End-stage renal disease hemodialysis blood pressure pattern recognition and intervention system and related devices

Abstract

This application relates to a blood pressure pattern recognition and intervention system for hemodialysis in end-stage renal disease. The method includes: a dialysis device for performing hemodialysis on a user; a data collection device; a memory and a processor, the processor loading a dialysis control program to execute the following steps: filtering and combining data collected by the data collection device and a historical database to obtain blood pressure-related features; generating classification factors and grouping patients into different groups; establishing training and testing sets, training a machine learning model to obtain a blood pressure prediction model, and validating and adjusting the model based on the testing set; inputting the data collected by the data collection device into the blood pressure prediction model, and obtaining dialysis device control parameters based on the blood pressure prediction model; and controlling the dialysate temperature, dialysate flow rate, and dialysis duration of the dialysis device based on the dialysis device control parameters. This application has the advantage of being able to calculate and adjust the parameters of the dialysis device according to the user's real-time condition.

Description

Technical Field

[0001] This application relates to the field of medical devices, and in particular to a blood pressure pattern recognition and intervention system and related apparatus for hemodialysis in end-stage renal disease. Background Technology

[0002] Hemodialysis, as an effective renal replacement therapy, has been widely used worldwide, especially for patients with chronic renal failure and acute kidney injury. Current dialysis techniques and devices have achieved relatively precise and automated control, but many challenges remain.

[0003] Hypotension is a common complication during dialysis, which can cause serious discomfort and danger to patients. This problem can usually be controlled by adjusting the infusion rate, infusion time, and infusion temperature, but this process is not simple. Because each patient's physical response can vary greatly, it requires very precise control and individualized adjustments.

[0004] In reality, many medical staff may lack experience or be overworked, making it difficult to accurately and promptly determine the appropriate infusion rate, infusion time, and infusion temperature. Patient vital signs during dialysis can change constantly; without real-time monitoring and intelligent assistance, treatment outcomes may be unsatisfactory, and the patient's health may even be endangered. Summary of the Invention

[0005] In order to calculate and adjust the parameters of the dialysis device according to the user's real-time situation, this application provides a blood pressure pattern recognition and intervention system and related device for hemodialysis in end-stage renal disease.

[0006] This application provides a blood pressure pattern recognition and intervention system for hemodialysis in end-stage renal disease, which adopts the following technical solution:

[0007] A blood pressure pattern recognition and intervention system for hemodialysis in end-stage renal disease includes:

[0008] Dialysis devices are used to perform hemodialysis on users;

[0009] Data collection device used to collect users' physiological data, medical record data, climate data, and dialysis treatment data during the operation of the dialysis device;

[0010] The memory stores the dialysis control program, and the processor loads the dialysis control program to perform the following steps:

[0011] S1. Upload the data collected by the data collection device to the cloud server, and control the cloud server to retrieve the historical database. Based on the data collected by the data collection device and the historical database, filter and combine the data to obtain blood pressure-related characteristics.

[0012] S2. Control the cloud server to generate classification factors based on the obtained blood pressure-related features, and use the classification factors to group patients into different groups;

[0013] S3. Control the cloud server to select patient data from each group to establish training and test sets, train the machine learning model to obtain the blood pressure prediction model, and verify and adjust the model based on the test set;

[0014] S4. Download and update the parameters of the blood pressure prediction model in the memory, input the data collected by the data collection device into the blood pressure prediction model, and obtain the control parameters of the dialysis device based on the blood pressure prediction model;

[0015] S5. Control the dialysate temperature, dialysate flow rate and dialysis duration of the dialysis device based on the control parameters of the dialysis device.

[0016] By employing the aforementioned technical solution, the system collects patients' physiological data, medical records, climate data, and dialysis device operating data through a data collection device. This allows for real-time monitoring of the patient's condition and personalized adjustments to dialysis fluid temperature, flow rate, and dialysis duration based on individual characteristics and environmental factors. Utilizing a cloud server for data filtering and combination, and training and validating machine learning models based on blood pressure-related characteristics, the system can accurately predict blood pressure changes and adjust dialysis device parameters accordingly. Through automated data analysis and dialysis parameter control, the system reduces the workload of medical staff, allowing them to focus more on other patient medical needs. Precise blood pressure prediction and personalized dialysis control reduce discomfort and complications during dialysis, improving the patient's treatment experience and quality of life. Furthermore, with the help of a cloud server, the system can also support telemedicine services and multi-site collaborative treatment, enhancing the accessibility and continuity of healthcare services.

[0017] Optionally, it also includes an operation screen for acquiring the user's physiological data, and the data collection device is capable of performing the following steps:

[0018] Acquire basic body feature parameters input to the operation screen for display, store and update in the body feature information database;

[0019] Real-time blood information is acquired using a data acquisition device, stored, and updated in a blood information database.

[0020] User body characteristic parameters are obtained based on historical information within the latest time period from a database of body characteristic information and blood information.

[0021] By adopting the above technical solution, patients and medical staff can more easily input and view basic physical characteristic parameters through the operating screen. This intuitive interaction enhances the system's user-friendliness and helps improve the collaborative participation of patients and medical staff. By combining the basic physical characteristic parameters input through the operating screen with real-time blood information acquired by the data acquisition device, the system can comprehensively monitor the user's physical characteristic parameters. This provides richer evidence for personalized treatment and precise intervention. The system can store and update data in the physical characteristic and blood information databases in real time. This real-time update mechanism helps capture subtle changes in the patient's condition, thereby enabling more timely and accurate treatment adjustments. Storing and updating data in the physical characteristic and blood information databases allows medical staff to easily review and analyze the patient's historical data. This helps to identify long-term trends and potential problems, thereby better understanding the patient's health condition and developing appropriate treatment plans.

