Dry body weight prediction method and system in hemodialysis process and electronic equipment

By introducing causal convolutional self-attention layer into the Transformer model, preprocessing and feature extraction of hemodialysis clinical data is solved, and the problem of poor dry weight prediction accuracy in hemodialysis is achieved, and more accurate prediction and personalized prescriptions are achieved. Improved dialysis efficacy and patient safety.

CN119092115APending Publication Date: 2024-12-06宁波工业互联网研究院有限公司
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411078794.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-07
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

In the prior art, the dry weight prediction accuracy during hemodialysis is poor, resulting in a single dialysis prescription and a lag in personalization. Traditional methods require professional equipment and technology, and the Transformer model fails to effectively process local context information, resulting in a high risk of misjudgment and reduces Model accuracy.

Method used

The improved Transformer model is adopted to introduce a causal convolutional self-attention layer. By pre-processing and feature extraction of hemodialysis clinical data, the feature data of the target object is obtained, and data conversion is carried out based on model parameters to improve the accuracy of dry weight prediction.

Benefits of technology

Improve the accuracy of dry weight prediction, assisting doctors in prescribing more reliable hemodialysis prescriptions, enhance the personalization and dynamicity of the prescriptions, and reduce the incidence of adverse events in dialysis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119092115A_ABST
    Figure CN119092115A_ABST
Patent Text Reader

Abstract

The invention relates to a dry body weight prediction method and system in a hemodialysis process and electronic equipment. The method comprises the following steps: acquiring clinical data of hemodialysis of a target object, and preprocessing the clinical data to obtain preprocessed clinical data; performing feature extraction on the preprocessed clinical data to obtain feature data of the target object; inputting the feature data of the target object into an improved Transformer model for dry weight prediction to obtain a dry weight prediction value of the target object; wherein a model parameter generated by causal convolution is introduced into a self-attention layer in the improved Transform model, and data conversion is performed on the feature data of the target object based on the model parameter. Through the method and the device, the technical problem of poor precision of dry body weight prediction in related technologies is solved, and the technical effect of improving the precision of dry body weight prediction is achieved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of dialysis-assisted diagnosis, and in particular to a method, system and electronic equipment for predicting dry weight during hemodialysis. Background Art

[0002] Hemodialysis (HD) is an extracorporeal blood purification treatment technology designed to replace damaged kidneys. Its main functions include removing excess water and solutes from the body by convection. Blood diffuses or convects through the membrane in the dialyzer, and most of the body's water, electrolytes, and middle-molecular substances are removed through the membrane. Then a similar volume of fluid and useful components such as plasma (called fresh fluid) are added to eliminate waste and excess water in the body.

[0003] Dry weight is a key parameter used to determine ultrafiltration volume before hemodialysis. In clinical practice, it is defined as the lowest post-dialysis weight that does not cause adverse symptoms during dialysis. However, in the medical field, the accurate measurement of dry weight remains "the missing piece of the puzzle". Due to the complexity and instability of existing dry weight assessment methods and significant individual differences between patients, ensuring timely and accurate assessment of dry weight poses a major challenge. Correctly evaluating the dry weight of hemodialysis patients is of great significance. If the dry weight is set too high, the patient's fluid load will increase the incidence of hypertension and cardiovascular and cerebrovascular complications such as left ventricular hypertrophy, heart failure, cerebral hemorrhage, etc., and affect the patient's long-term survival rate. If the dry weight is set too low, the incidence of adverse events during dialysis will increase, such as hypotension, dizziness, tinnitus, etc., affecting dialysis compliance and dialysis adequacy.

[0004] Traditional dry weight prediction still has the following defects: First, the prescription of dialysis patients is single, which is easy to cause dialysis complications; second, personalized dialysis prescription has a strong lag, which affects the dialysis efficacy; third, traditional dry weight prediction methods require professional equipment and professional technology to complete, such as the bioelectrical impedance spectroscopy (BIS) evaluation of dry weight is too high; fourth, most of the existing dialysis adequacy prediction methods use single dialysis data as input, which cannot reflect the dynamic nature of the prescription; fifth, the self-attention layer of the traditional Transformer model fails to effectively process local context information (by calculating weights point by point), which leads to a high risk of misjudging the actual situation of the prediction point, thereby reducing the overall accuracy of the model and bringing challenges to optimization.

