Peritoneal dialysis patient total cause death risk prediction method

Abstract

The invention relates to a peritoneal dialysis patient total cause death risk prediction method, belongs to the technical field of death risk prediction, and solves the problem of low prediction accuracy in the prior art. The method comprises the following steps: constructing a sample set in each data center; training a neural network model based on the sample sets of the plurality of data centers to obtain a risk prediction pre-training model; obtaining the feature representation of each sample based on a risk prediction pre-training model, and carrying out clustering to obtain the clustering type of each sample; performing fine adjustment on the risk prediction pre-training model based on the samples of each clustering type to obtain a death risk prediction model corresponding to each clustering type; inputting time sequence test data and baseline data of a to-be-predicted patient into the risk prediction pre-training model to obtain feature representation; and obtaining a clustering type corresponding to the to-be-predicted patient according to the feature representation, and obtaining a prediction result of the to-be-predicted patient based on a risk prediction model corresponding to the clustering type. The death risk prediction result can be quickly and accurately obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of all-cause mortality risk prediction, and in particular to a method for predicting all-cause mortality risk in peritoneal dialysis patients. Background Art

[0002] Peritoneal dialysis is one of the commonly used renal replacement therapies for patients with end-stage renal disease, but its patients face a high risk of all-cause mortality. Accurately predicting the risk of all-cause mortality in patients with peritoneal dialysis is of great significance for optimizing treatment plans and improving patient prognosis. Traditional risk prediction methods mostly rely on data from a single medical institution, and have problems such as limited data volume and uneven data distribution, resulting in insufficient accuracy and generalization of the prediction model; Although there are machine learning-based all-cause mortality risk prediction models, most of these models are based on a single data source and cannot fully utilize the advantages of multi-source data. In addition, traditional methods have shortcomings in data privacy protection and are difficult to meet the strict privacy requirements of medical data. Summary of the invention

[0003] In view of the above analysis, an embodiment of the present invention aims to provide a method for predicting the risk of all-cause mortality in peritoneal dialysis patients, so as to solve the problem of low existing prediction accuracy.

[0004] In one aspect, an embodiment of the present invention provides a method for predicting the risk of all-cause mortality in peritoneal dialysis patients, comprising the following steps: A sample set was constructed in each data center based on the patients’ time-series test data and baseline data; A risk prediction pre-training model is obtained by training a neural network model based on sample sets from multiple data centers; Obtaining a feature representation of each sample based on the risk prediction pre-training model, and clustering the feature representation to obtain a cluster type for each sample; Fine-tuning the risk prediction pre-training model based on samples of each cluster type to obtain a death risk prediction model corresponding to each cluster type; The time series test data and baseline data of the patient to be predicted are input into the risk prediction pre-training model to obtain the feature representation of the patient to be predicted; the cluster type corresponding to the patient to be predicted is obtained according to the feature representation, and the prediction result of the patient to be predicted is obtained based on the risk prediction model corresponding to the cluster type.

[0005] Based on the further improvement of the above method, the feature representation is clustered to obtain the cluster type of each sample, including: S31, randomly initialize K cluster centers; S32, sending the K cluster centers to each data center; each data center determines the cluster type of each sample based on the distance between each sample in the local sample set and the K cluster centers, and calculates the local cluster center; S33, updating the K clustering centers based on the local clustering center of each data center and the clustering weight of each data center; S34, determine the update accuracy of the K cluster centers. If all are less than a preset threshold, end clustering and obtain the cluster type of each sample; otherwise, return to step S32.

[0006] Based on the further improvement of the above method, the clustering weight of each data center is calculated in the following way, including: Each data center calculates the local data clustering effect based on the local clustering center; and obtains the first weight of each data center based on the local data clustering effect of each data center; Obtain K first pre-updated cluster centers based on the local cluster center of each data center; For each data center, obtain K second pre-updated cluster centers corresponding to the data center based on the local cluster centers of other data centers; Calculating a second weight of the data center based on the K first pre-updated cluster centers and the K second pre-updated cluster centers corresponding to the data center; Based on the first weight and the second weight of each data center, a clustering weight of each data center is obtained.

