Blood purification system control method and device based on neural network

Abstract

The invention relates to a blood purification system control method and device based on a neural network, and aims to overcome the defects of a traditional fixed treatment mode in processing special cases or coping with emergencies. According to the method, current treatment parameters, such as the blood flow rate, the dialysate flow and the ultrafiltration rate, of a patient detected by a blood purification system are obtained, data analysis is conducted on the parameters through a preset neural network model, control parameters are obtained, and the system is adjusted in real time. Based on the feedback mechanism, the method continuously optimizes the treatment regimen until the parameters of the patient meet the preset conditions. The system can dynamically respond to physical sign changes of the patient, and personalized and intelligent treatment management is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a blood purification system control method and device based on a neural network. Background Art

[0002] In the current development of medical technology, blood purification systems based on fixed treatment modes have been widely used in many clinical settings. This system operates according to a preset treatment plan and is designed to provide stable and reliable treatment effects. However, the fixed mode management method is often insufficient when dealing with special cases or responding to emergencies. This model lacks flexibility and may not be able to effectively cope with individual differences and complex clinical conditions, thus affecting the individualization and optimization of treatment effects. Therefore, it is urgent to introduce more advanced and flexible control methods into the existing treatment mode to improve the adaptability and response capabilities of the system. Summary of the invention

[0003] The main purpose of the present invention is to provide a blood purification system control method and device based on a neural network to solve the technical problem that the management method of the blood purification system based on a fixed treatment mode is often insufficient when dealing with special cases or responding to emergencies.

[0004] To achieve the above-mentioned purpose, the present invention provides a blood purification system control method based on a neural network, comprising the following steps: obtaining current treatment parameters of a target patient detected by the blood purification system, wherein the treatment parameters include at least any one of the following: blood flow rate, dialysate flow rate and ultrafiltration rate; using a preset neural network model to perform data analysis on at least the current treatment parameters to obtain control parameters of the blood purification system, and using the control parameters to adjust the blood purification system; and continuing to execute the step of "obtaining the current treatment parameters detected by the blood purification system" based on a feedback mechanism until the current treatment parameters detected by the blood purification system meet preset conditions.

[0005] The present invention also provides a blood purification system control device based on a neural network, comprising: an acquisition unit, used to acquire the current treatment parameters of the target patient detected by the blood purification system, wherein the treatment parameters include at least any one of the following: blood flow rate, dialysate flow rate and ultrafiltration rate; an adjustment unit, used to use a preset neural network model to perform data analysis on at least the current treatment parameters, obtain the control parameters of the blood purification system, and use the control parameters to adjust the blood purification system; a feedback unit, used to continue to execute the step of "obtaining the current treatment parameters detected by the blood purification system" based on a feedback mechanism until the current treatment parameters detected by the blood purification system meet preset conditions.

[0006] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.

[0007] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned methods are implemented.

[0008] The blood purification system control method and device based on neural network provided by the present invention can achieve more accurate control effect by automatically analyzing and adjusting the treatment parameters in the blood purification system using the neural network model, especially when dealing with special cases or responding to emergencies, the treatment parameters can be dynamically adjusted according to real-time feedback, avoiding the limitation of difficult to respond flexibly in the traditional fixed mode. The method uses a neural network to perform multi-dimensional analysis of treatment parameters such as blood flow rate, dialysate flow rate and ultrafiltration rate, and continuously optimizes the treatment plan based on the feedback mechanism, making the blood purification process more personalized and intelligent. In addition, the system can timely identify and respond to changes in patient signs through continuous parameter collection and analysis, and improve the safety and effectiveness of treatment. The present invention also supports the combination with computer equipment and readable storage media, and realizes automatic control through the execution of computer programs, which reduces the need for manual intervention, enhances the robustness and adaptability of the system, and is suitable for promotion and application in complex medical environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 is a schematic diagram of a blood purification system control method based on a neural network in one embodiment of the present invention; Figure 2 is a structural diagram of a blood purification system control device based on a neural network in one embodiment of the present invention; Figure 3 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0010] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0011] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0012] Reference Figure 1 The embodiment of the present invention provides a blood purification system control method based on a neural network, comprising the following steps: S1, obtaining current treatment parameters of the target patient detected by the blood purification system, wherein the treatment parameters include at least any one of the following: blood flow rate, dialysate flow rate and ultrafiltration rate.