[0022] Optionally, S1 includes the following sub-steps:

[0023] S11. Upload the physiological data, medical record data, climate data, and dialysis treatment data collected by the data collection device to the cloud server;

[0024] S12. Control the cloud server to retrieve historical databases related to the collected data;

[0025] S13. Filter the data obtained from the data collection device and historical database to remove irrelevant or invalid information;

[0026] S14. Combine or integrate the filtered data;

[0027] S15. Based on the filtered and combined data, extract features related to blood pressure.

[0028] By employing the above technical solutions and breaking down the processes of data collection, screening, combination, and feature extraction into specific sub-steps, the system can manage and process information more accurately. Sub-steps S11 and S12 allow the system to collect data from multiple sources (e.g., physiological data, medical record data, climate data, etc.) and integrate it with historical databases. Through the data screening step in S13, the system can eliminate irrelevant or invalid information. This process helps reduce noise and interference, improving data quality and the accuracy of analysis. The data combination or integration step in S14 optimizes data analysis, making it more suitable for blood pressure pattern recognition. This step ensures the correlation and consistency between selected features, thereby enhancing the robustness and interpretability of the blood pressure prediction model. The feature extraction step in S15 identifies individual characteristics related to blood pressure, thus supporting more personalized interventions and treatments.

[0029] Optionally, step S2 includes the following sub-steps:

[0030] S21. Analyze blood pressure-related characteristics and generate classification factors based on the analysis results;

[0031] S22. Preprocess the data to suit the clustering algorithm;

[0032] S23. Based on the generated classification factors, users are grouped into different groups, wherein the groups represent different blood pressure levels, risk factors, and disease stages;

[0033] S24. Evaluate and verify the results of the clustering.

[0034] By employing the aforementioned technical solutions, sub-step S21 analyzes blood pressure-related characteristics and generates categorical factors, contributing to a deeper understanding of the relationship between blood pressure and other physiological and medical factors. This precise analysis helps capture the complex and variable blood pressure patterns of patients, providing a more accurate basis for subsequent prediction and intervention. Sub-step S22's data preprocessing ensures the data is suitable for the clustering algorithm, improving data quality and enhancing the algorithm's stability and reliability through methods such as cleaning and normalization. Sub-step S23 groups users into groups representing different blood pressure levels, risk factors, and disease stages. This clustering enables the system to identify and target specific needs and risk factors for each group, achieving more personalized interventions and treatments. Through the integrated analysis of these sub-steps, the system can provide more comprehensive and accurate insights into patient blood pressure patterns. This not only helps doctors and nurses make better clinical decisions but also supports patient self-management and preventative interventions.

[0035] Optionally, step S3 includes the following sub-steps:

[0036] S31. Select data from each blood pressure classification group, and divide the data in each group into a training set and a test set based on a pre-selected partitioning strategy and partitioning ratio. The training set is used to train the model, and the test set is used to validate and adjust the model. The pre-selected partitioning strategy is random partitioning, hierarchical partitioning, or cross-validation.

[0037] S32. Select the multi-class logistic regression algorithm model as the machine learning algorithm model, and normalize the features in the training set;

[0038] S33. Train the multi-class logistic regression algorithm based on the training set, and update the parameters by minimizing the classification error in each iteration;

[0039] S34. Validate the trained model using test set data and adjust the model based on its performance on the test set.

[0040] By employing the aforementioned technical solutions and utilizing the data partitioning strategy in sub-step S31 (including random partitioning, stratified partitioning, or cross-validation), the model can be trained on the training set and validated on the test set. This ensures the model's generalization ability, making it applicable not only to training data but also to unseen data. Sub-steps S32 and S33 employ a multi-class logistic regression model, performing normalization and iterative training, enabling the model to handle multi-class problems and more accurately reflect multiple levels and classifications of blood pressure. Through precise analysis and model training of different blood pressure classification groups, the system can more accurately identify and predict each patient's blood pressure pattern, thereby achieving more personalized dialysis treatment.

[0041] Optionally, S4 includes the following sub-steps:

[0042] S41. Input the data collected by the data collection device, the standard dialysate temperature, the standard dialysate flow rate, and the standard dialysis duration into the blood pressure prediction model to predict future blood pressure changes;

[0043] S42. Input the future blood pressure change predicted by the blood pressure prediction model into a preset reward function for calculation and obtain a reward value. The output of the preset reward function is related to the difference between the future blood pressure change predicted by the blood pressure prediction model and the actual measured blood pressure change. The larger the difference, the lower the reward value; the smaller the difference, the lower the reward value.

[0044] S43. Based on the reinforcement learning algorithm and the change of reward value, adjust the dialysate volume, dialysate speed and dialysate temperature to increase the reward value, thereby obtaining new dialysate volume, dialysate speed and dialysate temperature. Substitute them into the previous step and execute until the dialysate volume, dialysate speed and dialysate temperature converge to a stable range, thereby obtaining the control parameters of the dialysis device.

[0045] By employing the above technical solution and using a preset reward function, the system compares the predicted blood pressure changes with the actual measured blood pressure changes and assigns reward values ​​based on the magnitude of the difference. This mechanism helps incentivize the algorithm to better fit the actual situation and promotes more accurate predictions.

[0046] Optionally, step S5 includes the following sub-steps:

[0047] S51. Receive dialysis device control parameters generated from the blood pressure prediction model;

[0048] S52. Based on the received parameters, set the dialysate temperature, dialysate flow rate, and dialysis duration of the dialysis device;

[0049] S53. Adjust the dialysis device in real time to match the set parameters;

[0050] S54. Continuously monitor the dialysis process and collect real-time blood information as feedback, and adjust the dialysate temperature, dialysate flow rate and dialysis duration of the dialysis device based on the feedback.

[0051] By employing the above technical solution, continuously monitoring and collecting real-time blood information as feedback helps to detect potential problems in a timely manner and allows the system to make real-time adjustments based on the feedback information. This feedback mechanism can enhance the system's adaptability, enabling it to better adapt to individual patient differences.

[0052] Secondly, the computer device provided in this application adopts the following technical solution:

[0053] A computer device comprising:

[0054] One or more processors;

[0055] Memory;

[0056] One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, the one or more applications being configured to: execute the dialysis control program described above.