[0007] Based on the further improvement of the above method, each data center uses the following formula to calculate the local data clustering effect based on the local cluster center: ; in, represents the clustering effect of the i-th data center, represents the cohesion of the jth sample in the i-th data center, represents the separation degree of the jth sample in the i-th data center, Represents the number of samples in the i-th data center.

[0008] Based on the further improvement of the above method, the first weight is calculated using the following formula: ; in, represents the first weight of the i-th data center, represents the clustering effect of the i-th data center, and N represents the number of data centers.

[0009] Based on the further improvement of the above method, the second weight of the data center is calculated based on the K first pre-updated cluster centers and the K second pre-updated cluster centers corresponding to the data center using the following formula: ; in, represents the jth first pre-updated cluster center, represents the jth second pre-updated cluster center corresponding to the i-th data center, represents the jth second pre-updated cluster center corresponding to the tth data center, N represents the number of data centers, represents the second weight of the i-th data center.

[0010] Based on the further improvement of the above method, the risk prediction pre-training model is fine-tuned based on the samples of each cluster type to obtain the risk prediction model corresponding to each cluster type, including: For each cluster type, determining a third weight of each data center based on samples in each data center belonging to the cluster type; Based on the third weight of each data center, federated learning is used to fine-tune the risk prediction pre-training model to obtain a risk prediction model corresponding to each clustering type.

[0011] Based on the further improvement of the above method, the third weight of each data center is determined by the following formula based on the samples belonging to the cluster type in each data center: ; in, represents the sample quantity weight of the i-th data center, represents the data quality weight of the i-th data center, and represents the adjustment parameters, Represents the third weight of the i-th number center.

[0012] Based on the further improvement of the above method, the sample set of each data center includes a support set and a query set; the neural network model is trained based on the sample sets of multiple data centers to obtain a risk prediction pre-training model, including: S20, constructing a neural network model, and sending the constructed neural network model to each data center; S21, record the parameters of the current neural network model as initial parameters; S22. For each data center, the neural network model is trained based on the local support set to update the parameters of the local neural network model; the loss of the updated local neural network model is calculated on the local query set to obtain the query loss; S23, updating the initial parameters based on the query loss of each data center, and updating the parameters of the neural network model of each data center to the updated initial parameters; S24. Determine whether the stopping condition is met. If so, stop training to obtain the risk prediction pre-training model. Otherwise, return to step S21.

[0013] Based on the further improvement of the above method, the following formula is used to update the initial parameters based on the query loss of each data center: ; ; in, represents the gradient of the total loss of N data centers, represents the total loss of N data centers, represents the query loss of the i-th data center, represents the training weight of the i-th data center, represents the learning rate, Represents the initial parameters.

[0014] Compared with the prior art, the present invention constructs a sample set in each data center, adopts sample sets from multiple data centers to make full use of data from multiple data sources, improves the accuracy of the training model, and ensures data security. After obtaining the pre-trained model, clustering is performed based on the characteristic representation of the sample, and the pre-trained model is fine-tuned based on the sample of each cluster type, thereby training the corresponding death risk prediction model for each type to improve the accuracy of the model. The pre-training fine-tuning method will not affect the performance of the model due to the small number of a certain type, thereby improving the accuracy of the death risk prediction model corresponding to each cluster type. By obtaining the cluster type of the patient to be predicted, its time series test data and baseline data are input into the risk prediction model corresponding to its cluster type, thereby quickly and accurately obtaining the death risk prediction result of the patient to be predicted.