[0013] S2, using a preset neural network model to perform data analysis on at least the current treatment parameters to obtain control parameters of the blood purification system, and using the control parameters to adjust the blood purification system.

[0014] S3, continuing to execute the step of "obtaining the current treatment parameters detected by the blood purification system" based on the feedback mechanism until the current treatment parameters detected by the blood purification system meet the preset conditions.

[0015] The core of this technical solution is to use advanced neural network models to dynamically adjust treatment parameters through analysis of real-time data, thereby overcoming the limitations of fixed patterns. First, in this technical solution, the system establishes a data foundation by acquiring the current treatment parameters of the target patient (such as blood flow rate, dialysate flow rate, and ultrafiltration rate). These parameters are key indicators in the blood purification process and are directly related to the effectiveness and safety of the treatment. By collecting these parameters in real time, the system can obtain accurate data on the current treatment status and provide necessary information for subsequent analysis and adjustment.

[0016] Next, the preset neural network model is used to perform data analysis on the current treatment parameters to obtain the control parameters of the blood purification system. This neural network model is trained to learn various complex patterns and relationships from historical data, so as to effectively predict and analyze new data. For example, during the model training phase, the system may use a convolutional neural network (CNN) to process imaging data and extract spatial features related to lesions, while using a recurrent neural network (RNN) to analyze time series data and capture the long-term dependencies of physiological indicators. Through the comprehensive application of these neural network models, the system can accurately predict treatment effects and potential complications and generate corresponding control parameters.

[0017] Based on these control parameters, the system adjusts the blood purification equipment to optimize the treatment process. The adjustment of control parameters can include changing the blood flow rate, adjusting the dialysate flow rate, or optimizing the ultrafiltration rate. This adjustment can not only improve the effect of treatment, but also better adapt to the individual differences and actual conditions of patients. For example, if the neural network model predicts that a patient may experience a decrease in blood flow, the system can immediately adjust the blood flow rate to prevent possible complications.

[0018] In order to ensure that the adjusted treatment parameters can continue to adapt to the actual situation of the patient, the system introduces a feedback mechanism. The core of the feedback mechanism is to dynamically adjust the system's real-time monitoring and parameter adjustment. Specifically, the system will continue to execute the steps of obtaining the current treatment parameters and compare them with the preset conditions. If the current parameters still do not meet the preset conditions, the system will continue to adjust the control parameters based on the latest data. This dynamic adjustment capability ensures that the system can respond to changes in the patient's status in a timely manner, further improving the personalization and optimization of treatment.

[0019] For example, in a specific case, the patient may show normal physiological indicators during the initial treatment phase. However, as treatment progresses, the patient's blood pressure may fluctuate. The system monitors these changes in real time and uses a neural network model to predict future trends in blood pressure. If the model predicts that blood pressure will continue to fluctuate, the system will automatically adjust treatment parameters, such as increasing the dialysate flow rate or changing the ultrafiltration rate, to prevent abnormal blood pressure from affecting the patient's health. This flexible adjustment capability greatly improves the adaptability of the system, enabling it to maintain efficient treatment effects in the face of complex and changing clinical situations.

[0020] In summary, the embodiment of the present invention effectively solves the shortcomings of fixed treatment modes in dealing with special cases and responding to emergencies by introducing a control method based on a neural network. Through real-time data acquisition, neural network model analysis, control parameter adjustment, and dynamic adjustment of the feedback mechanism, the system can achieve a higher level of personalized treatment and optimization, thereby significantly improving the adaptability and response capabilities of the blood purification system. Such a technical solution not only makes up for the shortcomings of traditional treatment modes, but also promotes the advancement of medical technology, and has broad application prospects and important clinical significance.

[0021] In one example, before using a preset neural network model to perform data analysis on the treatment parameters to obtain the control parameters of the blood purification system, the blood purification system control method also includes: obtaining medical data involved in the treatment process of the blood purification device, and integrating the medical data into a sample set, wherein the medical data includes: the treatment parameters of the patient detected by the blood purification device, and the patient's vital sign parameters and patient medical images detected by other treatment devices, and the multiple samples in the sample set correspond to one patient respectively; constructing a convolutional neural network, and automatically learning the sample set based on the convolutional neural network to learn to identify the spatial characteristics and local correlation of each sample data in the sample set , determine the lesion information corresponding to each sample data in the sample set; construct a recursive neural network, and automatically learn the sample set based on the recursive neural network to capture the long-term dependency of each sample data in the sample set, and determine the prediction information corresponding to each sample data in the sample set; integrate the trained convolutional neural network and recursive neural network to obtain an initial neural network model, train the initial neural network model based on the outputs of the convolutional neural network and the recursive neural network to fuse the outputs of the convolutional neural network and the recursive neural network, and output the control parameters of the blood purification system based on the fusion result to obtain the final preset neural network model.

