Intelligent decision-making system for peritoneal dialysis prescription

Abstract

The invention discloses an intelligent decision-making system for a peritoneal dialysis prescription, and relates to the technical field of data processing, and the system comprises a data collection module which is used for integrating a multi-source information system, and obtaining a glucose clearance rate sequence, a creatinine clearance rate sequence, an electrolyte concentration sequence and a blood pressure concentration sequence of a target patient in a preset window; the monitoring feature vector construction module is used for performing time sequence iteration multi-dimensional processing on the peritoneal dialysis monitoring data and constructing a monitoring feature vector; the element type matching module is used for obtaining a matched peritoneal dialysis prescription decision element type; and the display module is used for displaying the matched peritoneal dialysis prescription decision meta-type. The peritoneal dialysis prescription decision-making method and device solve the technical problem that in the prior art, analysis is not accurate due to the fact that reliability of data based on peritoneal dialysis prescription decision-making is poor, and the technical effects that multi-source data are automatically collected to construct the monitoring feature vector to conduct decision-making meta-type matching, and the peritoneal dialysis prescription decision-making accuracy and efficiency are improved are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to an intelligent peritoneal dialysis prescription decision system. BACKGROUND

[0002] At present, the formulation of peritoneal dialysis prescription mainly depends on personal experience and limited clinical data, which is difficult to comprehensively and dynamically grasp the complex conditions of the patient's peritoneal transport characteristics, residual renal function, fluid and electrolyte balance, etc., resulting in low precision of the prescription. Different experience differences also make the prescription lack of standardization and consistency. At the same time, artificial analysis and processing of patient data is time-consuming and laborious, and cannot timely adjust the prescription according to the real-time condition of the patient.

[0003] The prior art has the technical problem of poor reliability of data relied on by peritoneal dialysis prescription decision, resulting in inaccurate analysis. SUMMARY

[0004] The present application provides an intelligent peritoneal dialysis prescription decision system, which is used to solve the technical problem of poor reliability of data relied on by peritoneal dialysis prescription decision in the prior art, resulting in inaccurate analysis.

[0005] In view of the above problems, the present application provides an intelligent peritoneal dialysis prescription decision system.

[0006] The present application provides an intelligent peritoneal dialysis prescription decision system, which comprises:

[0007] A data acquisition module is used to integrate a multi-source information system to obtain a glucose clearance rate sequence, a creatinine clearance rate sequence, an electrolyte concentration sequence and a blood pressure concentration sequence of a target patient in a preset window; a monitoring feature vector construction module is used to iterate multidimensional processing of peritoneal dialysis monitoring data time sequence by traversing the glucose clearance rate sequence, the creatinine clearance rate sequence, the electrolyte concentration sequence and the blood pressure concentration sequence to construct a monitoring feature vector; a prototype matching module is used to perform multidimensional peritoneal dialysis prescription decision prototype matching based on the monitoring feature vector to obtain a matched peritoneal dialysis prescription decision prototype; and a display module is used to display the matched peritoneal dialysis prescription decision prototype on an explicit interface.

[0008] One or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0009] The data acquisition module integrates multi-source information systems to obtain the target patient's glucose clearance rate, creatinine clearance rate, electrolyte concentration, and blood pressure concentration sequences within a preset window. The monitoring feature vector construction module performs iterative multi-dimensional processing of peritoneal dialysis monitoring data to construct monitoring feature vectors. The metatype matching module performs multi-dimensional peritoneal dialysis prescription decision metatype matching to obtain matching peritoneal dialysis prescription decision metatypes. The display module displays the matching peritoneal dialysis prescription decision metatypes on an explicit interface. This achieves the technical effect of automatically collecting multi-source data to construct monitoring feature vectors for decision metatype matching, thereby improving the accuracy and efficiency of peritoneal dialysis prescription decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0011] Figure 1 A schematic diagram of the structure of an intelligent decision-making system for peritoneal dialysis prescription is provided for an embodiment of the present application;

[0012] Figure 2 A schematic diagram of the structure of a data acquisition module in an intelligent decision-making system for peritoneal dialysis prescription is provided for an embodiment of the present application.

[0013] Description of the accompanying drawings: data collection module 10, monitoring feature vector construction module 20, metatype matching module 30, display module 40. DETAILED DESCRIPTION

[0014] The present application provides an intelligent peritoneal dialysis prescription decision-making system to solve the technical problem in the prior art that the data based on which peritoneal dialysis prescription decisions are made are poorly reliable, resulting in inaccurate analysis.

[0015] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making any creative work are within the scope of protection of this application.

[0016] Examples, such as Figure 1 As shown, the present application provides a peritoneal dialysis prescription intelligent decision-making system, the system comprising:

[0017] The data acquisition module 10 is configured to integrate a multi-source information system to obtain a glucose clearance rate sequence, a creatinine clearance rate sequence, an electrolyte concentration sequence and a blood pressure concentration sequence of the target patient within a preset window.

[0018] Specifically, the data acquisition module 10 is configured to integrate a multi-source information system to obtain a glucose clearance rate sequence, a creatinine clearance rate sequence, an electrolyte concentration sequence and a blood pressure concentration sequence of the target patient within a preset window.

[0019] In one possible implementation, the peritoneal transport type, toxin level, edema information, nausea and vomiting information and the like of the patient are extracted from the multi-source information system to verify the feasibility of the matched peritoneal dialysis prescription decision pattern, and preferably, the verification is performed by experienced analysis by those skilled in the art. Thus, the technical effect of improving the reliability of decision analysis is achieved.

[0020] The monitoring feature vector construction module 20 is configured to perform peritoneal dialysis monitoring data time sequence iteration multi-dimensional processing on the glucose clearance rate sequence, the creatinine clearance rate sequence, the electrolyte concentration sequence and the blood pressure concentration sequence to construct a monitoring feature vector.

[0021] Specifically, based on the glucose clearance rate sequence, the creatinine clearance rate sequence, the electrolyte concentration sequence and the blood pressure concentration sequence obtained by the data acquisition module 10, the data value is deeply mined from multiple dimensions through time series iterative multi-dimensional processing. When processing the glucose clearance rate sequence and the creatinine clearance rate sequence, the trend attention recognition unit identifies the fluctuation trend thereof, determines the glucose clearance rate fluctuation trend factor and the creatinine clearance rate fluctuation trend factor, and then calculates the attention weight of the two through the attention weight recognition subunit. Then, the trend feature fusion subunit fuses these information to form a peritoneal transport monitoring feature subvector, which reflects the transport characteristics of the peritoneum to the material. For the electrolyte concentration sequence and the blood pressure concentration sequence, the trend feature enhancement analysis unit processes them respectively, obtains the long and short term trend feature subvector through the long and short trend feature recognition subunit, constructs a trend feature enhancement matrix and enhances the short trend feature subvector using the matrix, to obtain the electrolyte concentration monitoring feature subvector and the blood pressure concentration monitoring feature subvector. Finally, the monitoring feature vector integration unit integrates the above subvectors together to construct a comprehensive and representative monitoring feature vector. This monitoring feature vector comprehensively reflects the multiple aspects of the patient's peritoneal dialysis process, providing data support for the subsequent meta-type matching module 30 to perform peritoneal dialysis prescription decision meta-type matching, making the decision result more targeted and accurate.