[0015] In the present invention, the above-mentioned technical solutions can also be combined with each other to achieve more preferred combination solutions. Other features and advantages of the present invention will be described in the subsequent description, and some advantages can become obvious from the description, or can be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained through the contents particularly pointed out in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings are only used for the purpose of illustrating specific embodiments and are not to be considered as limiting the present invention. In the entire drawings, the same reference symbols represent the same components; Figure 1This is a flow chart of a method for predicting all-cause mortality risk in peritoneal dialysis patients according to an embodiment of the present invention. DETAILED DESCRIPTION

[0017] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0018] A specific embodiment of the present invention discloses a method for predicting the risk of all-cause mortality in peritoneal dialysis patients. Figure 1 As shown, the following steps are included: S1. Construct a sample set based on the patient's time series test data and baseline data in each data center; S2, training a neural network model based on sample sets from multiple data centers to obtain a risk prediction pre-training model; S3. Obtaining a feature representation of each sample based on the risk prediction pre-training model, and clustering the feature representation to obtain a cluster type for each sample; S4. Fine-tune the risk prediction pre-training model based on samples of each cluster type to obtain a death risk prediction model corresponding to each cluster type; S5. Input the time series test data and baseline data of the patient to be predicted into the risk prediction pre-training model to obtain the feature representation of the patient to be predicted; obtain the cluster type corresponding to the patient to be predicted based on the feature representation, and obtain the prediction result of the patient to be predicted based on the risk prediction model corresponding to the cluster type.

[0019] Compared with the prior art, the training method of the all-cause mortality risk prediction model for peritoneal dialysis patients provided in this embodiment constructs a sample set in each data center, adopts sample sets from multiple data centers to make full use of data from multiple data sources, improves the accuracy of the training model, and ensures data security. After obtaining the pre-trained model, clustering is performed based on the feature representation of the sample, and the pre-trained model is fine-tuned based on the sample of each cluster type, so as to train the corresponding mortality risk prediction model for each type and improve the accuracy of the model. The performance of the model will not be affected by the small number of a certain type through pre-training and fine-tuning, thereby improving the accuracy of the mortality risk prediction model corresponding to each cluster type. By obtaining the cluster type of the patient to be predicted, its time series test data and baseline data are input into the risk prediction model corresponding to its cluster type, so as to quickly and accurately obtain the mortality risk prediction result of the patient to be predicted.

[0020] During implementation, the patient's time series test data includes: blood routine time series data, biochemical time series data, vital signs time series data (such as blood pressure, heart rate, etc.) These data change dynamically over time. Baseline data includes relatively static data such as patient demographic information (such as age, gender, etc.), basic disease information, etc.

[0021] Patient data of each data center is collected to construct a sample set. The sample is labeled as whether it is dead or not, 1 indicates death, and 0 indicates not death.

[0022] During implementation, if some of the time points of a certain data item in the time series data have missing data, it can be filled by median filling, mean filling, etc. If a certain data item is missing, the default value is used instead.

[0023] During implementation, the sample set of each data center includes a support set and a query set, that is, a part of the samples in the sample set is divided into a support set, and a part of the samples constitutes a query set.

[0024] The neural network model is trained based on sample sets from multiple data centers to obtain a risk prediction pre-training model, including: S20, constructing a neural network model, and sending the constructed neural network model to each data center; S21, record the parameters of the current neural network model as initial parameters; S22. For each data center, the neural network model is trained based on the local support set to update the parameters of the local neural network model; the loss of the updated local neural network model is calculated on the local query set to obtain the query loss; S23, updating the initial parameters based on the query loss of each data center, and updating the parameters of the neural network model of each data center to the updated initial parameters; S24. Determine whether the stopping condition is met. If so, stop training to obtain the risk prediction pre-training model. Otherwise, return to step S21.

[0025] During implementation, a neural network model is first constructed, and the constructed neural network model is sent to each data center, and the current parameters of the neural network model are recorded as initial parameters. .