[0022] That is, by acquiring the medical data generated by the blood purification device during the treatment process, and integrating these data into a sample set. The medical data includes not only the patient treatment parameters detected by the blood purification device (such as blood flow rate, dialysate flow rate and ultrafiltration rate), but also the patient's vital signs parameters (such as blood pressure, heart rate, etc.) and patient medical images (such as vascular structure diagrams or tissue sections) detected by other treatment equipment. This sample set contains a large amount of patient data, and each sample data represents a patient's complete treatment record and related medical information. By integrating these data, the system can establish a comprehensive database to provide a sufficient information basis for subsequent analysis and model training.

[0023] Next, the system builds a convolutional neural network (CNN) and automatically learns and processes the sample set based on the network. Convolutional neural networks are good at extracting spatial features and local correlations from high-dimensional data, which is particularly important for processing medical data with complex structures (such as imaging data). CNN can automatically identify the spatial features of each sample data in the sample set and determine the lesion information from it. For example, by analyzing the patient's medical images, CNN can detect abnormal tissue structures or lesion areas, and use this information as an important basis for adjusting the treatment plan.

[0024] At the same time, the system also builds a recursive neural network (RNN) and uses it to learn and process the sample set. Recursive neural networks are particularly suitable for processing time series data and can capture long-term dependencies in the data. During blood purification treatment, the patient's physiological indicators (such as blood pressure, heart rate, etc.) change over time. By analyzing these time series data, RNN can identify the dynamic change trend of the patient's condition and predict possible physiological reactions in the future. For example, RNN can predict the patient's possible blood pressure changes in the next treatment, helping doctors adjust the treatment plan in advance to deal with possible complications.

[0025] Once both CNN and RNN are trained, the system integrates them to obtain an initial neural network model. This initial model combines the capabilities of CNN in spatial feature extraction and RNN in time series prediction, thereby comprehensively utilizing the advantages of both networks. Then, based on the outputs of CNN and RNN, the system further trains the initial neural network model. The core of this step is to fuse the output results of CNN and RNN to optimize the overall performance of the model. This fusion not only improves the ability to process complex data, but also improves the accuracy of predicting treatment parameters.

[0026] Finally, based on the fusion results, the system outputs the control parameters of the blood purification system and obtains a final preset neural network model. These control parameters are used to adjust the operation of the blood purification equipment in real time to optimize the treatment effect. Through this series of steps, the system can dynamically adjust the treatment parameters during the treatment process to adapt to the individual differences and real-time status changes of patients, thereby significantly improving the personalization and effect of treatment.

[0027] In general, the technical solution provided by the present invention combines the advantages of convolutional neural networks and recursive neural networks, fully utilizes the spatial characteristics and time series information in the patient's medical data, and can achieve more accurate and flexible treatment control.

[0028] In one example, a convolutional neural network is constructed, and the sample set is automatically learned based on the convolutional neural network to learn and identify the spatial features and local correlations of each sample data in the sample set, and determine the lesion information corresponding to each sample data in the sample set, including: dividing each sample data in the sample set into image data and time series detection data, denoising and contrast enhancement processing for the image data, and smoothing and normalizing the time series detection data; annotating lesion information and relationship feature annotation for the image data that has undergone the denoising and contrast enhancement processing and the time series detection data that has undergone the smoothing and normalization processing, and reintegrating into a sample set based on the annotated data; constructing the convolutional neural network, inputting the annotated sample set into the convolutional neural network, and using a back propagation algorithm to adjust the parameters of the convolutional neural network based on the annotated sample set to obtain a trained convolutional neural network.

[0029] In this example, the process of building a convolutional neural network (CNN) and how to perform data processing based on the network to identify lesion information of sample data can be explained in detail as follows: First, each sample data in the sample set is divided into two main types: imaging data and time series detection data. Imaging data may include medical images such as CT scan images or MRI images, while time series detection data involves physiological parameters continuously recorded during the treatment of patients, such as blood pressure, heart rate, etc. These two types of data are processed differently, so they need to be treated separately during the processing.