[0022] The meta-type matching module 30 is configured to perform multi-dimensional peritoneal dialysis prescription decision meta-type matching based on the monitoring feature vector, and obtain a matched peritoneal dialysis prescription decision meta-type.

[0023] Specifically, the meta-type matching module 30 is responsible for comparing the monitoring feature vector with historical experience data to obtain the most suitable peritoneal dialysis prescription decision meta-type. The module first extracts the stored historical multi-source peritoneal dialysis prescription decision set from the multi-source information system. These historical data contain a large number of dialysis prescriptions and related monitoring data of different patients. Then, the same type aggregation unit classifies and aggregates these historical data, classifies similar prescription decisions into one category, forms an aggregated historical multi-source peritoneal dialysis prescription decision cluster, and makes the data structure more clear for subsequent analysis. After that, the center identification unit traverses these decision clusters and determines a representative center from each decision cluster, that is, an aggregated historical multi-source peritoneal dialysis prescription decision meta-type, each meta-type contains an aggregated monitoring feature vector, which is a comprehensive embodiment of the characteristics of the patients corresponding to this type of prescription decision. Finally, the aggregated monitoring feature vector matching unit matches the monitoring feature vector generated by the monitoring feature vector construction module with each aggregated monitoring feature vector in the aggregated historical multi-source peritoneal dialysis prescription decision meta-type set, calculates the similarity, and finds the aggregated historical multi-source peritoneal dialysis prescription decision meta-type with the largest similarity. This meta-type is the most suitable prescription decision meta-type for the current patient's monitoring data.

[0024] The display module 40 is configured to display the matched peritoneal dialysis prescription decision pattern on an explicit interface.

[0025] Specifically, after completing the multi-dimensional peritoneal dialysis prescription decision pattern matching based on the monitoring feature vector, the display module 40 presents the matching results in an intuitive and easy-to-understand manner on an explicit interface, ensuring that medical personnel can clearly and quickly obtain key information. The display content covers a variety of important data related to peritoneal dialysis prescriptions, such as the recommended dialysis type for different transport types, the appropriate glucose concentration in the dialysate, and the dialysis fluid replacement frequency, single fluid volume adjustment suggestions based on residual renal function, fluid balance, and electrolyte balance. Through such display, direct decisions about peritoneal dialysis treatment plans can be made quickly based on these information combined with the actual physical condition of the patient, without the need to spend additional time on complex data interpretation and analysis. Not only does this improve the efficiency of medical decision-making, but it also reduces treatment deviations caused by human data interpretation errors, effectively ensuring that patients can receive accurate and effective peritoneal dialysis treatment in a timely manner.

[0026] In one possible implementation, as shown in Figure 2 The data acquisition module 10 further comprises:

[0027] A glucose clearance rate sequence acquisition unit is configured to extract the glucose concentration data in blood and peritoneal dialysis fluid monitored by the peritoneal dialysis monitoring device in the multi-source information system within a preset window to obtain a glucose clearance rate sequence.

[0028] A creatinine clearance rate sequence acquisition unit is configured to extract the synchronous creatinine concentration monitoring data of the portable dialysis analyzer and the portable blood analyzer in the multi-source information system within a preset window, and after mapping analysis, obtain a creatinine clearance rate sequence.

[0029] An electrolyte concentration sequence acquisition unit is configured to extract the electrolyte concentration of the portable electrolyte analyzer in the multi-source information system within a preset window to obtain an electrolyte concentration sequence.

[0030] A blood pressure concentration sequence acquisition unit is configured to extract blood pressure monitoring data of the blood pressure monitor in the multi-source information system within a preset window to obtain the blood pressure concentration sequence.

[0031] Specifically, the glucose clearance rate sequence acquisition unit aims to accurately obtain the glucose concentration data in blood and peritoneal dialysis fluid monitored by the peritoneal dialysis monitoring device within the preset window from the multi-source information system, and then generate a glucose clearance rate sequence. The specific algorithm process is as follows: first, all the blood and peritoneal dialysis fluid glucose concentration data recorded by the peritoneal dialysis monitoring device within the preset window are screened out from the multi-source information system to construct an original data set. Then, the data set is preprocessed to remove possible abnormal values and noise data, and abnormal values are identified and removed by setting a reasonable data range. Then, for each time point, the glucose clearance rate is calculated according to the formula: glucose clearance rate = (peritoneal dialysis fluid glucose concentration x peritoneal dialysis outflow) / blood glucose concentration. Finally, the glucose clearance rates calculated at each time point are arranged in chronological order to form a complete glucose clearance rate sequence.

[0032] The creatinine clearance rate sequence acquisition unit extracts creatinine concentration data synchronously monitored by the portable dialysis analyzer and the portable blood analyzer within the preset window from the multi-source information system. Due to the differences in measurement principle and accuracy between the two analyzers, the original data obtained cannot be directly used. Therefore, mapping analysis is carried out on these synchronous creatinine concentration monitoring data to unify the data of different analyzers to the same standard and dimension to eliminate errors caused by data differences. After such mapping analysis, the creatinine clearance rate at each time point is calculated according to the formula: creatinine clearance rate = (calibrated dialysis analyzer measured creatinine concentration x urine or peritoneal fluid volume) / blood analyzer measured creatinine concentration, the creatinine clearance rate at each time point within the preset window is calculated, and finally these creatinine clearance rates arranged in chronological order are combined into a creatinine clearance rate sequence. This sequence can effectively reflect the dynamic changes of the patient's renal function during peritoneal dialysis, providing a key basis for subsequent accurate assessment of the patient's renal function and development of personalized peritoneal dialysis prescriptions.

[0033] The electrolyte concentration sequence acquisition unit accurately extracts electrolyte concentration data recorded by the portable electrolyte analyzer within the preset window from the multi-source information system. The portable electrolyte analyzer can monitor the electrolyte concentration in the patient's body in real time and accurately. These dispersed electrolyte concentration data within the preset window are collected, preliminarily sorted and sorted, and converted into an electrolyte concentration sequence arranged in chronological order. This sequence contains dynamic change information of the patient's electrolyte concentration over time during peritoneal dialysis, and can intuitively reflect the patient's electrolyte balance state.