[0005] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0006] The embodiments of the present invention provide a method, system and electronic device for predicting dry weight during hemodialysis, so as to at least solve the technical problem of poor dry weight prediction accuracy in the related art.

[0007] According to one aspect of an embodiment of the present invention, a method for predicting dry weight during hemodialysis is provided, comprising: obtaining clinical data of a target object undergoing hemodialysis, and preprocessing the clinical data to obtain preprocessed clinical data; performing feature extraction on the preprocessed clinical data to obtain feature data of the target object; inputting the feature data of the target object into an improved Transformer model to perform dry weight prediction to obtain a predicted value of the dry weight of the target object; wherein the self-attention layer in the improved Transformer model introduces model parameters generated by causal convolution, and performs data conversion on the feature data of the target object based on the model parameters.

[0008] Optionally, the clinical data is preprocessed, including: screening out target subjects who have received hemodialysis treatment for at least several consecutive months; using the average value of the data recorded during the hemodialysis treatment and the data before and after the hemodialysis as characteristic variables for the target subject undergoing a hemodialysis treatment; excluding target subjects who have not received hemodialysis treatment, are younger than a preset age, exceed a preset hemodialysis interval, and have incomplete clinical data; and dividing the clinical data of the same target subject into clinical data of different periods.

[0009] Optionally, the self-attention layer includes a self-attention mechanism component and a multi-head attention mechanism component; wherein the expressions of the self-attention mechanism component and the multi-head attention mechanism component are respectively:

[0010]

[0011] MultiHead(Q,K,V)=Concat(head1…head h )W O

[0012] head i =Attention(QW i Q ,KW i K ,WV i V )

[0013] Where Q, K, and V represent the query, key, and value matrices respectively, and d k represents the dimension of the key, and W represents the mapping parameter matrix.

[0014] Optionally, the Transformer model is used to encode the position of each data point in the feature data of the target object into a vector feature and merge it into the Transformer model.

[0015] Optionally, the expression used to encode the position of each data point in the feature data of the target object as a vector feature and merge it into the Transformer model is:

[0016]

[0017] Among them, pos represents the position of each data point, i represents the dimension, and d model Indicates the dimension of the output.

[0018] Optionally, the expression of the causal convolution is:

[0019]

[0020] Among them, O(a,b) represents the output of causal convolution, a and b represent the weight parameters related to the convolution layer, respectively. represents the input sequence and k represents the number of hidden layers.

[0021] Optionally, the method also includes: dividing the preprocessed clinical data into a training set, a validation set and a test set; wherein the training set is used to train the improved Transformer model, the validation set is used to adjust the hyperparameters of the improved Transformer model and prevent overfitting, and the test set is used to evaluate the generalization ability of the improved Transformer model after training.

[0022] Optionally, the method also includes: obtaining the true value of the dry weight of the target object and the average value of the original data; performing mean square error calculation based on the true value of the dry weight of the target object and the predicted value of the dry weight to obtain the mean square error value of the target object; wherein the mean square error value is used to evaluate the average difference between the predicted value of the dry weight and the true value of the dry weight; performing goodness of fit calculation of the regression model based on the average value of the original data, the true value of the dry weight of the target object and the predicted value of the dry weight to obtain a goodness of fit result, wherein the goodness of fit result is used to evaluate the degree of fit of the regression model.

[0023] According to another aspect of an embodiment of the present invention, a dry weight prediction system during hemodialysis is provided, comprising: a preprocessing module, used to obtain clinical data of a target object undergoing hemodialysis, and preprocess the clinical data to obtain preprocessed clinical data; a feature extraction module, used to extract features from the preprocessed clinical data to obtain feature data of the target object; a dry weight prediction module, used to input the feature data of the target object into an improved Transformer model for dry weight prediction to obtain a predicted value of the dry weight of the target object; wherein the self-attention layer in the improved Transformer model introduces model parameters generated by causal convolution, and performs data conversion on the feature data of the target object based on the model parameters.