[0030] Image data processing includes denoising and contrast enhancement. The purpose of denoising is to remove noise in the image caused by equipment limitations or external factors to improve image quality. Contrast enhancement helps to highlight the key structures in the image, making the lesion area more prominent, thereby helping the convolutional neural network to better learn and identify lesion information in the image.

[0031] The processing of time series detection data includes smoothing and normalization. Smoothing reduces noise or abnormal fluctuations in the data, making the time series data more stable and convenient for subsequent analysis. Normalization scales the time series data to a standard range to eliminate differences between different data scales, making data processing more consistent and effective.

[0032] After completing the preliminary data processing, the image data and time series detection data are annotated with lesion information and relationship features. Lesion information annotation refers to marking the lesion areas in the image data so that the neural network can learn the characteristics of these areas. Relationship feature annotation refers to annotating the key trends and patterns in the time series data to help the network understand the important relationships in the data. Through these annotations, the data of the sample set is reintegrated to ensure that the annotation information can correctly reflect the actual situation of the sample.

[0033] Next, a convolutional neural network is constructed and the labeled sample set is input into the network. Convolutional neural networks are particularly effective when processing image data because they can automatically extract spatial features in the data, such as edges, textures, and shapes. By performing convolution operations on the input data, the network can extract and learn these features layer by layer and form a deep understanding of the image content.

[0034] During the training process, the back propagation algorithm is used to adjust the parameters of the convolutional neural network. The back propagation algorithm is an optimization technique that gradually improves the prediction accuracy of the network by calculating the error between the predicted result and the true annotation and adjusting the weights and biases in the network based on this error. After multiple iterations, the convolutional neural network will gradually adjust its parameters so that the network can accurately identify the lesion area in the image data.

[0035] Finally, through the above processing and training steps, a trained convolutional neural network is obtained. This network can effectively identify and classify lesion information in image data, providing support for subsequent medical analysis and decision-making. Through such a processing flow, the convolutional neural network can extract useful information from complex medical data and improve the accuracy and efficiency of diagnosis.

[0036] In other words, the advantages of convolutional neural networks (CNNs) in medical data processing are mainly reflected in data feature extraction and processing of complex multidimensional data. CNN can automatically extract spatial features and local correlations from high-dimensional data, and is particularly suitable for processing medical imaging data such as vascular structure diagrams and tissue sections, as well as multidimensional monitoring data such as time series of blood pressure and heart rate. Through convolution kernels, CNN can identify complex pathological patterns and predict treatment responses and the probability of complications. In addition, when processing multidimensional data generated during blood purification (such as blood parameter change diagrams and 3D medical imaging data), CNN can effectively perform feature extraction and pattern recognition, thereby improving the accuracy of diagnosis and treatment.

[0037] In one example, a recursive neural network is constructed, and the sample set is automatically learned based on the recursive neural network to capture the long-term dependency of each sample data in the sample set and determine the prediction information corresponding to each sample data in the sample set, including: dividing each sample data in the sample set into time series detection data, smoothing and normalizing the time series detection data, and reintegrating into a sample set based on the smoothed and normalized time series detection data; using a long short-term memory network to construct the recursive neural network, inputting the labeled sample set into the recursive neural network, using a combination of a forward propagation algorithm and a back propagation algorithm, and adjusting the parameters of the recursive neural network based on the labeled sample set to obtain a trained recursive neural network.

[0038] That is, the data in the sample set is divided into time series detection data. These data are records of changes over time, such as the patient's blood pressure, heart rate, blood oxygen saturation, etc. These data have temporal continuity and dependence. Therefore, in order to carry out effective processing and analysis, time series data needs to be specially processed.

[0039] The time series detection data is smoothed and normalized. The purpose of smoothing is to reduce noise or irregular fluctuations in the data, making the data more stable and easier to analyze. Common smoothing methods include moving average or low-pass filtering, which can remove random fluctuations in time series. Normalization converts the data to a standard range (for example, between 0 and 1) to eliminate differences in different magnitudes or units, making subsequent processing and analysis more consistent. The processed data is reintegrated into a sample set to ensure that the format and range of all time series data are consistent and ready for further analysis. The key to this step is to ensure that all data is unified in format and processing so that the recursive neural network can effectively learn and process the data.