[0034] The blood pressure concentration sequence acquisition unit is dedicated to extracting blood pressure monitoring data recorded by the blood pressure monitor in a preset window from the multi-source information system. The blood pressure monitor will continuously and regularly measure the blood pressure of the patient in this preset time period, and these scattered measurement data will be collected one by one. Then, the blood pressure data is arranged in chronological order of measurement, thereby forming a blood pressure concentration sequence. This sequence clearly presents the fluctuations of the patient's blood pressure over time during peritoneal dialysis, and is an important basis for reflecting the patient's cardiovascular system status and dialysis tolerance.

[0035] In one possible implementation, the monitoring feature vector construction module 20 further includes:

[0036] A trend attention identification unit is configured to identify the trend attention of the glucose clearance rate sequence and the creatinine clearance rate sequence, and obtain a peritoneal transport monitoring feature sub-vector.

[0037] A trend feature enhancement analysis unit is configured to respectively perform trend feature enhancement analysis on the electrolyte concentration sequence and the voltage concentration sequence, and obtain an electrolyte concentration monitoring feature sub-vector and a voltage concentration monitoring feature sub-vector.

[0038] A monitoring feature vector summarizing unit is configured to summarize the peritoneal transport monitoring feature sub-vector, the electrolyte concentration monitoring feature sub-vector, and the voltage concentration monitoring feature sub-vector into the monitoring feature vector.

[0039] Specifically, the trend attention identification unit is dedicated to the glucose clearance rate sequence and the creatinine clearance rate sequence. First, the fluctuation trend identification sub-unit analyzes the two sequences respectively, identifies the glucose clearance rate fluctuation trend factor and the creatinine clearance rate fluctuation trend factor, and reveals the change trend of the data. Then, the attention weight identification sub-unit calculates the proportion of the glucose clearance rate fluctuation trend factor in the sum of the glucose clearance rate fluctuation trend factor and the creatinine clearance rate fluctuation trend factor through the ratio calculation micro-unit, obtains the glucose clearance rate attention weight, and subtracts the weight from 1 to obtain the creatinine clearance rate attention weight, thereby measuring the relative importance of the two sequences to the peritoneal transport monitoring feature. Finally, the trend feature fusion sub-unit fuses the glucose clearance rate sequence and the creatinine clearance rate sequence according to the weights, thereby obtaining the peritoneal transport monitoring feature sub-vector.

[0040] The task of the monitoring feature vector summary unit is to effectively integrate different types of monitoring feature sub-vectors generated by other units in the system to form a monitoring feature vector that comprehensively reflects the peritoneal dialysis state of the patient. Among them, the peritoneal transport monitoring feature sub-vector is obtained by analyzing the glucose clearance rate sequence and the creatinine clearance rate sequence by the trend attention recognition unit, which contains important information about the patient's peritoneal transport capacity and can help determine the patient's peritoneal transport type. The electrolyte concentration monitoring feature sub-vector is obtained after long and short trend feature recognition, construction of trend feature enhancement matrix and enhancement of short trend feature sub-vector by the trend feature enhancement analysis unit, which presents the dynamic change trend of the patient's electrolyte concentration in detail and is of great significance for evaluating the electrolyte balance. The blood pressure concentration monitoring feature sub-vector also undergoes trend feature enhancement analysis and accurately reflects the changes of the patient's blood pressure during peritoneal dialysis, which is a key basis for determining the stability of the patient's cardiovascular system. The monitoring feature vector summary unit combines the three monitoring feature sub-vectors by weighting according to the importance of the information represented by each sub-vector. The final monitoring feature vector combines multiple key aspects of data during peritoneal dialysis of the patient, providing comprehensive and accurate data support for the subsequent meta-type matching module to carry out multi-dimensional peritoneal dialysis prescription decision meta-type matching, ensuring that the decision system can develop a more scientific and personalized peritoneal dialysis prescription based on the patient's actual situation.

[0041] The monitoring feature vector summary unit receives three important sub-vectors from different analysis links, which are the peritoneal transport monitoring feature sub-vector reflecting the patient's peritoneal transport characteristics of substances such as glucose and creatinine, the electrolyte concentration monitoring feature sub-vector reflecting the dynamic changes and balance of electrolytes such as sodium, potassium, and chlorine in the patient's body, and the blood pressure concentration monitoring feature sub-vector showing the fluctuation trend and stability of the patient's blood pressure. The monitoring feature vector summary unit uses weighted summation to combine them into a comprehensive monitoring feature vector according to the physiological significance and importance of each sub-vector to the dialysis prescription decision, covering multiple key physiological indicators of the patient during peritoneal dialysis, providing a comprehensive, accurate and representative data basis for the subsequent precise matching of the most suitable peritoneal dialysis prescription decision meta-type through the meta-type matching module.

[0042] In one possible implementation manner, the trend attention recognition unit further includes:

[0043] A fluctuation trend recognition sub-unit is configured to respectively perform fluctuation trend recognition on the glucose clearance rate sequence and the creatinine clearance rate sequence to obtain a glucose clearance rate fluctuation trend factor and a creatinine clearance rate fluctuation trend factor.

[0044] The attention weight identification subunit is configured to identify attention weights of the glucose clearance rate fluctuation trend factor and the creatinine clearance rate fluctuation trend factor, and obtain glucose clearance rate attention weights and creatinine clearance rate attention weights.

[0045] The trend feature fusion subunit is configured to perform trend feature fusion on the glucose clearance rate sequence and the creatinine clearance rate sequence based on the glucose clearance rate attention weights and the creatinine clearance rate attention weights, and obtain a peritoneal transport monitoring feature subvector.

[0046] Specifically, the fluctuation trend identification subunit adopts a method combining the moving average method and the slope analysis method when identifying the fluctuation trend of the glucose clearance rate sequence and the creatinine clearance rate sequence. For the glucose clearance rate sequence, a suitable moving window size is first selected, for example, 5 time points are set, and the 5-point moving average value of each time point is calculated to smooth the random noise in the sequence. Then, the difference between adjacent moving average values is calculated. If the difference is positive and reaches a certain threshold (for example, 0.1), it indicates that the glucose clearance rate in this time period is in an upward trend. If the difference is negative and exceeds the threshold, it is a downward trend. At the same time, the slope of the sequence in each sub-interval is calculated by linear regression. When the slope is positive and greater than a certain critical value (for example, 0.05), it is determined to be a significant increase. When the slope is negative and less than the critical value, it is determined to be a significant decrease. According to the duration, amplitude, etc. of these upward and downward trends, the glucose clearance rate fluctuation trend factor is quantitatively obtained. For the creatinine clearance rate sequence, the same processing procedure is adopted. First, the moving average is smoothed, then the difference and the slope are calculated, and the creatinine clearance rate fluctuation trend factor is determined according to the trend characteristics, providing key data for subsequent analysis.