[0026] Specifically, the constructed neural network model includes: Multi-channel time series feature extraction module, used to extract the features of time series inspection data; A baseline feature extraction module is used to extract features of baseline data; A feature fusion module is used to fuse the features of the time series test data with the features of the baseline data; The prediction module is used to predict the risk of death based on the fused features.

[0027] In implementation, the multi-channel temporal feature extraction module includes multiple channels, each channel is composed of GRU units, and each channel corresponds to a temporal test data, such as blood routine data. The baseline feature extraction module uses a learnable embedding matrix to map the baseline data to the same representation space as the temporal test data, that is, the feature dimension is the same.

[0028] During implementation, the feature fusion module can adopt the attention mechanism to dynamically adjust the weights of different features to improve the accuracy and interpretability of model predictions.

[0029] For each data center, the local neural network model is first trained based on the support set in the local data set, and the parameters of the local neural network model are updated. In implementation, multi-step training can be performed, that is, the parameters of the local neural network model are updated multiple times, and the parameters of the last update are recorded as , Represents the parameters of the local neural network model of the i-th data center.

[0030] After updating the parameters of the local neural network model, each data center calculates the loss on the local query set based on the updated local neural network model to obtain the query loss.

[0031] During implementation, for the i-th data center, the samples on the local query set are input into the local neural network model (the model parameters are ), calculate the loss based on the output of the model. When implementing, the cross entropy loss function can be used to calculate the loss of each sample, and the average loss of the samples in the query set can be calculated to obtain the query loss .

[0032] After obtaining the query loss of each data center, the following formula is used to update the initial parameters based on the query loss of each selected data center: ; ; in, represents the gradient of the total loss of N data centers, represents the total loss of N data centers, represents the query loss of the i-th data center, represents the training weight of the i-th data center, represents the learning rate, Represents the initial parameters.

[0033] During implementation, the training weight of a data center can be determined based on the number and quality of samples in the data center.

[0034] The parameters of the neural network model of each data center are updated to the updated initial parameters, thereby completing the back propagation of the neural network model based on multiple data centers.

[0035] After completing one back propagation, if the current stop condition is met, the training is terminated to obtain the risk prediction pre-training model. If not, return to step S21 to continue training until the stop condition of model training is met. During implementation, the stop condition can be reaching a preset number of training times or the total training loss reaches a preset accuracy.

[0036] The type of peritoneal transport, complications, nutritional status, etc. will affect the risk of all-cause mortality in peritoneal dialysis patients. Different groups differ in patient distribution and statistical characteristics, and therefore the predicted risks will differ. To improve the accuracy of risk prediction, after obtaining a risk prediction pre-training model, the present invention obtains a feature representation of each sample based on the risk prediction pre-training model, and clusters the feature representation to obtain a clustering type for each sample.

[0037] During implementation, the feature representation of each sample is the feature output by the feature fusion module after the sample's time series test data and baseline data are input into the risk prediction pre-training model.

[0038] Specifically, clustering the feature representation to obtain a clustering type for each sample includes: S31, randomly initialize K cluster centers; S32, sending the K cluster centers to each data center; each data center determines the cluster type of each sample based on the distance between each sample in the local sample set and the K cluster centers, and calculates the local cluster center; S33, updating the K clustering centers based on the local clustering center of each data center and the clustering weight of each data center; S34, determine the update accuracy of the K cluster centers. If all are less than a preset threshold, end clustering and obtain the cluster type of each sample; otherwise, return to step S32.

[0039] When implementing, first initialize K cluster centers, expressed as , each cluster center represents a cluster type. The number of K can be set empirically. The dimension of each cluster center is the same as the dimension of the feature representation.

[0040] First, send K cluster centers to each data center.

[0041] Each data center calculates the distance between each sample in the local sample set and the K cluster centers, and determines which cluster type each sample belongs to based on the distance. In implementation, the Euclidean distance can be used to calculate the distance between the sample and the cluster center. For each cluster type, the local cluster center of the data center is calculated based on the samples of that type in the data center. In implementation, the centroid or mean of all samples of that type can be calculated as the local cluster center corresponding to that type.