[0024] According to another aspect of an embodiment of the present invention, there is provided an electronic device, including: a processor, and a memory storing a program, wherein the program includes instructions, and when the instructions are executed by the processor, the processor executes the method described above.

[0025] In an embodiment of the present invention, clinical data of a target object undergoing hemodialysis is obtained, and the clinical data is preprocessed to obtain preprocessed clinical data; feature extraction is performed on the preprocessed clinical data to obtain feature data of the target object; the feature data of the target object is input into an improved Transformer model to perform dry weight prediction to obtain a predicted dry weight value of the target object; wherein the self-attention layer in the improved Transformer model introduces model parameters generated by causal convolution, and data conversion is performed on the feature data of the target object based on the model parameters, thereby solving the technical problem of poor dry weight prediction accuracy in the related art and achieving the technical effect of improving the dry weight prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other embodiments can be obtained based on these drawings without creative work.

[0027] Figure 1 A flowchart of a method for predicting dry weight during hemodialysis provided by an embodiment of the present invention;

[0028] Figure 2 A flowchart of data preprocessing provided by an embodiment of the present invention;

[0029] Figure 3An overall architecture diagram of a dry weight prediction method based on an improved Transformer provided in an embodiment of the present invention;

[0030] Figure 4 A comparison chart of experimental results of the improved Transformer model provided by an embodiment of the present invention and other existing models;

[0031] Figure 5 A schematic diagram of a dry weight prediction system during hemodialysis provided by an embodiment of the present invention;

[0032] Figure 6 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0033] Embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present invention are shown in the accompanying drawings, it should be understood that the embodiments of the present invention can be implemented in various forms and should not be construed as being limited to the embodiments described herein, which are instead provided to provide a more thorough and complete understanding of the embodiments of the present invention. It should be understood that the drawings and embodiments of the embodiments of the present invention are only for exemplary purposes and are not intended to limit the scope of protection of the embodiments of the present invention.

[0034] According to one aspect of an embodiment of the present invention, a method for predicting dry weight during hemodialysis is provided. Figure 1 A flowchart of a method for predicting dry weight during hemodialysis provided by an embodiment of the present invention is shown in FIG. Figure 1 As shown, the method comprises the following steps:

[0035] Step S102, obtaining clinical data of hemodialysis performed by the target subject, and preprocessing the clinical data to obtain preprocessed clinical data;

[0036] The above-mentioned target subjects include but are not limited to patients;

[0037] Step S104, extracting features from the preprocessed clinical data to obtain feature data of the target object;

[0038] The above-mentioned characteristic data include but are not limited to multi-dimensional time series characteristics or multi-dimensional time series physiological information;

[0039] Step S106, inputting the characteristic data of the target object into the improved Transformer model for dry weight prediction to obtain the dry weight prediction value of the target object; wherein the self-attention layer in the improved Transformer model introduces model parameters generated by causal convolution, and performs data conversion on the characteristic data of the target object based on the model parameters.

[0040] Optionally, a causal convolution with kernel k and stride 1 is used to generate Q and K for the current input. The above model parameters include but are not limited to Q and K, where Q and K represent the query and key matrices respectively.

[0041] It should be noted that the core basic components of the improved Transformer model architecture include but are not limited to fully connected layers, residual networks, self-attention and multi-head attention mechanisms; the use of causal convolution can capture more local contextual information, which is conducive to improving the accuracy of dry weight prediction.

[0042] In an embodiment of the present invention, clinical data of a target object undergoing hemodialysis is obtained, and the clinical data is preprocessed to obtain preprocessed clinical data; feature extraction is performed on the preprocessed clinical data to obtain feature data of the target object; the feature data of the target object is input into an improved Transformer model to perform dry weight prediction to obtain a predicted dry weight value of the target object; wherein the self-attention layer in the improved Transformer model introduces model parameters generated by causal convolution, and data conversion is performed on the feature data of the target object based on the model parameters, thereby solving the technical problem of poor dry weight prediction accuracy in the related art and achieving the technical effect of improving the dry weight prediction accuracy.