[0047] The attention weight identification subunit compares the glucose clearance rate fluctuation trend factor with the sum of the glucose clearance rate fluctuation trend factor and the creatinine clearance rate fluctuation trend factor. Specifically, the glucose clearance rate fluctuation trend factor is divided by the sum of the two, and the quotient obtained is the glucose clearance rate attention weight. This weight reflects the relative importance of the glucose clearance rate fluctuation trend in the overall trend when evaluating the peritoneal transport characteristics. The creatinine clearance rate attention weight is obtained by subtracting the glucose clearance rate attention weight from 1. In this way, the respective weights of the two fluctuation trend factors are determined, providing key evidence for subsequent trend feature fusion of the glucose clearance rate sequence and the creatinine clearance rate sequence to more accurately construct the peritoneal transport monitoring feature subvector, which helps the system to comprehensively and accurately evaluate the peritoneal transport function of the patient, and then develop a more reasonable peritoneal dialysis prescription.

[0048] When the trend feature fusion subunit is working, first, the glucose clearance rate attention weight and the creatinine clearance rate attention weight are determined, and the sum of the two weights is equal to 1. Then, the glucose clearance rate sequence and the creatinine clearance rate sequence are found, and the clearance rate data corresponding to different time points are arranged in order in the two sequences. Because the units, size ranges, etc. of the data in the two sequences may not be the same, normalization processing is first performed to make them on the same comparable scale. After processing, starting from the first data of the two sequences, the glucose clearance rate attention weight is multiplied by the first data of the glucose clearance rate sequence, and the creatinine clearance rate attention weight is multiplied by the first data of the creatinine clearance rate sequence, and the two results are added to obtain the first fused data. In the same way, the data of each corresponding time point in the two sequences are calculated to obtain a series of fused data. Finally, the fused data are arranged in order to form a new vector, and the new vector is the peritoneal transport monitoring feature subvector.

[0049] In one possible implementation manner, the attention weight identification subunit further includes:

[0050] The ratio calculation micro unit is configured to calculate the ratio of the glucose clearance rate fluctuation trend factor to the sum of the glucose clearance rate fluctuation trend factor and the creatinine clearance rate fluctuation trend factor, and obtain the glucose clearance rate attention weight.

[0051] The creatinine clearance rate attention weight acquisition micro unit is configured to take the difference between 1 and the glucose clearance rate attention weight as the creatinine clearance rate attention weight.

[0052] Specifically, the ratio calculation micro unit is configured to process the quantitative relationship between the glucose clearance rate fluctuation trend factor and the creatinine clearance rate fluctuation trend factor. When the two factors are obtained, the ratio calculation micro unit performs a division operation by taking the glucose clearance rate fluctuation trend factor as the numerator and the sum of the glucose clearance rate fluctuation trend factor and the creatinine clearance rate fluctuation trend factor as the denominator. Through such a calculation manner, the obtained result is the glucose clearance rate attention weight. The weight value represents the relative importance of the glucose clearance rate fluctuation trend in the process of evaluating the peritoneal transport characteristics of the patient, and the value range is between 0 and 1. The larger the value is, the more critical the glucose clearance rate fluctuation trend is in the overall evaluation.

[0053] The creatinine clearance rate attention weight is obtained by the creatinine clearance rate attention weight obtained by the ratio calculation micro unit. The creatinine clearance rate attention weight is obtained by subtracting the glucose clearance rate attention weight from 1. Since the sum of the two weights is 1, it means that when the influence of the fluctuation trend of glucose and creatinine clearance rate on the peritoneal transport property is comprehensively considered, the weights of the two are mutually restricted and complementary. The two micro units work together to lay a foundation for the subsequent trend feature fusion sub-unit to effectively fuse the glucose clearance rate sequence and the creatinine clearance rate sequence based on these weights, which helps to construct a more accurate peritoneal transport monitoring feature sub-vector, thereby providing strong data support for formulating individualized peritoneal dialysis prescriptions.

[0054] In a possible implementation manner, the trend feature enhancement analysis unit further includes:

[0055] The long-short trend feature identification sub-unit is configured to identify long-short trend features of the electrolyte concentration sequence, and obtain an electrolyte concentration change long trend feature sub-vector and an electrolyte concentration short trend feature sub-vector.

[0056] The trend feature enhancement matrix construction sub-unit is configured to identify, in a one-to-one manner, trend feature approximation degrees of elements in the electrolyte concentration change long trend feature sub-vector and the electrolyte concentration short trend feature sub-vector, perform normalization processing on the identification results, and fill the normalized results into an initially empty matrix to construct a trend feature enhancement matrix.

[0057] The feature sub-vector acquisition sub-unit is configured to enhance the electrolyte concentration short trend feature sub-vector by using the trend feature enhancement matrix, and obtain an electrolyte concentration monitoring feature sub-vector.

[0058] Specifically, the long-short trend feature recognition subunit deeply analyzes the electrolyte concentration sequence to mine the long-short trend features contained therein. A method similar to the slow channel in the slowfast network is adopted to analyze the electrolyte concentration sequence with two different step lengths. On the one hand, a longer step length is used for analysis. This approach can span more data points, thereby capturing the long-term trend of electrolyte concentration changes in a longer time range, and thus generating an electrolyte concentration change long trend feature subvector. This subvector reflects the long-term change law of electrolyte concentration in the patient's body, which helps to grasp the patient's electrolyte balance state from a macro perspective. On the other hand, a conventional step length is used to analyze the electrolyte concentration sequence. This approach can sensitively detect subtle fluctuations in electrolyte concentration in the short term, and thus obtain an electrolyte concentration short trend feature subvector. This subvector can provide a basis for timely feedback on recent changes in the patient's electrolyte concentration and rapid response and adjustment of treatment plans. These two subvectors quantify and characterize the changes in electrolyte concentration from different time scales, providing an important data basis for subsequent trend feature enhancement analysis and other steps.

[0059] The trend feature enhancement matrix construction subunit is mainly responsible for constructing the trend feature enhancement matrix based on the electrolyte concentration change long trend feature subvector and the electrolyte concentration short trend feature subvector. After receiving these two subvectors, a one-to-one trend feature approximation degree identification is performed between the elements. By comparing the trend features such as the change direction and amplitude of the elements at the same position in the long trend feature subvector and the short trend feature subvector, the trend feature approximation degree between each element pair is calculated using cosine similarity. After completing the approximation degree identification, the identification results are normalized. Since the original calculation results of the approximation degree may have large differences in ranges, normalization can map these results to a unified interval, such as [0, 1], making the approximation degrees between different element pairs comparable. Finally, the normalized results are sequentially filled into an initially empty matrix. The number of rows and columns of the matrix corresponds to the number of elements in the long trend feature subvector and the short trend feature subvector, respectively. Each element in the matrix represents the normalized trend feature approximation degree of the corresponding element pair. After this series of operations, the trend feature enhancement matrix is constructed, which can reflect the correlation degree between the long and short trend feature subvectors.