[0040] Next, a long short-term memory (LSTM) network is used to build a recursive neural network. LSTM is a special type of recursive neural network that is particularly suitable for processing and predicting long-term dependencies in time series data. The LSTM network dynamically controls the preservation and forgetting of information through its built-in gate mechanisms (such as input gates, forget gates, and output gates), thereby capturing long-span dependencies in the data. This feature is particularly important for patient data during blood purification, because the patient's status may not change significantly until hours or days later, and LSTM can effectively learn these long-term dependencies.

[0041] The processed sample set is input into the constructed LSTM network. The network's forward propagation process processes the input data through the network layer and generates a prediction result. During the training process, the network's output is compared with the actual labeled data to calculate the error. This error is used to guide the network's back-propagation process, that is, to adjust the parameters in the network (such as weights and biases) through the back-propagation algorithm. The back-propagation algorithm gradually reduces the prediction error and improves the accuracy of the model by calculating the gradient and updating the network parameters.

[0042] Through this training method that combines forward propagation and back propagation algorithms, the recursive neural network can gradually learn the long-term dependencies in time series data, thereby improving the ability to predict future data trends. Ultimately, the trained recursive neural network can accurately capture the long-term dependency information in the sample data, provide support for the decision-making of the blood purification system, and improve the prediction and personalization of treatment effects.

[0043] In one example, the trained convolutional neural network and recursive neural network are integrated to obtain an initial neural network model, the initial neural network model is trained based on the outputs of the convolutional neural network and the recursive neural network to fuse the outputs of the convolutional neural network and the recursive neural network, and the control parameters of the blood purification system are output based on the fusion result to obtain the final preset neural network model, including: output control parameter labeling processing of the standardized outputs of the trained convolutional neural network and the recursive neural network for the same sample data to obtain target sample data, and integrating them into a target sample set based on multiple target sample data; based on the standardized outputs of the trained convolutional neural network and the recursive neural network, an integrated initial neural network model is constructed, wherein the initial neural network model uses a cascade connection mechanism to connect the standardized outputs of the convolutional neural network and the recursive neural network together; the target sample set is input into the initial neural network model, and a forward propagation algorithm is used to calculate the loss function of the recursive neural network based on the labeled target sample set, and a back propagation algorithm is used to optimize the network parameters of the recursive neural network based on the labeled target sample set to obtain a trained neural network model.

[0044] That is, the trained convolutional neural network and recurrent neural network are applied to the same sample data, and their standardized outputs are annotated with control parameters. The purpose of this step is to obtain target sample data, which represents the output of the convolutional neural network and the recurrent neural network when processing the same data. Standardized output ensures that the outputs of different networks have the same scale or range, making subsequent integration and comparison more consistent. Through the annotation process, each sample data can be clearly associated with the corresponding control parameters, thus providing a clear basis for the integration step.

[0045] Next, based on these target sample data, an integrated initial neural network model is constructed. This initial model uses a cascade connection mechanism to connect the standardized outputs of the convolutional neural network and the recurrent neural network. The cascade connection mechanism refers to connecting or combining the outputs of the two networks in series so that these outputs can be used together in an integrated model. This connection method enables the model to simultaneously utilize the spatial feature extraction capabilities of the convolutional neural network and the time series prediction capabilities of the recurrent neural network, thereby forming a more comprehensive processing system.

[0046] Then, the target sample set is input into the constructed initial neural network model. These target sample sets include the data obtained from the previous annotation processing, covering the different processing results of the convolutional neural network and the recurrent neural network on the samples. Through the forward propagation algorithm, the initial neural network model will calculate the prediction results for these samples, compare them with the actual annotated data, and generate the loss function of the recurrent neural network. The loss function reflects the gap between the model prediction results and the actual annotated results, and is an important basis for optimizing the model.

[0047] Based on the calculated loss function, the back propagation algorithm is used to optimize the parameters of the recursive neural network. The back propagation algorithm calculates the gradient of the loss function and passes these gradients back to the network, gradually adjusting the parameters in the network (such as weights and biases) to reduce the prediction error. This process will be iterated until the output of the model can better match the actual labeled data.

[0048] Finally, the optimized neural network model will be used as the final preset neural network model. This model combines the advantages of convolutional neural networks and recurrent neural networks, and can process spatial features and time series data at the same time, providing more accurate control parameters for the blood purification system. Through this integrated processing, the model can combine spatial and temporal information when processing complex medical data to provide more personalized and effective treatment recommendations.