[0042] After obtaining the local cluster center corresponding to each data center, the K cluster centers are updated based on the local cluster center of each data center and the weight of each data center.

[0043] During implementation, the quantity and quality of data in each data center are different because they have different weights.

[0044] The clustering weight of each data center is calculated in the following way, including: S331, each data center calculates a local data clustering effect based on a local clustering center; and obtains a first weight of each data center based on the local data clustering effect of each data center; S332, obtaining K first pre-updated cluster centers based on the local cluster centers of each data center; for each data center, obtaining K second pre-updated cluster centers corresponding to the data center based on the local cluster centers of other data centers; S333, calculating a second weight of the data center based on the K first pre-updated cluster centers and the K second pre-updated cluster centers corresponding to the data center; S334. Obtain a clustering weight of each data center based on the first weight and the second weight of each data center.

[0045] During implementation, each data center uses the following formula to calculate the local data clustering effect based on the local cluster center: ; in, represents the clustering effect of the i-th data center, represents the cohesion of the jth sample in the i-th data center, represents the separation degree of the jth sample in the i-th data center, Represents the number of samples in the i-th data center.

[0046] The clustering effect represents the similarity of samples of the same type and the separation between different types.

[0047] During implementation, the cohesion of the jth sample in the i-th data center is the average of the distances from the jth sample to other samples in the cluster type to which it belongs.

[0048] The separation degree of the jth sample is calculated as follows: the average distance from the jth sample to the samples of each cluster type other than the cluster type to which it belongs is calculated, and the sample with the minimum average distance is selected as the separation degree of the jth sample.

[0049] From the above calculation formula, we can know that the clustering effect of the i-th data center is a value between -1 and +1. The closer to +1, the better the clustering effect, and the closer to -1, the worse the clustering effect. The data center with poor clustering effect has a greater impact on the clustering.

[0050] Therefore, after obtaining the local data clustering effect of each data center, the first weight is calculated using the following formula: ; in, Represents the first weight of the i-th data center.

[0051] In implementation, K first pre-updated cluster centers are obtained based on the local cluster center of each data center. In implementation, for K clusters, the mean of the corresponding cluster centers of the local cluster centers of all data centers is calculated to obtain K first updated cluster centers, which are expressed as .

[0052] For each data center, the K second pre-updated cluster centers corresponding to the data center are obtained based on the local cluster centers of other data centers. For example, for the i-th data center, the local cluster centers of other data centers except the data center are averaged to obtain the second pre-updated cluster center corresponding to the i-th data center, which is expressed as The difference between the first pre-updated cluster center and the second pre-updated cluster center indicates the degree of influence of the i-th data center on future clustering.

[0053] Therefore, based on the K first pre-updated cluster centers and the K second pre-updated cluster centers corresponding to the data center, the second weight of the data center is calculated using the following formula: ; in, represents the jth first pre-updated cluster center, represents the jth second pre-updated cluster center corresponding to the i-th data center, represents the jth second pre-updated cluster center corresponding to the tth data center, N represents the number of data centers, represents the second weight of the i-th data center.

[0054] For the i-th data center, the first weight and the second weight are averaged to obtain the corresponding clustering weight.

[0055] Afterwards, the K cluster centers are updated based on the local cluster center of each data center and the cluster weight of each data center, that is, the local cluster centers of all data centers are weighted and averaged to obtain new K cluster centers, and the added weight is the cluster weight.

[0056] During implementation, the clustering weights are not manually set or fixed, but change dynamically according to the data. Because they are more in line with the actual objective situation, the clustering accuracy is higher.