[0060] The feature sub-vector obtaining sub-unit takes the trend feature enhancement matrix as the basis, the matrix is obtained after trend feature approximation degree identification and normalization processing between elements of the long trend feature sub-vector of electrolyte concentration change and the short trend feature sub-vector of electrolyte concentration, and it contains the correlation information between the long trend feature and the short trend feature. The trend feature enhancement matrix and the short trend feature sub-vector of electrolyte concentration are multiplied. Through this operation, the long-term stable change information contained in the long trend feature is integrated into the short trend feature sub-vector, so that the trend features which are important in the long term and the short term are highlighted, and the interference brought by noise and accidental fluctuations is suppressed. Finally, after such enhancement processing, the electrolyte concentration monitoring feature sub-vector is obtained. This sub-vector integrates the advantages of long and short trend features, and more comprehensively and accurately reflects the dynamic change of electrolyte concentration in the patient's body. And the same principle is also used to obtain the voltage concentration monitoring feature sub-vector, to ensure effective monitoring and analysis of multiple key physiological indicators of the patient.

[0061] In one possible implementation manner, the meta-type matching module 30 further includes:

[0062] A decision set extraction unit is configured to extract a historical multi-source peritoneal dialysis prescription decision set stored in the multi-source information system.

[0063] A same-class aggregation unit is configured to perform same-class aggregation on the historical multi-source peritoneal dialysis prescription decision set to obtain an aggregated historical multi-source peritoneal dialysis prescription decision cluster.

[0064] A center identification unit is configured to perform center identification on the aggregated historical multi-source peritoneal dialysis prescription decision cluster to obtain an aggregated historical multi-source peritoneal dialysis prescription decision meta-type set, wherein each aggregated historical multi-source peritoneal dialysis prescription decision meta-type includes an aggregated monitoring feature vector.

[0065] An aggregated monitoring feature vector matching unit is configured to perform matching based on the monitoring feature vector and each aggregated monitoring feature vector corresponding to the aggregated historical multi-source peritoneal dialysis prescription decision meta-type set, and take the aggregated historical multi-source peritoneal dialysis prescription decision meta-type corresponding to the maximum matching similarity value as the matching peritoneal dialysis prescription decision meta-type.

[0066] Specifically, the decision set extraction unit extracts a historical multi-source peritoneal dialysis prescription decision set stored in the multi-source information system. This set covers the peritoneal dialysis related decision information of numerous patients in the past, and is the basic data source for subsequent analysis.

[0067] The same kind of aggregation unit receives a set of historical multi-source peritoneal dialysis prescription decisions extracted from a multi-source information system, which covers peritoneal dialysis prescription information of a plurality of patients at different times, including dialysate adjustment scheme, dialysis exchange frequency, single liquid volume setting, and various decision-making basis related to peritoneal transport characteristics, residual renal function, and liquid and electrolyte balance. The same kind of aggregation unit uses a clustering algorithm to divide the historical decision data into different groups according to the similar characteristics of the decision data, such as similar peritoneal transport types (high transporters or low transporters), similar residual renal function levels, similar liquid and electrolyte balance states, and similar dialysis prescriptions (such as similarities in dialysis types, dialysate glucose concentrations, etc.). Each group is an aggregated historical multi-source peritoneal dialysis prescription decision cluster. The decisions in the same cluster have high similarity in key features, while the decisions in different clusters have obvious differences. Through such same kind of aggregation operation, the core feature of each cluster is accurately determined for the subsequent center identification unit, thereby laying a solid foundation for constructing an aggregated historical multi-source peritoneal dialysis prescription decision pattern set, so that the system can more efficiently and accurately match appropriate peritoneal dialysis prescription decision patterns for the current patient.

[0068] The center identification unit iterates through each aggregated historical multi-source peritoneal dialysis prescription decision cluster to calculate the mean of all data points in the cluster to determine the center of each cluster. This center represents the typical features of all decisions in the cluster and is used as an independent aggregated historical multi-source peritoneal dialysis prescription decision pattern. Each aggregated historical multi-source peritoneal dialysis prescription decision pattern includes an aggregated monitoring feature vector, which is obtained by comprehensive processing of all monitoring feature vectors in the cluster. It integrates feature information related to peritoneal transport, electrolyte concentration, blood pressure, etc. Through the operation of the center identification unit, the originally dispersed aggregated historical multi-source peritoneal dialysis prescription decision clusters are transformed into a representative decision pattern set, which provides a basis for the subsequent aggregated monitoring feature vector matching unit to match the monitoring feature vector of the current patient with these patterns to find the most suitable peritoneal dialysis prescription decision pattern, and helps the system to more accurately develop a personalized dialysis plan for the patient.

[0069] The aggregation monitoring feature vector matching unit compares the monitoring feature vector of the current patient constructed by the system in the early stage with each of the aggregation monitoring feature vectors in the aggregation historical multi-source peritoneal dialysis prescription decision pattern set one by one. In the comparison process, a similarity calculation method such as a cosine similarity algorithm is used to measure the similarity between the two vectors by a quantitative value. For each aggregation monitoring feature vector, the similarity with the monitoring feature vector of the current patient is calculated. After the calculation and comparison of the similarity of all aggregation monitoring feature vectors, the aggregation historical multi-source peritoneal dialysis prescription decision pattern corresponding to the aggregation monitoring feature vector with the largest similarity value is found. This decision pattern is the peritoneal dialysis prescription decision pattern matched by the system for the current patient, which integrates the successful treatment experience and decision basis of similar patients in history, provides a reference for formulating a scientific and reasonable peritoneal dialysis prescription for the current patient, and helps to determine the key parameters such as the dialysate adjustment scheme, the exchange frequency and the single liquid capacity suitable for the patient, thereby improving the accuracy of peritoneal dialysis treatment.

[0070] In a possible implementation manner, the center identification unit further includes:

[0071] The peritoneal dialysis prescription decision set extraction sub-unit is configured to extract a first aggregation historical multi-source peritoneal dialysis prescription decision set from the aggregation historical multi-source peritoneal dialysis prescription decision cluster.