[0049] In one example, before using a preset neural network model to perform data analysis on at least the current treatment parameters to obtain the control parameters of the blood purification system, the blood purification system control method also includes: obtaining the patient vital sign parameters and patient medical images of the target patient detected by the other treatment equipment; and inputting the patient vital sign parameters and patient medical images of the target patient into the preset neural network model.

[0050] That is, the target patient's vital signs and medical imaging data are obtained from other treatment devices. These data come from other treatment devices that are different from blood purification equipment, such as devices that monitor vital signs such as heart rate, blood pressure, and blood oxygen saturation, as well as equipment for medical imaging examinations (such as CT, MRI, or ultrasound). These data usually contain rich information about the patient's physiological state and detailed medical images, which help to fully understand the patient's health status.

[0051] However, machine learning models are highly dependent on data and require a large amount of high-quality patient data for training. If the data is insufficient or the data quality is not high, the performance of the model may drop significantly. In addition, data privacy issues and difficulty in obtaining data are also important factors that limit the widespread application of the model. Another is the generalization ability of the model. Although the performance of the machine learning model on the training data may be excellent, its performance on new and unseen data may drop significantly, especially when the patient's condition is significantly different from the training data. This problem of insufficient generalization ability makes it possible for the model to be unable to effectively handle various complex treatment scenarios in practical applications. Based on this, the embodiments of the present application also provide two technical solutions, as follows: In one example, the blood purification system control method further includes: using online learning technology to update parameters of the neural network model based on input data and output data of the neural network model.

[0052] That is, by adopting online learning technology, the blood purification system can continuously update parameters based on the input and output data of the neural network model. This means that in actual clinical applications, the model no longer relies solely on pre-collected training data, but can continuously absorb new data as the patient's treatment progresses, and use this data to adjust the model's parameters. This continuous learning and adjustment mechanism significantly enhances the model's adaptive ability.

[0053] For example, suppose that in the actual application of the blood purification system, the condition of a patient is quite different from the typical patient in the training data. Traditional machine learning-based models may not perform well in dealing with such cases because the training process of the model does not include enough data of similar patients. However, after adopting online learning technology, the model can update parameters during the treatment of the patient based on the input data (such as blood flow rate, blood pressure, vital sign parameters, etc.) and output data (i.e., control parameters generated by the model) obtained in real time. By continuously adjusting the model parameters, the system can gradually adapt to the individual differences of the patient and output more accurate and personalized control decisions.

[0054] The advantage of this online learning is that it can dynamically adjust the model, making it more flexible when dealing with special cases. Unlike the traditional fixed model, the online learning mechanism provides the model with the opportunity to continuously learn and update, which effectively solves the problem of insufficient model generalization ability. Even when faced with new and unseen data, the model can improve its adaptability through this real-time adjustment mechanism to ensure that it always performs well on different patients.

[0055] For example, suppose that the initial neural network model of a blood purification system is trained based on a past data set, and the patients included in this data set are mainly young and middle-aged people. However, when the system is actually applied to an elderly patient, the physiological characteristics of the elderly patient may be significantly different from those of the young and middle-aged people in the original training data, such as changes in blood flow rate and organ function. At this time, if the neural network model is not learned online, it may not be able to provide the best treatment control parameters for the elderly patient. However, through online learning technology, the system can dynamically adjust the weights and parameters of the model after receiving the vital signs parameters and treatment feedback data of the elderly patient, so that the neural network can quickly learn and adapt to the individual differences of the elderly patients and optimize the treatment effect.

[0056] In addition, online learning can not only enhance the generalization ability of the model, but also help solve the problem of insufficient data. Since traditional model training usually requires large-scale, high-quality patient data, which is often difficult to obtain in practical applications, online learning provides a sustainable way to utilize data. By gradually accumulating new patient data during the operation of the system, the model can continuously optimize itself, significantly reducing its dependence on large-scale data sets. This mechanism enables the system to maintain good performance even when the amount of data is relatively limited, and provide more personalized treatment plans for each patient.

[0057] In general, the introduction of online learning technology can effectively solve the data dependence and insufficient generalization problems faced by neural network models when applied in blood purification systems. By updating model parameters in real time, the system can dynamically adapt to new patient data and improve its ability to handle complex treatment scenarios, thereby providing more accurate and personalized blood purification solutions.

[0058] In one example, the blood purification system control method also includes: obtaining a pre-trained model trained based on a medical data set, and using transfer learning technology to migrate the pre-trained model to the neural network model, so that the neural network model extracts features from the sample set based on the pre-trained model.