[0057] Thirdly, this application provides a computer-readable storage medium that adopts the following technical solution:

[0058] A computer-readable storage medium storing a computer program that can be loaded by a processor and executed as described above.

[0059] The storage medium stores at least one instruction, at least one program, code set, or instruction set, which is loaded and executed by the processor to implement, as described above, the dialysis control program. Attached Figure Description

[0060] Figure 1 A flowchart of a dialysis control procedure according to an embodiment of the present invention is shown.

[0061] Figure 2 A flowchart illustrating S1 in one embodiment of the present invention is shown.

[0062] Figure 3 A flowchart illustrating S2 in one embodiment of the present invention is shown.

[0063] Figure 4 A flowchart illustrating S3 in one embodiment of the present invention is shown.

[0064] Figure 5 A flowchart illustrating S4 in one embodiment of the present invention is shown.

[0065] Figure 6 A flowchart illustrating S5 in one embodiment of the present invention is shown. Detailed Implementation

[0066] The present application will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of the application and are not intended to limit the scope of the application.

[0067] Maintenance hemodialysis (MHD) is the primary treatment for end-stage renal disease, but it also brings with it the serious complication of intradialysis-induced hypotension (IDH). IDH is caused by an imbalance between fluid removal and plasma replenishment during dialysis, leading to a decrease in effective arterial blood volume, reduced cardiac filling, and decreased cardiac output. This not only causes discomfort such as nausea, vomiting, and dizziness, but in severe cases, it can also trigger clinical symptoms such as muscle cramps and difficulty breathing, and even lead to cardiovascular and cerebrovascular diseases and death.

[0068] CRRT is a treatment option for critically ill patients with hemodynamic instability. It can be performed continuously for 24 hours, thus minimizing disruption to the patient's circulatory system, allowing for better control of fluid balance and electrolyte homeostasis, and improving the patient's metabolic status. The working principle of CRRT mainly involves filtration, dialysis, and adsorption. The treatment intensity can be adjusted by changing the flow rates of blood and dialysate or replacement fluid, as well as by regulating filtration pressure and anticoagulation levels, enabling personalized treatment based on the patient's specific condition.

[0069] Despite the many advantages of CRRT, special attention must be paid to potential complications such as hypotension, coagulation, and filter blockage. These complications can negatively impact the patient's treatment outcome and overall health. Therefore, CRRT treatment needs to be administered continuously for 24 hours and requires dedicated nursing staff for monitoring and management to ensure the safety and effectiveness of the treatment.

[0070] The incidence of IDH varies significantly across different studies, ranging from 20% to 50%. This variation may be related to inconsistencies in diagnostic criteria. Many factors contribute to IDH, including excessively high dialysate temperature, low plasma albumin levels, and the activation of leukocytes through blood filtration via the dialyzer. Some physiological responses, such as the DeJager-Krogh phenomenon, can also lead to IDH.

[0071] In real-world medical settings, healthcare professionals may struggle to accurately and promptly monitor critical parameters during dialysis, such as infusion rate, duration, and temperature, due to inexperience or excessive workload. Newcomers to dialysis may require time to adapt to and master the complex treatment parameters and patient physiological responses. Even experienced professionals may face pressure from heavy workloads, potentially impacting their efficiency and decision-making abilities.

[0072] During dialysis, a patient's vital signs can change constantly. For example, they may suddenly experience low blood pressure, irregular heartbeat, or seizures. This requires medical staff to be highly alert and responsive, able to detect problems and take appropriate action immediately. However, because human energy and attention are limited, and medical staff may struggle to maintain optimal performance after prolonged periods of high-intensity work, the inability to monitor patient changes in real time can lead to suboptimal treatment outcomes and even endanger the patient's health.

[0073] To address these issues, hospitals and medical institutions can explore the introduction of intelligent assistive devices and systems. These devices and systems can monitor and record patients' physiological parameters in real time, such as blood pressure and heart rate, and can be programmed with alarm values. When a parameter exceeds the normal range, the system will automatically sound an alarm to alert medical staff for timely intervention. Furthermore, intelligent assistive systems can automatically adjust the infusion rate, infusion time, and infusion temperature during dialysis based on the patient's specific condition to ensure optimal treatment outcomes.

[0074] This application proposes a blood pressure pattern recognition and intervention system for hemodialysis in end-stage renal disease, which is used to monitor the patient's condition, generate and adjust dialysis control parameters during the dialysis process, and help reduce the probability of IDH.

[0075] This end-stage renal disease hemodialysis blood pressure pattern recognition and intervention system includes a dialysis device, a data collection device, a memory, and a processor. The dialysis device is used to perform hemodialysis on the user. The data collection device is used to collect the user's physiological data, medical record data, climate data, and dialysis treatment data during the operation of the dialysis device. The memory stores the dialysis control program, and the processor loads the dialysis control program.

[0076] Specifically, a hemodialysis device mainly consists of the following parts:

[0077] The main body of the dialysis equipment, known as a dialysis machine, includes a dialysate pump (responsible for propelling the flow of blood and dialysate), an artificial kidney (or dialyzer), a temperature sensor, and a temperature control system (used to maintain the appropriate temperature of the blood and dialysate).

[0078] Dialyzer: The dialyzer is the core component of a hemodialysis device. Its interior consists of numerous microtubes, all encased in a semi-permeable membrane. Blood flows inside the tubes, while dialysate flows in the external cavity.

[0079] Blood flow system: This includes infusion and drainage blood vessels, whose function is to guide blood out of the patient's body, process it through the dialyzer, and then return it to the patient's body.

[0080] Dialysis fluid system: Responsible for preparing the dialysate and pushing it into the dialyzer via a pump system.

[0081] Dialysis fluid pump: It is a key component for controlling the flow of fluid, and can accurately regulate the flow rate and pressure of the fluid.

[0082] Temperature sensor: Used to monitor liquid temperature and ensure that the liquid is within a suitable temperature range.