[0043] In addition, predicting the patient's dry weight based on the improved Transformer model can assist doctors in prescribing hemodialysis, thereby improving prescription reliability and doctors' work efficiency.

[0044] As an optional embodiment, the clinical data is preprocessed, including: screening out target subjects who have received hemodialysis treatment for at least several consecutive months; using the average value of the data recorded during the hemodialysis treatment and the data before and after the hemodialysis as the characteristic variable of the target subject undergoing a hemodialysis treatment; excluding target subjects who have not received hemodialysis treatment, are younger than a preset age, exceed a preset hemodialysis interval, and have incomplete clinical data; and dividing the clinical data of the same target subject into clinical data of different periods.

[0045] Figure 2 A flowchart of data preprocessing provided by an embodiment of the present invention is as follows: Figure 2As shown, first, patients who had received HD treatment for at least 3 consecutive months were screened to ensure that the selected patients were in a clinically stable state; second, the data recorded during dialysis treatment and the average of the data before and after dialysis were used as the characteristic variables of the patient's one HD treatment. In particular, these characteristic variables were considered constant before each prescription update and blood parameter retest, and the same was true for dry weight. Then, patients who did not receive HD treatment, were younger than 18 years old, had a dialysis interval of more than 1 week, and had incomplete clinical data were excluded. Finally, the data of the same patient were divided into two parts for separate experiments; among them, the first 2 months were defined as the contact period, and the last 2 months were defined as the stable period.

[0046] Optionally, a time series dry weight prediction algorithm based on the improved Transformer model transforms the dry weight assessment problem into a multivariate time series prediction task to reflect the potential dynamics of dry weight. As a model that uses the attention mechanism for data processing, the improved Transformer model is not restricted by distance when accessing historical data. Self-attention and multi-head attention mechanisms are designed to convert data into tensor form. Finally, data training and testing are performed. The core of the architecture is the basic components: self-attention and multi-head attention mechanisms.

[0047] As an optional embodiment, the self-attention layer includes a self-attention mechanism component and a multi-head attention mechanism component; wherein the expressions of the self-attention mechanism component and the multi-head attention mechanism component are respectively:

[0048]

[0049] MultiHead(Q,K,V)=Concat(head1…head h )W O

[0050] head i =Attention(QW i Q ,KW i K ,VW i V )

[0051] Where Q, K, and V represent the query, key, and value matrices respectively, and d k represents the dimension of the key, and W represents the mapping parameter matrix.

[0052] Furthermore, maintaining the temporal order of the time series is crucial for time series prediction tasks. To solve this problem, a time series dry weight prediction algorithm based on an improved Transformer model can be used to encode the position of each data point in the feature data of the target object as a vector feature and incorporate it into the model.

[0053] As an optional embodiment, the expression used to encode the position of each data point in the feature data of the target object as a vector feature and merge it into the Transformer model is:

[0054]

[0055] Among them, pos represents the position of each data point, i represents the dimension, and d model Indicates the dimension of the output.

[0056] Figure 3 The overall architecture diagram of an improved Transformer dry weight prediction method provided in an embodiment of the present invention is as follows: Figure 3 As shown in Figure 1, a causal convolution with kernel k and stride 1 is used to generate Q and K for the current input. Compared with the matrix multiplication method used in the normalized self-attention layer, this method can effectively capture relevant information of the recent time period. By utilizing causal convolution, the model's ability to integrate temporal context is enhanced.

[0057] As an optional embodiment, the expression of causal convolution in the Self-Attention layer is:

[0058]

[0059] Among them, O(a,b) represents the output of causal convolution, a and b represent the weight parameters related to the convolution layer, respectively. represents the input sequence and k represents the number of hidden layers.