[0072] The prescription decision pattern extraction sub-unit is configured to randomly extract an initial aggregation historical multi-source peritoneal dialysis prescription decision pattern from the first aggregation historical multi-source peritoneal dialysis prescription decision set.

[0073] The initial center coefficient calculation sub-unit is configured to calculate an initial center coefficient of the initial aggregation historical multi-source peritoneal dialysis prescription decision pattern.

[0074] The prescription decision pattern iteration sub-unit is configured to randomly extract an iteration aggregation historical multi-source peritoneal dialysis prescription decision pattern from the first aggregation historical multi-source peritoneal dialysis prescription decision set again.

[0075] The iteration center coefficient calculation sub-unit is configured to calculate an iteration center coefficient of the iteration aggregation historical multi-source peritoneal dialysis prescription decision pattern.

[0076] The iteration center coefficient judgment subunit is configured to determine whether the iteration center coefficient is greater than or equal to the initial center coefficient. If yes, the direction of updating the initial aggregated historical multi-source peritoneal dialysis prescription decision pattern to the iteration aggregated historical multi-source peritoneal dialysis prescription decision pattern is taken as an iteration direction, and the iteration aggregated historical multi-source peritoneal dialysis prescription decision pattern is updated according to a preset iteration step size until a preset maximum iteration number is met, and a first aggregated historical multi-source peritoneal dialysis prescription decision pattern is obtained. The first aggregated historical multi-source peritoneal dialysis prescription decision pattern includes a first aggregated monitoring feature vector.

[0077] The prescription decision cluster center identification subunit is configured to identify a center of the aggregated historical multi-source peritoneal dialysis prescription decision cluster, and obtain an aggregated historical multi-source peritoneal dialysis prescription decision pattern set.

[0078] Specifically, when the same type aggregation operation on the historical multi-source peritoneal dialysis prescription decision set is completed, the aggregated historical multi-source peritoneal dialysis prescription decision cluster is formed, and a large amount of historical peritoneal dialysis prescription decision data with similar features is contained in the cluster. The function of the prescription decision set extraction subunit is to filter the aggregated decision cluster and extract a specific subset, i.e., the first aggregated historical multi-source peritoneal dialysis prescription decision set. The subset is not randomly selected, but determined by the preset filtering rule of the system, for example, filtering according to the standards of data integrity, sample representativeness, etc. Through this operation, the first aggregated historical multi-source peritoneal dialysis prescription decision set extracted contains more valuable and representative data of the core features of the aggregated cluster, which helps to improve the accuracy and efficiency of the whole system analysis and decision.

[0079] Since the first aggregated historical multi-source peritoneal dialysis prescription decision set contains a large number of historical peritoneal dialysis prescription decision data, in order to determine a representative decision pattern from the set, a random extraction method is used to randomly select a data point in the set as an initial aggregated historical multi-source peritoneal dialysis prescription decision pattern. This random selection method can avoid the bias caused by the fixed selection mode to a certain extent, and ensure the comprehensiveness and objectivity of the subsequent analysis.

[0080] The initial center coefficient calculation subunit calculates the Euclidean distance of the initial decision pattern and each decision pattern in the set in the monitoring feature vector space, and finally obtains a value that can reflect the relative central position of the initial decision pattern in the set, which is the initial center coefficient. The size of the initial center coefficient represents the overall closeness of the initial aggregated historical multi-source peritoneal dialysis prescription decision pattern to other decision patterns in the set. The larger the coefficient, the closer the initial decision pattern is to the center of the set. In the subsequent iterative optimization process, this coefficient will be used as one of the judgment criteria to help the system continuously find more representative aggregated historical multi-source peritoneal dialysis prescription decision patterns, thereby improving the accuracy of peritoneal dialysis prescription decision.

[0081] When the initial center coefficient calculation subunit in the system calculates the initial center coefficient of the initial aggregated historical multi-source peritoneal dialysis prescription decision pattern, in order to further find a decision pattern that can better represent the characteristics of the first aggregated historical multi-source peritoneal dialysis prescription decision set, the prescription decision pattern iteration subunit returns to the first aggregated historical multi-source peritoneal dialysis prescription decision set again, and selects a new data point as the iterative aggregated historical multi-source peritoneal dialysis prescription decision pattern using random extraction. This random extraction is to expand the search range and avoid falling into a local optimal solution. The newly extracted iterative decision pattern will be used as the basis for subsequent iterative optimization and compared and analyzed with the initial decision pattern. Through continuous iteration and updating, the system gradually finds a decision pattern that better meets the center characteristics of the set, thereby providing strong support for formulating more accurate and personalized peritoneal dialysis prescriptions.

[0082] When the prescription decision pattern iteration subunit randomly extracts the iterative aggregated historical multi-source peritoneal dialysis prescription decision pattern from the first aggregated historical multi-source peritoneal dialysis prescription decision set, a similar method is used to calculate the initial center coefficient to comprehensively consider the relationship between the iterative decision pattern and other decision patterns in the first aggregated historical multi-source peritoneal dialysis prescription decision set. Calculate the Euclidean distance of the iterative decision pattern and each decision pattern in the feature space, and finally obtain a value using the weighted average algorithm, which is the iterative center coefficient of the iterative aggregated historical multi-source peritoneal dialysis prescription decision pattern. This coefficient can reflect the relative central degree of the iterative decision pattern in the entire set, and will be compared with the initial center coefficient in the subsequent iterative judgment process to determine whether to update the decision pattern in a certain direction, thereby continuously approaching the most representative aggregated historical multi-source peritoneal dialysis prescription decision pattern and helping the system to formulate more accurate peritoneal dialysis prescriptions.

[0083] When the iterative centering coefficient of the iterative aggregated historical multi-source peritoneal dialysis prescription decision pattern is obtained, the iterative centering coefficient judgment subunit compares the iterative centering coefficient with the initial centering coefficient calculated by the initial centering coefficient calculation subunit. If the iterative centering coefficient is greater than or equal to the initial centering coefficient, it means that the iterative decision pattern has made positive progress in the degree of proximity to the set center compared with the initial decision pattern. At this time, the subunit will determine that the direction from the initial aggregated historical multi-source peritoneal dialysis prescription decision pattern to the iterative aggregated historical multi-source peritoneal dialysis prescription decision pattern is the iterative direction. Then, the iterative aggregated historical multi-source peritoneal dialysis prescription decision pattern is updated according to the preset iterative step size of the system. This updating process will continue to repeat, and the iterative centering coefficient is recalculated after each update and compared with the initial centering coefficient until the preset maximum iteration number is reached. After multiple iterations and updates, the final decision pattern obtained is the first aggregated historical multi-source peritoneal dialysis prescription decision pattern, which contains the first aggregated monitoring feature vector that comprehensively reflects the characteristics of the decision pattern in peritoneal transport, electrolyte concentration and other aspects, and can provide an important reference for formulating more accurate and more patient-specific peritoneal dialysis prescriptions.