[0059] That is, obtain a pre-trained model trained on a medical data set and transfer the pre-trained model to a neural network model to extract features from a sample set. Transfer learning is a method that reduces the reliance on large amounts of labeled data by transferring the knowledge of a model in one field to another related field.

[0060] The core of transfer learning is to use pre-trained models that have been trained on large-scale datasets. These models have learned some common feature extraction capabilities. For example, a pre-trained model trained on a large-scale medical imaging dataset can recognize and extract some basic features in the image, such as edges, textures, and shapes, which are common in various medical images. Therefore, migrating these general feature extraction capabilities to specific blood purification treatment data can significantly improve the model's ability to process new data, and can effectively improve performance even when there is insufficient training data.

[0061] Specifically, transfer learning technology solves the problems of data dependency and insufficient generalization ability through the following steps. First, a pre-trained model is trained using a large-scale medical data set, which has performed well in feature extraction. Next, this pre-trained model is used as a basis to migrate to the neural network model of the target blood purification system. During the migration process, the weights of the first few layers of the pre-trained model can be frozen, and only the subsequent layers are adjusted and trained so that the model can focus on extracting features related to the blood purification system.

[0062] For example, suppose a pre-trained model is trained on a large dataset containing a variety of medical images and has learned to recognize common imaging features. When the pre-trained model is applied to new blood purification system data, it can use the previously learned feature extraction capabilities to quickly adapt to the new data, thereby improving the performance of the neural network model in the new environment. This approach not only reduces the need for large amounts of labeled data, but also improves the generalization ability of the model in practical applications, enabling it to better handle data from different patients and complex treatment scenarios.

[0063] In short, by transferring the knowledge of the pre-trained model to the new neural network model through transfer learning technology, the problems of data dependence and insufficient generalization ability can be effectively overcome. This method can reduce the dependence on a large amount of data while improving the adaptability and prediction accuracy of the model, thereby providing more accurate and personalized control parameters in actual blood purification treatment and improving the treatment effect.

[0064] like Figure 2 As shown, the embodiment of the present invention further provides a blood purification system control device based on a neural network, comprising: An acquisition unit 1 is used to acquire the current treatment parameters of the target patient detected by the blood purification system, wherein the treatment parameters include at least any one of the following: blood flow rate, dialysate flow rate and ultrafiltration rate; An adjustment unit 2, configured to use a preset neural network model to perform data analysis on at least the current treatment parameters, obtain control parameters of the blood purification system, and use the control parameters to adjust the blood purification system; The feedback unit 3 is used to continue to execute the step of "obtaining the current treatment parameters detected by the blood purification system" based on the feedback mechanism until the current treatment parameters detected by the blood purification system meet the preset conditions.

[0065] In this embodiment, for the specific implementation of each unit in the above device embodiment, please refer to the above method embodiment, which will not be repeated here.

[0066] Reference Figure 3 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0067] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0068] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0069] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0070] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0071] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A blood purification system control method based on neural network, characterized in that: The following steps are involved: Acquiring current treatment parameters of the target patient detected by the blood purification system, wherein the treatment parameters include at least any one of the following: blood flow rate, dialysate flow rate, and ultrafiltration rate; Using a preset neural network model to perform data analysis on at least the current treatment parameters to obtain control parameters of the blood purification system, and using the control parameters to adjust the blood purification system; The step of "obtaining the current treatment parameters detected by the blood purification system" is continued based on the feedback mechanism until the current treatment parameters detected by the blood purification system meet the preset conditions.

2. The blood purification system control method according to claim 1, characterized in that: Before using a preset neural network model to perform data analysis on the treatment parameters to obtain control parameters of the blood purification system, the blood purification system control method further includes: Acquire medical data involved in the treatment process of the blood purification device, and integrate the medical data into a sample set, wherein the medical data includes: treatment parameters of the patient detected by the blood purification device, and patient vital sign parameters and patient medical images detected by other treatment devices, and multiple samples in the sample set correspond to one patient respectively; Constructing a convolutional neural network, and automatically learning the sample set based on the convolutional neural network to learn and identify the spatial features and local correlation of each sample data in the sample set, and determine the lesion information corresponding to each sample data in the sample set; Constructing a recursive neural network, and automatically learning the sample set based on the recursive neural network to capture the long-term dependency of each sample data in the sample set, and determining the prediction information corresponding to each sample data in the sample set; The trained convolutional neural network and recursive neural network are integrated to obtain an initial neural network model, and the initial neural network model is trained based on the outputs of the convolutional neural network and the recursive neural network to fuse the outputs of the convolutional neural network and the recursive neural network, and the control parameters of the blood purification system are output based on the fusion result to obtain the final preset neural network model.