[0060] As an optional embodiment, the above method also includes: dividing the preprocessed clinical data into a training set, a validation set and a test set; wherein the training set is used to train the improved Transformer model, the validation set is used to adjust the hyperparameters of the improved Transformer model and prevent overfitting, and the test set is used to evaluate the generalization ability of the improved Transformer model after training.

[0061] During the entire training process, the validation set and test set should remain unchanged to ensure the fairness of the evaluation results. If the amount of data is small, in order to make more full use of the data, the cross-validation method can be used, that is, by repeatedly using different subsets of the data for training and validation to reduce the estimation bias caused by random segmentation.

[0062] As an optional embodiment, the above method also includes: obtaining the true value of the dry weight of the target object and the average value of the original data; performing mean square error calculation based on the true value of the dry weight of the target object and the predicted value of the dry weight to obtain the mean square error value of the target object; wherein the mean square error value is used to evaluate the average difference between the predicted value of the dry weight and the true value of the dry weight; performing fit calculation of the regression model based on the average value of the original data, the true value of the dry weight of the target object and the predicted value of the dry weight to obtain a fit result, wherein the fit result is used to evaluate the fit of the regression model.

[0063] The average value of the above raw data may be the average value calculated from the actual values ​​of the dry weight of multiple target objects.

[0064] To ensure the best performance of the causal convolutional transformer model, it is necessary to sample the patient data using an appropriate time window before providing the model with multivariate temporal information of HD patients. Given that hemodialysis is usually performed three times a week, an analysis was performed to examine the impact of different time window lengths on dry weight prediction results. Five different time window lengths (k = 3, 6, 9, 12, 15) were used to evaluate their impact. Different time window lengths have different results for the model. Shorter windows have less patient information and lower prediction accuracy. Longer windows may ignore the patient's condition. Specifically, k was selected as the input time step of the causal convolutional transformer model in order to predict the patient's dry weight during subsequent hemodialysis. To ensure fair comparison of model performance, the doctor's clinical dry weight assessment of the patient was considered as the true value for the study. In addition, two popular metrics were utilized: MSE (mean squared error) measuring the average difference between the predicted and true values, and R 2 (R Square) evaluates the goodness of fit of the regression model.

[0065]

[0066] Where n represents the number of samples, y represents the true value of dry weight, Denotes predicted dry weight.

[0067]

[0068] Where n represents the number of samples, y represents the true value of dry weight, represents the predicted dry weight, Represents the average value of the original data.

[0069] Figure 4 The experimental results comparison diagram of the improved Transformer model provided by the embodiment of the present invention and other existing models is as follows: Figure 4As shown in the figure, the prediction results of different time window lengths while keeping the rest of the variables unchanged. According to the experimental results, different time window lengths have different results for the model. Shorter windows have less patient information and lower prediction accuracy. Longer windows may ignore the patient's condition. The time window length of 12 shows the best performance in terms of MSE and R2. This indicates that dry weight can be better predicted using patient information for about one month. Therefore, the time window length of 12 was selected in subsequent experiments.

[0070] Furthermore, an improved Transformer model was built to predict dry weight, which had higher prediction accuracy than existing models such as lSTM and GRU; second, the introduction of causal convolution in the self-attention layer of the improved Tranformer model enhanced the model's prediction ability by enhancing the model's ability to capture local temporal information. Comparison with other models further clarified the effectiveness of the improved Tranformer model. In addition, compared with HD patients in the contact period, fluctuations in indicators were observed, and the model showed superior prediction effects on the dry weight of HD patients in the stable period. This finding further emphasizes the importance of classifying patients accordingly. This method can tailor personalized prediction models for individuals at different periods, thereby achieving more accurate and specific predictions.

[0071] According to another aspect of the embodiments of the present invention, a dry weight prediction system during hemodialysis is provided. Figure 5 A schematic diagram of a dry weight prediction system during hemodialysis provided by an embodiment of the present invention is shown in FIG. Figure 5 As shown, the dry weight prediction system during hemodialysis includes: a preprocessing module 502, a feature extraction module 504 and a dry weight prediction module 506. The dry weight prediction system during hemodialysis is described in detail below.