[0084] When the same-class aggregation unit completes the clustering of the historical multi-source peritoneal dialysis prescription decision set, a plurality of aggregated historical multi-source peritoneal dialysis prescription decision clusters are formed, and these clusters contain similar prescription decision information. The prescription decision cluster center identification subunit will perform in-depth analysis on each cluster, calculate the mean of all data points in the cluster, and determine the center position of each cluster. The decision information corresponding to this center position is defined as the aggregated historical multi-source peritoneal dialysis prescription decision pattern. Through such center identification operation on all aggregated historical multi-source peritoneal dialysis prescription decision clusters, the aggregated historical multi-source peritoneal dialysis prescription decision pattern set is finally obtained.

[0085] In one possible implementation manner, the iterative centering coefficient judgment subunit further includes:

[0086] The iterative direction acquisition micro-unit is configured to, when the iterative centering coefficient is less than the initial centering coefficient, take the direction from the iterative aggregated historical multi-source peritoneal dialysis prescription decision pattern to the initial aggregated historical multi-source peritoneal dialysis prescription decision pattern as the iterative direction, update the initial aggregated historical multi-source peritoneal dialysis prescription decision pattern according to the preset iterative step size, until the preset maximum iteration number is satisfied, and obtain the first aggregated historical multi-source peritoneal dialysis prescription decision pattern.

[0087] Specifically, after the iterative center coefficient calculation subunit obtains the iterative center coefficient of the iterative aggregated historical multi-source peritoneal dialysis prescription decision pattern, and compares it with the initial center coefficient obtained by the initial center coefficient calculation subunit, if it is found that the iterative center coefficient is less than the initial center coefficient, it indicates that the decision pattern obtained by the current iteration does not develop in the direction closer to the set center. At this time, it is determined that the direction from the iterative aggregated historical multi-source peritoneal dialysis prescription decision pattern to the initial aggregated historical multi-source peritoneal dialysis prescription decision pattern is the new iteration direction. Then, according to the preset iteration step of the system, the initial aggregated historical multi-source peritoneal dialysis prescription decision pattern is updated in the direction of the newly determined iteration direction. Each update is an attempt, and the system will continuously evaluate whether the updated decision pattern meets the requirements. After each update, it is checked whether the preset maximum iteration number is reached. Only when the preset maximum iteration number is met, the entire iterative updating process will stop, and the first aggregated historical multi-source peritoneal dialysis prescription decision pattern is finally obtained. This decision pattern is optimized through multiple adjustments, and contains the typical characteristics of a large amount of historical decision data, including the first aggregated monitoring feature vector, which can provide an important reference for formulating accurate peritoneal dialysis prescriptions.

[0088] In one possible implementation manner, the iterative center coefficient judgment subunit further includes:

[0089] The vector set mean calculation micro-unit is configured to calculate the mean of the monitoring feature vector set corresponding to the first aggregated historical multi-source peritoneal dialysis prescription decision set, and obtain the first aggregated monitoring feature vector.

[0090] Specifically, when the peritoneal dialysis prescription decision set extraction subunit extracts the first aggregated historical multi-source peritoneal dialysis prescription decision set from the aggregated historical multi-source peritoneal dialysis prescription decision cluster, the set contains a series of data related to peritoneal dialysis prescription decisions, and each data corresponds to a monitoring feature vector. These vectors together form a monitoring feature vector set. The vector set mean calculation micro-unit adds the elements of the same dimension of each vector in the set and divides the result by the number of vectors to obtain the mean of the dimension. By calculating all dimensions in this way, a comprehensive vector, i.e., the first aggregated monitoring feature vector, is finally obtained. This first aggregated monitoring feature vector can reflect the overall characteristics and trends of the first aggregated historical multi-source peritoneal dialysis prescription decision set, and provides important basic data and reference for subsequent system construction, matching and ultimately formulating scientific and reasonable peritoneal dialysis prescriptions.

[0091] It should be noted that the above-mentioned embodiment sequences of the present application are merely for description only, but not for representing the advantages and disadvantages of the embodiments. And the above-mentioned embodiments of the present specification have been described. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0092] The above only describes the preferred embodiments of the present application, and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0093] The specification and drawings are merely exemplary of the present application, and any and all modifications, variations, combinations or equivalents that are within the scope of the present application should be included. Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalents, the present application is intended to include these modifications and variations.

Claims

1. A peritoneal dialysis prescription intelligent decision system, characterized in that, The system comprises: a data acquisition module for integrating a multi-source information system to obtain a glucose clearance rate sequence, a creatinine clearance rate sequence, an electrolyte concentration sequence and a blood pressure concentration sequence of a target patient in a preset window; a monitoring feature vector construction module for performing peritoneal dialysis monitoring data time series iterative multi-dimensional processing on the glucose clearance rate sequence, the creatinine clearance rate sequence, the electrolyte concentration sequence and the blood pressure concentration sequence to construct a monitoring feature vector; a schema matching module for performing multi-dimensional peritoneal dialysis prescription decision schema matching based on the monitoring feature vector to obtain a matched peritoneal dialysis prescription decision schema; a display module for displaying the matched peritoneal dialysis prescription decision schema on an explicit interface.

2. A peritoneal dialysis prescription intelligent decision system as claimed in claim 1, wherein, The data acquisition module further comprises: a glucose clearance rate sequence acquisition unit for extracting glucose concentration data of blood and peritoneal dialysis fluid monitored by a peritoneal dialysis monitoring device in the multi-source information system in a preset window to obtain a glucose clearance rate sequence; a creatinine clearance rate sequence acquisition unit for extracting synchronous creatinine concentration monitoring data of a portable dialysis analyzer and a portable blood analyzer in the multi-source information system in a preset window, and performing mapping analysis to obtain a creatinine clearance rate sequence; an electrolyte concentration sequence acquisition unit for extracting electrolyte concentration of a portable electrolyte analyzer in the multi-source information system in a preset window to obtain an electrolyte concentration sequence; a blood pressure concentration sequence acquisition unit for extracting blood pressure monitoring data of a blood pressure monitor in the multi-source information system in a preset window to obtain the blood pressure concentration sequence.

3. A peritoneal dialysis prescription intelligent decision system as claimed in claim 1, wherein, The monitoring feature vector construction module further comprises: a trend attention recognition unit for performing trend attention recognition on the glucose clearance rate sequence and the creatinine clearance rate sequence to obtain a peritoneal transport monitoring feature sub-vector; a trend feature enhancement analysis unit for performing trend feature enhancement analysis on the electrolyte concentration sequence and the voltage concentration sequence respectively to obtain an electrolyte concentration monitoring feature sub-vector and a voltage concentration monitoring feature sub-vector; a monitoring feature vector summarizing unit for summarizing the peritoneal transport monitoring feature sub-vector, the electrolyte concentration monitoring feature sub-vector and the voltage concentration monitoring feature sub-vector into the monitoring feature vector.