3. The blood purification system control method according to claim 1, characterized in that: Constructing a convolutional neural network, and automatically learning the sample set based on the convolutional neural network to learn and identify the spatial features and local correlation of each sample data in the sample set, and determining the lesion information corresponding to each sample data in the sample set, including: Dividing each sample data in the sample set into image data and time series detection data, performing denoising and contrast enhancement processing on the image data, and performing smoothing and normalization processing on the time series detection data; Performing lesion information labeling and relationship feature labeling processing on the image data that has undergone the denoising and contrast enhancement processing and the time series detection data that has undergone the smoothing and normalization processing, and reintegrating the labeled data into a sample set; The convolutional neural network is constructed, the sample set after labeling is input into the convolutional neural network, and the back propagation algorithm is used to adjust the parameters of the convolutional neural network based on the sample set after labeling to obtain a trained convolutional neural network.

4. The blood purification system control method according to claim 1, characterized in that: Constructing a recursive neural network, and automatically learning the sample set based on the recursive neural network to capture the long-term dependency of each sample data in the sample set, and determining the prediction information corresponding to each sample data in the sample set, including: Dividing each sample data in the sample set into time series detection data, performing smoothing and normalization processing on the time series detection data, and reintegrating into a sample set based on the smoothed and normalized time series detection data; The recursive neural network is constructed using a long short-term memory network, the sample set that has been labeled is input into the recursive neural network, and a combination of a forward propagation algorithm and a back propagation algorithm is used, and the parameters of the recursive neural network are adjusted based on the sample set that has been labeled to obtain a trained recursive neural network.

5. The blood purification system control method according to claim 1, characterized in that: The trained convolutional neural network and recursive neural network are integrated to obtain an initial neural network model, the initial neural network model is trained based on the outputs of the convolutional neural network and the recursive neural network to fuse the outputs of the convolutional neural network and the recursive neural network, and the control parameters of the blood purification system are output based on the fusion result to obtain the final preset neural network model, including: Performing output control parameter labeling processing on the standardized outputs of the trained convolutional neural network and recurrent neural network for the same sample data to obtain target sample data, and integrating the target sample data into a target sample set based on a plurality of the target sample data; Based on the trained standardized outputs of the convolutional neural network and the recursive neural network, an integrated initial neural network model is constructed, wherein the initial neural network model uses a cascade connection mechanism to connect the standardized outputs of the convolutional neural network and the recursive neural network together; The target sample set is input into the initial neural network model, and the forward propagation algorithm is used to calculate the loss function of the recursive neural network based on the labeled target sample set, and the back propagation algorithm is used to optimize the network parameters of the recursive neural network based on the labeled target sample set to obtain a trained neural network model.

6. The blood purification system control method according to claim 2, characterized in that: Before using a preset neural network model to perform data analysis on at least the current treatment parameters to obtain control parameters of the blood purification system, the blood purification system control method further includes: Acquiring patient vital sign parameters and patient medical images detected by the other treatment equipment of the target patient; And the patient's vital sign parameters and patient medical images of the target patient are input into the preset neural network model.

7. The blood purification system control method according to claim 3, characterized in that: The blood purification system control method further includes: Adopting online learning technology to update the parameters of the neural network model based on the input data and output data of the neural network model; and / or, A pre-trained model trained based on a medical data set is obtained, and the pre-trained model is migrated to the neural network model using a transfer learning technique, so that the neural network model extracts features from the sample set based on the pre-trained model.

8. A blood purification system control device based on a neural network, characterized in that: include: An acquisition unit, configured to acquire current treatment parameters of the target patient detected by the blood purification system, wherein the treatment parameters include at least any one of the following: blood flow rate, dialysate flow rate, and ultrafiltration rate; An adjustment unit, configured to use a preset neural network model to perform data analysis on at least the current treatment parameters to obtain control parameters of the blood purification system, and use the control parameters to adjust the blood purification system; A feedback unit is used to continue to execute the step of "obtaining the current treatment parameters detected by the blood purification system" based on the feedback mechanism until the current treatment parameters detected by the blood purification system meet the preset conditions.

9. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images