[0057] After updating the K cluster centers, determine whether the end condition is met and whether the updated accuracy is less than the preset threshold. If the updated accuracy of the K cluster centers is less than the preset threshold, the clustering is terminated and the cluster type of each sample is obtained; otherwise, return to step S32 to continue clustering until the end condition is met. During implementation, the end condition can also be the number of iterations met.

[0058] After clustering, each data center determines which cluster type its sample belongs to based on the last K cluster centers.

[0059] Based on the samples of each cluster type, the risk prediction pre-training model is fine-tuned to obtain the risk prediction model corresponding to each cluster type, specifically including: S41. For each cluster type, determine a third weight of each data center based on samples belonging to the cluster type in each data center; S42. Fine-tune the risk prediction pre-training model based on the third weight of each data center to obtain a risk prediction model corresponding to each clustering type.

[0060] For the k-th clustering type, the number and quality of samples of this type in each data center are different because the impact weights on fine-tuning are different.

[0061] Specifically, the third weight of each data center is determined by using the following formula based on the samples belonging to the cluster type in each data center: ; in, represents the sample quantity weight of the i-th data center, represents the data quality weight of the i-th data center, and represents the adjustment parameters, Represents the third weight of the i-th number center.

[0062] During implementation, for the kth cluster type, the third weight of each data center is calculated first. The sample quantity weight of the i-th data center is the ratio of the number of samples belonging to the kth cluster type in the i-th data center to the total number of samples of the kth cluster type.

[0063] The data quality weight of the i-th data center is: the average accuracy of the samples belonging to the k-th cluster type in the i-th data center. The accuracy of the samples is ,in, represents the prediction result of the pre-training model, and y represents the sample label.

[0064] During implementation, each clustering type fine-tunes the risk prediction pre-trained model based on its corresponding samples to obtain its corresponding risk prediction model. Since data cannot be shared, the fine-tuning process can refer to the above steps S21~S24. The neural network model is the risk prediction pre-trained model. The samples belonging to the kth clustering type in each data center are also divided into a support set and a query set. The training weight of each data center is its third weight, so as to obtain the risk prediction model corresponding to each clustering type.

[0065] For the patient to be predicted, first obtain their time series test data and baseline data and input them into the pre-training model to obtain the feature representation of the patient to be predicted. According to their feature representation, the cluster type corresponding to the patient to be predicted can be obtained. Then, the time series test data and baseline data of the patient to be predicted are input into the risk prediction model corresponding to their cluster type, so as to quickly and accurately obtain the death risk prediction result of the patient to be predicted.

[0066] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0067] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for predicting the risk of all-cause mortality in peritoneal dialysis patients, characterized in that: The following steps are involved: A sample set was constructed in each data center based on the patients’ time-series test data and baseline data; A risk prediction pre-training model is obtained by training a neural network model based on sample sets from multiple data centers; Obtaining a feature representation of each sample based on the risk prediction pre-training model, and clustering the feature representation to obtain a cluster type for each sample; Fine-tuning the risk prediction pre-training model based on samples of each cluster type to obtain a death risk prediction model corresponding to each cluster type; The time series test data and baseline data of the patient to be predicted are input into the risk prediction pre-training model to obtain the feature representation of the patient to be predicted; the cluster type corresponding to the patient to be predicted is obtained according to the feature representation, and the prediction result of the patient to be predicted is obtained based on the risk prediction model corresponding to the cluster type.

2. The method for predicting all-cause mortality risk in peritoneal dialysis patients according to claim 1, characterized in that: Clustering the feature representations to obtain the clustering type of each sample includes: S31, randomly initialize K cluster centers; S32, sending the K cluster centers to each data center; each data center determines the cluster type of each sample based on the distance between each sample in the local sample set and the K cluster centers, and calculates the local cluster center; S33, updating the K clustering centers based on the local clustering center of each data center and the clustering weight of each data center; S34, determine the update accuracy of the K cluster centers. If all are less than a preset threshold, end clustering and obtain the cluster type of each sample; otherwise, return to step S32.