[0083] Temperature regulation system: When the temperature of the liquid exceeds the preset value, the temperature regulation system can adjust the temperature of the liquid to prevent discomfort to the patient due to excessively high or low temperatures.

[0084] During dialysis, blood is drawn from the patient through a blood flow system and enters the microtubules inside the dialyzer. Outside the microtubules is the dialysate, containing appropriate electrolytes and glucose. A semipermeable membrane separates the blood from the dialysate. Due to the concentration gradient, waste products and excess electrolytes in the blood diffuse through the semipermeable membrane into the dialysate. To remove excess water from the blood, the dialysis machine creates a pressure difference between the blood and the dialysate, forcing water from the blood to flow into the dialysate through the semipermeable membrane. After these processes are complete, the clean blood is returned to the patient through the blood flow system. The entire process is automatically controlled by the dialysis machine, involving factors such as the flow rate of blood and dialysate, the dialysate formulation and temperature, and the ultrafiltration rate to ensure dialysis effectiveness and patient comfort.

[0085] Dialysis equipment typically features either an LCD or an OLED display, both of which provide clear visuals. High-end equipment may use high-definition or ultra-high-definition displays for an even better visual experience. During hemodialysis, the display screen is used to show various important information, such as the current dialysate volume, dialysate rate, dialysate temperature, and the patient's physiological parameters. This information can be displayed in numerical, textual, or graphical form. When the patient needs to input information, such as basic physical characteristics, the graphical interface provides one or more input methods. Common input methods include virtual keyboards, drop-down menus, and sliders. Patients can select or input their choices by touching the screen. The user's touch actions are captured by the touch layer and transmitted to the processor. The processor then changes the displayed content or performs corresponding calculations based on the user's actions.

[0086] The data collection device is capable of performing the following steps 1-3.

[0087] Step 1: Obtain the basic body feature parameters input to the operation screen for display, storage and update in the body feature information database.

[0088] Assume that the pre-entered basic body characteristics parameters include the user's age, gender, weight, height, etc. The screen will display these parameters, for example: "Age: 35 years old, Gender: Male, Weight: 70 kg, Height: 175 cm".

[0089] Step 2: Acquire real-time blood information using a data acquisition device, store and update it in the blood information database.

[0090] Data acquisition devices, such as blood analyzers, can collect users' blood information, including parameters such as blood glucose, blood pressure, and hemoglobin. This data is updated in real time and saved to a blood information database, for example: "Time: 10:00, Blood glucose: 5.5 mmol / L, Blood pressure: 120 / 80 mmHg, Hemoglobin: 150 g / L".

[0091] Step 3: Obtain the user's physical characteristic parameters based on the latest historical information from the physical characteristic information and blood information database within the latest time period.

[0092] The system calculates or predicts other physical parameters of the user based on the user's physical characteristics (such as age, gender, weight, and height) and the latest historical information in the blood information database (such as recent blood sugar, blood pressure, and hemoglobin data). For example, the system may predict the user's insulin requirements or estimate the user's oxygen demand based on the user's hemoglobin level.

[0093] Basic physical characteristics parameters mainly refer to physiological or biochemical indicators related to an individual's health status, as well as parameters closely related to treatment efficacy. In a hemodialysis setting, the following parameters can be considered basic physical characteristics parameters:

[0094] Age: Age has a significant impact on the effectiveness of hemodialysis and the type of treatment required. Older people may require more attention and care.

[0095] Gender: Men and women may have different physiological responses and needs during hemodialysis.

[0096] Weight: Weight can affect the amount of dialysis fluid needed and the dosage of medications.

[0097] Blood pressure: Blood pressure is a key parameter in the hemodialysis process and needs to be monitored and adjusted in real time.

[0098] Heart rate: Changes in heart rate can be an early sign of dialysis complications, such as low blood pressure.

[0099] Blood biochemical parameters, including hemoglobin, blood urea nitrogen, serum creatinine, and electrolytes (such as potassium, sodium, calcium, and phosphorus), play an important role in assessing the effectiveness of hemodialysis and adjusting treatment plans.

[0100] Number of dialysis sessions: The frequency of hemodialysis and the patient's history of dialysis are also important factors to consider.

[0101] All these parameters may be entered before the equipment is operated and dynamically adjusted based on monitoring results during treatment.

[0102] Specifically, refer to Figure 1 The dialysis control procedure is used to perform the following steps:

[0103] S1. Upload the data collected by the data collection device to the cloud server, and control the cloud server to retrieve the historical database. Based on the data collected by the data collection device and the historical database, filter and combine the data to obtain blood pressure-related characteristics.

[0104] In some embodiments, refer to Figure 2 S1 can be achieved through the following sub-steps:

[0105] S11. Upload the physiological data, medical record data, climate data, and dialysis treatment data collected by the data collection device to the cloud server;

[0106] S12. Control the cloud server to retrieve historical databases related to the collected data;

[0107] S13. Filter the data obtained from the data collection device and historical database to remove irrelevant or invalid information;

[0108] S14. Combine or integrate the filtered data;

[0109] S15. Based on the filtered and combined data, extract features related to blood pressure.

[0110] For example, in a hospital dialysis center, a hemodialysis machine is connected to a data collection device. This device monitors the patient's physiological parameters (such as heart rate and blood pressure), medical records (such as past medical history and medication use), current weather conditions, and the dialysis machine's operational data (such as dialysate temperature and flow rate). This data is uploaded to a cloud server via a secure network connection. Upon receiving the data, the cloud server connects to a historical database via an API. This historical database may include the patient's past blood pressure records and detailed data from previous dialysis sessions, allowing for comparison and analysis with the currently collected data. A filtering algorithm running on the cloud server removes irrelevant or invalid information. For example, the algorithm might exclude outliers (such as abnormally high or low blood pressure readings due to sensor malfunction) or historical data unrelated to current dialysis treatment. The filtered data is further integrated or combined, for example, by combining physiological data, medical records, weather data, and dialysis treatment data through time series analysis or data fusion techniques. This creates a more complete and representative dataset for further analysis. Finally, based on the integrated dataset, feature engineering and machine learning methods can be used to extract blood pressure-related features. For example, statistical analysis can be used to determine which dialysis parameters are significantly related to blood pressure changes, or deep learning methods can be applied to automatically extract complex features. These features are crucial for understanding trends in blood pressure changes and their relationship to dialysis treatment.

