determine the weight of each risk factor for the breast cancer risk value; The risk value weights output by the random forest model and the risk value weights output by the multivariate regression model are input as new features into the trained decision tree fusion model, and the comprehensive risk weight assessment results are output; According to the correspondence between the comprehensive risk weight output by the decision tree fusion model, the risk value assessment range into which the breast cancer risk value falls, and the detection cycle, the breast cancer detection cycle is analyzed and determined.

8. A breast cancer risk management and control system according to claim 1, characterized in that: Data preprocessing includes the following steps: Clean the collected raw data, which are patient-related data; Standardize the data to eliminate the dimensional differences between different data types; The processed data were integrated into a format suitable for input into the breast cancer risk assessment model.

9. A computer storage medium, characterized in that: The method comprises a program capable of being loaded and executed by a processor to implement a breast cancer risk management method in a breast cancer risk management system according to any one of claims 1 to 8.

Images