3. The method for predicting all-cause mortality risk in peritoneal dialysis patients according to claim 2, characterized in that: The clustering weight of each data center is calculated in the following way, including: Each data center calculates the local data clustering effect based on the local clustering center; and obtains the first weight of each data center based on the local data clustering effect of each data center; Obtain K first pre-updated cluster centers based on the local cluster center of each data center; For each data center, obtain K second pre-updated cluster centers corresponding to the data center based on the local cluster centers of other data centers; Calculating a second weight of the data center based on the K first pre-updated cluster centers and the K second pre-updated cluster centers corresponding to the data center; Based on the first weight and the second weight of each data center, a clustering weight of each data center is obtained.

4. The method for predicting all-cause mortality risk in peritoneal dialysis patients according to claim 3, characterized in that: Each data center uses the following formula to calculate the local data clustering effect based on the local cluster center: ； in, represents the clustering effect of the i-th data center, represents the cohesion of the jth sample in the i-th data center, represents the separation degree of the jth sample in the i-th data center, Represents the number of samples in the i-th data center.

5. The method for predicting all-cause mortality risk in peritoneal dialysis patients according to claim 3, characterized in that: The first weight is calculated using the following formula: ； in, represents the first weight of the i-th data center, represents the clustering effect of the i-th data center, and N represents the number of data centers.

6. The method for predicting all-cause mortality risk in peritoneal dialysis patients according to claim 3, characterized in that: The second weight of the data center is calculated based on the K first pre-updated cluster centers and the K second pre-updated cluster centers corresponding to the data center using the following formula: ； in, represents the jth first pre-updated cluster center, represents the jth second pre-updated cluster center corresponding to the i-th data center, represents the jth second pre-updated cluster center corresponding to the tth data center, N represents the number of data centers, represents the second weight of the i-th data center.

7. The method for predicting all-cause mortality risk in peritoneal dialysis patients according to claim 1, characterized in that: The risk prediction pre-training model is fine-tuned based on samples of each cluster type to obtain a risk prediction model corresponding to each cluster type, including: For each cluster type, determining a third weight of each data center based on samples in each data center belonging to the cluster type; Based on the third weight of each data center, federated learning is used to fine-tune the risk prediction pre-training model to obtain a risk prediction model corresponding to each clustering type.

8. The method for predicting all-cause mortality risk in peritoneal dialysis patients according to claim 7, characterized in that: The third weight of each data center is determined based on the samples belonging to this cluster type in each data center using the following formula: ； in, represents the sample quantity weight of the i-th data center, represents the data quality weight of the i-th data center, and represents the adjustment parameters, Represents the third weight of the i-th number center.

9. The method for predicting all-cause mortality risk in peritoneal dialysis patients according to claim 1, characterized in that: The sample set of each data center includes a support set and a query set. The neural network model is trained based on the sample sets of multiple data centers to obtain a risk prediction pre-training model, including: S20, constructing a neural network model, and sending the constructed neural network model to each data center; S21, record the parameters of the current neural network model as initial parameters; S22. For each data center, the neural network model is trained based on the local support set to update the parameters of the local neural network model; the loss of the updated local neural network model is calculated on the local query set to obtain the query loss; S23, updating the initial parameters based on the query loss of each data center, and updating the parameters of the neural network model of each data center to the updated initial parameters; S24. Determine whether the stopping condition is met. If so, stop training to obtain the risk prediction pre-training model. Otherwise, return to step S21.

10. The method for predicting all-cause mortality risk in peritoneal dialysis patients according to claim 9, characterized in that: The initial parameters are updated based on the query loss of each data center using the following formula: ； ； in, represents the gradient of the total loss of N data centers, represents the total loss of N data centers, represents the query loss of the i-th data center, represents the training weight of the i-th data center, represents the learning rate, Represents the initial parameters.

Images