[0111] S2. Control the cloud server to generate classification factors based on the obtained blood pressure-related features, and use the classification factors to group patients into different groups.

[0112] In some embodiments, refer to Figure 3 S2 can be achieved through the following sub-steps:

[0113] S21. Analyze blood pressure-related characteristics and generate classification factors based on the analysis results.

[0114] For example, suppose that in step S15, the following blood pressure-related features are extracted: the patient's age, gender, weight, history of hypertension, medication records, and specific parameters during dialysis (such as dialysate temperature, flow rate, etc.).

[0115] In step S21, multiple linear regression, decision trees, or other specific analytical methods can be used to study how these characteristics affect changes in blood pressure.

[0116] Multiple linear regression: By creating a regression model, the degree of association between each feature and blood pressure can be quantified. For example, how many millimeters of mercury might increase blood pressure for every 1 kilogram of weight gain, or what kind of change in blood pressure might occur for every 1°C increase in dialysate temperature.

[0117] Decision tree analysis: Decision trees can also be used to identify which features play a decisive role in blood pressure classification. For example, analysis might reveal that patients with a history of hypertension experience greater fluctuations in blood pressure during dialysis.

[0118] Correlation analysis: By calculating correlation coefficients such as Pearson or Spearman, it is possible to identify which features have a significant linear or nonlinear association with blood pressure levels.

[0119] The results of these analyses can be used to generate categorical factors, which are key parameters used in subsequent steps to group patients. For example, if a significant relationship is found between weight and a history of hypertension and changes in blood pressure, these characteristics may be used as primary categorical factors to help group patients into groups based on different blood pressure levels, risk factors, and disease stages.

[0120] S22. Preprocess the data to suit the clustering algorithm.

[0121] For example, S22 can be achieved through the following steps:

[0122] Missing value handling: Some patients may not provide complete medical records or certain measurements. In such cases, various techniques (such as mean imputation, regression imputation, etc.) can be used to estimate the missing values.

[0123] Data standardization / normalization: Since the scales of different features can vary greatly (e.g., age ranges from 0-100, while weight can range from 40-200 kg), it may be necessary to scale all features to the same scale before clustering. This can be achieved, for example, using Z-score normalization or Min-Max normalization.

[0124] Outlier Handling: Outliers are common in real-world data. Values ​​that are too large or too small can distort clustering results. Outliers can be identified and handled using methods such as IQR ranges or Z-scores.

[0125] Converting non-numerical data: If the categorical factors contain class data, such as gender (male / female) or history of hypertension (yes / no), it may be necessary to convert them to numerical form. For example, one-hot encoding or label encoding can be used.

[0126] Feature selection / dimensionality reduction: If there are many features, feature selection or dimensionality reduction techniques (such as principal component analysis) can be used to reduce the dimensionality, making the algorithm run more efficiently and reducing the impact of noise.

[0127] After these preprocessing steps are completed, the dataset will be ready for use by the clustering algorithm. This ensures that the algorithm can more effectively and accurately identify meaningful patterns and clusters among patients, reflecting different blood pressure levels, risk factors, and disease stages.

[0128] S23. Based on the generated classification factors, users are grouped into different groups, wherein the groups represent different blood pressure levels, risk factors, and disease stages.

[0129] For example, S23 can be achieved through the following steps:

[0130] Choosing an appropriate clustering algorithm: Select a clustering algorithm suitable for the characteristics of your data. Commonly used clustering algorithms include K-means, hierarchical clustering, and DBSCAN. Let's assume we choose the K-means algorithm for clustering.

[0131] Determining the K value: The K value represents how many groups the data should be divided into. The elbow rule or silhouette coefficient can be used to determine the optimal K value. For example, the elbow rule might find that K=3 is the most suitable number of groups.

[0132] Run the clustering algorithm: Use the chosen K-means algorithm, K=3, and input the preprocessed data into the algorithm. The algorithm will attempt to find the best way to divide the data into three groups, which may represent high, moderate, and low blood pressure risk groups.

[0133] Interpreting Groups: Each group may represent different blood pressure levels and risk factors based on the characteristics of the patients it includes. For example:

[0134] Group 1 may include young, normal-weight patients with no family history of hypertension, representing a low-risk group.

[0135] Group 2 may include patients who are of middle age, overweight, and have a family history of hypertension, representing a medium-risk group.

[0136] Group 3 may include older, obese patients and those with a history of heart disease, representing a high-risk group.

[0137] For further analysis: These groups can be used for further analysis and intervention, such as developing personalized treatment and prevention plans for patients at different risk levels.

[0138] S24. Evaluate and verify the results of the clustering.

[0139] For example, S24 can be achieved through the following steps:

[0140] The previous step used the K-means algorithm to divide patients with high blood pressure into three distinct groups, representing different blood pressure levels, risk factors, and disease stages. The next step is to evaluate and validate the results of this grouping.

[0141] Internal assessment:

[0142] SSE (Sum of Squared Errors): Calculates the sum of squared distances from each point to the center of its group. The smaller the SSE, the more similar the members within the group.

[0143] Silhouette coefficient: Measures the difference between the similarity of each sample to other samples in its own group and the similarity to its nearest neighbor group. The closer the value is to 1, the better the sample fits its group.

[0144] External evaluation (if real group labels are available):

[0145] Adjusted RAND index: measures the similarity of two data segments, taking into account the effect of random assignment. The value ranges from -1 to 1. The larger the value, the more the clustering result matches the actual label.