[0072] A preprocessing module 502 is used to obtain clinical data of hemodialysis performed by a target subject and preprocess the clinical data to obtain preprocessed clinical data;

[0073] The feature extraction module 504 is connected to the preprocessing module 502 and is used to extract features from the preprocessed clinical data to obtain feature data of the target object;

[0074] The dry weight prediction module 506 is connected to the feature extraction module 504, and is used to input the feature data of the target object into the improved Transformer model for dry weight prediction to obtain the dry weight prediction value of the target object; wherein the self-attention layer in the improved Transformer model introduces the model parameters generated by the causal convolution, and performs data conversion on the feature data of the target object based on the model parameters.

[0075] In an embodiment of the present invention, the system obtains clinical data of a target object undergoing hemodialysis, and preprocesses the clinical data to obtain preprocessed clinical data; performs feature extraction on the preprocessed clinical data to obtain feature data of the target object; inputs the feature data of the target object into an improved Transformer model to perform dry weight prediction to obtain a predicted dry weight value of the target object; wherein the self-attention layer in the improved Transformer model introduces model parameters generated by causal convolution, and performs data conversion on the feature data of the target object based on the model parameters, thereby solving the technical problem of poor dry weight prediction accuracy in related technologies and achieving the technical effect of improving dry weight prediction accuracy.

[0076] It should be noted here that the above-mentioned preprocessing module 502, feature extraction module 504 and dry weight prediction module 506 correspond to steps S102 to S106 in the method embodiment, and the examples and application scenarios implemented by the above-mentioned modules and corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned method embodiment.

[0077] As an optional embodiment, the preprocessing module 502 includes: a first preprocessing unit, used to screen out target subjects who have received hemodialysis treatment for at least several consecutive months; a second preprocessing unit, used to use the average value of the data recorded during the hemodialysis treatment and the data before and after the hemodialysis as the characteristic variable of the target subject undergoing a hemodialysis treatment; a third preprocessing unit, used to exclude target subjects who have not received hemodialysis treatment, are younger than a preset age, exceed a preset hemodialysis interval, and have incomplete clinical data; and a fourth preprocessing unit, used to divide the clinical data of the same target subject into clinical data of different periods.

[0078] As an optional embodiment, the self-attention layer includes a self-attention mechanism component and a multi-head attention mechanism component; wherein the expressions of the self-attention mechanism component and the multi-head attention mechanism component are respectively:

[0079]

[0080] MultiHead(Q,K,V)=Concat(head1…head h )W O

[0081] head i =Attention(QW i Q ,KW i K ,VW i V )

[0082] Where Q, K, and V represent the query, key, and value matrices respectively, and d k represents the dimension of the key, and W represents the mapping parameter matrix.

[0083] As an optional embodiment, the Transformer model is used to encode the position of each data point in the feature data of the target object into a vector feature and merge it into the Transformer model.

[0084] As an optional embodiment, the expression used to encode the position of each data point in the feature data of the target object as a vector feature and merge it into the Transformer model is:

[0085]

[0086] Among them, pos represents the position of each data point, i represents the dimension, and d model Indicates the dimension of the output.

[0087] It should be noted that n corresponding to 2i is an even number, and n corresponding to 2i+1 is an odd number.

[0088] As an optional embodiment, the expression of causal convolution is:

[0089]

[0090] Among them, O(a,b) represents the output of causal convolution, a and b represent the weight parameters related to the convolution layer, respectively. represents the input sequence and k represents the number of hidden layers.

[0091] As an optional embodiment, the above system also includes: a segmentation module, used to segment the preprocessed clinical data into a training set, a validation set and a test set; wherein the training set is used to train the improved Transformer model, the validation set is used to adjust the hyperparameters of the improved Transformer model and prevent overfitting, and the test set is used to evaluate the generalization ability of the improved Transformer model after training.