4. A peritoneal dialysis prescription intelligent decision system as claimed in claim 3, wherein, The trend attention recognition unit further comprises: a fluctuation trend recognition sub-unit for performing fluctuation trend recognition on the glucose clearance rate sequence and the creatinine clearance rate sequence respectively to obtain a glucose clearance rate fluctuation trend factor and a creatinine clearance rate fluctuation trend factor; an attention weight recognition sub-unit for performing attention weight recognition on the glucose clearance rate fluctuation trend factor and the creatinine clearance rate fluctuation trend factor to obtain a glucose clearance rate attention weight and a creatinine clearance rate attention weight; a trend feature fusion sub-unit for performing trend feature fusion on the glucose clearance rate sequence and the creatinine clearance rate sequence based on the glucose clearance rate attention weight and the creatinine clearance rate attention weight to obtain a peritoneal transport monitoring feature sub-vector.

5. A peritoneal dialysis prescription intelligent decision system as claimed in claim 4, wherein, The attention weight recognition sub-unit further comprises: A ratio calculation micro unit is configured to calculate a ratio of a glucose clearance rate fluctuation trend factor to a sum of the glucose clearance rate fluctuation trend factor and a creatinine clearance rate fluctuation trend factor, and obtain a glucose clearance rate attention weight; A creatinine clearance rate attention weight acquisition micro unit is configured to take a difference between 1 and the glucose clearance rate attention weight as the creatinine clearance rate attention weight.

6. A peritoneal dialysis prescription intelligent decision system as claimed in claim 3, wherein, The trend feature enhancement analysis unit further includes: A long-short trend feature identification subunit is configured to identify long-short trend features of the electrolyte concentration sequence, and obtain an electrolyte concentration change long trend feature subvector and an electrolyte concentration short trend feature subvector; A trend feature enhancement matrix construction subunit is configured to identify a trend feature approximation degree between elements in the electrolyte concentration change long trend feature subvector and the electrolyte concentration short trend feature subvector in a one-to-one manner, normalize the identification result, and fill the normalized result into an initially empty matrix to construct a trend feature enhancement matrix; A feature subvector acquisition subunit is configured to enhance the electrolyte concentration short trend feature subvector by using the trend feature enhancement matrix, and obtain an electrolyte concentration monitoring feature subvector.

7. A peritoneal dialysis prescription intelligent decision system as in claim 1, wherein, The prototype matching module further includes: A decision set extraction unit is configured to extract a historical multi-source peritoneal dialysis prescription decision set stored in the multi-source information system; A same-class aggregation unit is configured to aggregate the historical multi-source peritoneal dialysis prescription decision set in the same class, and obtain an aggregated historical multi-source peritoneal dialysis prescription decision cluster; A center identification unit is configured to traverse the aggregated historical multi-source peritoneal dialysis prescription decision cluster to identify a center, and obtain an aggregated historical multi-source peritoneal dialysis prescription decision prototype set, wherein each aggregated historical multi-source peritoneal dialysis prescription decision prototype includes an aggregated monitoring feature vector; An aggregated monitoring feature vector matching unit is configured to match the monitoring feature vector with each aggregated monitoring feature vector corresponding to the aggregated historical multi-source peritoneal dialysis prescription decision prototype set, and take an aggregated historical multi-source peritoneal dialysis prescription decision prototype corresponding to a maximum matching similarity as the matching peritoneal dialysis prescription decision prototype.

8. A peritoneal dialysis prescription intelligent decision system as in claim 7, wherein, The center identification unit further includes: A peritoneal dialysis prescription decision set extraction subunit is configured to extract a first aggregated historical multi-source peritoneal dialysis prescription decision set from the aggregated historical multi-source peritoneal dialysis prescription decision cluster; A prescription decision prototype extraction subunit is configured to randomly extract an initial aggregated historical multi-source peritoneal dialysis prescription decision prototype from the first aggregated historical multi-source peritoneal dialysis prescription decision set; An initial center coefficient calculation subunit is configured to calculate an initial center coefficient of the initial aggregated historical multi-source peritoneal dialysis prescription decision prototype; A prescription decision prototype iteration subunit is configured to randomly extract an iteration aggregated historical multi-source peritoneal dialysis prescription decision prototype from the first aggregated historical multi-source peritoneal dialysis prescription decision set again; An iteration center coefficient calculation subunit is configured to calculate an iteration center coefficient of the iteration aggregated historical multi-source peritoneal dialysis prescription decision prototype. The iteration center coefficient judgment subunit is configured to judge whether the iteration center coefficient is greater than or equal to the initial center coefficient. If yes, the direction from the initial aggregated historical multi-source peritoneal dialysis prescription decision pattern to the iteration aggregated historical multi-source peritoneal dialysis prescription decision pattern is taken as an iteration direction, the iteration aggregated historical multi-source peritoneal dialysis prescription decision pattern is updated according to a preset iteration step, until a preset maximum iteration number is met, and a first aggregated historical multi-source peritoneal dialysis prescription decision pattern is obtained. The first aggregated historical multi-source peritoneal dialysis prescription decision pattern includes a first aggregated monitoring feature vector. The prescription decision cluster center identification subunit is configured to identify a center of the aggregated historical multi-source peritoneal dialysis prescription decision cluster, and obtain an aggregated historical multi-source peritoneal dialysis prescription decision pattern set.

9. A peritoneal dialysis prescription intelligent decision system as in claim 8, wherein, The iteration center coefficient judgment subunit further includes: The iteration direction acquisition micro-unit is configured to, when the iteration center coefficient is less than the initial center coefficient, take a direction from the iteration aggregated historical multi-source peritoneal dialysis prescription decision pattern to the initial aggregated historical multi-source peritoneal dialysis prescription decision pattern as an iteration direction, update the initial aggregated historical multi-source peritoneal dialysis prescription decision pattern according to a preset iteration step, until a preset maximum iteration number is met, and obtain a first aggregated historical multi-source peritoneal dialysis prescription decision pattern.

10. A peritoneal dialysis prescription intelligent decision system as in claim 8, wherein, The iteration center coefficient judgment subunit further includes: The vector set mean value calculation micro-unit is configured to calculate a mean value of a monitoring feature vector set corresponding to the first aggregated historical multi-source peritoneal dialysis prescription decision set, and obtain a first aggregated monitoring feature vector.

Images