[0146] The Fowlkes-Mallows index measures the consistency between the clustering results and the actual categories. The closer the value is to 1, the better the consistency.

[0147] Visual analysis:

[0148] Scatter plot: Visualize different groups by using different colors or markers to represent different groups, so as to intuitively show the effect of clustering.

[0149] Distribution plot: Shows the distribution of important features among the various groups, helping to understand the differences between the groups.

[0150] Practical application evaluation:

[0151] Patient feedback: The treatment plan after grouping was applied in practice, and feedback was collected from patients and doctors to understand whether grouping really helps to provide more personalized treatment.

[0152] Long-term follow-up: Track the health status of patients after grouping to see if it matches the expected group characteristics (such as high risk, low risk, etc.).

[0153] These methods allow for a comprehensive evaluation and validation of the clustering effect, ensuring that the clustering is not only mathematically sound but also meaningful and valuable in practical applications. Such evaluation and validation are an indispensable part of the data analysis process, contributing to improved reliability and accuracy of the final model.

[0154] S3. Control the cloud server to select patient data from each group to establish training and test sets, train machine learning models to obtain blood pressure prediction models, and verify and adjust the models based on the test sets.

[0155] In some embodiments, refer to Figure 4 S3 can be achieved through the following sub-steps:

[0156] S31. Select data from each blood pressure classification group, and divide the data in each group into a training set and a test set based on a pre-selected partitioning strategy and partitioning ratio. The training set is used to train the model, and the test set is used to validate and adjust the model. The pre-selected partitioning strategy is random partitioning, hierarchical partitioning, or cross-validation.

[0157] For example, S31 can be implemented through the following steps:

[0158] Suppose that patients with high blood pressure have been divided into three groups, representing low, moderate, and high blood pressure levels, respectively. The task now is to take a subset of data from each of these three groups to train and test a machine learning model.

[0159] Splitting strategy selection: Choose a splitting strategy, such as random splitting. In random splitting, samples are randomly selected from each group, regardless of any relationships between the samples.

[0160] Determine the split ratio: Determine the ratio of training set to test set, for example, 80% of the data is used for training and 20% of the data is used for testing.

[0161] Specific division process:

[0162] Low blood pressure group: If the group has 1000 samples, 800 samples are randomly selected as the training set and the remaining 200 samples are used as the test set.

[0163] For the moderate blood pressure group: In the same way, if there are 1,500 samples in the moderate blood pressure group, 1,200 are used for training and 300 are used for testing.

[0164] Hypertension group: If the hypertension group has 500 samples, 400 are used for training and 100 are used for testing.

[0165] Optional cross-validation: To improve the robustness of the model, cross-validation can also be used. For example, with 5-fold cross-validation, the training set for each group is further divided into 5 parts, 4 of which are used in each training iteration, and the remaining parts are used for validation.

[0166] Ensure stratification: When the sample size of the hypertension group is small, stratified sampling can be used to ensure that the proportion of hypertension samples in the training set and the test set is consistent with the proportion in the original group.

[0167] S32. Select a multi-class logistic regression algorithm model as the machine learning algorithm model, and normalize the features in the training set.

[0168] For example, S32 can be implemented through the following steps:

[0169] Choosing a multi-class logistic regression model: Since the blood pressure classification problem is a multi-class problem (e.g., low, moderate, and high blood pressure), multi-class logistic regression is chosen as the classification model. This is a linear model widely used for multi-class classification problems.

[0170] Prepare training features: Assume the following features are used for training: age, gender, weight, height, heart rate, cholesterol level, etc.

[0171] Normalization: Since the dimensions and ranges of these features may vary greatly, they need to be normalized so that the mean of each feature is 0 and the standard deviation is 1. This helps optimization algorithms such as gradient descent converge faster.

[0172] Calculate the mean and standard deviation: For each feature, calculate the mean and standard deviation of that feature in the training set.

[0173] Apply normalization: Using the mean and standard deviation mentioned above, transform each feature in the training set into (feature value - mean) / standard deviation.

[0174] Building a multi-class logistic regression model: Using normalized features, construct a multi-class logistic regression model. This can typically be achieved using machine learning libraries such as Scikit-Learn by calling the corresponding multi-class logistic regression function.

[0175] Setting model parameters: Depending on the specific data, you may need to set some model parameters, such as the strength of the regularization term and the maximum number of iterations.

[0176] S33. Train the multi-class logistic regression algorithm based on the training set, and update the parameters by minimizing the classification error in each iteration.

[0177] For example, S33 can be implemented through the following steps:

[0178] Initialize parameters: The parameters of multi-class logistic regression consist of weights and biases. They can be initialized randomly or set to zero.

[0179] Defining a loss function: To train a model, a loss function is needed to measure the difference between the model's predictions and the true labels. In multi-class logistic regression, the cross-entropy loss function is typically used.

[0180] Choosing an optimization algorithm: An optimization algorithm is needed to tune the model's parameters to minimize the loss function. Commonly used optimization algorithms include stochastic gradient descent (SGD), Adam, etc.

[0181] Training the model:

[0182] Iterative training: Set the number of iterations, for example, 1000. Perform the following operations in each iteration:

[0183] Forward propagation: Calculates the predicted values ​​for all samples on the training set using the current parameters.

[0184] Calculate the loss: Use the predicted values ​​and the true labels to calculate the value of the loss function.

[0185] Backpropagation: Calculate the gradient of the loss function with respect to each parameter.

[0186] Update parameters: Update the model parameters based on the calculated gradient and the selected learning rate.

[0187] Monitor the training process: Monitor the training loss and validation loss after each iteration or a set of iterations to understand whether the model is learning and to ensure that there is no overfitting.

[0188] Evaluate the model: Use the trained model to evaluate it on the validation or test set to check the model's generalization ability.

[0189] Save the model: If the model performs satisfactorily, the trained parameters can be saved for future use.

[0190] S34. Validate the trained model using test set data and adjust the model based on its performance on the test set.