[0092] As an optional embodiment, the above system also includes: an acquisition module, used to obtain the actual value of the dry weight of the target object and the average value of the original data; a first calculation module, used to calculate the mean square error based on the actual value of the dry weight and the predicted value of the dry weight of the target object, and obtain the mean square error value of the target object; wherein the mean square error value is used to evaluate the average difference between the predicted value of the dry weight and the actual value of the dry weight; a second calculation module, used to calculate the fit of the regression model based on the average value of the original data, the actual value of the dry weight and the predicted value of the dry weight, and obtain a fit result, wherein the fit result is used to evaluate the fit of the regression model.

[0093] According to another aspect of an embodiment of the present invention, an electronic device is provided, comprising: a processor, and a memory storing a program, wherein the program comprises instructions, and when the instructions are executed by the processor, the processor executes the method for predicting dry weight during hemodialysis according to an embodiment of the present invention.

[0094] According to another aspect of an embodiment of the present invention, a non-transitory machine-readable medium storing a computer program is provided, wherein the computer program, when executed by a processor of a computer, is used to enable the computer to perform the method for predicting dry weight during hemodialysis according to an embodiment of the present invention.

[0095] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a computer program, wherein when the computer program is executed by a processor of a computer, the computer is used to enable the computer to execute the method for predicting dry weight during hemodialysis according to an embodiment of the present invention.

[0096] refer to Figure 6 , a block diagram of an electronic device that can be used as a server or client of an embodiment of the present invention will now be described, which is an example of a hardware device that can be applied to various aspects of the present invention. Electronic devices are intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0097] like Figure 6As shown, the electronic device includes a computing unit 601, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 602 or a computer program loaded from a storage unit 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the electronic device can also be stored. The computing unit 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0098] Multiple components in the electronic device are connected to the I / O interface 605, including: an input unit 606, an output unit 607, a storage unit 608, and a communication unit 609. The input unit 606 can be any type of device that can input information to the electronic device, and the input unit 606 can receive input digital or character information, and generate key signal input related to user settings and / or function control of the electronic device. The output unit 607 can be any type of device that can present information, and can include but is not limited to a display, a speaker, a video / audio output terminal, a vibrator, and / or a printer. The storage unit 608 can include but is not limited to a disk, an optical disk. The communication unit 609 allows the electronic device to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks, and can include but is not limited to a modem, a network card, an infrared communication device, a wireless communication transceiver, and / or a chipset, such as a Bluetooth device, a WiFi device, a WiMax device, a cellular communication device, and / or the like.

[0099] The computing unit 601 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a CPU, a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 601 performs the various methods and processes described above. For example, in some embodiments, the method embodiments of the present invention may be implemented as a computer program, which is tangibly contained in a machine-readable medium, such as a storage unit 608. In some embodiments, part or all of the computer program may be loaded and / or installed on an electronic device via a ROM 602 and / or a communication unit 609. In some embodiments, the computing unit 601 may be configured to perform the above-described method in any other appropriate manner (e.g., by means of firmware).

[0100] The computer programs for implementing the methods of the embodiments of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer programs are executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer programs may be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0101] In the context of an embodiment of the present invention, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable signal medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0102] It should be noted that the term "including" and its variations used in the embodiments of the present invention are open inclusions, that is, "including but not limited to". The term "based on" means "based at least in part on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one other embodiment"; the term "some embodiments" means "at least some embodiments". The modifications of "one" and "multiple" mentioned in the embodiments of the present invention are illustrative rather than restrictive. Those skilled in the art should understand that unless the context clearly indicates otherwise, it should be understood as "one or more".

[0103] The user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in the embodiments of the present invention are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0104] The various steps described in the method implementation methods provided in the embodiments of the present invention may be performed in different orders and / or in parallel. In addition, the method implementation methods may include additional steps and / or omit the steps shown. The scope of protection of the present invention is not limited in this respect.

[0105] The term "embodiment" in this specification refers to specific features, structures or characteristics described in conjunction with the embodiment that can be included in at least one embodiment of the present invention. The appearance of this phrase in various places in the specification does not necessarily mean the same embodiment, nor does it mean that it is mutually exclusive with other embodiments and is independent or optional. The various embodiments in this specification are described in a related manner, and the same or similar parts between the various embodiments refer to each other. In particular, for the device, equipment, and system embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts refer to the partial description of the method embodiment.