[0191] For example, S34 can be implemented through the following steps:

[0192] Forward propagation: Using the weights and bias parameters obtained during training, predictive values ​​are computed on the test set. This involves inputting the features of the test set into the model and obtaining a blood pressure category prediction for each sample through the model.

[0193] Calculate evaluation metrics: Select appropriate evaluation metrics, such as accuracy, precision, and recall, to measure the model's performance on the test set.

[0194] These metrics are calculated and compared with real blood pressure categories to obtain a performance evaluation of the model on unseen data.

[0195] Error analysis: Analyze the model's incorrect predictions on the test set in an attempt to understand the model's flaws and limitations.

[0196] For example, it might be found that the model performs poorly on hypertension samples, which may suggest that we need more training samples on hypertension or improvements in feature engineering.

[0197] Model tuning: Based on performance and error analysis on the test set, determine whether model tuning is necessary. Possible tuning options include changing hyperparameters (e.g., learning rate, regularization parameters), adding more features, and modifying preprocessing steps.

[0198] If adjustments are made, return to the training steps to retrain the model and validate it again using the test set.

[0199] Final evaluation and reporting: After completing all necessary adjustments, provide a complete report on the final performance of the model on the test set.

[0200] This may include not only metrics such as accuracy, but also confusion matrices, ROC curves, and other metrics to comprehensively demonstrate the model's performance.

[0201] Model deployment preparation: If the test results are satisfactory, the model can be prepared for actual deployment, which may involve further optimization and encapsulation for use in the actual dialysis device control environment.

[0202] S4. Download and update the parameters of the blood pressure prediction model in the memory, input the data collected by the data collection device into the blood pressure prediction model, and obtain the control parameters of the dialysis device based on the blood pressure prediction model.

[0203] In some embodiments, refer to Figure 5 S4 can be implemented through the following sub-steps:

[0204] S41. Input the data collected by the data collection device, the standard dialysate temperature, the standard dialysate flow rate, and the standard dialysis duration into the blood pressure prediction model to predict future blood pressure changes.

[0205] For example, in a dialysis center, dialysis machines collect patients' physiological data in real time (such as heart rate, blood flow rate, etc.), dialysate temperature, flow rate, and dialysis duration via data collection devices. This data is sent in real time to a blood pressure prediction model on a cloud server. The model analyzes this data and, based on previous training, predicts blood pressure changes over a future period. For example, it predicts the possible upward or downward trend of blood pressure within the next 30 minutes.

[0206] S42. Input the future blood pressure change predicted by the blood pressure prediction model into a preset reward function for calculation and obtain a reward value. The output of the preset reward function is related to the difference between the future blood pressure change predicted by the blood pressure prediction model and the actual measured blood pressure change. The larger the difference, the lower the reward value, and the smaller the difference, the lower the reward value.

[0207] For example, the blood pressure changes predicted by the predictive model are fed into a pre-defined reward function. This reward function might be a mathematical equation whose purpose is to assign a reward value based on the difference between the prediction and the actual measurement. For instance, a high reward value is given if the prediction is accurate, and a low reward value is given if the prediction is inaccurate. In real-world scenarios, this can be used to adjust dialysis parameters to keep the patient's blood pressure within an ideal range.

[0208] S43. Based on the reinforcement learning algorithm and the change of reward value, adjust the dialysate volume, dialysate speed and dialysate temperature to increase the reward value, thereby obtaining new dialysate volume, dialysate speed and dialysate temperature. Substitute them into the previous step and execute until the dialysate volume, dialysate speed and dialysate temperature converge to a stable range, thereby obtaining the control parameters of the dialysis device.

[0209] S5. Control the dialysate temperature, dialysate flow rate and dialysis duration of the dialysis device based on the control parameters of the dialysis device.

[0210] In some embodiments, refer to Figure 6 S5 can be implemented through the following sub-steps:

[0211] S51. Receive dialysis device control parameters generated from the blood pressure prediction model.

[0212] S52. Based on the received parameters, set the dialysate temperature, dialysate flow rate, and dialysis duration of the dialysis device.

[0213] For example, a dialysis machine can communicate with a cloud server. The cloud server runs a blood pressure prediction model that analyzes large amounts of data, such as physiological data, dialysate temperature, dialysate flow rate, and dialysis duration, to predict future blood pressure changes and generate control parameters for the dialysis device accordingly.

[0214] In a specific dialysis procedure, the patient begins dialysis treatment, and the dialysis machine establishes a connection with the cloud server. As the dialysis process progresses, the blood pressure prediction model continuously receives new data and dynamically calculates new control parameters for the dialysis device based on methods such as reinforcement learning and the previously described reward function. These parameters may include the temperature and flow rate of the dialysate, and the duration of dialysis.

[0215] The dialysis machine receives these newly calculated control parameters in real time and automatically adjusts the dialysis process accordingly to ensure stable blood pressure for the patient. For example, if a blood pressure prediction model predicts that the patient's blood pressure may drop, the dialysis unit may respond by increasing the dialysate flow rate.

[0216] S53. Adjust the dialysis device in real time to match the set parameters.

[0217] S54. Continuously monitor the dialysis process and collect real-time blood information as feedback, and adjust the dialysate temperature, dialysate flow rate and dialysis duration of the dialysis device based on the feedback.

[0218] For example, the system can use devices such as cameras to determine if a user is experiencing tremors, and if so, assess the severity of the tremors. Based on the severity of the tremors, the current dialysate temperature, and the output flow rate, it calculates the necessary adjustments to the dialysate temperature and output flow rate. According to the adjustment values ​​calculated in the previous step, the dialysate temperature and output flow rate are adjusted, and the user's tremors are continuously monitored to see if they improve. If so, the dialysate temperature and output flow rate are further adjusted until the user's tremor severity is below a standard level. A concrete example: Suppose there is a simple rule that when the weighted average of the frequency and amplitude of tremors exceeds a certain threshold (e.g., 10), the tremors are considered severe, and in this case, the dialysate temperature might need to be increased by 0.2 degrees Celsius and the dialysate flow rate increased by 10%. If the weighted average of the frequency and amplitude of tremors is below the threshold, but tremors still occur, for example, a value of 5, then only a 0.1 degree Celsius increase in dialysate temperature and a 5% increase in dialysate flow rate might be needed. This rule is a very simple example; in practical applications, more complex rules or models may be needed to determine how to adjust the dialysate temperature and output flow rate.

[0219] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0220] In one embodiment, a computer device is provided, which may be a server. The computer device includes a processor, memory, a network interface, and a database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database contains data related to a dialysis control program. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a dialysis control program.

[0221] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the dialysis control program of the above embodiment.

[0222] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the dialysis control program of the above embodiment.

[0223] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments of this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0224] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0225] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. An end-stage renal disease hemodialysis blood pressure pattern recognition and intervention system, comprising: include: Dialysis devices are used to perform hemodialysis on users; Data collection device used to collect users' physiological data, medical record data, climate data, and dialysis treatment data during the operation of the dialysis device; The memory stores the dialysis control program, and the processor loads the dialysis control program to perform the following steps: S1. Upload the data collected by the data collection device to the cloud server, and control the cloud server to retrieve the historical database. Based on the data collected by the data collection device and the historical database, filter and combine the data to obtain blood pressure-related characteristics. S2. Control the cloud server to generate classification factors based on the obtained blood pressure-related features, and use the classification factors to group patients into different groups; S3. Control the cloud server to select patient data from each group to establish training and test sets, train the machine learning model to obtain the blood pressure prediction model, and verify and adjust the model based on the test set; S4. Download and update the parameters of the blood pressure prediction model in the memory, input the data collected by the data collection device into the blood pressure prediction model, and obtain the control parameters of the dialysis device based on the blood pressure prediction model; S5. Control the dialysate temperature, dialysate flow rate, and dialysis duration of the dialysis device based on the dialysis device control parameters; S4 includes the following sub-steps: S41. Input the data collected by the data collection device, the standard dialysate temperature, the standard dialysate flow rate, and the standard dialysis duration into the blood pressure prediction model to predict future blood pressure changes; S42. Input the future blood pressure change predicted by the blood pressure prediction model into a preset reward function for calculation and obtain a reward value. The output of the preset reward function is related to the difference between the future blood pressure change predicted by the blood pressure prediction model and the actual measured blood pressure change. The larger the difference, the lower the reward value; the smaller the difference, the higher the reward value. S43. Based on the reinforcement learning algorithm and the change of reward value, adjust the dialysate volume, dialysate speed and dialysate temperature to increase the reward value, thereby obtaining new dialysate volume, dialysate speed and dialysate temperature. Substitute them into the previous step and execute until the dialysate volume, dialysate speed and dialysate temperature converge to a stable range, thereby obtaining the control parameters of the dialysis device.

2. The end-stage renal disease hemodialysis blood pressure pattern recognition and intervention system of claim 1, wherein, It also includes an operation screen for acquiring the user's physiological data, and the data collection device is capable of performing the following steps: Acquire basic body feature parameters input to the operation screen for display, store and update in the body feature information database; Real-time blood information is acquired using a data acquisition device, stored, and updated in a blood information database. User body characteristic parameters are obtained based on historical information within the latest time period from a database of body characteristic information and blood information.

3. The end-stage renal disease hemodialysis blood pressure pattern recognition and intervention system of claim 2, wherein, S1 includes the following sub-steps: S11. Upload the physiological data, medical record data, climate data, and dialysis treatment data collected by the data collection device to the cloud server; S12. Control the cloud server to retrieve historical databases related to the collected data; S13. Filter the data obtained from the data collection device and historical database to remove irrelevant or invalid information; S14. Combine or integrate the filtered data; S15. Based on the filtered and combined data, extract features related to blood pressure.

4. The end-stage renal disease hemodialysis blood pressure pattern recognition and intervention system of claim 3, wherein, S2 includes the following sub-steps: S21. Analyze blood pressure-related characteristics and generate classification factors based on the analysis results; S22. Preprocess the data to suit the clustering algorithm; S23. Based on the generated classification factors, users are grouped into different groups, wherein the groups represent different blood pressure levels, risk factors, and disease stages; S24. Evaluate and verify the results of the clustering.

5. The blood pressure pattern recognition and intervention system for hemodialysis in end-stage renal disease according to claim 4, characterized in that, S3 includes the following sub-steps: S31. Select data from each blood pressure classification group, and divide the data in each group into a training set and a test set based on a pre-selected partitioning strategy and partitioning ratio. The training set is used to train the model, and the test set is used to validate and adjust the model. The pre-selected partitioning strategy is random partitioning, hierarchical partitioning, or cross-validation. S32. Select the multi-class logistic regression algorithm model as the machine learning algorithm model, and normalize the features in the training set; S33. Train the multi-class logistic regression algorithm based on the training set, and update the parameters by minimizing the classification error in each iteration; S34. Validate the trained model using test set data and adjust the model based on its performance on the test set.

6. The blood pressure pattern recognition and intervention system for hemodialysis in end-stage renal disease according to claim 5, characterized in that, S5 includes the following sub-steps: S51. Receive dialysis device control parameters generated from the blood pressure prediction model; S52. Based on the received parameters, set the dialysate temperature, dialysate flow rate, and dialysis duration of the dialysis device; S53. Adjust the dialysis device in real time to match the set parameters; S54. Continuously monitor the dialysis process and collect real-time blood information as feedback, and adjust the dialysate temperature, dialysate flow rate and dialysis duration of the dialysis device based on the feedback.

7. A computer device, characterized in that, It includes: One or more processors; Memory; One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, the one or more applications being configured to: execute the dialysis control program according to any one of claims 1 to 6.

8. A computer-readable storage medium, characterized in that, The storage medium stores at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, the at least one program, the code set, or instruction set is loaded and executed by a processor to implement: the dialysis control program as described in any one of claims 1 to 6.

Images