[0106] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of patent protection. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the attached claims.

Claims

1. A method for predicting dry weight during hemodialysis, characterized in that: include: Acquiring clinical data of hemodialysis performed by a target subject, and preprocessing the clinical data to obtain preprocessed clinical data; Performing feature extraction on the preprocessed clinical data to obtain feature data of the target object; The characteristic data of the target object is input into the improved Transformer model for dry weight prediction to obtain the dry weight prediction value of the target object; wherein the self-attention layer in the improved Transformer model introduces model parameters generated by causal convolution, and performs data conversion on the characteristic data of the target object based on the model parameters.

2. The method according to claim 1, characterized in that Preprocessing the clinical data includes: Target subjects who have received hemodialysis treatment for at least several consecutive months were selected; The average value of the data recorded during the hemodialysis treatment and the data before and after the hemodialysis is used as the characteristic variable of the target subject undergoing a hemodialysis treatment; Target subjects who did not receive hemodialysis treatment, were younger than the preset age, exceeded the preset hemodialysis interval, and had incomplete clinical data were excluded; Divide the clinical data of the same target object into clinical data of different periods.

3. The method according to claim 1, characterized in that The self-attention layer includes a self-attention mechanism component and a multi-head attention mechanism component; wherein the expressions of the self-attention mechanism component and the multi-head attention mechanism component are respectively: MultiHead(Q,K,V)=Concat(head1…head h )W O head i =Attention(QW i Q ,KW i K ,WV i V ) Where Q, K, and V represent the query, key, and value matrices respectively, and d k represents the dimension of the key, and W represents the mapping parameter matrix.

4. The method according to claim 1, characterized in that: The Transformer model is used to encode the position of each data point in the feature data of the target object into a vector feature and merge it into the Transformer model.

5. The method according to claim 4, characterized in that The expression used to encode the position of each data point in the feature data of the target object as a vector feature and merge it into the Transformer model is: Among them, pos represents the position of each data point, i represents the dimension, and d model Indicates the dimension of the output.

6. The method according to claim 1, characterized in that The expression of the causal convolution is: Among them, O(a,b) represents the output of causal convolution, a and b represent the weight parameters related to the convolution layer, respectively. represents the input sequence and k represents the number of hidden layers.

7. The method according to claim 1, characterized in that The method further comprises: The preprocessed clinical data is divided into a training set, a validation set and a test set; wherein the training set is used to train the improved Transformer model, the validation set is used to adjust the hyperparameters of the improved Transformer model and prevent overfitting, and the test set is used to evaluate the generalization ability of the improved Transformer model after training.

8. The method according to any one of claims 1 to 7, characterized in that: The method further comprises: Obtaining the actual value of the dry weight of the target object and the average value of the original data; Performing mean square error calculation based on the actual dry weight value and the predicted dry weight value of the target object to obtain the mean square error value of the target object; wherein the mean square error value is used to evaluate the average difference between the predicted dry weight value and the actual dry weight value; The goodness of fit of the regression model is calculated based on the average value of the original data, the actual value of the dry weight of the target object and the predicted value of the dry weight to obtain a goodness of fit result; wherein the goodness of fit result is used to evaluate the degree of fit of the regression model.

9. A dry weight prediction system during hemodialysis, characterized in that: include: A preprocessing module, used to obtain clinical data of hemodialysis performed by a target subject, and preprocess the clinical data to obtain preprocessed clinical data; A feature extraction module, used to extract features from the preprocessed clinical data to obtain feature data of the target object; A dry weight prediction module is used to input the characteristic data of the target object into an improved Transformer model for dry weight prediction to obtain a predicted dry weight value of the target object; wherein the self-attention layer in the improved Transformer model introduces model parameters generated by causal convolution, and performs data conversion on the characteristic data of the target object based on the model parameters.

10. An electronic device comprising: A processor and a memory storing a program, wherein the program comprises instructions, and when the instructions are executed by the processor, the processor executes the method according to any one of claims 1 to 8.